osmo-bsc/src/osmo-bsc/bsc_subscr_conn_fsm.c

1357 lines
42 KiB
C
Raw Normal View History

introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
/* (C) 2017 by Harald Welte <laforge@gnumonks.org>
* All Rights Reserved
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License as published by
* the Free Software Foundation; either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Affero General Public License for more details.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*
*/
#include <limits.h>
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
#include <osmocom/core/fsm.h>
#include <osmocom/core/logging.h>
#include <osmocom/gsm/gsm0808.h>
#include <osmocom/sigtran/sccp_sap.h>
#include <osmocom/gsm/gsm0808_utils.h>
#include <osmocom/bsc/debug.h>
#include <osmocom/bsc/a_reset.h>
#include <osmocom/bsc/gsm_08_08.h>
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
#include <osmocom/bsc/gsm_data.h>
#include <osmocom/bsc/bts.h>
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
#include <osmocom/bsc/handover_fsm.h>
#include <osmocom/bsc/lchan_fsm.h>
#include <osmocom/bsc/lchan_rtp_fsm.h>
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
#include <osmocom/bsc/bsc_subscriber.h>
#include <osmocom/bsc/osmo_bsc_sigtran.h>
#include <osmocom/bsc/osmo_bsc_lcls.h>
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
#include <osmocom/bsc/bsc_subscr_conn_fsm.h>
#include <osmocom/bsc/osmo_bsc.h>
#include <osmocom/bsc/penalty_timers.h>
#include <osmocom/bsc/bsc_rll.h>
#include <osmocom/bsc/abis_rsl.h>
#include <osmocom/core/tdef.h>
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
#include <osmocom/bsc/gsm_04_08_rr.h>
#include <osmocom/bsc/assignment_fsm.h>
#include <osmocom/bsc/codec_pref.h>
#include <osmocom/mgcp_client/mgcp_client_endpoint_fsm.h>
#include <osmocom/mgcp_client/mgcp_client_pool.h>
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
#include <osmocom/core/byteswap.h>
#include <osmocom/bsc/lb.h>
#include <osmocom/bsc/lcs_loc_req.h>
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
#define S(x) (1 << (x))
#define MGCP_MGW_TIMEOUT 4 /* in seconds */
#define MGCP_MGW_TIMEOUT_TIMER_NR 1
#define MGCP_MGW_HO_TIMEOUT 4 /* in seconds */
#define MGCP_MGW_HO_TIMEOUT_TIMER_NR 2
enum gscon_fsm_states {
ST_INIT,
/* wait for initial BSSMAP after the MSC opened a new SCCP connection */
ST_WAIT_INITIAL_USER_DATA,
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
/* waiting for CC from MSC */
ST_WAIT_CC,
/* active connection */
ST_ACTIVE,
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
ST_ASSIGNMENT,
ST_HANDOVER,
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
/* BSSMAP CLEAR has been received */
ST_WAIT_CLEAR_CMD,
ST_WAIT_SCCP_RLSD,
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
};
static const struct value_string gscon_fsm_event_names[] = {
{GSCON_EV_A_CONN_IND, "MT-CONNECT.ind"},
{GSCON_EV_A_INITIAL_USER_DATA, "A_INITIAL_USER_DATA"},
{GSCON_EV_MO_COMPL_L3, "MO_COMPL_L3"},
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
{GSCON_EV_A_CONN_CFM, "MO-CONNECT.cfm"},
{GSCON_EV_A_CLEAR_CMD, "CLEAR_CMD"},
{GSCON_EV_A_DISC_IND, "DISCONNET.ind"},
{GSCON_EV_A_COMMON_ID_IND, "COMMON_ID.ind"},
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
{GSCON_EV_ASSIGNMENT_START, "ASSIGNMENT_START"},
{GSCON_EV_ASSIGNMENT_END, "ASSIGNMENT_END"},
{GSCON_EV_HANDOVER_START, "HANDOVER_START"},
{GSCON_EV_HANDOVER_END, "HANDOVER_END"},
{GSCON_EV_RSL_CONN_FAIL, "RSL_CONN_FAIL"},
{GSCON_EV_MO_DTAP, "MO_DTAP"},
{GSCON_EV_MT_DTAP, "MT_DTAP"},
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
{GSCON_EV_TX_SCCP, "TX_SCCP"},
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
{GSCON_EV_MGW_MDCX_RESP_MSC, "MGW_MDCX_RESP_MSC"},
{GSCON_EV_LCLS_FAIL, "LCLS_FAIL"},
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
{GSCON_EV_FORGET_LCHAN, "FORGET_LCHAN"},
{GSCON_EV_FORGET_MGW_ENDPOINT, "FORGET_MGW_ENDPOINT"},
{GSCON_EV_LCS_LOC_REQ_END, "LCS_LOC_REQ_END"},
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
{}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
};
struct osmo_tdef_state_timeout conn_fsm_timeouts[32] = {
[ST_WAIT_INITIAL_USER_DATA] = { .T = -25 },
[ST_WAIT_CC] = { .T = -3210 },
[ST_WAIT_CLEAR_CMD] = { .T = -4 },
[ST_WAIT_SCCP_RLSD] = { .T = -4 },
};
/* Transition to a state, using the T timer defined in conn_fsm_timeouts.
* The actual timeout value is in turn obtained from network->T_defs.
* Assumes local variable 'conn' exists. */
#define conn_fsm_state_chg(state) \
osmo_tdef_fsm_inst_state_chg(conn->fi, state, \
conn_fsm_timeouts, \
conn->network->T_defs, \
-1)
/* forward MT DTAP from BSSAP side to RSL side */
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
static inline void submit_dtap(struct gsm_subscriber_connection *conn, struct msgb *msg)
{
OSMO_ASSERT(msg);
OSMO_ASSERT(conn);
gscon_submit_rsl_dtap(conn, msg, OBSC_LINKID_CB(msg), 1);
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
static void gscon_dtap_queue_flush(struct gsm_subscriber_connection *conn, int send);
int gscon_sigtran_send(struct gsm_subscriber_connection *conn, struct msgb *msg)
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
{
int rc;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (!msg)
return -ENOMEM;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
/* Make sure that we only attempt to send SCCP messages if we have
* a live SCCP connection. Otherwise drop the message. */
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (conn->fi->state == ST_INIT || conn->fi->state == ST_WAIT_CC) {
LOGPFSML(conn->fi, LOGL_ERROR, "No active SCCP connection, dropping message\n");
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
msgb_free(msg);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
return -ENODEV;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
rc = osmo_bsc_sigtran_send(conn, msg);
if (rc < 0)
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
LOGPFSML(conn->fi, LOGL_ERROR, "Unable to deliver SCCP message\n");
return rc;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
void gscon_bssmap_clear(struct gsm_subscriber_connection *conn, enum gsm0808_cause cause)
{
/* already clearing? */
switch (conn->fi->state) {
case ST_WAIT_CLEAR_CMD:
case ST_WAIT_SCCP_RLSD:
return;
default:
break;
}
conn->clear_cause = cause;
conn_fsm_state_chg(ST_WAIT_CLEAR_CMD);
}
static void gscon_fsm_wait_clear_cmd_onenter(struct osmo_fsm_inst *fi, uint32_t prev_state)
{
struct msgb *resp;
int rc;
struct gsm_subscriber_connection *conn = fi->priv;
enum gsm0808_cause cause = conn->clear_cause;
if (!conn->sccp.msc) {
LOGPFSML(fi, LOGL_ERROR, "Unable to deliver BSSMAP Clear Request message, no MSC for this conn\n");
goto nothing_sent;
}
LOGPFSML(fi, LOGL_DEBUG, "Tx BSSMAP CLEAR REQUEST(%s) to MSC\n", gsm0808_cause_name(cause));
resp = gsm0808_create_clear_rqst(cause);
if (!resp) {
LOGPFSML(fi, LOGL_ERROR, "Unable to compose BSSMAP Clear Request message\n");
goto nothing_sent;
}
rate_ctr_inc(rate_ctr_group_get_ctr(conn->sccp.msc->msc_ctrs, MSC_CTR_BSSMAP_TX_DT1_CLEAR_RQST));
rc = osmo_bsc_sigtran_send(conn, resp);
if (rc < 0) {
LOGPFSML(conn->fi, LOGL_ERROR, "Unable to deliver BSSMAP Clear Request message\n");
goto nothing_sent;
}
return;
nothing_sent:
/* Normally, we request a CLEAR from the MSC and terminate as soon as the CLEAR COMMAND has been issued by the
* MSC. But if we are trying to clear without being able to send anything to the MSC, we might as well shut down
* the conn right away now. */
conn_fsm_state_chg(ST_WAIT_SCCP_RLSD);
}
void gscon_fsm_wait_sccp_rlsd_onenter(struct osmo_fsm_inst *fi, uint32_t prev_state)
{
struct gsm_subscriber_connection *conn = fi->priv;
/* According to 3GPP 48.008 3.1.9.1. "The BSS need not wait for the radio channel
* release to be completed or for the guard timer to expire before returning the
* CLEAR COMPLETE message" */
if (!gscon_sigtran_send(conn, gsm0808_create_clear_complete()))
rate_ctr_inc(rate_ctr_group_get_ctr(conn->sccp.msc->msc_ctrs, MSC_CTR_BSSMAP_TX_DT1_CLEAR_COMPLETE));
/* Give the handover_fsm a chance to book this as handover success before tearing down everything,
* making it look like a sudden death failure. */
if (conn->ho.fi)
osmo_fsm_inst_dispatch(conn->ho.fi, HO_EV_CONN_RELEASING, NULL);
if (conn->lcs.loc_req)
osmo_fsm_inst_dispatch(conn->lcs.loc_req->fi, LCS_LOC_REQ_EV_CONN_CLEAR, NULL);
gscon_release_lchans(conn, true, bsc_gsm48_rr_cause_from_gsm0808_cause(conn->clear_cause));
osmo_mgcpc_ep_clear(conn->user_plane.mgw_endpoint);
/* If there is no SCCP connection at all, then no need to wait for an SCCP RLSD. */
if (!conn->sccp.msc || conn->sccp.state != SUBSCR_SCCP_ST_CONNECTED)
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_REGULAR, NULL);
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
/* forward MO DTAP from RSL side to BSSAP side */
static void forward_dtap(struct gsm_subscriber_connection *conn, struct msgb *msg, struct osmo_fsm_inst *fi)
{
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
struct msgb *resp = NULL;
OSMO_ASSERT(msg);
OSMO_ASSERT(conn);
resp = gsm0808_create_dtap(msg, OBSC_LINKID_CB(msg));
rate_ctr_inc(rate_ctr_group_get_ctr(conn->sccp.msc->msc_ctrs, MSC_CTR_BSSMAP_TX_DT1_DTAP));
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
gscon_sigtran_send(conn, resp);
}
/* Release an lchan in such a way that it doesn't fire events back to the conn. */
static void gscon_release_lchan(struct gsm_subscriber_connection *conn, struct gsm_lchan *lchan,
bool do_rr_release, bool err, enum gsm48_rr_cause cause_rr)
{
if (!lchan || !conn)
return;
if (lchan->conn == conn)
lchan_forget_conn(lchan);
if (conn->lchan == lchan)
conn->lchan = NULL;
if (conn->ho.fi && conn->ho.new_lchan == lchan)
conn->ho.new_lchan = NULL;
if (conn->assignment.new_lchan == lchan)
conn->assignment.new_lchan = NULL;
lchan_release(lchan, do_rr_release, err, cause_rr,
gscon_last_eutran_plmn(conn));
}
RR Channel Release: pass Cause code from BSSMAP Clear to the BTS In lchan.release, add 'cause_rr', and set RR Channel Release message's cause value to lchan.release.cause_rr. In lchan_release(), do not set lchan.release.rsl_error_cause to the RR cause value, these are unrelated. Store in new lchan.release.cause_rr instead. The rsl_error_cause is apparently only used for logging, except for one place in lchan_fsm_wait_activ_ack() that compares it to RSL_ERR_RCH_ALR_ACTV_ALLOC, so there should not be a functional difference by this fix. Propagate the BSSMAP Clear Command cause to the RR Channel Release: Add struct gscon_clear_cmd_data as event data for GSCON_EV_A_CLEAR_CMD -- so far it sent the is_csfb flag, add the gsm0808_cause; invoking the event happens in bssmap_handle_clear_cmd(). Adjust event handling in gscon_fsm_allstate(); there, pass the cause to gscon_release_lchans(). In gscon_release_lchans(), pass the cause to gscon_release_lchan(), and then lchan_release(), which sets the new lchan.release.cause_rr to the passed cause value. As soon as the lchan FSM enters the proper state, it calls gsm48_send_rr_release(). There, set the cause value in the encoded message to lchan.release.cause_rr. Interworking with osmo-msc: so far, osmo-msc fails to set the Clear Command cause code for normal release, it just passes 0 which amounts to GSM0808_CAUSE_RADIO_INTERFACE_MESSAGE_FAILURE. Before this patch, osmo-bsc always sent GSM48_RR_CAUSE_NORMAL in the RR Channel Release, and after this patch it will receive 0 == GSM0808_CAUSE_RADIO_INTERFACE_MESSAGE_FAILURE from osmo-msc and more accurately translate that to GSM48_RR_CAUSE_PROT_ERROR_UNSPC. This means in practice that we will now see an error cause in RR Channel Release instead of GSM48_RR_CAUSE_NORMAL when working with osmo-msc. For changing osmo-msc to send GSM0808_CAUSE_CALL_CONTROL instead (which translates to GSM48_RR_CAUSE_NORMAL), see OS#4664 and change-id I1347ed72ae7d7ea73a557b866e764819c5ef8c42 (osmo-msc). A test for this is in Ie6c99f28b610a67f2d59ec00b3541940e882251b (osmo-ttcn3-hacks). Related: SYS#4872 Change-Id: I734cc55c501d61bbdadee81a223b26f9df57f959
2020-07-09 13:54:40 +00:00
void gscon_release_lchans(struct gsm_subscriber_connection *conn, bool do_rr_release, enum gsm48_rr_cause cause_rr)
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
{
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (conn->ho.fi)
handover_end(conn, HO_RESULT_CONN_RELEASE);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
assignment_reset(conn);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
RR Channel Release: pass Cause code from BSSMAP Clear to the BTS In lchan.release, add 'cause_rr', and set RR Channel Release message's cause value to lchan.release.cause_rr. In lchan_release(), do not set lchan.release.rsl_error_cause to the RR cause value, these are unrelated. Store in new lchan.release.cause_rr instead. The rsl_error_cause is apparently only used for logging, except for one place in lchan_fsm_wait_activ_ack() that compares it to RSL_ERR_RCH_ALR_ACTV_ALLOC, so there should not be a functional difference by this fix. Propagate the BSSMAP Clear Command cause to the RR Channel Release: Add struct gscon_clear_cmd_data as event data for GSCON_EV_A_CLEAR_CMD -- so far it sent the is_csfb flag, add the gsm0808_cause; invoking the event happens in bssmap_handle_clear_cmd(). Adjust event handling in gscon_fsm_allstate(); there, pass the cause to gscon_release_lchans(). In gscon_release_lchans(), pass the cause to gscon_release_lchan(), and then lchan_release(), which sets the new lchan.release.cause_rr to the passed cause value. As soon as the lchan FSM enters the proper state, it calls gsm48_send_rr_release(). There, set the cause value in the encoded message to lchan.release.cause_rr. Interworking with osmo-msc: so far, osmo-msc fails to set the Clear Command cause code for normal release, it just passes 0 which amounts to GSM0808_CAUSE_RADIO_INTERFACE_MESSAGE_FAILURE. Before this patch, osmo-bsc always sent GSM48_RR_CAUSE_NORMAL in the RR Channel Release, and after this patch it will receive 0 == GSM0808_CAUSE_RADIO_INTERFACE_MESSAGE_FAILURE from osmo-msc and more accurately translate that to GSM48_RR_CAUSE_PROT_ERROR_UNSPC. This means in practice that we will now see an error cause in RR Channel Release instead of GSM48_RR_CAUSE_NORMAL when working with osmo-msc. For changing osmo-msc to send GSM0808_CAUSE_CALL_CONTROL instead (which translates to GSM48_RR_CAUSE_NORMAL), see OS#4664 and change-id I1347ed72ae7d7ea73a557b866e764819c5ef8c42 (osmo-msc). A test for this is in Ie6c99f28b610a67f2d59ec00b3541940e882251b (osmo-ttcn3-hacks). Related: SYS#4872 Change-Id: I734cc55c501d61bbdadee81a223b26f9df57f959
2020-07-09 13:54:40 +00:00
gscon_release_lchan(conn, conn->lchan, do_rr_release, false, cause_rr);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
}
static int validate_initial_user_data(struct osmo_fsm_inst *fi, struct msgb *msg)
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
{
struct bssmap_header *bs;
enum BSS_MAP_MSG_TYPE bssmap_type;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
msg->l3h = msgb_l2(msg);
if (!msgb_l3(msg)) {
LOGPFSML(fi, LOGL_ERROR, "internal error: no l3 in msg\n");
return -EINVAL;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (msgb_l3len(msg) < sizeof(*bs)) {
LOGPFSML(fi, LOGL_ERROR, "message too short for BSSMAP header (%u < %zu)\n",
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
msgb_l3len(msg), sizeof(*bs));
return -EINVAL;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
bs = (struct bssmap_header*)msgb_l3(msg);
if (msgb_l3len(msg) < (bs->length + sizeof(*bs))) {
LOGPFSML(fi, LOGL_ERROR,
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
"message too short for length indicated in BSSMAP header (%u < %u)\n",
msgb_l3len(msg), bs->length);
return -EINVAL;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
switch (bs->type) {
case BSSAP_MSG_BSS_MANAGEMENT:
break;
default:
LOGPFSML(fi, LOGL_ERROR,
"message type not allowed for initial BSSMAP: %s\n", gsm0808_bssap_name(bs->type));
return -EINVAL;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
msg->l4h = &msg->l3h[sizeof(*bs)];
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
/* Validate initial message type. See also BSC_Tests.TC_outbound_connect. */
bssmap_type = msg->l4h[0];
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
switch (bssmap_type) {
case BSS_MAP_MSG_HANDOVER_RQST:
case BSS_MAP_MSG_PERFORM_LOCATION_RQST:
return 0;
default:
LOGPFSML(fi, LOGL_ERROR, "No support for initial BSSMAP: %s: %s\n",
gsm0808_bssap_name(bs->type), gsm0808_bssmap_name(bssmap_type));
return -EINVAL;
}
}
static void handle_initial_user_data(struct osmo_fsm_inst *fi, struct msgb *msg)
{
struct gsm_subscriber_connection *conn = fi->priv;
struct bssmap_header *bs;
enum BSS_MAP_MSG_TYPE bssmap_type;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
/* validate_initial_user_data() must be called before this */
OSMO_ASSERT(msgb_l4(msg));
bs = msgb_l3(msg);
bssmap_type = msg->l4h[0];
/* FIXME: Extract optional IMSI and update FSM using osmo_fsm_inst_set_id() (OS#2969) */
LOGPFSML(fi, LOGL_DEBUG, "Rx initial BSSMAP: %s: %s\n", gsm0808_bssap_name(bs->type),
gsm0808_bssmap_name(bssmap_type));
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
switch (bssmap_type) {
case BSS_MAP_MSG_HANDOVER_RQST:
rate_ctr_inc(&conn->sccp.msc->msc_ctrs->ctr[MSC_CTR_BSSMAP_RX_DT1_HANDOVER_RQST]);
/* Inter-BSC incoming Handover Request, another BSS is handovering to us. */
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
handover_start_inter_bsc_in(conn, msg);
return;
case BSS_MAP_MSG_PERFORM_LOCATION_RQST:
rate_ctr_inc(&conn->sccp.msc->msc_ctrs->ctr[MSC_CTR_BSSMAP_RX_DT1_PERFORM_LOCATION_REQUEST]);
/* Location Services: MSC asks for location of an IDLE subscriber */
conn_fsm_state_chg(ST_ACTIVE);
lcs_loc_req_start(conn, msg);
return;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
default:
LOGPFSML(fi, LOGL_ERROR, "No support for initial BSSMAP: %s: %s\n",
gsm0808_bssap_name(bs->type), gsm0808_bssmap_name(bssmap_type));
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_REGULAR, NULL);
return;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
static void handle_sccp_n_connect(struct osmo_fsm_inst *fi, struct osmo_scu_prim *scu_prim)
{
struct gsm_subscriber_connection *conn = fi->priv;
struct msgb *msg = scu_prim->oph.msg;
/* Make sure the conn FSM will osmo_sccp_tx_disconn() on term */
conn->sccp.state = SUBSCR_SCCP_ST_CONNECTED;
msg->l3h = msgb_l2(msg);
/* If (BSSMAP) user data is included, validate it before accepting the connection */
if (msgb_l3(msg) && msgb_l3len(msg)) {
if (validate_initial_user_data(fi, msg)) {
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_REGULAR, NULL);
return;
}
}
/* accept the new conn. */
if (osmo_sccp_tx_conn_resp(conn->sccp.msc->a.sccp_user, scu_prim->u.connect.conn_id,
&scu_prim->u.connect.called_addr, NULL, 0)) {
LOGPFSML(fi, LOGL_ERROR, "Cannot send SCCP CONN RESP\n");
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_REGULAR, NULL);
return;
}
/* The initial user data may already be included in this N-Connect, or it may come later in a separate message.
* If it is already included, also go to ST_WAIT_INITIAL_USER_DATA now, so that we don't have to tend to two
* separate code paths doing the same thing (handling of HANDOVER_END). */
OSMO_ASSERT(conn_fsm_state_chg(ST_WAIT_INITIAL_USER_DATA) == 0);
/* It is usually a bad idea to continue using a fi after a state change, because the fi might terminate during
* the state change. In this case it is certain that the fi stays around for the initial user data. */
if (msgb_l3(msg) && msgb_l3len(msg)) {
handle_initial_user_data(fi, msg);
} else {
LOGPFSML(fi, LOGL_DEBUG, "N-Connect does not contain user data (no BSSMAP message included)\n");
}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
static void gscon_fsm_init(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
struct gsm_subscriber_connection *conn = fi->priv;
struct osmo_scu_prim *scu_prim = NULL;
struct msgb *msg = NULL;
int rc;
switch (event) {
case GSCON_EV_MO_COMPL_L3:
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
/* RLL ESTABLISH IND with initial L3 Message */
msg = data;
rc = osmo_bsc_sigtran_open_conn(conn, msg);
if (rc < 0) {
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_ERROR, NULL);
} else {
/* SCCP T(conn est) is 1-2 minutes, way too long. The MS will timeout
* using T3210 (20s), T3220 (5s) or T3230 (10s) */
conn_fsm_state_chg(ST_WAIT_CC);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
gscon_update_id(conn);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
case GSCON_EV_A_CONN_IND:
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
gscon_update_id(conn);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
scu_prim = data;
if (!conn->sccp.msc) {
LOGPFSML(fi, LOGL_NOTICE, "N-CONNECT.ind from unknown MSC %s\n",
osmo_sccp_addr_dump(&scu_prim->u.connect.calling_addr));
/* We cannot find a way to the sccp_user without the MSC, so we cannot
* use osmo_sccp_tx_disconn() :( */
//osmo_sccp_tx_disconn(conn->sccp.msc->a.sccp_user, scu_prim->u.connect.conn_id, &scu_prim->u.connect.called_addr, 0);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_REGULAR, NULL);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
return;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
handle_sccp_n_connect(fi, scu_prim);
break;
default:
OSMO_ASSERT(false);
}
}
static void gscon_fsm_wait_initial_user_data(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
struct gsm_subscriber_connection *conn = fi->priv;
struct msgb *msg = data;
enum handover_result ho_result;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
switch (event) {
case GSCON_EV_A_INITIAL_USER_DATA:
if (validate_initial_user_data(fi, msg)) {
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_REGULAR, NULL);
return;
}
handle_initial_user_data(fi, msg);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
break;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
case GSCON_EV_HANDOVER_END:
ho_result = HO_RESULT_ERROR;
if (data)
ho_result = *(enum handover_result*)data;
LOGPFSML(fi, LOGL_DEBUG, "Handover result: %s\n", handover_result_name(ho_result));
if (ho_result == HO_RESULT_OK) {
/* In this case the ho struct should still be populated. */
if (conn->ho.scope & HO_INTER_BSC_IN) {
/* Done with establishing a conn where we accept another BSC's MS via
* inter-BSC handover */
osmo_fsm_inst_state_chg(fi, ST_ACTIVE, 0, 0);
gscon_dtap_queue_flush(conn, 1);
return;
}
LOG_HO(conn, LOGL_ERROR,
"Conn is in state %s, the only accepted handover kind is inter-BSC incoming handover\n",
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
osmo_fsm_inst_state_name(conn->fi));
}
gscon_bssmap_clear(conn, GSM0808_CAUSE_EQUIPMENT_FAILURE);
return;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
default:
OSMO_ASSERT(false);
}
}
/* We've sent the CONNECTION.req to the SCCP provider and are waiting for CC from MSC */
static void gscon_fsm_wait_cc(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
struct gsm_subscriber_connection *conn = fi->priv;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
switch (event) {
case GSCON_EV_A_CONN_CFM:
/* MSC has confirmed the connection */
if (!conn->lchan) {
/* If associated lchan was released while we were waiting for the
confirmed connection, then instead simply drop the connection */
LOGPFSML(fi, LOGL_INFO,
"Connection confirmed but lchan was dropped previously, clearing conn\n");
gscon_bssmap_clear(conn, GSM0808_CAUSE_EQUIPMENT_FAILURE);
break;
}
/* We now change into the active state and wait there for further operations. */
conn_fsm_state_chg(ST_ACTIVE);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
/* if there's user payload, forward it just like EV_MT_DTAP */
/* FIXME: Question: if there's user payload attached to the CC, forward it like EV_MT_DTAP? */
break;
default:
OSMO_ASSERT(false);
}
}
/* We're on an active subscriber connection, passing DTAP back and forth */
static void gscon_fsm_active(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
struct gsm_subscriber_connection *conn = fi->priv;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
struct gsm_bts *bts;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
switch (event) {
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
case GSCON_EV_ASSIGNMENT_START:
bts = conn->lchan? conn->lchan->ts->trx->bts : NULL;
if (!bts) {
LOGPFSML(fi, LOGL_ERROR, "Cannot do assignment, no active BTS\n");
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
return;
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
/* Rely on assignment_fsm timeout */
osmo_fsm_inst_state_chg(fi, ST_ASSIGNMENT, 0, 0);
assignment_fsm_start(conn, bts, data);
return;
case GSCON_EV_HANDOVER_START:
/* Rely on handover_fsm timeout */
if (osmo_fsm_inst_state_chg(fi, ST_HANDOVER, 0, 0))
LOGPFSML(fi, LOGL_ERROR, "Cannot transition to HANDOVER state, discarding\n");
else
handover_start(data);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
case GSCON_EV_MO_DTAP:
forward_dtap(conn, (struct msgb *)data, fi);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
case GSCON_EV_MT_DTAP:
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
submit_dtap(conn, (struct msgb *)data);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
case GSCON_EV_TX_SCCP:
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
gscon_sigtran_send(conn, (struct msgb *)data);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
case GSCON_EV_LCS_LOC_REQ_END:
/* On the A-interface, there is nothing to do. If there still is an lchan, the conn should stay open. If
* not, it is up to the MSC to send a Clear Command.
* On the Lb-interface, tear down the SCCP connection. */
lb_close_conn(conn);
break;
case GSCON_EV_MO_COMPL_L3:
/* It is possible to have an A-interface conn already established without an lchan being active, during
* a Perform Location Request (LCS). */
/* RLL ESTABLISH IND with initial L3 Message */
gscon_sigtran_send(conn, (struct msgb*)data);
break;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
default:
OSMO_ASSERT(false);
}
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
static void gscon_fsm_assignment(struct osmo_fsm_inst *fi, uint32_t event, void *data)
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
{
struct gsm_subscriber_connection *conn = fi->priv;
switch (event) {
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
case GSCON_EV_ASSIGNMENT_END:
osmo_fsm_inst_state_chg(fi, ST_ACTIVE, 0, 0);
gscon_dtap_queue_flush(conn, 1);
return;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
case GSCON_EV_MO_DTAP:
forward_dtap(conn, (struct msgb *)data, fi);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
case GSCON_EV_MT_DTAP:
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
submit_dtap(conn, (struct msgb *)data);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
case GSCON_EV_TX_SCCP:
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
gscon_sigtran_send(conn, (struct msgb *)data);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
default:
OSMO_ASSERT(false);
}
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
static void gscon_fsm_handover(struct osmo_fsm_inst *fi, uint32_t event, void *data)
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
{
struct gsm_subscriber_connection *conn = fi->priv;
switch (event) {
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
case GSCON_EV_HANDOVER_END:
osmo_fsm_inst_state_chg(fi, ST_ACTIVE, 0, 0);
gscon_dtap_queue_flush(conn, 1);
return;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
case GSCON_EV_MO_DTAP:
forward_dtap(conn, (struct msgb *)data, fi);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
case GSCON_EV_MT_DTAP:
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
/* cache until handover is done */
submit_dtap(conn, (struct msgb *)data);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
case GSCON_EV_TX_SCCP:
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
gscon_sigtran_send(conn, (struct msgb *)data);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
default:
OSMO_ASSERT(false);
}
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
static bool same_mgw_info(const struct mgcp_conn_peer *a, const struct mgcp_conn_peer *b)
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
{
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (!a || !b)
return false;
if (a == b)
return true;
if (strcmp(a->addr, b->addr))
return false;
if (a->port != b->port)
return false;
if (a->call_id != b->call_id)
return false;
return true;
}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
/* Make sure a conn->user_plane.mgw_endpoint is allocated with the proper mgw endpoint name. For
* SCCPlite, pass in msc_assigned_cic the CIC received upon BSSMAP Assignment Command or BSSMAP Handover
* Request form the MSC (which is only stored in conn->user_plane after success). Ignored for AoIP. */
struct osmo_mgcpc_ep *gscon_ensure_mgw_endpoint(struct gsm_subscriber_connection *conn,
uint16_t msc_assigned_cic, struct gsm_lchan *for_lchan)
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
{
const char *epname;
struct mgcp_client *mgcp_client = NULL;
if (!conn) {
LOG_LCHAN(for_lchan, LOGL_ERROR, "no conn!\n");
return NULL;
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (conn->user_plane.mgw_endpoint)
return conn->user_plane.mgw_endpoint;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
if (gscon_is_sccplite(conn) || gscon_is_aoip(conn)) {
/* Get MGCP client from pool */
mgcp_client = mgcp_client_pool_get(conn->network->mgw.mgw_pool);
if (!mgcp_client) {
LOGPFSML(conn->fi, LOGL_ERROR,
"cannot ensure MGW endpoint -- no MGW configured, check configuration!\n");
conn->user_plane.mgw_endpoint = NULL;
return NULL;
}
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (gscon_is_sccplite(conn)) {
/* derive endpoint name from CIC on A interface side */
conn->user_plane.mgw_endpoint =
osmo_mgcpc_ep_alloc(conn->fi, GSCON_EV_FORGET_MGW_ENDPOINT,
mgcp_client,
conn->network->mgw.tdefs,
conn->fi->id,
"%x@%s", msc_assigned_cic,
mgcp_client_endpoint_domain(mgcp_client));
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
LOGPFSML(conn->fi, LOGL_DEBUG, "MGW endpoint name derived from CIC 0x%x: %s\n",
msc_assigned_cic, osmo_mgcpc_ep_name(conn->user_plane.mgw_endpoint));
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
} else if (gscon_is_aoip(conn)) {
if (is_ipaccess_bts(for_lchan->ts->trx->bts))
/* use dynamic RTPBRIDGE endpoint allocation in MGW */
epname = mgcp_client_rtpbridge_wildcard(mgcp_client);
else {
epname = mgcp_client_e1_epname(conn, mgcp_client, for_lchan->ts->e1_link.e1_nr,
for_lchan->ts->e1_link.e1_ts, 16,
for_lchan->ts->e1_link.e1_ts_ss*2);
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
conn->user_plane.mgw_endpoint =
osmo_mgcpc_ep_alloc(conn->fi, GSCON_EV_FORGET_MGW_ENDPOINT, mgcp_client,
conn->network->mgw.tdefs, conn->fi->id, "%s", epname);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
} else {
LOGPFSML(conn->fi, LOGL_ERROR, "Conn is neither SCCPlite nor AoIP!?\n");
return NULL;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
return conn->user_plane.mgw_endpoint;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
bool gscon_connect_mgw_to_msc(struct gsm_subscriber_connection *conn,
struct gsm_lchan *for_lchan,
const char *addr, uint16_t port,
struct osmo_fsm_inst *notify,
uint32_t event_success, uint32_t event_failure,
void *notify_data,
struct osmo_mgcpc_ep_ci **created_ci)
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
{
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
int rc;
struct osmo_mgcpc_ep_ci *ci;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
struct mgcp_conn_peer mgw_info;
enum mgcp_verb verb;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (created_ci)
*created_ci = NULL;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (gscon_is_sccplite(conn)) {
/* SCCPlite connection uses an MGW endpoint created by the MSC, so there is nothing to do
* here. */
if (notify)
osmo_fsm_inst_dispatch(notify, event_success, notify_data);
return true;
}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
mgw_info = (struct mgcp_conn_peer){
.port = port,
.call_id = conn->sccp.conn_id,
.ptime = 20,
.x_osmo_osmux_use = conn->assignment.req.use_osmux,
.x_osmo_osmux_cid = conn->assignment.req.osmux_cid,
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
};
mgcp_pick_codec(&mgw_info, for_lchan, false);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
rc = osmo_strlcpy(mgw_info.addr, addr, sizeof(mgw_info.addr));
if (rc <= 0 || rc >= sizeof(mgw_info.addr)) {
LOGPFSML(conn->fi, LOGL_ERROR, "Failed to compose MGW endpoint address for MGW -> MSC\n");
return false;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
ci = conn->user_plane.mgw_endpoint_ci_msc;
if (ci) {
const struct mgcp_conn_peer *prev_crcx_info = osmo_mgcpc_ep_ci_get_remote_rtp_info(ci);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (!conn->user_plane.mgw_endpoint) {
LOGPFSML(conn->fi, LOGL_ERROR, "Internal error: conn has a CI but no endpoint\n");
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
return false;
}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (!prev_crcx_info) {
LOGPFSML(conn->fi, LOGL_ERROR, "There already is an MGW connection for the MSC side,"
" but it seems to be broken. Will not CRCX another one (%s)\n",
osmo_mgcpc_ep_ci_name(ci));
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
return false;
}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (same_mgw_info(&mgw_info, prev_crcx_info)) {
LOGPFSML(conn->fi, LOGL_DEBUG,
"MSC side MGW endpoint ci is already configured to %s\n",
osmo_mgcpc_ep_ci_name(ci));
/* Immediately dispatch the success event */
osmo_fsm_inst_dispatch(notify, event_success, notify_data);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
return true;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
verb = MGCP_VERB_MDCX;
} else
verb = MGCP_VERB_CRCX;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
gscon_ensure_mgw_endpoint(conn, for_lchan->activate.info.msc_assigned_cic, for_lchan);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (!conn->user_plane.mgw_endpoint) {
LOGPFSML(conn->fi, LOGL_ERROR, "Unable to allocate endpoint info\n");
return false;
}
if (!ci) {
ci = osmo_mgcpc_ep_ci_add(conn->user_plane.mgw_endpoint, "to-MSC");
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (created_ci)
*created_ci = ci;
conn->user_plane.mgw_endpoint_ci_msc = ci;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (!ci) {
LOGPFSML(conn->fi, LOGL_ERROR, "Unable to allocate endpoint CI info\n");
return false;
}
osmo_mgcpc_ep_ci_request(ci, verb, &mgw_info, notify, event_success, event_failure, notify_data);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
return true;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
#define EV_TRANSPARENT_SCCP S(GSCON_EV_TX_SCCP) | S(GSCON_EV_MO_DTAP) | S(GSCON_EV_MT_DTAP)
static const struct osmo_fsm_state gscon_fsm_states[] = {
[ST_INIT] = {
.name = "INIT",
.in_event_mask = S(GSCON_EV_MO_COMPL_L3) | S(GSCON_EV_A_CONN_IND),
.out_state_mask = 0
| S(ST_WAIT_INITIAL_USER_DATA)
| S(ST_WAIT_CC) | S(ST_ACTIVE) | S(ST_WAIT_CLEAR_CMD) | S(ST_WAIT_SCCP_RLSD),
.action = gscon_fsm_init,
},
[ST_WAIT_INITIAL_USER_DATA] = {
.name = "WAIT_INITIAL_USER_DATA",
.in_event_mask = 0
| S(GSCON_EV_A_INITIAL_USER_DATA)
| S(GSCON_EV_HANDOVER_END)
,
.out_state_mask = S(ST_WAIT_CC) | S(ST_ACTIVE) | S(ST_WAIT_CLEAR_CMD) | S(ST_WAIT_SCCP_RLSD),
.action = gscon_fsm_wait_initial_user_data,
},
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
[ST_WAIT_CC] = {
.name = "WAIT_CC",
.in_event_mask = S(GSCON_EV_A_CONN_CFM),
.out_state_mask = S(ST_ACTIVE) | S(ST_WAIT_CLEAR_CMD) | S(ST_WAIT_SCCP_RLSD),
.action = gscon_fsm_wait_cc,
},
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
[ST_ACTIVE] = {
.name = "ACTIVE",
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
.in_event_mask = EV_TRANSPARENT_SCCP | S(GSCON_EV_ASSIGNMENT_START) |
S(GSCON_EV_HANDOVER_START)
| S(GSCON_EV_LCS_LOC_REQ_END)
| S(GSCON_EV_MO_COMPL_L3)
,
.out_state_mask = S(ST_WAIT_CLEAR_CMD) | S(ST_WAIT_SCCP_RLSD) | S(ST_ASSIGNMENT) |
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
S(ST_HANDOVER),
.action = gscon_fsm_active,
},
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
[ST_ASSIGNMENT] = {
.name = "ASSIGNMENT",
.in_event_mask = EV_TRANSPARENT_SCCP | S(GSCON_EV_ASSIGNMENT_END),
.out_state_mask = S(ST_ACTIVE) | S(ST_WAIT_CLEAR_CMD) | S(ST_WAIT_SCCP_RLSD),
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
.action = gscon_fsm_assignment,
},
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
[ST_HANDOVER] = {
.name = "HANDOVER",
.in_event_mask = EV_TRANSPARENT_SCCP | S(GSCON_EV_HANDOVER_END),
.out_state_mask = S(ST_ACTIVE) | S(ST_WAIT_CLEAR_CMD) | S(ST_WAIT_SCCP_RLSD),
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
.action = gscon_fsm_handover,
},
[ST_WAIT_CLEAR_CMD] = {
.name = "WAIT_CLEAR_CMD",
.onenter = gscon_fsm_wait_clear_cmd_onenter,
.out_state_mask = S(ST_WAIT_SCCP_RLSD),
},
[ST_WAIT_SCCP_RLSD] = {
.name = "WAIT_SCCP_RLSD",
.onenter = gscon_fsm_wait_sccp_rlsd_onenter,
.in_event_mask = S(GSCON_EV_HANDOVER_END),
},
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
};
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
void gscon_change_primary_lchan(struct gsm_subscriber_connection *conn, struct gsm_lchan *new_lchan)
{
/* On release, do not receive release events that look like the primary lchan is gone. */
struct gsm_lchan *old_lchan = conn->lchan;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
OSMO_ASSERT(new_lchan);
if (old_lchan == new_lchan)
return;
if (!old_lchan)
LOGPFSML(conn->fi, LOGL_DEBUG, "setting primary lchan for this conn to %s\n",
new_lchan->fi? osmo_fsm_inst_name(new_lchan->fi) : gsm_lchan_name(new_lchan));
else
LOGPFSML(conn->fi, LOGL_DEBUG, "primary lchan for this conn changes from %s to %s\n",
old_lchan->fi? osmo_fsm_inst_name(old_lchan->fi) : gsm_lchan_name(old_lchan),
new_lchan->fi? osmo_fsm_inst_name(new_lchan->fi) : gsm_lchan_name(new_lchan));
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
conn->lchan = new_lchan;
conn->lchan->conn = conn;
if (conn->lchan->fi_rtp)
osmo_fsm_inst_dispatch(conn->lchan->fi_rtp, LCHAN_RTP_EV_ESTABLISHED, 0);
if (old_lchan && (old_lchan != new_lchan))
gscon_release_lchan(conn, old_lchan, false, false, GSM48_RR_CAUSE_NORMAL);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
}
void gscon_lchan_releasing(struct gsm_subscriber_connection *conn, struct gsm_lchan *lchan)
{
if (!lchan)
return;
if (conn->assignment.new_lchan == lchan) {
if (conn->assignment.fi)
osmo_fsm_inst_dispatch(conn->assignment.fi, ASSIGNMENT_EV_LCHAN_ERROR, lchan);
lchan_forget_conn(conn->assignment.new_lchan);
conn->assignment.new_lchan = NULL;
}
if (conn->ho.new_lchan == lchan) {
if (conn->ho.fi)
osmo_fsm_inst_dispatch(conn->ho.fi, HO_EV_LCHAN_ERROR, lchan);
}
if (conn->lchan == lchan) {
lchan_forget_conn(conn->lchan);
conn->lchan = NULL;
}
/* If the conn has no lchan anymore, it was released by the BTS and needs to Clear towards MSC.
* However, if a Location Request is still busy, do not send Clear Request. */
if (!conn->lchan && !conn->lcs.loc_req) {
switch (conn->fi->state) {
case ST_WAIT_CC:
/* The SCCP connection was not yet confirmed by a CC, the BSSAP is not fully established
yet so we cannot release it. First wait for the CC, and release in gscon_fsm_wait_cc(). */
break;
default:
gscon_bssmap_clear(conn, GSM0808_CAUSE_EQUIPMENT_FAILURE);
break;
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
}
}
/* An lchan was deallocated. */
void gscon_forget_lchan(struct gsm_subscriber_connection *conn, struct gsm_lchan *lchan)
{
const char *detach_label = NULL;
if (!conn)
return;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (!lchan)
return;
if (conn->assignment.new_lchan == lchan) {
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
conn->assignment.new_lchan = NULL;
detach_label = "assignment.new_lchan";
}
if (conn->ho.new_lchan == lchan) {
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
conn->ho.new_lchan = NULL;
detach_label = "ho.new_lchan";
}
if (conn->lchan == lchan) {
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
conn->lchan = NULL;
detach_label = "primary lchan";
}
/* Log for both lchan FSM and conn FSM to ease reading the log in case of problems */
if (detach_label) {
LOGPFSML(conn->fi, LOGL_DEBUG, "conn detaches lchan %s\n",
lchan->fi? osmo_fsm_inst_name(lchan->fi) : gsm_lchan_name(lchan));
if (lchan->fi)
LOGPFSML(lchan->fi, LOGL_DEBUG, "conn %s detaches lchan (%s)\n",
osmo_fsm_inst_name(conn->fi), detach_label);
}
if (!conn->lchan
&& !conn->ho.new_lchan
&& !conn->assignment.new_lchan
&& !conn->lcs.loc_req)
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
gscon_bssmap_clear(conn, GSM0808_CAUSE_EQUIPMENT_FAILURE);
}
static void gscon_forget_mgw_endpoint(struct gsm_subscriber_connection *conn)
{
struct mgcp_client *mgcp_client;
/* Put MGCP client back into MGW pool */
mgcp_client = osmo_mgcpc_ep_client(conn->user_plane.mgw_endpoint);
mgcp_client_pool_put(mgcp_client);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
conn->user_plane.mgw_endpoint = NULL;
conn->user_plane.mgw_endpoint_ci_msc = NULL;
conn->ho.created_ci_for_msc = NULL;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
lchan_forget_mgw_endpoint(conn->lchan);
lchan_forget_mgw_endpoint(conn->assignment.new_lchan);
lchan_forget_mgw_endpoint(conn->ho.new_lchan);
}
void gscon_forget_mgw_endpoint_ci(struct gsm_subscriber_connection *conn, struct osmo_mgcpc_ep_ci *ci)
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
{
if (conn->ho.created_ci_for_msc == ci)
conn->ho.created_ci_for_msc = NULL;
if (conn->user_plane.mgw_endpoint_ci_msc == ci)
conn->user_plane.mgw_endpoint_ci_msc = NULL;
if (conn->assignment.created_ci_for_msc == ci)
conn->assignment.created_ci_for_msc = NULL;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
static void gscon_fsm_allstate(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
struct gsm_subscriber_connection *conn = fi->priv;
const struct tlv_parsed *tp;
struct osmo_mobile_identity mi_imsi;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
/* Regular allstate event processing */
switch (event) {
case GSCON_EV_A_CLEAR_CMD:
RR Channel Release: pass Cause code from BSSMAP Clear to the BTS In lchan.release, add 'cause_rr', and set RR Channel Release message's cause value to lchan.release.cause_rr. In lchan_release(), do not set lchan.release.rsl_error_cause to the RR cause value, these are unrelated. Store in new lchan.release.cause_rr instead. The rsl_error_cause is apparently only used for logging, except for one place in lchan_fsm_wait_activ_ack() that compares it to RSL_ERR_RCH_ALR_ACTV_ALLOC, so there should not be a functional difference by this fix. Propagate the BSSMAP Clear Command cause to the RR Channel Release: Add struct gscon_clear_cmd_data as event data for GSCON_EV_A_CLEAR_CMD -- so far it sent the is_csfb flag, add the gsm0808_cause; invoking the event happens in bssmap_handle_clear_cmd(). Adjust event handling in gscon_fsm_allstate(); there, pass the cause to gscon_release_lchans(). In gscon_release_lchans(), pass the cause to gscon_release_lchan(), and then lchan_release(), which sets the new lchan.release.cause_rr to the passed cause value. As soon as the lchan FSM enters the proper state, it calls gsm48_send_rr_release(). There, set the cause value in the encoded message to lchan.release.cause_rr. Interworking with osmo-msc: so far, osmo-msc fails to set the Clear Command cause code for normal release, it just passes 0 which amounts to GSM0808_CAUSE_RADIO_INTERFACE_MESSAGE_FAILURE. Before this patch, osmo-bsc always sent GSM48_RR_CAUSE_NORMAL in the RR Channel Release, and after this patch it will receive 0 == GSM0808_CAUSE_RADIO_INTERFACE_MESSAGE_FAILURE from osmo-msc and more accurately translate that to GSM48_RR_CAUSE_PROT_ERROR_UNSPC. This means in practice that we will now see an error cause in RR Channel Release instead of GSM48_RR_CAUSE_NORMAL when working with osmo-msc. For changing osmo-msc to send GSM0808_CAUSE_CALL_CONTROL instead (which translates to GSM48_RR_CAUSE_NORMAL), see OS#4664 and change-id I1347ed72ae7d7ea73a557b866e764819c5ef8c42 (osmo-msc). A test for this is in Ie6c99f28b610a67f2d59ec00b3541940e882251b (osmo-ttcn3-hacks). Related: SYS#4872 Change-Id: I734cc55c501d61bbdadee81a223b26f9df57f959
2020-07-09 13:54:40 +00:00
OSMO_ASSERT(data);
conn->clear_cause = *(const enum gsm0808_cause *)data;
conn_fsm_state_chg(ST_WAIT_SCCP_RLSD);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
case GSCON_EV_A_DISC_IND:
/* MSC or SIGTRAN network has hard-released SCCP connection, terminate the FSM now.
* Cleanup is done in gscon_pre_term() and gscon_cleanup(). */
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_REGULAR, data);
break;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
case GSCON_EV_FORGET_MGW_ENDPOINT:
gscon_forget_mgw_endpoint(conn);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
case GSCON_EV_RSL_CONN_FAIL:
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (conn->lchan) {
conn->lchan->release.in_error = true;
conn->lchan->release.rsl_error_cause = data ? *(uint8_t*)data : RSL_ERR_IE_ERROR;
conn->lchan->release.rr_cause =
bsc_gsm48_rr_cause_from_rsl_cause(conn->lchan->release.rsl_error_cause);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
}
/* Request BSSMAP Clear, but do not abort an ongoing Location Request */
if (!conn->lcs.loc_req)
gscon_bssmap_clear(conn, GSM0808_CAUSE_RADIO_INTERFACE_FAILURE);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
case GSCON_EV_MGW_MDCX_RESP_MSC:
LOGPFSML(fi, LOGL_DEBUG, "Rx MDCX of MSC side (LCLS?)\n");
break;
case GSCON_EV_LCLS_FAIL:
break;
case GSCON_EV_A_COMMON_ID_IND:
OSMO_ASSERT(data);
tp = data;
if (osmo_mobile_identity_decode(&mi_imsi, TLVP_VAL(tp, GSM0808_IE_IMSI), TLVP_LEN(tp, GSM0808_IE_IMSI), false)
|| mi_imsi.type != GSM_MI_TYPE_IMSI) {
LOGPFSML(fi, LOGL_ERROR, "CommonID: could not parse IMSI\n");
return;
}
if (!conn->bsub)
conn->bsub = bsc_subscr_find_or_create_by_imsi(conn->network->bsc_subscribers, mi_imsi.imsi,
BSUB_USE_CONN);
else {
/* we already have a bsc_subscr associated; maybe that subscriber has no IMSI yet? */
if (!conn->bsub->imsi[0])
bsc_subscr_set_imsi(conn->bsub, mi_imsi.imsi);
}
if (TLVP_PRESENT(tp, GSM0808_IE_LAST_USED_EUTRAN_PLMN_ID)) {
conn->fast_return.allowed = true; /* Always allowed for CSFB */
conn->fast_return.last_eutran_plmn_valid = true;
osmo_plmn_from_bcd(TLVP_VAL(tp, GSM0808_IE_LAST_USED_EUTRAN_PLMN_ID), &conn->fast_return.last_eutran_plmn);
LOGPFSML(fi, LOGL_DEBUG, "subscr comes from E-UTRAN PLMN %s\n",
osmo_plmn_name(&conn->fast_return.last_eutran_plmn));
}
gscon_update_id(conn);
break;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
default:
OSMO_ASSERT(false);
break;
}
}
static void gscon_cleanup(struct osmo_fsm_inst *fi, enum osmo_fsm_term_cause cause)
{
struct gsm_subscriber_connection *conn = fi->priv;
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
lchan_forget_conn(conn->lchan);
lchan_forget_conn(conn->assignment.new_lchan);
lchan_forget_conn(conn->ho.new_lchan);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
lb_close_conn(conn);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
if (conn->sccp.state != SUBSCR_SCCP_ST_NONE) {
LOGPFSML(fi, LOGL_DEBUG, "Disconnecting SCCP\n");
struct bsc_msc_data *msc = conn->sccp.msc;
/* FIXME: include a proper cause value / error message? */
osmo_sccp_tx_disconn(msc->a.sccp_user, conn->sccp.conn_id, &msc->a.bsc_addr, 0);
conn->sccp.state = SUBSCR_SCCP_ST_NONE;
}
if (conn->bsub) {
LOGPFSML(fi, LOGL_DEBUG, "Putting bsc_subscr\n");
bsc_subscr_put(conn->bsub, BSUB_USE_CONN);
conn->bsub = NULL;
}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
llist_del(&conn->entry);
talloc_free(conn);
}
static void gscon_pre_term(struct osmo_fsm_inst *fi, enum osmo_fsm_term_cause cause)
{
struct gsm_subscriber_connection *conn = fi->priv;
struct mgcp_client *mgcp_client;
/* Put MGCP client back into MGW pool */
mgcp_client = osmo_mgcpc_ep_client(conn->user_plane.mgw_endpoint);
mgcp_client_pool_put(mgcp_client);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
osmo_mgcpc_ep_clear(conn->user_plane.mgw_endpoint);
conn->user_plane.mgw_endpoint = NULL;
conn->user_plane.mgw_endpoint = NULL;
if (conn->ho.fi)
osmo_fsm_inst_dispatch(conn->ho.fi, HO_EV_CONN_RELEASING, NULL);
if (conn->lcs.loc_req)
osmo_fsm_inst_dispatch(conn->lcs.loc_req->fi, LCS_LOC_REQ_EV_CONN_CLEAR, NULL);
if (conn->lcls.fi) {
/* request termination of LCLS FSM */
osmo_fsm_inst_term(conn->lcls.fi, cause, NULL);
conn->lcls.fi = NULL;
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
LOGPFSML(fi, LOGL_DEBUG, "Releasing all lchans (if any) because this conn is terminating\n");
gscon_release_lchans(conn, true, bsc_gsm48_rr_cause_from_gsm0808_cause(conn->clear_cause));
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
/* drop pending messages */
gscon_dtap_queue_flush(conn, 0);
penalty_timers_clear(&conn->hodec2.penalty_timers, NULL);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
}
static int gscon_timer_cb(struct osmo_fsm_inst *fi)
{
struct gsm_subscriber_connection *conn = fi->priv;
switch (fi->T) {
case -3210:
gscon_release_lchan(conn, conn->lchan, true, true, GSM48_RR_CAUSE_ABNORMAL_TIMER);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
/* MSC has not responded/confirmed connection with CC, this
* could indicate a bad SCCP connection. We now inform the the
* FSM that controls the BSSMAP reset about the event. Maybe
* a BSSMAP reset is necessary. */
a_reset_conn_fail(conn->sccp.msc);
/* Since we could not reach the MSC, we give up and terminate
* the FSM instance now (N-DISCONNET.req is sent in
* gscon_cleanup() above) */
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_REGULAR, NULL);
break;
case -4:
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
/* The MSC has sent a BSSMAP Clear Command, we acknowledged that, but the conn was never
* disconnected. */
LOGPFSML(fi, LOGL_ERROR, "Long after expecting %s, the conn is still not"
" released. For sanity, discarding this conn now.\n",
fi->state == ST_WAIT_CLEAR_CMD ? "BSSMAP Clear Command" : "SCCP RLSD");
a_reset_conn_fail(conn->sccp.msc);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_ERROR, NULL);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
break;
default:
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
LOGPFSML(fi, LOGL_ERROR, "Unknown timer %d expired\n", fi->T);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
OSMO_ASSERT(false);
}
return 0;
}
static struct osmo_fsm gscon_fsm = {
.name = "SUBSCR_CONN",
.states = gscon_fsm_states,
.num_states = ARRAY_SIZE(gscon_fsm_states),
.allstate_event_mask = S(GSCON_EV_A_DISC_IND) | S(GSCON_EV_A_CLEAR_CMD) | S(GSCON_EV_A_COMMON_ID_IND) |
S(GSCON_EV_RSL_CONN_FAIL) |
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
S(GSCON_EV_LCLS_FAIL) |
S(GSCON_EV_FORGET_LCHAN) |
S(GSCON_EV_FORGET_MGW_ENDPOINT),
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
.allstate_action = gscon_fsm_allstate,
.cleanup = gscon_cleanup,
.pre_term = gscon_pre_term,
.timer_cb = gscon_timer_cb,
.log_subsys = DMSC,
.event_names = gscon_fsm_event_names,
};
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
void bsc_subscr_conn_fsm_init()
{
OSMO_ASSERT(osmo_fsm_register(&gscon_fsm) == 0);
OSMO_ASSERT(osmo_fsm_register(&lcls_fsm) == 0);
}
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
/* Allocate a subscriber connection and its associated FSM */
struct gsm_subscriber_connection *bsc_subscr_con_allocate(struct gsm_network *net)
{
struct gsm_subscriber_connection *conn;
conn = talloc_zero(net, struct gsm_subscriber_connection);
if (!conn)
return NULL;
conn->network = net;
INIT_LLIST_HEAD(&conn->dtap_queue);
INIT_LLIST_HEAD(&conn->hodec2.penalty_timers);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
conn->sccp.conn_id = -1;
/* Default clear cause (on RR translates to GSM48_RR_CAUSE_ABNORMAL_UNSPEC) */
conn->clear_cause = GSM0808_CAUSE_EQUIPMENT_FAILURE;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
/* don't allocate from 'conn' context, as gscon_cleanup() will call talloc_free(conn) before
* libosmocore will call talloc_free(conn->fi), i.e. avoid use-after-free during cleanup */
conn->fi = osmo_fsm_inst_alloc(&gscon_fsm, net, conn, LOGL_DEBUG, NULL);
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
if (!conn->fi) {
talloc_free(conn);
return NULL;
}
/* indicate "IE not [yet] received" */
conn->lcls.config = GSM0808_LCLS_CFG_NA;
conn->lcls.control = GSM0808_LCLS_CSC_NA;
conn->lcls.fi = osmo_fsm_inst_alloc_child(&lcls_fsm, conn->fi, GSCON_EV_LCLS_FAIL);
if (!conn->lcls.fi) {
osmo_fsm_inst_term(conn->fi, OSMO_FSM_TERM_ERROR, NULL);
return NULL;
}
conn->lcls.fi->priv = conn;
introduce an osmo_fsm for gsm_subscriber_connection In the current implementation of osmo-bsc, the subscriber connection is not handled (very) statefully. However, there is some state keeping in the code that handles the mgcp connection, but there are still to much loose ends which allow odd situations to happen, which then lead severe error situations (see also closes tags at the end) This commit adds a number of improvements to fix those problems. - Use an osmo-fsm to control the gsm_subscriber_connection state and make sure that certain operations can only take place at certain states (e.g let connection oriented SCCP traffic only pass when an SCCP connection actually exists. Remove the old osmo_bsc_mgcp.c code. Use the recently developed MGCP client FSM to handle the MGCP connections. Also make sure that stuff that already works does not break. This in particular refers to the internal handover capability and the respective unit-tests. See also OS#2823, OS#2768 and OS#2898 - Fix logic to permit assignment to a signalling channel. (OS#2762) - Introduce T993210 to release lchan + subscr_conn if MSC fails to respond The GSM specs don't have an explicit timer for this, so let's introdcue a custom timer (hence starting with 99). This timeout catches the following situation: * we send a SCCP CR with COMPL_L3_INFO from the MS to the MSC, * the MSC doesn't respond (e.g. SCCP routing failure, program down, ...) The MS is supposed to timeout with T3210, 3220 or 3230. But the BSC shouldn't trust the MS but have some timer on its own. SCCP would have a timer T(conn est), but that one is specified to be 1-2min and hence rather long. See also: OS#2775 - Terminate bsc_subscr_conn_fsm on SCCP N-DISC.ind from MSC If the MSC is disconnecting the SCCP channel, we must terminate the FSM which in turn will release all lchan's and other state. This makes TC_chan_rel_hard_rlsd pass, see also OS#2731 As a side-effect, this fixes TC_chan_act_ack_est_ind_refused(), where the MSC is answering with CREF to our CR/COMPL_L3. - Release subscriber connection on RLL RELEASE IND of SAPI0 on main DCCH The subscriber connection isn't really useful for anything after the SAPI0 main signalling link has been released. We could try to re-establish, but our best option is probably simply releasing the subscriber_conn and anything related to it. This will make TC_chan_rel_rll_rel_ind pass, see also OS#2730 This commit has been tested using the BSC_Tests TTCN3 testsuit and the following tests were passed: TC_chan_act_noreply TC_chan_act_ack_noest TC_chan_act_ack_est_ind_noreply TC_chan_act_ack_est_ind_refused TC_chan_act_nack TC_chan_exhaustion TC_ctrl TC_chan_rel_conn_fail TC_chan_rel_hard_clear TC_chan_rel_hard_rlsd TC_chan_rel_a_reset TC_rll_est_ind_inact_lchan TC_rll_est_ind_inval_sapi1 TC_rll_est_ind_inval_sapi3 TC_rll_est_ind_inval_sacch TC_assignment_cic_only TC_assignment_csd TC_assignment_ctm TC_assignment_fr_a5_0 TC_assignment_fr_a5_1_codec_missing TC_assignment_fr_a5_1 TC_assignment_fr_a5_3 TC_assignment_fr_a5_4 TC_paging_imsi_nochan TC_paging_tmsi_nochan TC_paging_tmsi_any TC_paging_tmsi_sdcch TC_paging_tmsi_tch_f TC_paging_tmsi_tch_hf TC_paging_imsi_nochan_cgi TC_paging_imsi_nochan_lac_ci TC_paging_imsi_nochan_ci TC_paging_imsi_nochan_lai TC_paging_imsi_nochan_lac TC_paging_imsi_nochan_all TC_paging_imsi_nochan_plmn_lac_rnc TC_paging_imsi_nochan_rnc TC_paging_imsi_nochan_lac_rnc TC_paging_imsi_nochan_lacs TC_paging_imsi_nochan_lacs_empty TC_paging_imsi_a_reset TC_paging_counter TC_rsl_drop_counter TC_classmark TC_unsol_ass_fail TC_unsol_ass_compl TC_unsol_ho_fail TC_err_82_short_msg TC_ho_int Authors: Harald Welte <laforge@gnumonks.org> Philipp Maier <pmaier@sysmocom.de> Neels Hofmeyr <neels@hofmeyr.de> Closes: OS#2730 Closes: OS#2731 Closes: OS#2762 Closes: OS#2768 Closes: OS#2775 Closes: OS#2823 Closes: OS#2898 Closes: OS#2936 Change-Id: I68286d26e2014048b054f39ef29c35fef420cc97
2018-01-28 02:04:16 +00:00
llist_add_tail(&conn->entry, &net->subscr_conns);
return conn;
}
static void gsm0808_send_rsl_dtap(struct gsm_subscriber_connection *conn,
struct msgb *msg, int link_id, int allow_sacch);
#define GSCON_DTAP_QUEUE_MSGB_CB_LINK_ID 0
#define GSCON_DTAP_QUEUE_MSGB_CB_ALLOW_SACCH 1
static void gscon_dtap_queue_add(struct gsm_subscriber_connection *conn, struct msgb *msg,
int link_id, bool allow_sacch)
{
if (conn->dtap_queue_len >= 8) {
LOGP(DMSC, LOGL_ERROR, "%s: Cannot queue more DTAP messages,"
" already reached sane maximum of %u queued messages\n",
bsc_subscr_name(conn->bsub), conn->dtap_queue_len);
msgb_free(msg);
return;
}
conn->dtap_queue_len ++;
LOGP(DMSC, LOGL_DEBUG, "%s: Queueing DTAP message during handover/assignment (%u)\n",
bsc_subscr_name(conn->bsub), conn->dtap_queue_len);
msg->cb[GSCON_DTAP_QUEUE_MSGB_CB_LINK_ID] = (unsigned long)link_id;
msg->cb[GSCON_DTAP_QUEUE_MSGB_CB_ALLOW_SACCH] = allow_sacch ? 1 : 0;
msgb_enqueue(&conn->dtap_queue, msg);
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
static void gscon_dtap_queue_flush(struct gsm_subscriber_connection *conn, int send)
{
struct msgb *msg;
unsigned int flushed_count = 0;
while ((msg = msgb_dequeue(&conn->dtap_queue))) {
conn->dtap_queue_len --;
flushed_count ++;
if (send) {
int link_id = (int)msg->cb[GSCON_DTAP_QUEUE_MSGB_CB_LINK_ID];
bool allow_sacch = !!msg->cb[GSCON_DTAP_QUEUE_MSGB_CB_ALLOW_SACCH];
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
LOGPFSML(conn->fi, LOGL_DEBUG,
"%s: Sending queued DTAP message after handover/assignment (%u/%u)\n",
bsc_subscr_name(conn->bsub), flushed_count, conn->dtap_queue_len);
gsm0808_send_rsl_dtap(conn, msg, link_id, allow_sacch);
} else
msgb_free(msg);
}
}
static void rll_ind_cb(struct gsm_lchan *lchan, uint8_t link_id, void *_data, enum bsc_rllr_ind rllr_ind)
{
struct msgb *msg = _data;
/*
* There seems to be a small window that the RLL timer can
* fire after a lchan_release call and before the S_CHALLOC_FREED
* is called. Check if a conn is set before proceeding.
*/
if (!lchan->conn) {
msgb_free(msg);
return;
}
switch (rllr_ind) {
case BSC_RLLR_IND_EST_CONF:
rsl_data_request(msg, link_id);
break;
case BSC_RLLR_IND_REL_IND:
bsc_sapi_n_reject(lchan->conn, RSL_LINK_ID2DLCI(link_id),
GSM0808_CAUSE_MS_NOT_EQUIPPED);
msgb_free(msg);
break;
case BSC_RLLR_IND_ERR_IND:
case BSC_RLLR_IND_TIMEOUT:
bsc_sapi_n_reject(lchan->conn, RSL_LINK_ID2DLCI(link_id),
GSM0808_CAUSE_BSS_NOT_EQUIPPED);
msgb_free(msg);
break;
default:
LOGPLCHAN(lchan, DRLL, LOGL_NOTICE, "Received unknown rllr_ind %u\n", rllr_ind);
msgb_free(msg);
break;
}
}
/*! \brief process incoming 08.08 DTAP from MSC (send via BTS to MS) */
static void gsm0808_send_rsl_dtap(struct gsm_subscriber_connection *conn,
struct msgb *msg, int link_id, int allow_sacch)
{
uint8_t sapi;
int rc;
if (!conn->lchan) {
LOGP(DMSC, LOGL_ERROR,
"%s Called submit dtap without an lchan.\n",
bsc_subscr_name(conn->bsub));
msgb_free(msg);
rc = -EINVAL;
goto failed_to_send;
}
sapi = link_id & 0x7;
msg->lchan = conn->lchan;
/* If we are on a TCH and need to submit a SMS (on SAPI=3) we need to use the SACH */
if (allow_sacch && sapi != 0) {
if (conn->lchan->type == GSM_LCHAN_TCH_F || conn->lchan->type == GSM_LCHAN_TCH_H)
link_id |= 0x40;
}
msg->l3h = msg->data;
/* is requested SAPI already up? */
if (conn->lchan->sapis[sapi] == LCHAN_SAPI_UNUSED) {
/* Establish L2 for additional SAPI */
OBSC_LINKID_CB(msg) = link_id;
rc = rll_establish(msg->lchan, sapi, rll_ind_cb, msg);
if (rc) {
msgb_free(msg);
bsc_sapi_n_reject(conn, RSL_LINK_ID2DLCI(link_id), GSM0808_CAUSE_BSS_NOT_EQUIPPED);
goto failed_to_send;
}
return;
} else {
/* Directly forward via RLL/RSL to BTS */
rc = rsl_data_request(msg, link_id);
if (rc)
goto failed_to_send;
}
return;
failed_to_send:
LOGPFSML(conn->fi, LOGL_ERROR, "Tx BSSMAP CLEAR REQUEST to MSC\n");
gscon_bssmap_clear(conn, GSM0808_CAUSE_EQUIPMENT_FAILURE);
}
void gscon_submit_rsl_dtap(struct gsm_subscriber_connection *conn,
struct msgb *msg, int link_id, int allow_sacch)
{
/* buffer message during assignment / handover */
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
if (conn->fi->state != ST_ACTIVE) {
gscon_dtap_queue_add(conn, msg, link_id, !! allow_sacch);
return;
}
gsm0808_send_rsl_dtap(conn, msg, link_id, allow_sacch);
}
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
/* Compose an FSM ID, if possible from the current subscriber information */
void gscon_update_id(struct gsm_subscriber_connection *conn)
{
osmo_fsm_inst_update_id_f(conn->fi, "msc%u-conn%u%s%s",
conn->sccp.msc ? conn->sccp.msc->nr : UINT_MAX,
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
conn->sccp.conn_id,
conn->bsub? "_" : "",
conn->bsub? bsc_subscr_id(conn->bsub) : "");
}
bool gscon_is_aoip(struct gsm_subscriber_connection *conn)
{
if (!conn || !conn->sccp.msc)
return false;
return msc_is_aoip(conn->sccp.msc);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
}
bool gscon_is_sccplite(struct gsm_subscriber_connection *conn)
{
if (!conn || !conn->sccp.msc)
return false;
return msc_is_sccplite(conn->sccp.msc);
large refactoring: use FSMs for lchans; add inter-BSC HO Add FSMs: - timeslot_fsm: handle dynamic timeslots and OML+RSL availability. - lchan_fsm: handle an individual lchan activation, RTP stream and release, signal the appropriate calling FSMs on success, failure, release. - mgw_endpoint_fsm: handle one entire endpoint with several CI. - assignment_fsm: BSSMAP Assignment Request. - handover_fsm: all of intra, inter-MO and inter-MT handover. Above FSMs absorb large parts of the gscon FSM. The gscon FSM was surpassing the maximum amount events (32), and it is more logical to treat assignment, handover and MGW procedures in separate FSMs. - Add logging macros for each FSM type: - LOG_TS() - LOG_LCHAN() - LOG_MGWEP(), LOG_CI() - LOG_ASSIGNMENT() - LOG_HO() These log with the osmo_fsm_inst where present. New style decision: logging without a final newline char is awkward, especially for gsmtap logging and when other logs interleave LOGPC() calls; we have various cases where the final \n goes missing, and also this invokes the log category checking N times instead of once. So I decided to make these macros *always* append a newline, but only if there is no final newline yet. I hope that the compiler optimizes the strlen() of the constant format strings away. Thus I can log with or without typing "\n" and always get an \n termination anyway. General: - replace osmo_timers, state enums and program-wide osmo_signal_dispatch() with dedicated FSM timeouts, states and events. - introduce a common way to handle Tnnn timers: gsm_timers.h/.c: struct T_def. These can be used (with some macro magic) to define a state's timeout once, and not make mistakes for each osmo_fsm_inst_state_chg(). Details: bsc_subscr_conn_fsm.c: - move most states of this FSM to lchan_fsm, assignment_fsm, handover_fsm and mgw_endpoint_fsm. - There is exactly one state for an ongoing Assignment, with all details handled in conn->assignment.fi. The state relies on the assignment_fsm's timeout. - There is one state for an ongoing Handover; except for an incoming Handover from a remote BSS, the gscon remains in ST_INIT until the new lchan and conn are both established. - move bssmap_add_lcls_status() to osmo_bsc_lcls.c abis_rsl.c: - move all dynamic timeslot logic away into timeslot_fsm. Only keep plain send/receive functions in abis_rsl.c - reduce some rsl functions to merely send a message, rename to "_tx_". - rsl_ipacc_mdcx(): add '_tx_' in the name; move parts that change the lchan state out into the lchan_fsm, the lchan->abis_ip.* are now set there prior to invoking this function. - move all timers and error/release handling away into various FSMs. - tweak ipa_smod_s_for_lchan() and ipa_rtp_pt_for_lchan() to not require an lchan passed, but just mode,type that they require. Rename to ipacc_speech_mode*() and ipacc_payload_type(). - add rsl_forward_layer3_info, used for inter-BSC HO MO, to just send the RR message received during BSSMAP Handover Command. - move various logging to LOG_LCHAN() in order to log with the lchan FSM instance. One drawback is that the lchan FSM is limited to one logging category, i.e. this moves some logging from DRR to DRSL. It might actually make sense to combine those categories. - lose LOGP...LOGPC logging cascades: they are bad for gsmtap logging and for performance. - handle_classmark_chg(): change logging, move cm2 len check out of the cm3 condition (I hope that's correct). - gsm48_send_ho_cmd(): split off gsm48_make_ho_cmd() which doesn't send right away, so that during inter-bsc HO we can make an RR Handover Command to send via the MSC to the remote BSS. assignment_fsm.c: - the Chan Mode Modify in case of re-using the same lchan is not implemented yet, because this was also missing in the previous implementation (OS#3357). osmo_bsc_api.c: - simplify bsc_mr_config() and move to lchan_fsm.c, the only caller; rename to lchan_mr_config(). (bsc_mr_config() used to copy the values to mr_bts_lv twice, once by member assignment and then again with a memcpy.) - During handover, we used to copy the MR config from the old lchan. Since we may handover between FR and HR, rather set the MR Config anew every time, so that FR rates are always available on FR lchans, and never on HR lchans. Depends: I03ee7ce840ecfa0b6a33358e7385528aabd4873f (libosmocore), I1f2918418c38918c5ac70acaa51a47adfca12b5e (libosmocore) Change-Id: I82e3f918295daa83274a4cf803f046979f284366
2018-05-14 16:14:15 +00:00
}