osmo-msc/src/libmsc/msc_a.c

1819 lines
54 KiB
C
Raw Normal View History

large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
/* Code to manage a subscriber's MSC-A role */
/*
* (C) 2019 by sysmocom - s.m.f.c. GmbH <info@sysmocom.de>
* All Rights Reserved
*
* SPDX-License-Identifier: AGPL-3.0+
*
* Author: Neels Hofmeyr
*
* 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 <osmocom/core/utils.h>
#include <osmocom/core/tdef.h>
#include <osmocom/core/rate_ctr.h>
#include <osmocom/core/signal.h>
#include <osmocom/msc/gsm_data.h>
#include <osmocom/msc/msc_roles.h>
#include <osmocom/msc/msub.h>
#include <osmocom/msc/msc_a.h>
#include <osmocom/msc/msc_t.h>
#include <osmocom/msc/msc_i.h>
#include <osmocom/msc/paging.h>
#include <osmocom/msc/signal.h>
#include <osmocom/msc/vlr.h>
#include <osmocom/msc/transaction.h>
#include <osmocom/msc/ran_peer.h>
#include <osmocom/msc/ran_msg_a.h>
#include <osmocom/msc/ran_msg_iu.h>
#include <osmocom/msc/sgs_iface.h>
#include <osmocom/msc/gsm_04_08.h>
#include <osmocom/msc/gsm_09_11.h>
#include <osmocom/msc/gsm_04_14.h>
#include <osmocom/msc/call_leg.h>
#include <osmocom/msc/rtp_stream.h>
#include <osmocom/msc/msc_ho.h>
#define MSC_A_USE_WAIT_CLEAR_COMPLETE "wait-Clear-Complete"
static struct osmo_fsm msc_a_fsm;
static const struct osmo_tdef_state_timeout msc_a_fsm_timeouts[32] = {
[MSC_A_ST_VALIDATE_L3] = { .T = -1 },
[MSC_A_ST_AUTH_CIPH] = { .keep_timer = true },
[MSC_A_ST_WAIT_CLASSMARK_UPDATE] = { .keep_timer = true },
[MSC_A_ST_AUTHENTICATED] = { .keep_timer = true },
[MSC_A_ST_RELEASING] = { .T = -2 },
[MSC_A_ST_RELEASED] = { .T = -2 },
};
/* Transition to a state, using the T timer defined in msc_a_fsm_timeouts.
* The actual timeout value is in turn obtained from network->T_defs.
* Assumes local variable fi exists. */
#define msc_a_state_chg_always(msc_a, state) \
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
osmo_tdef_fsm_inst_state_chg((msc_a)->c.fi, state, msc_a_fsm_timeouts, (msc_a)->c.ran->tdefs, 5)
/* Same as msc_a_state_chg_always() but ignore if the msc_a already is in the target state. */
#define msc_a_state_chg(msc_a, STATE) do { \
if ((msc_a)->c.fi->state != STATE) \
msc_a_state_chg_always(msc_a, STATE); \
} while(0)
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
struct gsm_network *msc_a_net(const struct msc_a *msc_a)
{
return msub_net(msc_a->c.msub);
}
struct vlr_subscr *msc_a_vsub(const struct msc_a *msc_a)
{
if (!msc_a)
return NULL;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return msub_vsub(msc_a->c.msub);
}
struct msc_i *msc_a_msc_i(const struct msc_a *msc_a)
{
if (!msc_a)
return NULL;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return msub_msc_i(msc_a->c.msub);
}
struct msc_t *msc_a_msc_t(const struct msc_a *msc_a)
{
if (!msc_a)
return NULL;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return msub_msc_t(msc_a->c.msub);
}
struct msc_a *msc_a_fi_priv(struct osmo_fsm_inst *fi)
{
OSMO_ASSERT(fi);
OSMO_ASSERT(fi->fsm == &msc_a_fsm);
OSMO_ASSERT(fi->priv);
return fi->priv;
}
bool msc_a_require_ciphering(const struct msc_a *msc_a)
{
struct gsm_network *net = msc_a_net(msc_a);
bool is_utran = (msc_a->c.ran->type == OSMO_RAT_UTRAN_IU);
if (is_utran)
return net->uea_encryption_mask > (1 << OSMO_UTRAN_UEA0);
else
return net->a5_encryption_mask > 0x1;
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
static void update_counters(struct osmo_fsm_inst *fi, bool conn_accepted)
{
struct msc_a *msc_a = fi->priv;
struct gsm_network *net = msc_a_net(msc_a);
switch (msc_a->complete_layer3_type) {
case COMPLETE_LAYER3_LU:
rate_ctr_inc(rate_ctr_group_get_ctr(net->msc_ctrs, conn_accepted ? MSC_CTR_LOC_UPDATE_COMPLETED : MSC_CTR_LOC_UPDATE_FAILED));
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
break;
case COMPLETE_LAYER3_CM_SERVICE_REQ:
rate_ctr_inc(rate_ctr_group_get_ctr(net->msc_ctrs, conn_accepted ? MSC_CTR_CM_SERVICE_REQUEST_ACCEPTED : MSC_CTR_CM_SERVICE_REQUEST_REJECTED));
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
break;
case COMPLETE_LAYER3_PAGING_RESP:
rate_ctr_inc(rate_ctr_group_get_ctr(net->msc_ctrs, conn_accepted ? MSC_CTR_PAGING_RESP_ACCEPTED : MSC_CTR_PAGING_RESP_REJECTED));
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
break;
case COMPLETE_LAYER3_CM_RE_ESTABLISH_REQ:
rate_ctr_inc(rate_ctr_group_get_ctr(net->msc_ctrs,
conn_accepted ? MSC_CTR_CM_RE_ESTABLISH_REQ_ACCEPTED
: MSC_CTR_CM_RE_ESTABLISH_REQ_REJECTED));
break;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
default:
break;
}
}
static void evaluate_acceptance_outcome(struct osmo_fsm_inst *fi, bool conn_accepted)
{
struct msc_a *msc_a = fi->priv;
struct vlr_subscr *vsub = msc_a_vsub(msc_a);
update_counters(fi, conn_accepted);
if (conn_accepted) {
/* Record the Cell ID seen in Complete Layer 3 Information in the VLR, so that it also shows in vty
* 'show' output. */
vsub->cgi = msc_a->via_cell;
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
/* Trigger transactions that we paged for */
if (msc_a->complete_layer3_type == COMPLETE_LAYER3_PAGING_RESP) {
if (conn_accepted)
paging_response(msc_a);
else
paging_expired(vsub);
}
if (conn_accepted)
osmo_signal_dispatch(SS_SUBSCR, S_SUBSCR_ATTACHED, msc_a_vsub(msc_a));
if (msc_a->complete_layer3_type == COMPLETE_LAYER3_LU)
msc_a_put(msc_a, MSC_A_USE_LOCATION_UPDATING);
if (msc_a->complete_layer3_type == COMPLETE_LAYER3_CM_RE_ESTABLISH_REQ) {
/* Trigger new Assignment to recommence the voice call. A little dance here because normally we verify
* that no CC trans is already active. */
struct gsm_trans *cc_trans = msc_a->cc.active_trans;
msc_a->cc.active_trans = NULL;
osmo_fsm_inst_dispatch(msc_a->c.fi, MSC_A_EV_TRANSACTION_ACCEPTED, cc_trans);
msc_a_try_call_assignment(cc_trans);
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
}
bool msc_a_is_accepted(const struct msc_a *msc_a)
{
if (!msc_a || !msc_a->c.fi)
return false;
return msc_a->c.fi->state == MSC_A_ST_AUTHENTICATED
|| msc_a->c.fi->state == MSC_A_ST_COMMUNICATING;
}
bool msc_a_in_release(struct msc_a *msc_a)
{
if (!msc_a)
return true;
if (msc_a->c.fi->state == MSC_A_ST_RELEASING)
return true;
if (msc_a->c.fi->state == MSC_A_ST_RELEASED)
return true;
return false;
}
static int msc_a_ran_dec(struct msc_a *msc_a, const struct an_apdu *an_apdu, enum msc_role from_role)
{
int rc;
struct msc_a_ran_dec_data d = {
.from_role = from_role,
.an_apdu = an_apdu,
};
msc_a_get(msc_a, __func__);
rc = msc_role_ran_decode(msc_a->c.fi, an_apdu, msc_a_ran_decode_cb, &d);
msc_a_put(msc_a, __func__);
return rc;
};
static void msc_a_fsm_validate_l3(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
struct msc_a *msc_a = fi->priv;
const struct an_apdu *an_apdu;
switch (event) {
case MSC_A_EV_FROM_I_COMPLETE_LAYER_3:
case MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST:
case MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST:
an_apdu = data;
msc_a_ran_dec(msc_a, an_apdu, MSC_ROLE_I);
return;
case MSC_A_EV_COMPLETE_LAYER_3_OK:
msc_a_state_chg(msc_a, MSC_A_ST_AUTH_CIPH);
return;
case MSC_A_EV_MO_CLOSE:
case MSC_A_EV_CN_CLOSE:
evaluate_acceptance_outcome(fi, false);
/* fall through */
case MSC_A_EV_UNUSED:
msc_a_state_chg(msc_a, MSC_A_ST_RELEASING);
return;
default:
OSMO_ASSERT(false);
}
}
/* Figure out whether to first send a Classmark Request to the MS to figure out algorithm support. */
static bool msc_a_need_classmark_for_ciphering(struct msc_a *msc_a)
{
struct gsm_network *net = msc_a_net(msc_a);
struct vlr_subscr *vsub = msc_a_vsub(msc_a);
int i = 0;
bool request_classmark = false;
/* Only on GERAN-A do we ever need Classmark Information for Ciphering. */
if (msc_a->c.ran->type != OSMO_RAT_GERAN_A)
return false;
for (i = 0; i < 8; i++) {
int supported;
/* A5/n permitted by osmo-msc.cfg? */
if (!(net->a5_encryption_mask & (1 << i)))
continue;
/* A5/n supported by MS? */
supported = osmo_gsm48_classmark_supports_a5(&vsub->classmark, i);
if (supported < 0) {
LOG_MSC_A(msc_a, LOGL_DEBUG, "For A5/%d, we still need Classmark %d\n", i, -supported);
request_classmark = true;
}
}
return request_classmark;
}
static int msc_a_ran_enc_ciphering(struct msc_a *msc_a, bool umts_aka, bool retrieve_imeisv);
/* VLR callback for ops.set_ciph_mode() */
int msc_a_vlr_set_cipher_mode(void *_msc_a, bool umts_aka, bool retrieve_imeisv)
{
struct msc_a *msc_a = _msc_a;
struct vlr_subscr *vsub;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
if (!msc_a) {
LOGP(DMSC, LOGL_ERROR, "Insufficient info to start ciphering: "
"MSC-A role is NULL?!?\n");
return -EINVAL;
}
vsub = msc_a_vsub(msc_a);
if (!vsub || !vsub->last_tuple) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Insufficient info to start ciphering: "
"vlr_subscr is NULL?!?\n");
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return -EINVAL;
}
if (msc_a_need_classmark_for_ciphering(msc_a)) {
int rc;
struct ran_msg msg = {
.msg_type = RAN_MSG_CLASSMARK_REQUEST,
};
rc = msc_a_ran_down(msc_a, MSC_ROLE_I, &msg);
if (rc) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Cannot send Classmark Request\n");
return -EIO;
}
msc_a->state_before_classmark_update = msc_a->c.fi->state;
msc_a->action_on_classmark_update = (struct msc_a_action_on_classmark_update){
.type = MSC_A_CLASSMARK_UPDATE_THEN_CIPHERING,
.ciphering = {
.umts_aka = umts_aka,
.retrieve_imeisv = retrieve_imeisv,
},
};
msc_a_state_chg(msc_a, MSC_A_ST_WAIT_CLASSMARK_UPDATE);
return 0;
}
return msc_a_ran_enc_ciphering(msc_a, umts_aka, retrieve_imeisv);
}
static uint8_t filter_a5(uint8_t a5_mask, bool umts_aka)
{
/* With GSM AKA: allow A5/0, 1, 3 = 0b00001011 = 0xb.
* UMTS aka: allow A5/0, 1, 3, 4 = 0b00011011 = 0x1b.
*/
return a5_mask & (umts_aka ? 0x1b : 0x0b);
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
static int msc_a_ran_enc_ciphering(struct msc_a *msc_a, bool umts_aka, bool retrieve_imeisv)
{
struct gsm_network *net;
struct vlr_subscr *vsub;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
struct ran_msg msg;
if (!msc_a) {
LOGP(DMSC, LOGL_ERROR, "Insufficient info to start ciphering: "
"MSC-A role is NULL?!?\n");
return -EINVAL;
}
net = msc_a_net(msc_a);
vsub = msc_a_vsub(msc_a);
if (!net || !vsub || !vsub->last_tuple) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Insufficient info to start ciphering: "
"gsm_network and/or vlr_subscr is NULL?!?\n");
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return -EINVAL;
}
msg = (struct ran_msg){
.msg_type = RAN_MSG_CIPHER_MODE_COMMAND,
.cipher_mode_command = {
.vec = vsub->last_tuple ? &vsub->last_tuple->vec : NULL,
.classmark = &vsub->classmark,
.geran = {
.umts_aka = umts_aka,
.retrieve_imeisv = retrieve_imeisv,
.a5_encryption_mask = filter_a5(net->a5_encryption_mask, umts_aka),
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
/* for ran_a.c to store the GERAN key that is actually used */
.chosen_key = &msc_a->geran_encr,
},
.utran = {
.uea_encryption_mask = net->uea_encryption_mask,
},
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
},
};
if (msc_a_ran_down(msc_a, MSC_ROLE_I, &msg)) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Sending Cipher Mode Command failed\n");
/* Returning error to the VLR ops.set_ciph_mode() will cancel the attach. Other callers need to take
* care of the return value. */
return -EINVAL;
}
if (msc_a->geran_encr.key_len)
LOG_MSC_A(msc_a, LOGL_DEBUG, "RAN encoding chose ciphering: A5/%d kc %s kc128 %s\n",
msc_a->geran_encr.alg_id - 1,
osmo_hexdump_nospc_c(OTC_SELECT, msc_a->geran_encr.key, msc_a->geran_encr.key_len),
msc_a->geran_encr.kc128_present ?
osmo_hexdump_nospc_c(OTC_SELECT, msc_a->geran_encr.kc128, sizeof(msc_a->geran_encr.kc128))
: "-");
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return 0;
}
static void msc_a_fsm_auth_ciph(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
struct msc_a *msc_a = fi->priv;
/* If accepted, transition the state, all other cases mean failure. */
switch (event) {
case MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST:
case MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST:
msc_a_ran_dec(msc_a, data, MSC_ROLE_I);
return;
case MSC_A_EV_AUTHENTICATED:
msc_a_state_chg(msc_a, MSC_A_ST_AUTHENTICATED);
return;
case MSC_A_EV_UNUSED:
msc_a_state_chg(msc_a, MSC_A_ST_RELEASING);
return;
case MSC_A_EV_MO_CLOSE:
case MSC_A_EV_CN_CLOSE:
evaluate_acceptance_outcome(fi, false);
msc_a_state_chg(msc_a, MSC_A_ST_RELEASING);
return;
default:
OSMO_ASSERT(false);
}
}
static void msc_a_fsm_wait_classmark_update(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
struct msc_a *msc_a = fi->priv;
switch (event) {
case MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST:
case MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST:
msc_a_ran_dec(msc_a, data, MSC_ROLE_I);
return;
case MSC_A_EV_CLASSMARK_UPDATE:
switch (msc_a->action_on_classmark_update.type) {
case MSC_A_CLASSMARK_UPDATE_THEN_CIPHERING:
msc_a_state_chg(msc_a, MSC_A_ST_AUTH_CIPH);
if (msc_a_ran_enc_ciphering(msc_a,
msc_a->action_on_classmark_update.ciphering.umts_aka,
msc_a->action_on_classmark_update.ciphering.retrieve_imeisv)) {
LOG_MSC_A(msc_a, LOGL_ERROR,
"After Classmark Update, still failed to send Cipher Mode Command\n");
msc_a_state_chg(msc_a, MSC_A_ST_RELEASING);
}
return;
default:
LOG_MSC_A(msc_a, LOGL_ERROR, "Internal error: After Classmark Update, don't know what to do\n");
msc_a_state_chg(msc_a, msc_a->state_before_classmark_update);
return;
}
case MSC_A_EV_UNUSED:
/* Seems something detached / aborted in the middle of auth+ciph. */
evaluate_acceptance_outcome(fi, false);
msc_a_state_chg(msc_a, MSC_A_ST_RELEASING);
return;
case MSC_A_EV_MO_CLOSE:
case MSC_A_EV_CN_CLOSE:
evaluate_acceptance_outcome(fi, false);
msc_a_state_chg(msc_a, MSC_A_ST_RELEASING);
return;
default:
OSMO_ASSERT(false);
}
}
static bool msc_a_fsm_has_active_transactions(struct osmo_fsm_inst *fi)
{
struct msc_a *msc_a = fi->priv;
struct vlr_subscr *vsub = msc_a_vsub(msc_a);
struct gsm_trans *trans;
if (osmo_use_count_by(&msc_a->use_count, MSC_A_USE_SILENT_CALL)) {
LOG_MSC_A(msc_a, LOGL_DEBUG, "%s: silent call still active\n", __func__);
return true;
}
if (osmo_use_count_by(&msc_a->use_count, MSC_A_USE_CM_SERVICE_CC)) {
LOG_MSC_A(msc_a, LOGL_DEBUG, "%s: still awaiting MO CC request after a CM Service Request\n",
__func__);
return true;
}
if (osmo_use_count_by(&msc_a->use_count, MSC_A_USE_CM_SERVICE_SMS)) {
LOG_MSC_A(msc_a, LOGL_DEBUG, "%s: still awaiting MO SMS after a CM Service Request\n",
__func__);
return true;
}
if (osmo_use_count_by(&msc_a->use_count, MSC_A_USE_CM_SERVICE_SS)) {
LOG_MSC_A(msc_a, LOGL_DEBUG, "%s: still awaiting MO SS after a CM Service Request\n",
__func__);
return true;
}
if (vsub && !llist_empty(&vsub->cs.requests)) {
struct paging_request *pr;
llist_for_each_entry(pr, &vsub->cs.requests, entry) {
LOG_MSC_A(msc_a, LOGL_DEBUG, "%s: still active: %s\n", __func__, pr->label);
}
return true;
}
if ((trans = trans_has_conn(msc_a))) {
LOG_MSC_A(msc_a, LOGL_DEBUG, "connection still has active transaction: %s\n",
trans_type_name(trans->type));
return true;
}
return false;
}
static void msc_a_fsm_authenticated_enter(struct osmo_fsm_inst *fi, uint32_t prev_state)
{
struct msc_a *msc_a = fi->priv;
struct vlr_subscr *vsub = msc_a_vsub(msc_a);
/* Stop Location Update expiry for this subscriber. While the subscriber
* has an open connection the LU expiry timer must remain disabled.
* Otherwise we would kick the subscriber off the network when the timer
* expires e.g. during a long phone call.
* The LU expiry timer will restart once the connection is closed. */
if (vsub)
vsub->expire_lu = VLR_SUBSCRIBER_NO_EXPIRATION;
evaluate_acceptance_outcome(fi, true);
}
static void msc_a_fsm_authenticated(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
struct msc_a *msc_a = fi->priv;
switch (event) {
case MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST:
case MSC_A_EV_FROM_I_PREPARE_SUBSEQUENT_HANDOVER_REQUEST:
case MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST:
msc_a_ran_dec(msc_a, data, MSC_ROLE_I);
return;
case MSC_A_EV_COMPLETE_LAYER_3_OK:
/* When Authentication is off, we may already be in the Accepted state when the code
* evaluates the Compl L3. Simply ignore. This just cosmetically mutes the error log
* about the useless event. */
return;
case MSC_A_EV_TRANSACTION_ACCEPTED:
msc_a_state_chg(msc_a, MSC_A_ST_COMMUNICATING);
return;
case MSC_A_EV_MO_CLOSE:
case MSC_A_EV_CN_CLOSE:
case MSC_A_EV_UNUSED:
msc_a_state_chg(msc_a, MSC_A_ST_RELEASING);
return;
default:
OSMO_ASSERT(false);
}
}
/* The MGW has given us a local IP address for the RAN side. Ready to start the Assignment of a voice channel. */
static void msc_a_call_leg_ran_local_addr_available(struct msc_a *msc_a)
{
struct ran_msg msg;
struct gsm_trans *cc_trans = msc_a->cc.active_trans;
struct gsm0808_channel_type channel_type;
if (!cc_trans) {
LOG_MSC_A(msc_a, LOGL_ERROR, "No CC transaction active\n");
call_leg_release(msc_a->cc.call_leg);
return;
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
/* Once a CI is known, we could also CRCX the CN side of the MGW endpoint, but it makes sense to wait for the
* codec to be determined by the Assignment Complete message, first. */
if (mncc_bearer_cap_to_channel_type(&channel_type, &cc_trans->bearer_cap)) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Cannot compose Channel Type from bearer capabilities\n");
trans_free(cc_trans);
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return;
}
/* The RAN side RTP address is known, so the voice Assignment can commence. */
msg = (struct ran_msg){
.msg_type = RAN_MSG_ASSIGNMENT_COMMAND,
.assignment_command = {
.cn_rtp = &msc_a->cc.call_leg->rtp[RTP_TO_RAN]->local,
.channel_type = &channel_type,
.osmux_present = msc_a->cc.call_leg->rtp[RTP_TO_RAN]->use_osmux,
.osmux_cid = msc_a->cc.call_leg->rtp[RTP_TO_RAN]->local_osmux_cid,
.call_id_present = true,
.call_id = cc_trans->callref,
.lcls = cc_trans->cc.lcls,
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
},
};
if (msc_a_ran_down(msc_a, MSC_ROLE_I, &msg)) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Cannot send Assignment\n");
trans_free(cc_trans);
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return;
}
}
static void msc_a_call_leg_cn_local_addr_available(struct msc_a *msc_a, struct gsm_trans *cc_trans)
{
if (gsm48_tch_rtp_create(cc_trans)) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Cannot inform MNCC of RTP address\n");
trans_free(cc_trans);
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return;
}
}
static struct gsm_trans *find_waiting_call(struct msc_a *msc_a)
{
struct gsm_trans *trans;
struct gsm_network *net = msc_a_net(msc_a);
llist_for_each_entry(trans, &net->trans_list, entry) {
if (trans->msc_a != msc_a)
continue;
if (trans->type != TRANS_CC)
continue;
if (trans->msc_a->cc.active_trans == trans)
continue;
return trans;
}
return NULL;
}
static void msc_a_cleanup_rtp_streams(struct msc_a *msc_a, uint32_t event, void *data)
{
switch (event) {
case MSC_EV_CALL_LEG_TERM:
msc_a->cc.call_leg = NULL;
if (msc_a->cc.mncc_forwarding_to_remote_ran)
msc_a->cc.mncc_forwarding_to_remote_ran->rtps = NULL;
if (msc_a->ho.new_cell.mncc_forwarding_to_remote_ran)
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
msc_a->ho.new_cell.mncc_forwarding_to_remote_ran->rtps = NULL;
return;
case MSC_MNCC_EV_CALL_ENDED:
msc_a->cc.mncc_forwarding_to_remote_ran = NULL;
return;
default:
return;
}
}
static void msc_a_fsm_communicating(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
struct msc_a *msc_a = fi->priv;
struct rtp_stream *rtps;
struct gsm_trans *waiting_trans;
struct an_apdu *an_apdu;
msc_a_cleanup_rtp_streams(msc_a, event, data);
switch (event) {
case MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST:
case MSC_A_EV_FROM_I_PREPARE_SUBSEQUENT_HANDOVER_REQUEST:
case MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST:
an_apdu = data;
msc_a_ran_dec(msc_a, an_apdu, MSC_ROLE_I);
return;
case MSC_A_EV_FROM_T_PREPARE_HANDOVER_RESPONSE:
case MSC_A_EV_FROM_T_PREPARE_HANDOVER_FAILURE:
case MSC_A_EV_FROM_T_PROCESS_ACCESS_SIGNALLING_REQUEST:
case MSC_A_EV_FROM_T_SEND_END_SIGNAL_REQUEST:
an_apdu = data;
msc_a_ran_dec(msc_a, an_apdu, MSC_ROLE_T);
return;
case MSC_A_EV_TRANSACTION_ACCEPTED:
/* no-op */
return;
case MSC_EV_CALL_LEG_RTP_LOCAL_ADDR_AVAILABLE:
rtps = data;
if (!rtps) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Invalid data for %s\n", osmo_fsm_event_name(fi->fsm, event));
return;
}
if (!msc_a->cc.call_leg) {
LOG_MSC_A(msc_a, LOGL_ERROR, "No call leg active\n");
return;
}
if (!osmo_sockaddr_str_is_nonzero(&rtps->local)) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Invalid RTP address received from MGW: " OSMO_SOCKADDR_STR_FMT "\n",
OSMO_SOCKADDR_STR_FMT_ARGS(&rtps->local));
call_leg_release(msc_a->cc.call_leg);
return;
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
LOG_MSC_A(msc_a, LOGL_DEBUG,
"MGW endpoint's RTP address available for the CI %s: " OSMO_SOCKADDR_STR_FMT " (osmux=%s:%d)\n",
rtp_direction_name(rtps->dir), OSMO_SOCKADDR_STR_FMT_ARGS(&rtps->local),
rtps->use_osmux ? "yes" : "no", rtps->local_osmux_cid);
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
switch (rtps->dir) {
case RTP_TO_RAN:
msc_a_call_leg_ran_local_addr_available(msc_a);
return;
case RTP_TO_CN:
msc_a_call_leg_cn_local_addr_available(msc_a, rtps->for_trans);
return;
default:
LOG_MSC_A(msc_a, LOGL_ERROR, "Invalid data for %s\n", osmo_fsm_event_name(fi->fsm, event));
return;
}
case MSC_EV_CALL_LEG_RTP_COMPLETE:
/* Nothing to do. */
return;
case MSC_MNCC_EV_CALL_ENDED:
/* Cleaned up above */
return;
case MSC_EV_CALL_LEG_TERM:
/* RTP streams cleaned up above */
msc_a_get(msc_a, __func__);
if (msc_a->cc.active_trans)
trans_free(msc_a->cc.active_trans);
/* If there is another call still waiting to be activated, this is the time when the mgcp_ctx is
* available again and the other call can start assigning. */
waiting_trans = find_waiting_call(msc_a);
if (waiting_trans) {
LOG_MSC_A(msc_a, LOGL_DEBUG, "(ti %02x) Call waiting: starting Assignment\n",
waiting_trans->transaction_id);
msc_a_try_call_assignment(waiting_trans);
}
msc_a_put(msc_a, __func__);
return;
case MSC_A_EV_HANDOVER_REQUIRED:
msc_ho_start(msc_a, (struct ran_handover_required*)data);
return;
case MSC_A_EV_HANDOVER_END:
/* Termination event of the msc_ho_fsm. No action needed, it's all done in the msc_ho_fsm cleanup. This
* event only exists because osmo_fsm_inst_alloc_child() requires a parent term event; and maybe
* interesting for logging. */
return;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
case MSC_A_EV_MO_CLOSE:
case MSC_A_EV_CN_CLOSE:
case MSC_A_EV_UNUSED:
msc_a_state_chg(msc_a, MSC_A_ST_RELEASING);
return;
default:
OSMO_ASSERT(false);
}
}
static int msc_a_fsm_timer_cb(struct osmo_fsm_inst *fi)
{
struct msc_a *msc_a = fi->priv;
if (msc_a_in_release(msc_a)) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Timeout while releasing, discarding right now\n");
msc_a_put_all(msc_a, MSC_A_USE_WAIT_CLEAR_COMPLETE);
msc_a_state_chg(msc_a, MSC_A_ST_RELEASED);
} else {
enum gsm48_reject_value cause = GSM48_REJECT_CONGESTION;
osmo_fsm_inst_dispatch(fi, MSC_A_EV_CN_CLOSE, &cause);
}
return 0;
}
static void msc_a_fsm_releasing_onenter(struct osmo_fsm_inst *fi, uint32_t prev_state)
{
struct msc_a *msc_a = fi->priv;
struct vlr_subscr *vsub = msc_a_vsub(msc_a);
int i;
char buf[128];
const char * const use_counts_to_cancel[] = {
MSC_A_USE_LOCATION_UPDATING,
MSC_A_USE_CM_SERVICE_CC,
MSC_A_USE_CM_SERVICE_SMS,
MSC_A_USE_CM_SERVICE_SS,
MSC_A_USE_PAGING_RESPONSE,
};
LOG_MSC_A(msc_a, LOGL_DEBUG, "Releasing: msc_a use is %s\n",
osmo_use_count_name_buf(buf, sizeof(buf), &msc_a->use_count));
if (vsub) {
vlr_subscr_get(vsub, __func__);
/* Cancel all VLR FSMs, if any */
vlr_subscr_cancel_attach_fsm(vsub, OSMO_FSM_TERM_ERROR, GSM48_REJECT_CONGESTION);
/* The subscriber has no active connection anymore.
* Restart the periodic Location Update expiry timer for this subscriber. */
vlr_subscr_enable_expire_lu(vsub);
}
/* If we're closing in a middle of a trans, we need to clean up */
trans_conn_closed(msc_a);
call_leg_release(msc_a->cc.call_leg);
/* Cancel use counts for pending CM Service / Paging */
for (i = 0; i < ARRAY_SIZE(use_counts_to_cancel); i++) {
const char *use = use_counts_to_cancel[i];
int32_t count = osmo_use_count_by(&msc_a->use_count, use);
if (!count)
continue;
LOG_MSC_A(msc_a, LOGL_DEBUG, "Releasing: canceling still pending use: %s (%d)\n", use, count);
osmo_use_count_get_put(&msc_a->use_count, use, -count);
}
if (msc_a->c.ran->type == OSMO_RAT_EUTRAN_SGS) {
sgs_iface_tx_release(vsub);
/* In SGsAP there is no confirmation of a release. */
msc_a_state_chg(msc_a, MSC_A_ST_RELEASED);
} else {
struct ran_msg msg = {
.msg_type = RAN_MSG_CLEAR_COMMAND,
.clear_command = {
/* "Call Control" is the only cause code listed in 3GPP TS 48.008 3.2.1.21 CLEAR COMMAND
* that qualifies for a normal release situation. (OS#4664) */
.gsm0808_cause = GSM0808_CAUSE_CALL_CONTROL,
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
.csfb_ind = (vsub && vsub->sgs_fsm->state == SGS_UE_ST_ASSOCIATED),
},
};
msc_a_get(msc_a, MSC_A_USE_WAIT_CLEAR_COMPLETE);
msc_a_ran_down(msc_a, MSC_ROLE_I, &msg);
/* The connection is cleared. The MS will now go back to 4G,
Switch the RAN type back to SGS. */
if (vsub && vsub->sgs_fsm->state == SGS_UE_ST_ASSOCIATED)
vsub->cs.attached_via_ran = OSMO_RAT_EUTRAN_SGS;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
}
if (vsub)
vlr_subscr_put(vsub, __func__);
}
static void msc_a_fsm_releasing(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
struct msc_a *msc_a = fi->priv;
msc_a_cleanup_rtp_streams(msc_a, event, data);
switch (event) {
case MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST:
case MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST:
msc_a_ran_dec(msc_a, data, MSC_ROLE_I);
return;
case MSC_A_EV_MO_CLOSE:
case MSC_A_EV_CN_CLOSE:
case MSC_A_EV_UNUSED:
/* Already releasing */
return;
case MSC_EV_CALL_LEG_TERM:
case MSC_MNCC_EV_CALL_ENDED:
/* RTP streams cleaned up above */
return;
case MSC_A_EV_HANDOVER_END:
/* msc_ho_fsm does cleanup. */
return;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
default:
OSMO_ASSERT(false);
}
}
static void msc_a_fsm_released_onenter(struct osmo_fsm_inst *fi, uint32_t prev_state)
{
struct msc_a *msc_a = msc_a_fi_priv(fi);
char buf[128];
LOG_MSC_A(msc_a, LOGL_DEBUG, "Released: msc_a use is %s\n",
osmo_use_count_name_buf(buf, sizeof(buf), &msc_a->use_count));
if (osmo_use_count_total(&msc_a->use_count) == 0)
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_REGULAR, fi);
}
static void msc_a_fsm_released(struct osmo_fsm_inst *fi, uint32_t event, void *data)
{
if (event == MSC_A_EV_UNUSED)
osmo_fsm_inst_term(fi, OSMO_FSM_TERM_REGULAR, fi);
}
void msc_a_fsm_cleanup(struct osmo_fsm_inst *fi, enum osmo_fsm_term_cause cause)
{
struct msc_a *msc_a = msc_a_fi_priv(fi);
trans_conn_closed(msc_a);
if (msc_a_fsm_has_active_transactions(fi))
LOG_MSC_A(msc_a, LOGL_ERROR, "Deallocating active transactions failed\n");
LOG_MSC_A_CAT(msc_a, DREF, LOGL_DEBUG, "max total use count was %d\n", msc_a->max_total_use_count);
}
const struct value_string msc_a_fsm_event_names[] = {
OSMO_VALUE_STRING(MSC_REMOTE_EV_RX_GSUP),
OSMO_VALUE_STRING(MSC_EV_CALL_LEG_RTP_LOCAL_ADDR_AVAILABLE),
OSMO_VALUE_STRING(MSC_EV_CALL_LEG_RTP_COMPLETE),
OSMO_VALUE_STRING(MSC_EV_CALL_LEG_TERM),
OSMO_VALUE_STRING(MSC_MNCC_EV_NEED_LOCAL_RTP),
OSMO_VALUE_STRING(MSC_MNCC_EV_CALL_PROCEEDING),
OSMO_VALUE_STRING(MSC_MNCC_EV_CALL_COMPLETE),
OSMO_VALUE_STRING(MSC_MNCC_EV_CALL_ENDED),
OSMO_VALUE_STRING(MSC_A_EV_FROM_I_COMPLETE_LAYER_3),
OSMO_VALUE_STRING(MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST),
OSMO_VALUE_STRING(MSC_A_EV_FROM_I_PREPARE_SUBSEQUENT_HANDOVER_REQUEST),
OSMO_VALUE_STRING(MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST),
OSMO_VALUE_STRING(MSC_A_EV_FROM_T_PROCESS_ACCESS_SIGNALLING_REQUEST),
OSMO_VALUE_STRING(MSC_A_EV_FROM_T_PREPARE_HANDOVER_RESPONSE),
OSMO_VALUE_STRING(MSC_A_EV_FROM_T_PREPARE_HANDOVER_FAILURE),
OSMO_VALUE_STRING(MSC_A_EV_FROM_T_SEND_END_SIGNAL_REQUEST),
OSMO_VALUE_STRING(MSC_A_EV_COMPLETE_LAYER_3_OK),
OSMO_VALUE_STRING(MSC_A_EV_CLASSMARK_UPDATE),
OSMO_VALUE_STRING(MSC_A_EV_AUTHENTICATED),
OSMO_VALUE_STRING(MSC_A_EV_TRANSACTION_ACCEPTED),
OSMO_VALUE_STRING(MSC_A_EV_CN_CLOSE),
OSMO_VALUE_STRING(MSC_A_EV_MO_CLOSE),
OSMO_VALUE_STRING(MSC_A_EV_UNUSED),
OSMO_VALUE_STRING(MSC_A_EV_HANDOVER_REQUIRED),
OSMO_VALUE_STRING(MSC_A_EV_HANDOVER_END),
{}
};
#define S(x) (1 << (x))
static const struct osmo_fsm_state msc_a_fsm_states[] = {
[MSC_A_ST_VALIDATE_L3] = {
.name = OSMO_STRINGIFY(MSC_A_ST_VALIDATE_L3),
.in_event_mask = 0
| S(MSC_A_EV_FROM_I_COMPLETE_LAYER_3)
| S(MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST)
| S(MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST)
| S(MSC_A_EV_COMPLETE_LAYER_3_OK)
| S(MSC_A_EV_MO_CLOSE)
| S(MSC_A_EV_CN_CLOSE)
| S(MSC_A_EV_UNUSED)
,
.out_state_mask = 0
| S(MSC_A_ST_VALIDATE_L3)
| S(MSC_A_ST_AUTH_CIPH)
| S(MSC_A_ST_RELEASING)
,
.action = msc_a_fsm_validate_l3,
},
[MSC_A_ST_AUTH_CIPH] = {
.name = OSMO_STRINGIFY(MSC_A_ST_AUTH_CIPH),
.in_event_mask = 0
| S(MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST)
| S(MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST)
| S(MSC_A_EV_AUTHENTICATED)
| S(MSC_A_EV_MO_CLOSE)
| S(MSC_A_EV_CN_CLOSE)
| S(MSC_A_EV_UNUSED)
,
.out_state_mask = 0
| S(MSC_A_ST_WAIT_CLASSMARK_UPDATE)
| S(MSC_A_ST_AUTHENTICATED)
| S(MSC_A_ST_RELEASING)
,
.action = msc_a_fsm_auth_ciph,
},
[MSC_A_ST_WAIT_CLASSMARK_UPDATE] = {
.name = OSMO_STRINGIFY(MSC_A_ST_WAIT_CLASSMARK_UPDATE),
.in_event_mask = 0
| S(MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST)
| S(MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST)
| S(MSC_A_EV_CLASSMARK_UPDATE)
| S(MSC_A_EV_MO_CLOSE)
| S(MSC_A_EV_CN_CLOSE)
,
.out_state_mask = 0
| S(MSC_A_ST_AUTH_CIPH)
| S(MSC_A_ST_RELEASING)
,
.action = msc_a_fsm_wait_classmark_update,
},
[MSC_A_ST_AUTHENTICATED] = {
.name = OSMO_STRINGIFY(MSC_A_ST_AUTHENTICATED),
/* allow everything to release for any odd behavior */
.in_event_mask = 0
| S(MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST)
| S(MSC_A_EV_FROM_I_PREPARE_SUBSEQUENT_HANDOVER_REQUEST)
| S(MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST)
| S(MSC_A_EV_TRANSACTION_ACCEPTED)
| S(MSC_A_EV_MO_CLOSE)
| S(MSC_A_EV_CN_CLOSE)
| S(MSC_A_EV_UNUSED)
,
.out_state_mask = 0
| S(MSC_A_ST_RELEASING)
| S(MSC_A_ST_COMMUNICATING)
,
.onenter = msc_a_fsm_authenticated_enter,
.action = msc_a_fsm_authenticated,
},
[MSC_A_ST_COMMUNICATING] = {
.name = OSMO_STRINGIFY(MSC_A_ST_COMMUNICATING),
/* allow everything to release for any odd behavior */
.in_event_mask = 0
| S(MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST)
| S(MSC_A_EV_FROM_I_PREPARE_SUBSEQUENT_HANDOVER_REQUEST)
| S(MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST)
| S(MSC_A_EV_FROM_T_PREPARE_HANDOVER_RESPONSE)
| S(MSC_A_EV_FROM_T_PREPARE_HANDOVER_FAILURE)
| S(MSC_A_EV_FROM_T_PROCESS_ACCESS_SIGNALLING_REQUEST)
| S(MSC_A_EV_FROM_T_SEND_END_SIGNAL_REQUEST)
| S(MSC_A_EV_TRANSACTION_ACCEPTED)
| S(MSC_A_EV_MO_CLOSE)
| S(MSC_A_EV_CN_CLOSE)
| S(MSC_A_EV_UNUSED)
| S(MSC_EV_CALL_LEG_RTP_LOCAL_ADDR_AVAILABLE)
| S(MSC_EV_CALL_LEG_RTP_COMPLETE)
| S(MSC_EV_CALL_LEG_TERM)
| S(MSC_MNCC_EV_CALL_ENDED)
| S(MSC_A_EV_HANDOVER_REQUIRED)
| S(MSC_A_EV_HANDOVER_END)
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
,
.out_state_mask = 0
| S(MSC_A_ST_RELEASING)
,
.action = msc_a_fsm_communicating,
},
[MSC_A_ST_RELEASING] = {
.name = OSMO_STRINGIFY(MSC_A_ST_RELEASING),
.in_event_mask = 0
| S(MSC_A_EV_FROM_I_PROCESS_ACCESS_SIGNALLING_REQUEST)
| S(MSC_A_EV_FROM_I_SEND_END_SIGNAL_REQUEST)
| S(MSC_A_EV_UNUSED)
| S(MSC_EV_CALL_LEG_TERM)
| S(MSC_MNCC_EV_CALL_ENDED)
| S(MSC_A_EV_HANDOVER_END)
| S(MSC_A_EV_CN_CLOSE)
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
,
.out_state_mask = 0
| S(MSC_A_ST_RELEASED)
,
.onenter = msc_a_fsm_releasing_onenter,
.action = msc_a_fsm_releasing,
},
[MSC_A_ST_RELEASED] = {
.name = OSMO_STRINGIFY(MSC_A_ST_RELEASED),
.in_event_mask = 0
| S(MSC_A_EV_UNUSED)
,
.onenter = msc_a_fsm_released_onenter,
.action = msc_a_fsm_released,
},
};
static struct osmo_fsm msc_a_fsm = {
.name = "msc_a",
.states = msc_a_fsm_states,
.num_states = ARRAY_SIZE(msc_a_fsm_states),
.log_subsys = DMSC,
.event_names = msc_a_fsm_event_names,
.timer_cb = msc_a_fsm_timer_cb,
.cleanup = msc_a_fsm_cleanup,
};
static __attribute__((constructor)) void msc_a_fsm_init()
{
OSMO_ASSERT(osmo_fsm_register(&msc_a_fsm) == 0);
}
static int msc_a_use_cb(struct osmo_use_count_entry *e, int32_t old_use_count, const char *file, int line)
{
struct msc_a *msc_a = e->use_count->talloc_object;
char buf[128];
int32_t total;
int level;
if (!e->use)
return -EINVAL;
total = osmo_use_count_total(&msc_a->use_count);
if (total == 0
|| (total == 1 && old_use_count == 0 && e->count == 1))
level = LOGL_INFO;
else
level = LOGL_DEBUG;
LOG_MSC_A_CAT_SRC(msc_a, DREF, level, file, line, "%s %s: now used by %s\n",
(e->count - old_use_count) > 0? "+" : "-", e->use,
osmo_use_count_name_buf(buf, sizeof(buf), &msc_a->use_count));
if (e->count < 0)
return -ERANGE;
msc_a->max_total_use_count = OSMO_MAX(msc_a->max_total_use_count, total);
if (total == 0)
osmo_fsm_inst_dispatch(msc_a->c.fi, MSC_A_EV_UNUSED, NULL);
return 0;
}
struct msc_a *msc_a_alloc(struct msub *msub, struct ran_infra *ran)
{
struct msc_a *msc_a = msub_role_alloc(msub, MSC_ROLE_A, &msc_a_fsm, struct msc_a, ran);
msc_a->use_count = (struct osmo_use_count){
.talloc_object = msc_a,
.use_cb = msc_a_use_cb,
};
osmo_use_count_make_static_entries(&msc_a->use_count, msc_a->use_count_buf, ARRAY_SIZE(msc_a->use_count_buf));
/* Start timeout for first state */
msc_a_state_chg_always(msc_a, MSC_A_ST_VALIDATE_L3);
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return msc_a;
}
bool msc_a_is_establishing_auth_ciph(const struct msc_a *msc_a)
{
if (!msc_a || !msc_a->c.fi)
return false;
return msc_a->c.fi->state == MSC_A_ST_AUTH_CIPH;
}
const struct value_string complete_layer3_type_names[] = {
{ COMPLETE_LAYER3_NONE, "NONE" },
{ COMPLETE_LAYER3_LU, "LU" },
{ COMPLETE_LAYER3_CM_SERVICE_REQ, "CM_SERVICE_REQ" },
{ COMPLETE_LAYER3_PAGING_RESP, "PAGING_RESP" },
{ COMPLETE_LAYER3_CM_RE_ESTABLISH_REQ, "CM_RE_ESTABLISH_REQ" },
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
{ 0, NULL }
};
#define _msc_a_update_id(MSC_A, FMT, ARGS ...) \
do { \
if (osmo_fsm_inst_update_id_f(msc_a->c.fi, FMT ":%s:%s", \
## ARGS, \
msub_ran_conn_name(msc_a->c.msub), \
complete_layer3_type_name(msc_a->complete_layer3_type)) \
== 0) { \
struct vlr_subscr *_vsub = msc_a_vsub(MSC_A); \
if (_vsub) { \
if (_vsub->lu_fsm) \
osmo_fsm_inst_update_id(_vsub->lu_fsm, (MSC_A)->c.fi->id); \
if (_vsub->auth_fsm) \
osmo_fsm_inst_update_id(_vsub->auth_fsm, (MSC_A)->c.fi->id); \
if (_vsub->proc_arq_fsm) \
osmo_fsm_inst_update_id(_vsub->proc_arq_fsm, (MSC_A)->c.fi->id); \
} \
LOG_MSC_A(MSC_A, LOGL_DEBUG, "Updated ID\n"); \
} \
/* otherwise osmo_fsm_inst_update_id_f() will log an error. */ \
} while (0)
/* Compose an ID almost like gsm48_mi_to_string(), but print the MI type along, and print a TMSI as hex. */
void msc_a_update_id_from_mi(struct msc_a *msc_a, const struct osmo_mobile_identity *mi)
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
{
_msc_a_update_id(msc_a, "%s", osmo_mobile_identity_to_str_c(OTC_SELECT, mi));
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
}
/* Update msc_a->fi id string from current msc_a->vsub and msc_a->complete_layer3_type. */
void msc_a_update_id(struct msc_a *msc_a)
{
_msc_a_update_id(msc_a, "%s", vlr_subscr_name(msc_a_vsub(msc_a)));
}
/* Iterate all msc_a instances that are relevant for this subscriber, and update FSM ID strings for all of the FSM
* instances. */
void msc_a_update_id_for_vsub(struct vlr_subscr *for_vsub)
{
struct msub *msub;
llist_for_each_entry(msub, &msub_list, entry) {
struct vlr_subscr *vsub = msub_vsub(msub);
if (vsub != for_vsub)
continue;
msc_a_update_id(msub_msc_a(msub));
}
}
static bool msg_is_initially_permitted(const struct gsm48_hdr *hdr)
{
uint8_t pdisc = gsm48_hdr_pdisc(hdr);
uint8_t msg_type = gsm48_hdr_msg_type(hdr);
switch (pdisc) {
case GSM48_PDISC_MM:
switch (msg_type) {
case GSM48_MT_MM_LOC_UPD_REQUEST:
case GSM48_MT_MM_CM_SERV_REQ:
case GSM48_MT_MM_CM_REEST_REQ:
case GSM48_MT_MM_AUTH_RESP:
case GSM48_MT_MM_AUTH_FAIL:
case GSM48_MT_MM_ID_RESP:
case GSM48_MT_MM_TMSI_REALL_COMPL:
case GSM48_MT_MM_IMSI_DETACH_IND:
return true;
default:
break;
}
break;
case GSM48_PDISC_RR:
switch (msg_type) {
/* GSM48_MT_RR_CIPH_M_COMPL is actually handled in bssmap_rx_ciph_compl() and gets redirected in the
* BSSAP layer to ran_conn_cipher_mode_compl() (before this here is reached) */
case GSM48_MT_RR_PAG_RESP:
case GSM48_MT_RR_CIPH_M_COMPL:
return true;
default:
break;
}
break;
default:
break;
}
return false;
}
/* Main entry point for GSM 04.08/44.008 Layer 3 data (e.g. from the BSC). */
int msc_a_up_l3(struct msc_a *msc_a, struct msgb *msg)
{
struct gsm48_hdr *gh;
uint8_t pdisc;
int rc;
struct vlr_subscr *vsub = msc_a_vsub(msc_a);
int is_r99;
OSMO_ASSERT(msg->l3h);
OSMO_ASSERT(msg);
gh = msgb_l3(msg);
pdisc = gsm48_hdr_pdisc(gh);
LOG_MSC_A_CAT(msc_a, DRLL, LOGL_DEBUG, "Dispatching 04.08 message: %s %s\n",
gsm48_pdisc_name(pdisc), gsm48_pdisc_msgtype_name(pdisc, gsm48_hdr_msg_type(gh)));
/* To evaluate the 3GPP TS 24.007 Duplicate Detection, we need Classmark information on whether the MS is R99
* capable. If the subscriber is already actively connected, the Classmark information is stored with the
* vlr_subscr. Otherwise, this *must* be a Complete Layer 3 with Classmark info. */
if (vsub)
is_r99 = osmo_gsm48_classmark_is_r99(&vsub->classmark) ? 1 : 0;
else
is_r99 = compl_l3_msg_is_r99(msg);
if (is_r99 < 0) {
LOG_MSC_A(msc_a, LOGL_ERROR,
"No Classmark Information, dropping non-Complete-Layer3 message: %s\n",
gsm48_pdisc_msgtype_name(pdisc, gsm48_hdr_msg_type(gh)));
return -EACCES;
}
if (is_r99 >= 0
&& ran_dec_dtap_undup_is_duplicate(msc_a->c.fi, msc_a->n_sd_next, is_r99 ? true : false, msg)) {
LOG_MSC_A(msc_a, LOGL_DEBUG, "Dropping duplicate message"
" (3GPP TS 24.007 11.2.3.2 Message Type Octet / Duplicate Detection)\n");
return 0;
}
if (!msc_a_is_accepted(msc_a)
&& !msg_is_initially_permitted(gh)) {
LOG_MSC_A(msc_a, LOGL_ERROR,
"Message not permitted for initial conn: %s\n",
gsm48_pdisc_msgtype_name(pdisc, gsm48_hdr_msg_type(gh)));
return -EACCES;
}
if (vsub && vsub->cs.attached_via_ran != msc_a->c.ran->type) {
LOG_MSC_A(msc_a, LOGL_ERROR,
"Illegal situation: RAN type mismatch:"
" attached via %s, received message via %s\n",
osmo_rat_type_name(vsub->cs.attached_via_ran),
osmo_rat_type_name(msc_a->c.ran->type));
return -EACCES;
}
#if 0
if (silent_call_reroute(conn, msg))
return silent_call_rx(conn, msg);
#endif
switch (pdisc) {
case GSM48_PDISC_CC:
rc = gsm0408_rcv_cc(msc_a, msg);
break;
case GSM48_PDISC_MM:
rc = gsm0408_rcv_mm(msc_a, msg);
break;
case GSM48_PDISC_RR:
rc = gsm0408_rcv_rr(msc_a, msg);
break;
case GSM48_PDISC_SMS:
rc = gsm0411_rcv_sms(msc_a, msg);
break;
case GSM48_PDISC_MM_GPRS:
case GSM48_PDISC_SM_GPRS:
LOG_MSC_A_CAT(msc_a, DRLL, LOGL_NOTICE, "Unimplemented "
"GSM 04.08 discriminator 0x%02x\n", pdisc);
rc = -ENOTSUP;
break;
case GSM48_PDISC_NC_SS:
rc = gsm0911_rcv_nc_ss(msc_a, msg);
break;
case GSM48_PDISC_TEST:
rc = gsm0414_rcv_test(msc_a, msg);
break;
default:
LOG_MSC_A_CAT(msc_a, DRLL, LOGL_NOTICE, "Unknown "
"GSM 04.08 discriminator 0x%02x\n", pdisc);
rc = -EINVAL;
break;
}
return rc;
}
static void msc_a_up_call_assignment_complete(struct msc_a *msc_a, const struct ran_msg *ac)
{
struct gsm_trans *cc_trans = msc_a->cc.active_trans;
struct rtp_stream *rtps_to_ran = msc_a->cc.call_leg ? msc_a->cc.call_leg->rtp[RTP_TO_RAN] : NULL;
if (!rtps_to_ran) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Rx Assignment Complete, but no RTP stream is set up\n");
return;
}
if (!cc_trans) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Rx Assignment Complete, but CC transaction is active\n");
return;
}
if (rtps_to_ran->use_osmux != ac->assignment_complete.osmux_present) {
LOG_MSC_A_CAT(msc_a, DCC, LOGL_ERROR, "Osmux usage ass request and complete don't match: %d vs %d\n",
rtps_to_ran->use_osmux, ac->assignment_complete.osmux_present);
call_leg_release(msc_a->cc.call_leg);
return;
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
/* Update RAN-side endpoint CI: */
rtp_stream_set_codec(rtps_to_ran, ac->assignment_complete.codec);
rtp_stream_set_remote_addr(rtps_to_ran, &ac->assignment_complete.remote_rtp);
if (rtps_to_ran->use_osmux)
rtp_stream_set_remote_osmux_cid(rtps_to_ran,
ac->assignment_complete.osmux_cid);
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
rtp_stream_commit(rtps_to_ran);
/* Setup CN side endpoint CI:
* Now that
* - the first CI has been created and a definitive endpoint name is assigned to the call_leg's MGW
* endpoint,
* - the Assignment has chosen a speech codec
* go on to create the CN side RTP stream's CI. */
if (call_leg_ensure_ci(msc_a->cc.call_leg, RTP_TO_CN, cc_trans->callref, cc_trans,
&ac->assignment_complete.codec, NULL)) {
LOG_MSC_A_CAT(msc_a, DCC, LOGL_ERROR, "Error creating MGW CI towards CN\n");
call_leg_release(msc_a->cc.call_leg);
return;
}
}
static void msc_a_up_call_assignment_failure(struct msc_a *msc_a, const struct ran_msg *af)
{
struct gsm_trans *trans;
/* For a normal voice call, there will be an rtp_stream FSM. */
if (msc_a->cc.call_leg && msc_a->cc.call_leg->rtp[RTP_TO_RAN]) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Assignment Failure, releasing call\n");
rtp_stream_release(msc_a->cc.call_leg->rtp[RTP_TO_RAN]);
return;
}
/* Otherwise, a silent call might be active */
trans = trans_find_by_type(msc_a, TRANS_SILENT_CALL);
if (trans) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Assignment Failure, releasing silent call\n");
trans_free(trans);
return;
}
/* Neither a voice call nor silent call assignment. Assume the worst and detach. */
msc_a_release_cn(msc_a);
}
static void msc_a_up_classmark_update(struct msc_a *msc_a, const struct osmo_gsm48_classmark *classmark,
struct osmo_gsm48_classmark *dst)
{
if (!dst) {
struct vlr_subscr *vsub = msc_a_vsub(msc_a);
if (!vsub)
dst = &msc_a->temporary_classmark;
else
dst = &vsub->classmark;
}
LOG_MSC_A(msc_a, LOGL_DEBUG, "A5 capabilities received from Classmark Update: %s\n",
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
osmo_gsm48_classmark_a5_name(classmark));
osmo_gsm48_classmark_update(dst, classmark);
/* bump subscr conn FSM in case it is waiting for a Classmark Update */
if (msc_a->c.fi->state == MSC_A_ST_WAIT_CLASSMARK_UPDATE)
osmo_fsm_inst_dispatch(msc_a->c.fi, MSC_A_EV_CLASSMARK_UPDATE, NULL);
}
static void msc_a_up_sapi_n_reject(struct msc_a *msc_a, const struct ran_msg *msg)
{
int sapi = msg->sapi_n_reject.dlci & 0x7;
if (sapi == UM_SAPI_SMS)
gsm411_sapi_n_reject(msc_a);
}
static int msc_a_up_ho(struct msc_a *msc_a, const struct msc_a_ran_dec_data *d, uint32_t ho_fi_event)
{
if (!msc_a->ho.fi) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Rx Handover message, but no Handover ongoing: %s\n", d->ran_dec->msg_name);
return -EINVAL;
}
return osmo_fsm_inst_dispatch(msc_a->ho.fi, ho_fi_event, (void*)d);
}
int msc_a_ran_dec_from_msc_i(struct msc_a *msc_a, struct msc_a_ran_dec_data *d)
{
struct vlr_subscr *vsub = msc_a_vsub(msc_a);
struct gsm_network *net = msc_a_net(msc_a);
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
const struct ran_msg *msg = d->ran_dec;
int rc = -99;
switch (msg->msg_type) {
case RAN_MSG_COMPL_L3:
/* In case the cell_id from Complete Layer 3 Information lacks a PLMN, write the configured PLMN code
* into msc_a->via_cell. Then overwrite with those bits obtained from Complete Layer 3 Information. */
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
msc_a->via_cell = (struct osmo_cell_global_id){
.lai.plmn = msc_a_net(msc_a)->plmn,
};
gsm0808_cell_id_to_cgi(&msc_a->via_cell, msg->compl_l3.cell_id);
rc = msc_a_up_l3(msc_a, msg->compl_l3.msg);
if (!rc) {
struct ran_conn *conn = msub_ran_conn(msc_a->c.msub);
if (conn)
ran_peer_cells_seen_add(conn->ran_peer, msg->compl_l3.cell_id);
}
break;
case RAN_MSG_DTAP:
rc = msc_a_up_l3(msc_a, msg->dtap);
break;
case RAN_MSG_CLEAR_REQUEST:
rc = osmo_fsm_inst_dispatch(msc_a->c.fi, MSC_A_EV_MO_CLOSE, NULL);
break;
case RAN_MSG_CLEAR_COMPLETE:
switch (msc_a->c.fi->state) {
case MSC_A_ST_RELEASING:
msc_a_put_all(msc_a, MSC_A_USE_WAIT_CLEAR_COMPLETE);
msc_a_state_chg(msc_a, MSC_A_ST_RELEASED);
break;
case MSC_A_ST_RELEASED:
break;
default:
LOG_MSC_A(msc_a, LOGL_ERROR, "Received Clear Complete event, but did not send Clear Command\n");
msc_a_state_chg(msc_a, MSC_A_ST_RELEASING);
break;
}
rc = 0;
break;
case RAN_MSG_CLASSMARK_UPDATE:
msc_a_up_classmark_update(msc_a, msg->classmark_update.classmark, NULL);
rc = 0;
break;
case RAN_MSG_CIPHER_MODE_COMPLETE:
/* Remember what Ciphering was negotiated (e.g. for Handover) */
if (msg->cipher_mode_complete.alg_id) {
msc_a->geran_encr.alg_id = msg->cipher_mode_complete.alg_id;
LOG_MSC_A(msc_a, LOGL_DEBUG, "Cipher Mode Complete: chosen encryption algorithm: A5/%u\n",
msc_a->geran_encr.alg_id - 1);
}
if (msc_a->c.ran->type == OSMO_RAT_UTRAN_IU) {
int16_t utran_encryption;
/* utran: ensure chosen ciphering mode is allowed
* If the IE is missing (utran_encryption == -1), parse it as no encryption */
utran_encryption = msg->cipher_mode_complete.utran_encryption;
if (utran_encryption == -1)
utran_encryption = 0;
if ((net->uea_encryption_mask & (1 << utran_encryption)) == 0) {
/* cipher disallowed */
LOG_MSC_A(msc_a, LOGL_ERROR, "Cipher Mode Complete: RNC chosen forbidden ciphering UEA%d\n",
msg->cipher_mode_complete.utran_encryption);
vlr_subscr_rx_ciph_res(vsub, VLR_CIPH_REJECT);
rc = 0;
break;
}
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
vlr_subscr_rx_ciph_res(vsub, VLR_CIPH_COMPL);
rc = 0;
fix error on BSSMAP Cipher Mode Complete L3 msg IE When an MS returns the IMEISV in the BSSMAP Cipher Mode Complete message in the Layer 3 Message Contents IE, do not re-invoke the decode_cb() a second time, but instead point to it from the ran_msg.cipher_mode_complete struct. When the MSC-A decodes the Ciphering Mode Complete message, it always wants to also decode the enclosed DTAP from the Layer 3 Message Contents IE. However, when the MSC-I preliminarily decodes messages, it often just wants to identify specific messages without fully acting on them, let alone dispatching RAN_UP_L2 events more than once. So leave it up to the supplied decode_cb passed to ran_dec_l2() implementations to decide whether to decode the DTAP. In msc_a.c hence evaluate the DTAP by passing a msgb to msc_a_up_l3(), which will evaluate the RR Ciphering Mode Complete message found in the BSSMAP Cipher Mode Complete's Layer 3 Message Contents IE. Particularly, the previous choice of calling the decode_cb a second time for the enclosed DTAP caused a header/length parsing error: the second decode_cb call tried to mimick DTAP by overwriting the l3h pointer and truncating the length of the msgb, but subsequently ran_a_decode_l2() would again derive the l3h from the l2h, obliterating the intended re-interpretation as DTAP, and hence the previous truncation caused error messages on each and every Cipher Mode Complete message, like: DBSSAP ERROR libmsc/ran_msg_a.c:764 msc_a(IMSI-26242340300XXXX:MSISDN-XXXX:TMSI-0xA73E055A:GERAN-A-77923:LU)[0x5563947521e0]{MSC_A_ST_AUTH_CIPH}: RAN decode: BSSMAP: BSSMAP data truncated, discarding message This error was seen a lot at CCCamp2019. Modifying the msgb was a bad idea to begin with, the approach taken in this patch is much cleaner. Note that apparently many phones include the IMEISV in the Cipher Mode Complete message even though the BSSMAP Cipher Mode Command did not include the Cipher Response Mode IE. So, even though we did not specifically ask for the Cipher Mode Complete to include any identity, many MS default to including the IMEISV of their own accord. Reproduce: attach to osmo-msc with ciphering enabled using a Samsung Galaxy S4mini. Related: OS#4168 Change-Id: Icd8dad18d6dda24d075dd8da72c3d6db1302090d
2019-08-28 22:10:49 +00:00
/* Evaluate enclosed L3 message, typically Identity Response (IMEISV) */
if (msg->cipher_mode_complete.l3_msg) {
unsigned char *data = (unsigned char*)(msg->cipher_mode_complete.l3_msg->val);
uint16_t len = msg->cipher_mode_complete.l3_msg->len;
struct msgb *dtap = msgb_alloc(len, "DTAP from Cipher Mode Complete");
unsigned char *pos = msgb_put(dtap, len);
memcpy(pos, data, len);
dtap->l3h = pos;
rc = msc_a_up_l3(msc_a, dtap);
msgb_free(dtap);
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
break;
case RAN_MSG_CIPHER_MODE_REJECT:
vlr_subscr_rx_ciph_res(vsub, VLR_CIPH_REJECT);
rc = 0;
break;
case RAN_MSG_ASSIGNMENT_COMPLETE:
msc_a_up_call_assignment_complete(msc_a, msg);
rc = 0;
break;
case RAN_MSG_ASSIGNMENT_FAILURE:
msc_a_up_call_assignment_failure(msc_a, msg);
rc = 0;
break;
case RAN_MSG_SAPI_N_REJECT:
msc_a_up_sapi_n_reject(msc_a, msg);
rc = 0;
break;
case RAN_MSG_HANDOVER_PERFORMED:
/* The BSS lets us know that a handover happened within the BSS, which doesn't concern us. */
LOG_MSC_A(msc_a, LOGL_ERROR, "'Handover Performed' handling not implemented\n");
break;
case RAN_MSG_HANDOVER_REQUIRED:
/* The BSS lets us know that it wants to handover to a different cell */
rc = osmo_fsm_inst_dispatch(msc_a->c.fi, MSC_A_EV_HANDOVER_REQUIRED, (void*)&msg->handover_required);
break;
case RAN_MSG_HANDOVER_FAILURE:
rc = msc_a_up_ho(msc_a, d, MSC_HO_EV_RX_FAILURE);
break;
case RAN_MSG_LCLS_STATUS:
/* The BSS sends us LCLS_STATUS. We do nothing for now, but it is not an error. */
LOG_MSC_A(msc_a, LOGL_DEBUG, "LCLS_STATUS (%s) received from MSC-I\n",
gsm0808_lcls_status_name(msg->lcls_status.status));
rc = 0;
break;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
default:
LOG_MSC_A(msc_a, LOGL_ERROR, "Message from MSC-I not implemented: %s\n", ran_msg_type_name(msg->msg_type));
rc = -ENOTSUP;
break;
}
return rc;
}
static int msc_a_ran_dec_from_msc_t(struct msc_a *msc_a, struct msc_a_ran_dec_data *d)
{
struct msc_t *msc_t = msc_a_msc_t(msc_a);
int rc = -99;
if (!msc_t) {
LOG_MSC_A(msc_a, LOGL_ERROR, "Rx message from MSC-T role, but I have no active MSC-T role.\n");
return -EINVAL;
}
OSMO_ASSERT(d->ran_dec);
switch (d->ran_dec->msg_type) {
case RAN_MSG_CLEAR_REQUEST:
rc = osmo_fsm_inst_dispatch(msc_t->c.fi, MSC_T_EV_MO_CLOSE, NULL);
break;
case RAN_MSG_CLEAR_COMPLETE:
rc = osmo_fsm_inst_dispatch(msc_t->c.fi, MSC_T_EV_CLEAR_COMPLETE, NULL);
break;
case RAN_MSG_CLASSMARK_UPDATE:
msc_a_up_classmark_update(msc_a, d->ran_dec->classmark_update.classmark, &msc_t->classmark);
rc = 0;
break;
case RAN_MSG_HANDOVER_REQUEST_ACK:
/* new BSS accepts Handover */
rc = msc_a_up_ho(msc_a, d, MSC_HO_EV_RX_REQUEST_ACK);
break;
case RAN_MSG_HANDOVER_DETECT:
/* new BSS signals the MS is DETECTed on the new lchan */
rc = msc_a_up_ho(msc_a, d, MSC_HO_EV_RX_DETECT);
break;
case RAN_MSG_HANDOVER_COMPLETE:
/* new BSS signals the MS has fully moved to the new lchan */
rc = msc_a_up_ho(msc_a, d, MSC_HO_EV_RX_COMPLETE);
break;
case RAN_MSG_HANDOVER_FAILURE:
rc = msc_a_up_ho(msc_a, d, MSC_HO_EV_RX_FAILURE);
break;
default:
LOG_MSC_A(msc_a, LOGL_ERROR, "Message from MSC-T not implemented: %s\n",
ran_msg_type_name(d->ran_dec->msg_type));
rc = -ENOTSUP;
break;
}
return rc;
}
int msc_a_ran_decode_cb(struct osmo_fsm_inst *msc_a_fi, void *data, const struct ran_msg *msg)
{
struct msc_a *msc_a = msc_a_fi_priv(msc_a_fi);
struct msc_a_ran_dec_data *d = data;
int rc = -99;
d->ran_dec = msg;
switch (d->from_role) {
case MSC_ROLE_I:
LOG_MSC_A(msc_a, LOGL_DEBUG, "RAN decode: %s\n", msg->msg_name ? : ran_msg_type_name(msg->msg_type));
rc = msc_a_ran_dec_from_msc_i(msc_a, d);
break;
case MSC_ROLE_T:
LOG_MSC_A(msc_a, LOGL_DEBUG, "RAN decode from MSC-T: %s\n",
msg->msg_name ? : ran_msg_type_name(msg->msg_type));
rc = msc_a_ran_dec_from_msc_t(msc_a, d);
break;
default:
LOG_MSC_A(msc_a, LOGL_ERROR, "Message from invalid role %s: %s\n", msc_role_name(d->from_role),
ran_msg_type_name(msg->msg_type));
return -ENOTSUP;
}
if (rc)
LOG_MSC_A(msc_a, LOGL_ERROR, "RAN decode error (rc=%d) for %s from %s\n", rc, ran_msg_type_name(msg->msg_type),
msc_role_name(d->from_role));
return rc;
}
/* Your typical DTAP via FORWARD_ACCESS_SIGNALLING_REQUEST */
int _msc_a_ran_down(struct msc_a *msc_a, enum msc_role to_role, const struct ran_msg *ran_msg,
const char *file, int line)
{
return _msc_a_msg_down(msc_a, to_role, msub_role_to_role_event(msc_a->c.msub, MSC_ROLE_A, to_role),
ran_msg, file, line);
}
/* To transmit more complex events than just FORWARD_ACCESS_SIGNALLING_REQUEST, e.g. an
* MSC_T_EV_FROM_A_PREPARE_HANDOVER_REQUEST */
int _msc_a_msg_down(struct msc_a *msc_a, enum msc_role to_role, uint32_t to_role_event,
const struct ran_msg *ran_msg,
const char *file, int line)
{
struct an_apdu an_apdu = {
.an_proto = msc_a->c.ran->an_proto,
.msg = msc_role_ran_encode(msc_a->c.fi, ran_msg),
};
if (!an_apdu.msg)
return -EIO;
return _msub_role_dispatch(msc_a->c.msub, to_role, to_role_event, &an_apdu, file, line);
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
}
int msc_a_tx_dtap_to_i(struct msc_a *msc_a, struct msgb *dtap)
{
struct ran_msg ran_msg;
struct gsm48_hdr *gh = msgb_l3(dtap) ? : dtap->data;
uint8_t pdisc = gsm48_hdr_pdisc(gh);
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
fix segfault: don't send CC REL on NULL msc_a Apparently, if a conn disappears during an ongoing call, the CC code tried to send a CC REL on a NULL msc_a during cleanup, which lead to a crash (cccamp2019). Guard against that. Crash: #0 msc_a_tx_dtap_to_i (msc_a=0x0, dtap=0x55a4bf2fa0f0) at ../../../../src/osmo-msc/src/libmsc/msc_a.c:1565 #1 0x000055a4be1bb03c in trans_tx_gsm48 (trans=0x55a4bf2d52a0, trans=0x55a4bf2d52a0, trans=0x55a4bf2d52a0, msg=<optimized out>) at ../../../../src/osmo-msc/src/libmsc/gsm_04_08_cc.c:82 #2 gsm48_cc_tx_release (trans=trans@entry=0x55a4bf2d52a0, arg=arg@entry=0x7ffdd731a0e0) at ../../../../src/osmo-msc/src/libmsc/gsm_04_08_cc.c:1101 #3 0x000055a4be1bee65 in _gsm48_cc_trans_free (trans=trans@entry=0x55a4bf2d52a0) at ../../../../src/osmo-msc/src/libmsc/gsm_04_08_cc.c:278 #4 0x000055a4be1ab654 in trans_free (trans=trans@entry=0x55a4bf2d52a0) at ../../../../src/osmo-msc/src/libmsc/transaction.c:170 #5 0x000055a4be1bd091 in mncc_tx_to_gsm_cc (net=<optimized out>, msg=msg@entry=0x55a4bf2d3b68) at ../../../../src/osmo-msc/src/libmsc/gsm_04_08_cc.c:1971 #6 0x000055a4be1bf1e5 in mncc_tx_to_cc (net=<optimized out>, arg=arg@entry=0x55a4bf2d3b68) at ../../../../src/osmo-msc/src/libmsc/gsm_04_08_cc.c:2049 #7 0x000055a4be18ed63 in mncc_sock_read (bfd=0x55a4bf2563b8, bfd=0x55a4bf2563b8) at ../../../../src/osmo-msc/src/libmsc/mncc_sock.c:121 #8 mncc_sock_cb (bfd=0x55a4bf2563b8, flags=1) at ../../../../src/osmo-msc/src/libmsc/mncc_sock.c:189 #9 0x00007fcfad607ce1 in osmo_fd_disp_fds (_eset=0x7ffdd731a9a0, _wset=0x7ffdd731a920, _rset=0x7ffdd731a8a0) at ../../../src/libosmocore/src/select.c:223 #10 osmo_select_main (polling=<optimized out>) at ../../../src/libosmocore/src/select.c:263 #11 0x000055a4be17dd56 in main (argc=3, argv=<optimized out>) at ../../../../src/osmo-msc/src/osmo-msc/msc_main.c:723 Change-Id: Ia1bb0410ad0618c182a5f6da06af342b6d483eff
2019-08-21 14:56:41 +00:00
if (!msc_a) {
LOGP(DMSC, LOGL_ERROR, "Attempt to send DTAP to NULL MSC-A, dropping message: %s %s\n",
gsm48_pdisc_name(pdisc), gsm48_pdisc_msgtype_name(pdisc, gsm48_hdr_msg_type(gh)));
msgb_free(dtap);
return -EIO;
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
if (msc_a->c.ran->type == OSMO_RAT_EUTRAN_SGS) {
/* The SGs connection to the MME always is at the MSC-A. */
return sgs_iface_tx_dtap_ud(msc_a, dtap);
}
LOG_MSC_A(msc_a, LOGL_DEBUG, "Sending DTAP: %s %s\n",
gsm48_pdisc_name(pdisc), gsm48_pdisc_msgtype_name(pdisc, gsm48_hdr_msg_type(gh)));
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
ran_msg = (struct ran_msg){
.msg_type = RAN_MSG_DTAP,
.dtap = dtap,
};
return msc_a_ran_down(msc_a, MSC_ROLE_I, &ran_msg);
}
struct msc_a *msc_a_for_vsub(const struct vlr_subscr *vsub, bool valid_conn_only)
{
struct msc_a *msc_a = msub_msc_a(msub_for_vsub(vsub));
if (valid_conn_only && !msc_a_is_accepted(msc_a))
return NULL;
return msc_a;
}
int msc_tx_common_id(struct msc_a *msc_a, enum msc_role to_role)
{
struct vlr_subscr *vsub = msc_a_vsub(msc_a);
if (vsub == NULL)
return -ENODEV;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
struct ran_msg msg = {
.msg_type = RAN_MSG_COMMON_ID,
.common_id = {
.imsi = vsub->imsi,
.last_eutran_plmn_present = vsub->sgs.last_eutran_plmn_present,
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
},
};
if (vsub->sgs.last_eutran_plmn_present) {
memcpy(&msg.common_id.last_eutran_plmn, &vsub->sgs.last_eutran_plmn,
sizeof(vsub->sgs.last_eutran_plmn));
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return msc_a_ran_down(msc_a, to_role, &msg);
}
static int msc_a_start_assignment(struct msc_a *msc_a, struct gsm_trans *cc_trans)
{
struct call_leg *cl = msc_a->cc.call_leg;
struct msc_i *msc_i = msc_a_msc_i(msc_a);
struct gsm_network *net = msc_a_net(msc_a);
enum mgcp_codecs codec, *codec_ptr;
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
OSMO_ASSERT(!msc_a->cc.active_trans);
msc_a->cc.active_trans = cc_trans;
OSMO_ASSERT(cc_trans && cc_trans->type == TRANS_CC);
if (!cl) {
cl = msc_a->cc.call_leg = call_leg_alloc(msc_a->c.fi,
MSC_EV_CALL_LEG_TERM,
MSC_EV_CALL_LEG_RTP_LOCAL_ADDR_AVAILABLE,
MSC_EV_CALL_LEG_RTP_COMPLETE);
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
OSMO_ASSERT(cl);
/* HACK: We put the connection in loopback mode at the beginning to
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
* trick the hNodeB into doing the IuUP negotiation with itself.
* This is a hack we need because osmo-mgw does not support IuUP yet, see OS#2459. */
if (msc_a->c.ran->type == OSMO_RAT_UTRAN_IU)
cl->crcx_conn_mode[RTP_TO_RAN] = MGCP_CONN_LOOPBACK;
}
if (net->use_osmux != OSMUX_USAGE_OFF) {
msc_i = msc_a_msc_i(msc_a);
if (msc_i->c.remote_to) {
/* TODO: investigate what to do in this case */
LOG_MSC_A(msc_a, LOGL_ERROR, "Osmux not yet supported for inter-MSC");
} else {
cl->ran_peer_supports_osmux = msc_i->ran_conn->ran_peer->remote_supports_osmux;
}
}
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
/* This will lead to either MSC_EV_CALL_LEG_LOCAL_ADDR_AVAILABLE or MSC_EV_CALL_LEG_TERM.
* If the local address is already known, then immediately trigger. */
if (call_leg_local_ip(cl, RTP_TO_RAN))
return osmo_fsm_inst_dispatch(msc_a->c.fi, MSC_EV_CALL_LEG_RTP_LOCAL_ADDR_AVAILABLE, cl->rtp[RTP_TO_RAN]);
if (msc_a->c.ran->type == OSMO_RAT_UTRAN_IU) {
codec = CODEC_IUFP;
codec_ptr = &codec;
} else {
codec_ptr = NULL;
}
return call_leg_ensure_ci(msc_a->cc.call_leg, RTP_TO_RAN, cc_trans->callref, cc_trans, codec_ptr, NULL);
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
}
int msc_a_try_call_assignment(struct gsm_trans *cc_trans)
{
struct msc_a *msc_a = cc_trans->msc_a;
OSMO_ASSERT(cc_trans->type == TRANS_CC);
if (msc_a->cc.active_trans == cc_trans) {
LOG_MSC_A(msc_a, LOGL_DEBUG, "Assignment for this trans already started earlier\n");
large refactoring: support inter-BSC and inter-MSC Handover 3GPP TS 49.008 '4.3 Roles of MSC-A, MSC-I and MSC-T' defines distinct roles: - MSC-A is responsible for managing subscribers, - MSC-I is the gateway to the RAN. - MSC-T is a second transitory gateway to another RAN during Handover. After inter-MSC Handover, the MSC-I is handled by a remote MSC instance, while the original MSC-A retains the responsibility of subscriber management. MSC-T exists in this patch but is not yet used, since Handover is only prepared for, not yet implemented. Facilitate Inter-MSC and inter-BSC Handover by the same internal split of MSC roles. Compared to inter-MSC Handover, mere inter-BSC has the obvious simplifications: - all of MSC-A, MSC-I and MSC-T roles will be served by the same osmo-msc instance, - messages between MSC-A and MSC-{I,T} don't need to be routed via E-interface (GSUP), - no call routing between MSC-A and -I via MNCC necessary. This is the largest code bomb I have submitted, ever. Out of principle, I apologize to everyone trying to read this as a whole. Unfortunately, I see no sense in trying to split this patch into smaller bits. It would be a huge amount of work to introduce these changes in separate chunks, especially if each should in turn be useful and pass all test suites. So, unfortunately, we are stuck with this code bomb. The following are some details and rationale for this rather huge refactoring: * separate MSC subscriber management from ran_conn struct ran_conn is reduced from the pivotal subscriber management entity it has been so far to a mere storage for an SCCP connection ID and an MSC subscriber reference. The new pivotal subscriber management entity is struct msc_a -- struct msub lists the msc_a, msc_i, msc_t roles, the vast majority of code paths however use msc_a, since MSC-A is where all the interesting stuff happens. Before handover, msc_i is an FSM implementation that encodes to the local ran_conn. After inter-MSC Handover, msc_i is a compatible but different FSM implementation that instead forwards via/from GSUP. Same goes for the msc_a struct: if osmo-msc is the MSC-I "RAN proxy" for a remote MSC-A role, the msc_a->fi is an FSM implementation that merely forwards via/from GSUP. * New SCCP implementation for RAN access To be able to forward BSSAP and RANAP messages via the GSUP interface, the individual message layers need to be cleanly separated. The IuCS implementation used until now (iu_client from libosmo-ranap) did not provide this level of separation, and needed a complete rewrite. It was trivial to implement this in such a way that both BSSAP and RANAP can be handled by the same SCCP code, hence the new SCCP-RAN layer also replaces BSSAP handling. sccp_ran.h: struct sccp_ran_inst provides an abstract handler for incoming RAN connections. A set of callback functions provides implementation specific details. * RAN Abstraction (BSSAP vs. RANAP) The common SCCP implementation did set the theme for the remaining refactoring: make all other MSC code paths entirely RAN-implementation-agnostic. ran_infra.c provides data structures that list RAN implementation specifics, from logging to RAN de-/encoding to SCCP callbacks and timers. A ran_infra pointer hence allows complete abstraction of RAN implementations: - managing connected RAN peers (BSC, RNC) in ran_peer.c, - classifying and de-/encoding RAN PDUs, - recording connected LACs and cell IDs and sending out Paging requests to matching RAN peers. * RAN RESET now also for RANAP ran_peer.c absorbs the reset_fsm from a_reset.c; in consequence, RANAP also supports proper RESET semantics now. Hence osmo-hnbgw now also needs to provide proper RESET handling, which it so far duly ignores. (TODO) * RAN de-/encoding abstraction The RAN abstraction mentioned above serves not only to separate RANAP and BSSAP implementations transparently, but also to be able to optionally handle RAN on distinct levels. Before Handover, all RAN messages are handled by the MSC-A role. However, after an inter-MSC Handover, a standalone MSC-I will need to decode RAN PDUs, at least in order to manage Assignment of RTP streams between BSS/RNC and MNCC call forwarding. ran_msg.h provides a common API with abstraction for: - receiving events from RAN, i.e. passing RAN decode from the BSC/RNC and MS/UE: struct ran_dec_msg represents RAN messages decoded from either BSSMAP or RANAP; - sending RAN events: ran_enc_msg is the counterpart to compose RAN messages that should be encoded to either BSSMAP or RANAP and passed down to the BSC/RNC and MS/UE. The RAN-specific implementations are completely contained by ran_msg_a.c and ran_msg_iu.c. In particular, Assignment and Ciphering have so far been distinct code paths for BSSAP and RANAP, with switch(via_ran){...} statements all over the place. Using RAN_DEC_* and RAN_ENC_* abstractions, these are now completely unified. Note that SGs does not qualify for RAN abstraction: the SGs interface always remains with the MSC-A role, and SGs messages follow quite distinct semantics from the fairly similar GERAN and UTRAN. * MGW and RTP stream management So far, managing MGW endpoints via MGCP was tightly glued in-between GSM-04.08-CC on the one and MNCC on the other side. Prepare for switching RTP streams between different RAN peers by moving to object-oriented implementations: implement struct call_leg and struct rtp_stream with distinct FSMs each. For MGW communication, use the osmo_mgcpc_ep API that has originated from osmo-bsc and recently moved to libosmo-mgcp-client for this purpose. Instead of implementing a sequence of events with code duplication for the RAN and CN sides, the idea is to manage each RTP stream separately by firing and receiving events as soon as codecs and RTP ports are negotiated, and letting the individual FSMs take care of the MGW management "asynchronously". The caller provides event IDs and an FSM instance that should be notified of RTP stream setup progress. Hence it becomes possible to reconnect RTP streams from one GSM-04.08-CC to another (inter-BSC Handover) or between CC and MNCC RTP peers (inter-MSC Handover) without duplicating the MGCP code for each transition. The number of FSM implementations used for MGCP handling may seem a bit of an overkill. But in fact, the number of perspectives on RTP forwarding are far from trivial: - an MGW endpoint is an entity with N connections, and MGCP "sessions" for configuring them by talking to the MGW; - an RTP stream is a remote peer connected to one of the endpoint's connections, which is asynchronously notified of codec and RTP port choices; - a call leg is the higher level view on either an MT or MO side of a voice call, a combination of two RTP streams to forward between two remote peers. BSC MGW PBX CI CI [MGW-endpoint] [--rtp_stream--] [--rtp_stream--] [----------------call_leg----------------] * Use counts Introduce using the new osmo_use_count API added to libosmocore for this purpose. Each use token has a distinct name in the logging, which can be a globally constant name or ad-hoc, like the local __func__ string constant. Use in the new struct msc_a, as well as change vlr_subscr to the new osmo_use_count API. * FSM Timeouts Introduce using the new osmo_tdef API, which provides a common VTY implementation for all timer numbers, and FSM state transitions with the correct timeout. Originated in osmo-bsc, recently moved to libosmocore. Depends: Ife31e6798b4e728a23913179e346552a7dd338c0 (libosmocore) Ib9af67b100c4583342a2103669732dab2e577b04 (libosmocore) Id617265337f09dfb6ddfe111ef5e578cd3dc9f63 (libosmocore) Ie9e2add7bbfae651c04e230d62e37cebeb91b0f5 (libosmo-sccp) I26be5c4b06a680f25f19797407ab56a5a4880ddc (osmo-mgw) Ida0e59f9a1f2dd18efea0a51680a67b69f141efa (osmo-mgw) I9a3effd38e72841529df6c135c077116981dea36 (osmo-mgw) Change-Id: I27e4988e0371808b512c757d2b52ada1615067bd
2018-12-07 13:47:34 +00:00
return 0;
}
if (msc_a->cc.active_trans) {
LOG_MSC_A(msc_a, LOGL_INFO, "Another call is already ongoing, not assigning yet\n");
return 0;
}
LOG_MSC_A(msc_a, LOGL_DEBUG, "Starting call assignment\n");
return msc_a_start_assignment(msc_a, cc_trans);
}
const char *msc_a_cm_service_type_to_use(enum osmo_cm_service_type cm_service_type)
{
switch (cm_service_type) {
case GSM48_CMSERV_MO_CALL_PACKET:
case GSM48_CMSERV_EMERGENCY:
return MSC_A_USE_CM_SERVICE_CC;
case GSM48_CMSERV_SMS:
return MSC_A_USE_CM_SERVICE_SMS;
case GSM48_CMSERV_SUP_SERV:
return MSC_A_USE_CM_SERVICE_SS;
default:
return NULL;
}
}
void msc_a_release_cn(struct msc_a *msc_a)
{
osmo_fsm_inst_dispatch(msc_a->c.fi, MSC_A_EV_CN_CLOSE, NULL);
}
void msc_a_release_mo(struct msc_a *msc_a, enum gsm48_gsm_cause gsm_cause)
{
osmo_fsm_inst_dispatch(msc_a->c.fi, MSC_A_EV_MO_CLOSE, NULL);
}