osmo-msc/src/libmsc/ran_msg_a.c

1424 lines
47 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
/* BSSAP/BSSMAP encoding and decoding for MSC */
/*
* (C) 2019 by sysmocom - s.m.f.c. GmbH <info@sysmocom.de>
* All Rights Reserved
*
* Author: Neels Hofmeyr
*
* SPDX-License-Identifier: GPL-2.0+
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 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 General Public License for more details.
*
* You should have received a copy of the GNU General Public License along
* with this program; if not, write to the Free Software Foundation, Inc.,
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
*/
#include <osmocom/core/byteswap.h>
#include <osmocom/crypt/auth.h>
#include <osmocom/crypt/kdf.h>
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
#include <osmocom/gsm/tlv.h>
#include <osmocom/gsm/gsm0808.h>
#include <osmocom/gsm/mncc.h>
#include <osmocom/gsm/gsm48.h>
#include <osmocom/msc/debug.h>
#include <osmocom/msc/ran_msg_a.h>
#include <osmocom/msc/sccp_ran.h>
#include <osmocom/msc/gsm_data.h>
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
#define LOG_RAN_A_DEC(RAN_DEC, level, fmt, args...) \
LOG_RAN_DEC(RAN_DEC, DBSSAP, level, "BSSMAP: " fmt, ## args)
/* Assumes presence of struct ran_dec *ran_dec and ran_dec_msg.msg_name (set) in the local scope. */
#define LOG_RAN_A_DEC_MSG(level, fmt, args...) \
LOG_RAN_DEC(ran_dec, DBSSAP, level, "%s: " fmt, ran_dec_msg.msg_name, ## args)
#define LOG_RAN_A_ENC(FI, level, fmt, args...) \
LOG_RAN_ENC(FI, DBSSAP, level, "BSSMAP: " fmt, ## args)
static int ran_a_decode_l3_compl(struct ran_dec *ran_dec, struct msgb *msg, struct tlv_parsed *tp)
{
struct gsm0808_cell_id_list2 cil;
struct gsm0808_cell_id cell_id;
struct tlv_p_entry *ie_cell_id = TLVP_GET(tp, GSM0808_IE_CELL_IDENTIFIER);
struct tlv_p_entry *ie_l3_info = TLVP_GET(tp, GSM0808_IE_LAYER_3_INFORMATION);
struct tlv_p_entry *ie_codec_list_bss_supported = TLVP_GET(tp, GSM0808_IE_SPEECH_CODEC_LIST);
struct gsm0808_speech_codec_list codec_list_bss_supported;
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 ran_dec_msg = {
.msg_type = RAN_MSG_COMPL_L3,
.msg_name = "BSSMAP 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
.compl_l3 = {
.cell_id = &cell_id,
.msg = msg,
},
};
int rc;
if (!ie_cell_id) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Mandatory CELL IDENTIFIER not present, discarding message\n");
return -EINVAL;
}
if (!ie_l3_info) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Mandatory LAYER 3 INFORMATION not present, discarding message\n");
return -EINVAL;
}
/* Parse Cell ID element -- this should yield a cell identifier "list" with 1 element. */
rc = gsm0808_dec_cell_id_list2(&cil, ie_cell_id->val, ie_cell_id->len);
if (rc < 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Decoding CELL IDENTIFIER gave rc=%d\n", rc);
return -EINVAL;
}
if (cil.id_list_len != 1) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Unable to parse element CELL IDENTIFIER, discarding message\n");
return -EINVAL;
}
/* Sanity check the Cell Identity */
switch (cil.id_discr) {
case CELL_IDENT_WHOLE_GLOBAL:
case CELL_IDENT_LAI_AND_LAC:
case CELL_IDENT_LAC_AND_CI:
case CELL_IDENT_LAC:
break;
case CELL_IDENT_CI:
case CELL_IDENT_NO_CELL:
case CELL_IDENT_BSS:
default:
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "CELL IDENTIFIER does not specify a LAC, discarding message: %s\n",
gsm0808_cell_id_list_name(&cil));
return -EINVAL;
}
cell_id = (struct gsm0808_cell_id){
.id_discr = cil.id_discr,
.id = cil.id_list[0],
};
/* Parse Layer 3 Information element */
msg->l3h = (uint8_t*)ie_l3_info->val;
msgb_l3trim(msg, ie_l3_info->len);
if (msgb_l3len(msg) < sizeof(struct gsm48_hdr)) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "too short L3 info (%d), discarding message\n", msgb_l3len(msg));
return -ENODATA;
}
/* Decode Codec List (BSS Supported) */
if (ie_codec_list_bss_supported) {
rc = gsm0808_dec_speech_codec_list(&codec_list_bss_supported,
ie_codec_list_bss_supported->val, ie_codec_list_bss_supported->len);
if (rc < 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR,
"Complete Layer 3 Information: unable to decode IE Codec List (BSS Supported)"
" (rc=%d), continuing anyway\n", rc);
/* This IE is not critical, do not abort with error. */
} else
ran_dec_msg.compl_l3.codec_list_bss_supported = &codec_list_bss_supported;
}
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 ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_clear_request(struct ran_dec *ran_dec, struct msgb *msg, struct tlv_parsed *tp)
{
struct tlv_p_entry *ie_cause = TLVP_GET(tp, GSM0808_IE_CAUSE);
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_CLEAR_REQUEST,
.msg_name = "BSSMAP Clear Request",
};
if (!ie_cause) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Cause code is missing, using GSM0808_CAUSE_EQUIPMENT_FAILURE\n");
ran_dec_msg.clear_request.bssap_cause = GSM0808_CAUSE_EQUIPMENT_FAILURE;
} else {
ran_dec_msg.clear_request.bssap_cause = ie_cause->val[0];
}
return ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_clear_complete(struct ran_dec *ran_dec, struct msgb *msg, struct tlv_parsed *tp)
{
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_CLEAR_COMPLETE,
.msg_name = "BSSMAP Clear Complete",
};
return ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_classmark_update(struct ran_dec *ran_dec, struct msgb *msg, struct tlv_parsed *tp)
{
struct tlv_p_entry *ie_cm2 = TLVP_GET(tp, GSM0808_IE_CLASSMARK_INFORMATION_T2);
struct tlv_p_entry *ie_cm3 = TLVP_GET(tp, GSM0808_IE_CLASSMARK_INFORMATION_T3);
struct osmo_gsm48_classmark cm = {};
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_CLASSMARK_UPDATE,
.msg_name = "BSSMAP Classmark Update",
.classmark_update = {
.classmark = &cm,
},
};
if (!ie_cm2) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "mandatory Classmark Information Type 2 not present, discarding message\n");
return -EINVAL;
}
cm.classmark2_len = OSMO_MIN(sizeof(cm.classmark2), ie_cm2->len);
memcpy(&cm.classmark2, ie_cm2->val, cm.classmark2_len);
if (ie_cm3) {
cm.classmark3_len = OSMO_MIN(sizeof(cm.classmark3), ie_cm3->len);
memcpy(&cm.classmark3, ie_cm3->val, cm.classmark3_len);
}
return ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_cipher_mode_complete(struct ran_dec *ran_dec, struct msgb *msg, struct tlv_parsed *tp)
{
struct tlv_p_entry *ie_chosen_encr_alg = TLVP_GET(tp, GSM0808_IE_CHOSEN_ENCR_ALG);
struct tlv_p_entry *ie_l3_msg = TLVP_GET(tp, GSM0808_IE_LAYER_3_MESSAGE_CONTENTS);
int rc;
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_CIPHER_MODE_COMPLETE,
.msg_name = "BSSMAP Ciphering Mode Complete",
};
if (ie_chosen_encr_alg) {
uint8_t ie_val = ie_chosen_encr_alg->val[0];
/* 3GPP TS 48.008 3.2.2.44 Chosen Encryption Algorithm encodes as 1 = no encryption, 2 = A5/1, 4 = A5/3.
* Internally we handle without this weird off-by-one. */
if (ie_val < 1 || ie_val > 8)
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Unsupported value for 3.2.2.44 Chosen Encryption Algorithm: %u\n",
ie_val);
else
ran_dec_msg.cipher_mode_complete.alg_id = ie_chosen_encr_alg->val[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
if (ie_l3_msg)
ran_dec_msg.cipher_mode_complete.l3_msg = ie_l3_msg;
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 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
rc = ran_decoded(ran_dec, &ran_dec_msg);
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 rc;
}
static int ran_a_decode_cipher_mode_reject(struct ran_dec *ran_dec, struct msgb *msg, struct tlv_parsed *tp)
{
int rc;
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_CIPHER_MODE_REJECT,
.msg_name = "BSSMAP Ciphering Mode Reject",
};
rc = gsm0808_get_cause(tp);
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 (rc < 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "failed to extract Cause\n");
ran_dec_msg.cipher_mode_reject.bssap_cause = GSM0808_CAUSE_EQUIPMENT_FAILURE;
} else {
ran_dec_msg.cipher_mode_reject.bssap_cause = (enum gsm0808_cause)rc;
}
return ran_decoded(ran_dec, &ran_dec_msg);
}
enum mgcp_codecs ran_a_mgcp_codec_from_sc(const struct gsm0808_speech_codec *sc)
{
switch (sc->type) {
case GSM0808_SCT_FR1:
return CODEC_GSM_8000_1;
break;
case GSM0808_SCT_FR2:
return CODEC_GSMEFR_8000_1;
break;
case GSM0808_SCT_FR3:
return CODEC_AMR_8000_1;
break;
case GSM0808_SCT_FR4:
return CODEC_AMRWB_16000_1;
break;
case GSM0808_SCT_FR5:
return CODEC_AMRWB_16000_1;
break;
case GSM0808_SCT_HR1:
return CODEC_GSMHR_8000_1;
break;
case GSM0808_SCT_HR3:
return CODEC_AMR_8000_1;
break;
case GSM0808_SCT_HR4:
return CODEC_AMRWB_16000_1;
break;
case GSM0808_SCT_HR6:
return CODEC_AMRWB_16000_1;
break;
default:
return CODEC_PCMU_8000_1;
break;
}
}
static int ran_a_decode_assignment_complete(struct ran_dec *ran_dec, struct msgb *msg, struct tlv_parsed *tp)
{
struct tlv_p_entry *ie_aoip_transp_addr = TLVP_GET(tp, GSM0808_IE_AOIP_TRASP_ADDR);
struct tlv_p_entry *ie_speech_codec = TLVP_GET(tp, GSM0808_IE_SPEECH_CODEC);
struct tlv_p_entry *ie_codec_list_bss_supported = TLVP_GET(tp, GSM0808_IE_SPEECH_CODEC_LIST);
struct tlv_p_entry *ie_osmux_cid = TLVP_GET(tp, GSM0808_IE_OSMO_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
struct sockaddr_storage rtp_addr;
struct gsm0808_speech_codec sc;
struct gsm0808_speech_codec_list codec_list_bss_supported;
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 rc;
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_ASSIGNMENT_COMPLETE,
.msg_name = "BSSMAP Assignment Complete",
};
if (ie_aoip_transp_addr) {
/* Decode AoIP transport address element */
rc = gsm0808_dec_aoip_trasp_addr(&rtp_addr, ie_aoip_transp_addr->val, ie_aoip_transp_addr->len);
if (rc < 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Unable to decode AoIP Transport Layer Address\n");
return -EINVAL;
}
if (osmo_sockaddr_str_from_sockaddr(&ran_dec_msg.assignment_complete.remote_rtp, &rtp_addr)) {
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_RAN_A_DEC_MSG(LOGL_ERROR, "Assignment Complete: unable to decode remote RTP IP address\n");
return -EINVAL;
}
}
if (ie_osmux_cid) {
rc = gsm0808_dec_osmux_cid(&ran_dec_msg.assignment_complete.osmux_cid, ie_osmux_cid->val, ie_osmux_cid->len);
if (rc < 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Unable to decode Osmux CID\n");
return -EINVAL;
}
ran_dec_msg.assignment_complete.osmux_present = true;
}
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 (ie_speech_codec) {
/* Decode Speech Codec (Chosen) element */
rc = gsm0808_dec_speech_codec(&sc, ie_speech_codec->val, ie_speech_codec->len);
if (rc < 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Assignment Complete: unable to decode IE Speech Codec (Chosen)"
" (rc=%d).\n", rc);
return -EINVAL;
}
ran_dec_msg.assignment_complete.codec_present = true;
ran_dec_msg.assignment_complete.codec = ran_a_mgcp_codec_from_sc(&sc);
}
if (ie_codec_list_bss_supported) {
/* Decode Codec List (BSS Supported) */
rc = gsm0808_dec_speech_codec_list(&codec_list_bss_supported,
ie_codec_list_bss_supported->val, ie_codec_list_bss_supported->len);
if (rc < 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR,
"Assignment Complete: unable to decode IE Codec List (BSS Supported)"
" (rc=%d), continuing anyway\n", rc);
/* This IE is not critical, do not abort with error. */
} else
ran_dec_msg.assignment_complete.codec_list_bss_supported = &codec_list_bss_supported;
}
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 ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_assignment_failure(struct ran_dec *ran_dec, struct msgb *msg, struct tlv_parsed *tp)
{
struct tlv_p_entry *ie_cause = TLVP_GET(tp, GSM0808_IE_CAUSE);
struct tlv_p_entry *ie_rr_cause = TLVP_GET(tp, GSM0808_IE_RR_CAUSE);
struct tlv_p_entry *ie_speech_codec_list = TLVP_GET(tp, GSM0808_IE_SPEECH_CODEC_LIST);
struct gsm0808_speech_codec_list scl;
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_ASSIGNMENT_FAILURE,
.msg_name = "BSSMAP Assignment Failure",
.assignment_failure = {
.bssap_cause = GSM0808_CAUSE_EQUIPMENT_FAILURE,
.rr_cause = GSM48_RR_CAUSE_ABNORMAL_UNSPEC,
},
};
if (ie_cause)
ran_dec_msg.assignment_failure.bssap_cause = ie_cause->val[0];
if (ie_rr_cause)
ran_dec_msg.assignment_failure.rr_cause = ie_rr_cause->val[0];
if (ie_speech_codec_list
&& gsm0808_dec_speech_codec_list(&scl, ie_speech_codec_list->val, ie_speech_codec_list->len) == 0)
ran_dec_msg.assignment_failure.scl_bss_supported = &scl;
return ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_sapi_n_reject(struct ran_dec *ran_dec, struct msgb *msg, struct tlv_parsed *tp)
{
struct tlv_p_entry *ie_cause = TLVP_GET(tp, GSM0808_IE_CAUSE);
struct tlv_p_entry *ie_dlci = TLVP_GET(tp, GSM0808_IE_DLCI);
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_SAPI_N_REJECT,
.msg_name = "BSSMAP SAPI-N Reject",
};
/* Note: The MSC code seems not to care about the cause code, but by
* the specification it is mandatory, so we check its presence. See
* also 3GPP TS 48.008 3.2.1.34 SAPI "n" REJECT */
if (!ie_cause) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "SAPI-N Reject: cause code IE is missing, discarding message\n");
return -EINVAL;
}
ran_dec_msg.sapi_n_reject.bssap_cause = ie_cause->val[0];
if (!ie_dlci) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "SAPI-N Reject: DLCI IE is missing, discarding message\n");
return -EINVAL;
}
ran_dec_msg.sapi_n_reject.dlci = ie_dlci->val[0];
return ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_lcls_notification(struct ran_dec *ran_dec, const struct msgb *msg, const struct tlv_parsed *tp)
{
const struct tlv_p_entry *ie_lcls_bss_status = TLVP_GET(tp, GSM0808_IE_LCLS_BSS_STATUS);
const struct tlv_p_entry *ie_lcls_break_req = TLVP_GET(tp, GSM0808_IE_LCLS_BREAK_REQ);
struct ran_msg ran_dec_msg;
/* Either §3.2.2.119 LCLS-BSS-Status or §3.2.2.120 LCLS-Break-Request shall be present */
if (ie_lcls_bss_status && !ie_lcls_break_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
ran_dec_msg = (struct ran_msg){
.msg_type = RAN_MSG_LCLS_STATUS,
.msg_name = "BSSMAP LCLS Notification (LCLS Status)",
.lcls_status = {
.status = ie_lcls_bss_status->len ?
ie_lcls_bss_status->val[0] : GSM0808_LCLS_STS_NA,
},
};
return ran_decoded(ran_dec, &ran_dec_msg);
} else if (ie_lcls_break_req && !ie_lcls_bss_status) {
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_dec_msg = (struct ran_msg){
.msg_type = RAN_MSG_LCLS_BREAK_REQ,
.msg_name = "BSSMAP LCLS Notification (LCLS Break Req)",
.lcls_break_req = {
.todo = 23,
},
};
return ran_decoded(ran_dec, &ran_dec_msg);
}
LOG_RAN_A_DEC(ran_dec, LOGL_ERROR, "Ignoring broken LCLS Notification message\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;
}
static int ran_a_decode_handover_required(struct ran_dec *ran_dec, const struct msgb *msg, const struct tlv_parsed *tp)
{
const struct tlv_p_entry *ie_cause = TLVP_GET(tp, GSM0808_IE_CAUSE);
const struct tlv_p_entry *ie_cil = TLVP_GET(tp, GSM0808_IE_CELL_IDENTIFIER_LIST);
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_HANDOVER_REQUIRED,
.msg_name = "BSSMAP Handover Required",
};
/* On decoding failures, dispatch an invalid RAN_MSG_HANDOVER_REQUIRED so msc_a can pass down a
* BSS_MAP_MSG_HANDOVER_REQUIRED_REJECT message. */
if (ie_cause)
ran_dec_msg.handover_required.cause = ie_cause->val[0];
else
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Cause IE missing\n");
if (!ie_cil
|| gsm0808_dec_cell_id_list2(&ran_dec_msg.handover_required.cil, ie_cil->val, ie_cil->len) <= 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "No or invalid Cell Identifier List IE\n");
ran_dec_msg.handover_required.cil = (struct gsm0808_cell_id_list2){};
}
return ran_decoded(ran_dec, &ran_dec_msg);
}
static uint8_t a5_encryption_mask_from_gsm0808_chosen_enc_alg(enum gsm0808_chosen_enc_alg val)
{
return 1 << val;
}
static int ran_a_decode_handover_request(struct ran_dec *ran_dec, const struct msgb *msg, const struct tlv_parsed *tp)
{
struct osmo_gsm48_classmark classmark = {};
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_HANDOVER_REQUEST,
.msg_name = "BSSMAP Handover Request",
.handover_request = {
.classmark = &classmark,
},
};
struct ran_handover_request *r = &ran_dec_msg.handover_request;
const struct tlv_p_entry *ie_channel_type = TLVP_GET(tp, GSM0808_IE_CHANNEL_TYPE);
const struct tlv_p_entry *ie_encryption_information = TLVP_GET(tp, GSM0808_IE_ENCRYPTION_INFORMATION);
const struct tlv_p_entry *ie_classmark1 = TLVP_GET(tp, GSM0808_IE_CLASSMARK_INFORMATION_TYPE_1);
const struct tlv_p_entry *ie_classmark2 = TLVP_GET(tp, GSM0808_IE_CLASSMARK_INFORMATION_T2);
const struct tlv_p_entry *ie_cell_id_serving = TLVP_GET(&tp[0], GSM0808_IE_CELL_IDENTIFIER);
const struct tlv_p_entry *ie_cell_id_target = TLVP_GET(&tp[1], GSM0808_IE_CELL_IDENTIFIER);
const struct tlv_p_entry *ie_cause = TLVP_GET(tp, GSM0808_IE_CAUSE);
const struct tlv_p_entry *ie_classmark3 = TLVP_GET(tp, GSM0808_IE_CLASSMARK_INFORMATION_T3);
const struct tlv_p_entry *ie_current_channel_type_1 = TLVP_GET(tp, GSM0808_IE_CURRENT_CHANNEL_TYPE_1);
const struct tlv_p_entry *ie_speech_version_used = TLVP_GET(tp, GSM0808_IE_SPEECH_VERSION);
const struct tlv_p_entry *ie_chosen_encr_alg_serving = TLVP_GET(tp, GSM0808_IE_CHOSEN_ENCR_ALG);
const struct tlv_p_entry *ie_old_bss_to_new_bss_info = TLVP_GET(tp, GSM0808_IE_OLD_BSS_TO_NEW_BSS_INFORMATION);
const struct tlv_p_entry *ie_imsi = TLVP_GET(tp, GSM0808_IE_IMSI);
const struct tlv_p_entry *ie_aoip_transp_addr = TLVP_GET(tp, GSM0808_IE_AOIP_TRASP_ADDR);
const struct tlv_p_entry *ie_codec_list_msc_preferred = TLVP_GET(tp, GSM0808_IE_SPEECH_CODEC_LIST);
const struct tlv_p_entry *ie_call_id = TLVP_GET(tp, GSM0808_IE_CALL_ID);
const struct tlv_p_entry *ie_kc128 = TLVP_GET(tp, GSM0808_IE_KC_128);
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 tlv_p_entry *ie_global_call_ref = TLVP_GET(tp, GSM0808_IE_GLOBAL_CALL_REF);
struct gsm0808_channel_type channel_type;
struct gsm0808_encrypt_info encr_info;
struct gsm0808_speech_codec_list scl;
struct geran_encr geran_encr = {};
struct osmo_sockaddr_str rtp_ran_local;
if (!ie_channel_type) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Mandatory IE missing: Channel Type\n");
return -EINVAL;
}
if (gsm0808_dec_channel_type(&channel_type, ie_channel_type->val, ie_channel_type->len) <= 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Failed to decode Channel Type IE\n");
return -EINVAL;
}
r->geran.channel_type = &channel_type;
if (ie_encryption_information) {
int i;
if (gsm0808_dec_encrypt_info(&encr_info, ie_encryption_information->val, ie_encryption_information->len)
<= 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Failed to decode Encryption Information IE\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;
}
for (i = 0; i < encr_info.perm_algo_len; i++) {
r->geran.a5_encryption_mask |=
a5_encryption_mask_from_gsm0808_chosen_enc_alg(encr_info.perm_algo[i]);
}
if (encr_info.key_len > sizeof(geran_encr.key)) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Failed to decode Encryption Information IE:"
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
" encryption key is too long: %u\n", geran_encr.key_len);
return -EINVAL;
}
if (encr_info.key_len) {
memcpy(geran_encr.key, encr_info.key, encr_info.key_len);
geran_encr.key_len = encr_info.key_len;
}
if (ie_kc128) {
memcpy(geran_encr.kc128, ie_kc128->val, 16);
geran_encr.kc128_present = true;
}
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
r->geran.chosen_encryption = &geran_encr;
}
if (!ie_classmark1 && !ie_classmark2) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Mandatory IE missing: either Classmark Information 1"
" or Classmark Information 2 must be included\n");
return -EINVAL;
}
if (ie_classmark1) {
if (ie_classmark1->len != sizeof(classmark.classmark1)) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Invalid size for Classmark 1: %u, expected %zu\n",
ie_classmark1->len, sizeof(classmark.classmark1));
return -EINVAL;
}
memcpy((uint8_t*)&classmark.classmark1, ie_classmark1->val, ie_classmark1->len);
classmark.classmark1_set = true;
}
if (ie_classmark2) {
uint8_t len = OSMO_MIN(ie_classmark2->len, sizeof(classmark.classmark2));
memcpy((uint8_t*)&classmark.classmark2, ie_classmark2->val, len);
classmark.classmark2_len = len;
}
if (!ie_cell_id_serving) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Mandatory IE missing: Cell Identifier (Serving)\n");
return -EINVAL;
}
if (gsm0808_dec_cell_id(&r->cell_id_serving, ie_cell_id_serving->val,
ie_cell_id_serving->len) <= 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Failed to decode Cell Identifier (Serving) IE\n");
return -EINVAL;
}
if (!ie_cell_id_target) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Mandatory IE missing: Cell Identifier (Target)\n");
return -EINVAL;
}
if (gsm0808_dec_cell_id(&r->cell_id_target, ie_cell_id_target->val,
ie_cell_id_target->len) <= 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "Failed to decode Cell Identifier (Target) IE\n");
return -EINVAL;
}
if (ie_cause)
r->bssap_cause = ie_cause->val[0];
if (ie_classmark3) {
uint8_t len = OSMO_MIN(ie_classmark3->len, sizeof(classmark.classmark3));
memcpy(classmark.classmark3, ie_classmark3->val, len);
classmark.classmark3_len = len;
}
if (ie_current_channel_type_1) {
r->current_channel_type_1 = ie_current_channel_type_1->val[0];
r->current_channel_type_1_present = true;
}
if (ie_speech_version_used) {
r->speech_version_used = ie_speech_version_used->val[0];
}
if (ie_chosen_encr_alg_serving && ie_chosen_encr_alg_serving->len) {
geran_encr.alg_id = ie_chosen_encr_alg_serving->val[0];
r->geran.chosen_encryption = &geran_encr;
}
if (ie_old_bss_to_new_bss_info) {
r->old_bss_to_new_bss_info_raw = ie_old_bss_to_new_bss_info->val;
r->old_bss_to_new_bss_info_raw_len = ie_old_bss_to_new_bss_info->len;
}
if (ie_imsi) {
struct osmo_mobile_identity mi;
if (osmo_mobile_identity_decode(&mi, ie_imsi->val, ie_imsi->len, false)
|| mi.type != GSM_MI_TYPE_IMSI)
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "IE IMSI: cannot decode IMSI identity\n");
else
r->imsi = mi.imsi;
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 (ie_aoip_transp_addr) {
struct sockaddr_storage rtp_addr;
if (gsm0808_dec_aoip_trasp_addr(&rtp_addr, ie_aoip_transp_addr->val, ie_aoip_transp_addr->len) < 0)
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "unable to decode AoIP transport address\n");
else if (osmo_sockaddr_str_from_sockaddr(&rtp_ran_local, &rtp_addr) < 0)
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "unable to decode remote RTP IP address\n");
else
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
r->rtp_ran_local = &rtp_ran_local;
}
if (ie_codec_list_msc_preferred
&& gsm0808_dec_speech_codec_list(&scl, ie_codec_list_msc_preferred->val,
ie_codec_list_msc_preferred->len) == 0)
r->codec_list_msc_preferred = &scl;
if (ie_call_id && ie_call_id->len == 4) {
r->call_id = osmo_load32le(ie_call_id->val);
r->call_id_present = true;
}
if (ie_global_call_ref) {
r->global_call_reference = ie_global_call_ref->val;
r->global_call_reference_len = ie_global_call_ref->len;
}
return ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_handover_request_ack(struct ran_dec *ran_dec, const struct msgb *msg, const struct tlv_parsed *tp)
{
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_HANDOVER_REQUEST_ACK,
.msg_name = "BSSMAP Handover Request Acknowledge",
};
const struct tlv_p_entry *ie_l3_info = TLVP_GET(tp, GSM0808_IE_LAYER_3_INFORMATION);
const struct tlv_p_entry *ie_aoip_transp_addr = TLVP_GET(tp, GSM0808_IE_AOIP_TRASP_ADDR);
const struct tlv_p_entry *ie_speech_codec = TLVP_GET(tp, GSM0808_IE_SPEECH_CODEC);
const struct tlv_p_entry *ie_chosen_channel = TLVP_GET(tp, GSM0808_IE_CHOSEN_CHANNEL);
const struct tlv_p_entry *ie_chosen_encr_alg = TLVP_GET(tp, GSM0808_IE_CHOSEN_ENCR_ALG);
const struct tlv_p_entry *ie_chosen_speech_version = TLVP_GET(tp, GSM0808_IE_SPEECH_VERSION);
/* On missing mandatory IEs, dispatch an invalid RAN_MSG_HANDOVER_REQUEST_ACK so msc_a can act on the failure. */
if (ie_l3_info) {
ran_dec_msg.handover_request_ack.rr_ho_command = ie_l3_info->val;
ran_dec_msg.handover_request_ack.rr_ho_command_len = ie_l3_info->len;
}
if (ie_chosen_channel) {
ran_dec_msg.handover_request_ack.chosen_channel_present = true;
ran_dec_msg.handover_request_ack.chosen_channel = *ie_chosen_channel->val;
}
if (ie_chosen_encr_alg) {
ran_dec_msg.handover_request_ack.chosen_encr_alg = *ie_chosen_encr_alg->val;
if (ran_dec_msg.handover_request_ack.chosen_encr_alg < 1
|| ran_dec_msg.handover_request_ack.chosen_encr_alg > 8) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "invalid Chosen Encryption Algorithm: %u\n",
ran_dec_msg.handover_request_ack.chosen_encr_alg);
}
}
if (ie_chosen_speech_version) {
struct gsm0808_speech_codec sc;
ran_dec_msg.handover_request_ack.chosen_speech_version = ie_chosen_speech_version->val[0];
/* the codec may be extrapolated from this Speech Version or below from Speech Codec */
gsm0808_speech_codec_from_chan_type(&sc, ran_dec_msg.handover_request_ack.chosen_speech_version);
ran_dec_msg.handover_request_ack.codec_present = true;
ran_dec_msg.handover_request_ack.codec = ran_a_mgcp_codec_from_sc(&sc);
}
if (ie_aoip_transp_addr) {
struct sockaddr_storage rtp_addr;
if (gsm0808_dec_aoip_trasp_addr(&rtp_addr, ie_aoip_transp_addr->val, ie_aoip_transp_addr->len) < 0) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "unable to decode AoIP transport address\n");
} else if (osmo_sockaddr_str_from_sockaddr(&ran_dec_msg.handover_request_ack.remote_rtp,
&rtp_addr)) {
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "unable to decode remote RTP IP address\n");
ran_dec_msg.handover_request_ack.remote_rtp = (struct osmo_sockaddr_str){};
}
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 (ie_speech_codec) {
struct gsm0808_speech_codec sc;
if (gsm0808_dec_speech_codec(&sc, ie_speech_codec->val, ie_speech_codec->len) < 0)
LOG_RAN_A_DEC_MSG(LOGL_ERROR, "unable to decode IE Speech Codec (Chosen)\n");
else {
/* the codec may be extrapolated from above Speech Version or from this Speech Codec */
ran_dec_msg.handover_request_ack.codec_present = true;
ran_dec_msg.handover_request_ack.codec = ran_a_mgcp_codec_from_sc(&sc);
}
}
return ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_handover_detect(struct ran_dec *ran_dec, const struct msgb *msg, const struct tlv_parsed *tp)
{
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_HANDOVER_DETECT,
.msg_name = "BSSMAP Handover Detect",
};
return ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_handover_succeeded(struct ran_dec *ran_dec, const struct msgb *msg, const struct tlv_parsed *tp)
{
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_HANDOVER_SUCCEEDED,
.msg_name = "BSSMAP Handover Succeeded",
};
return ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_handover_complete(struct ran_dec *ran_dec, const struct msgb *msg, const struct tlv_parsed *tp)
{
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_HANDOVER_COMPLETE,
.msg_name = "BSSMAP Handover Complete",
};
return ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_handover_failure(struct ran_dec *ran_dec, const struct msgb *msg, const struct tlv_parsed *tp)
{
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_HANDOVER_FAILURE,
.msg_name = "BSSMAP Handover Failure",
};
return ran_decoded(ran_dec, &ran_dec_msg);
}
static int ran_a_decode_bssmap(struct ran_dec *ran_dec, struct msgb *bssmap)
{
struct tlv_parsed tp[2];
int rc;
struct bssmap_header *h = msgb_l2(bssmap);
uint8_t msg_type;
bssmap->l3h = bssmap->l2h + sizeof(*h);
if (msgb_l3len(bssmap) < 1) {
LOG_RAN_A_DEC(ran_dec, LOGL_ERROR, "No data received, discarding message\n");
return -1;
}
if (msgb_l3len(bssmap) < h->length) {
LOG_RAN_A_DEC(ran_dec, LOGL_ERROR, "BSSMAP data truncated, discarding message:"
" msgb_l3len(bssmap) == %u < bssmap_header->length == %u\n",
msgb_l3len(bssmap), h->length);
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 -1;
}
if (msgb_l3len(bssmap) > h->length) {
LOG_RAN_A_DEC(ran_dec, LOGL_NOTICE, "There are %u extra bytes after the BSSMAP data, truncating:"
" msgb_l3len(bssmap) == %u > bssmap_header->length == %u\n",
msgb_l3len(bssmap) - h->length,
msgb_l3len(bssmap), h->length);
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
msgb_l3trim(bssmap, h->length);
}
/* h->type == BSSAP_MSG_BSS_MANAGEMENT; h->length is the data length,
* which starts with the MAP msg_type, followed by IEs. */
msg_type = bssmap->l3h[0];
rc = osmo_bssap_tlv_parse2(tp, ARRAY_SIZE(tp), bssmap->l3h + 1, h->length - 1);
if (rc < 0) {
LOG_RAN_A_DEC(ran_dec, LOGL_ERROR, "Failed parsing TLV, discarding message\n");
return -EINVAL;
}
LOG_RAN_A_DEC(ran_dec, LOGL_DEBUG, "%s\n", gsm0808_bssmap_name(msg_type));
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 (msg_type) {
case BSS_MAP_MSG_COMPLETE_LAYER_3:
return ran_a_decode_l3_compl(ran_dec, bssmap, tp);
case BSS_MAP_MSG_CLEAR_RQST:
return ran_a_decode_clear_request(ran_dec, bssmap, tp);
case BSS_MAP_MSG_CLEAR_COMPLETE:
return ran_a_decode_clear_complete(ran_dec, bssmap, tp);
case BSS_MAP_MSG_CLASSMARK_UPDATE:
return ran_a_decode_classmark_update(ran_dec, bssmap, tp);
case BSS_MAP_MSG_CIPHER_MODE_COMPLETE:
return ran_a_decode_cipher_mode_complete(ran_dec, bssmap, tp);
case BSS_MAP_MSG_CIPHER_MODE_REJECT:
return ran_a_decode_cipher_mode_reject(ran_dec, bssmap, tp);
case BSS_MAP_MSG_ASSIGMENT_COMPLETE:
rc = ran_a_decode_assignment_complete(ran_dec, bssmap, tp);
if (rc < 0) {
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_ASSIGNMENT_FAILURE,
.msg_name = "BSSMAP Assignment Complete but failed to decode",
.clear_request = {
.bssap_cause = GSM0808_CAUSE_EQUIPMENT_FAILURE,
},
};
ran_decoded(ran_dec, &ran_dec_msg);
}
return rc;
case BSS_MAP_MSG_ASSIGMENT_FAILURE:
return ran_a_decode_assignment_failure(ran_dec, bssmap, tp);
case BSS_MAP_MSG_SAPI_N_REJECT:
return ran_a_decode_sapi_n_reject(ran_dec, bssmap, tp);
case BSS_MAP_MSG_LCLS_NOTIFICATION:
return ran_a_decode_lcls_notification(ran_dec, bssmap, tp);
/* From current RAN peer, the Handover origin: */
case BSS_MAP_MSG_HANDOVER_REQUIRED:
return ran_a_decode_handover_required(ran_dec, bssmap, tp);
/* From current MSC to remote handover target MSC */
case BSS_MAP_MSG_HANDOVER_RQST:
return ran_a_decode_handover_request(ran_dec, bssmap, tp);
/* From potential new RAN peer, the Handover target: */
case BSS_MAP_MSG_HANDOVER_RQST_ACKNOWLEDGE:
return ran_a_decode_handover_request_ack(ran_dec, bssmap, tp);
case BSS_MAP_MSG_HANDOVER_DETECT:
return ran_a_decode_handover_detect(ran_dec, bssmap, tp);
case BSS_MAP_MSG_HANDOVER_SUCCEEDED:
return ran_a_decode_handover_succeeded(ran_dec, bssmap, tp);
case BSS_MAP_MSG_HANDOVER_COMPLETE:
return ran_a_decode_handover_complete(ran_dec, bssmap, tp);
/* From any Handover peer: */
case BSS_MAP_MSG_HANDOVER_FAILURE:
return ran_a_decode_handover_failure(ran_dec, bssmap, tp);
default:
LOG_RAN_A_DEC(ran_dec, LOGL_ERROR, "Unimplemented msg type: %s\n", gsm0808_bssmap_name(msg_type));
return -EINVAL;
}
return -EINVAL;
}
static int ran_a_decode_l3(struct ran_dec *ran_dec, struct msgb *l3)
{
struct dtap_header *dtap = msgb_l2(l3);
struct ran_msg ran_dec_msg = {
.msg_type = RAN_MSG_DTAP,
.msg_name = "BSSAP DTAP",
.dtap = l3,
};
l3->l3h = l3->l2h + sizeof(struct dtap_header);
OMSC_LINKID_CB(l3) = dtap->link_id;
return ran_decoded(ran_dec, &ran_dec_msg);
}
int ran_a_decode_l2(struct ran_dec *ran_dec, struct msgb *bssap)
{
uint8_t bssap_type;
OSMO_ASSERT(bssap);
if (!msgb_l2(bssap) || !msgb_l2len(bssap)) {
LOG_RAN_A_DEC(ran_dec, LOGL_ERROR, "Cannot decode L2, msg->l2h is unset / empty: %s\n",
msgb_hexdump(bssap));
return -EINVAL;
}
if (msgb_l2len(bssap) < sizeof(struct bssmap_header)) {
LOG_RAN_A_DEC(ran_dec, LOGL_ERROR, "The header is too short -- discarding message\n");
return -EINVAL;
}
bssap_type = bssap->l2h[0];
switch (bssap_type) {
case BSSAP_MSG_BSS_MANAGEMENT:
return ran_a_decode_bssmap(ran_dec, bssap);
case BSSAP_MSG_DTAP:
return ran_a_decode_l3(ran_dec, bssap);
default:
LOG_RAN_A_DEC(ran_dec, LOGL_ERROR, "Unimplemented BSSAP msg type: %s\n", gsm0808_bssap_name(bssap_type));
return -EINVAL;
}
}
static struct msgb *ran_a_wrap_dtap(struct msgb *dtap)
{
struct msgb *an_apdu;
dtap->l3h = dtap->data;
an_apdu = gsm0808_create_dtap(dtap, OMSC_LINKID_CB(dtap));
an_apdu->l2h = an_apdu->data;
msgb_free(dtap);
return an_apdu;
}
static int ran_a_channel_type_to_speech_codec_list(struct gsm0808_speech_codec_list *scl, const struct gsm0808_channel_type *ct)
{
unsigned int i;
int rc;
memset(scl, 0, sizeof(*scl));
for (i = 0; i < ct->perm_spch_len; i++) {
rc = gsm0808_speech_codec_from_chan_type(&scl->codec[i], ct->perm_spch[i]);
if (rc != 0)
return -EINVAL;
}
scl->len = i;
return 0;
}
static void _gsm0808_assignment_extend_osmux(struct msgb *msg, uint8_t cid)
{
OSMO_ASSERT(msg->l3h[1] == msgb_l3len(msg) - 2); /*TL not in len */
msgb_tv_put(msg, GSM0808_IE_OSMO_OSMUX_CID, cid);
msg->l3h[1] = msgb_l3len(msg) - 2;
}
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
/* Compose a BSSAP Assignment Command.
* Passing an RTP address is optional.
* The msub is passed merely for error logging. */
static struct msgb *ran_a_make_assignment_command(struct osmo_fsm_inst *log_fi,
const struct ran_assignment_command *ac)
{
struct gsm0808_speech_codec_list scl;
struct gsm0808_speech_codec_list *use_scl = NULL;
struct sockaddr_storage rtp_addr;
struct sockaddr_storage *use_rtp_addr = NULL;
struct msgb *msg;
const uint32_t *call_id = 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 rc;
if (!ac->channel_type) {
LOG_RAN_A_ENC(log_fi, LOGL_ERROR, "Assignment Command: missing Channel Type\n");
return NULL;
}
if (ac->channel_type->ch_indctr == GSM0808_CHAN_SPEECH) {
rc = ran_a_channel_type_to_speech_codec_list(&scl, ac->channel_type);
if (rc < 0) {
LOG_RAN_A_ENC(log_fi, LOGL_ERROR, "Assignment Command: Cannot translate Channel Type to Speech Codec List\n");
return NULL;
}
use_scl = &scl;
/* Package RTP-Address data */
if (osmo_sockaddr_str_is_nonzero(ac->cn_rtp)) {
struct sockaddr_in *sin;
struct sockaddr_in6 *sin6;
int family = osmo_ip_str_type(ac->cn_rtp->ip);
switch (family) {
case AF_INET:
sin = (struct sockaddr_in *)&rtp_addr;
sin->sin_family = AF_INET;
sin->sin_port = osmo_htons(ac->cn_rtp->port);
if (inet_pton(AF_INET, ac->cn_rtp->ip, &sin->sin_addr) != 1) {
LOG_RAN_A_ENC(log_fi, LOGL_ERROR,
"Assignment Command: Invalid RTP-Address %s\n",
ac->cn_rtp->ip);
return NULL;
}
if (sin->sin_port == 0) {
LOG_RAN_A_ENC(log_fi, LOGL_ERROR,
"Assignment Command: Invalid RTP-Port\n");
return NULL;
}
break;
case AF_INET6:
sin6 = (struct sockaddr_in6 *)&rtp_addr;
sin6->sin6_family = AF_INET6;
sin6->sin6_port = osmo_htons(ac->cn_rtp->port);
if (inet_pton(AF_INET6, ac->cn_rtp->ip, &sin6->sin6_addr) != 1) {
LOG_RAN_A_ENC(log_fi, LOGL_ERROR,
"Assignment Command: Invalid RTP-Address %s\n",
ac->cn_rtp->ip);
return NULL;
}
if (sin6->sin6_port == 0) {
LOG_RAN_A_ENC(log_fi, LOGL_ERROR,
"Assignment Command: Invalid RTP-Port\n");
return NULL;
}
break;
default:
LOG_RAN_A_ENC(log_fi, LOGL_ERROR,
"Assignment Command: Invalid RTP-Address type for %s\n",
ac->cn_rtp->ip);
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 NULL;
}
use_rtp_addr = &rtp_addr;
}
}
if(ac->call_id_present == true)
call_id = &ac->call_id;
msg = gsm0808_create_ass(ac->channel_type, NULL, use_rtp_addr, use_scl, call_id);
if (ac->osmux_present)
_gsm0808_assignment_extend_osmux(msg, ac->osmux_cid);
return msg;
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 an A5/N number a5_n set dst to the matching GSM0808_ALG_ID_A5_<n>. */
static int a5_n_to_gsm0808_chosen_enc_alg(uint8_t *dst, int a5_n)
{
switch (a5_n) {
case 0:
*dst = GSM0808_ALG_ID_A5_0;
return 0;
case 1:
*dst = GSM0808_ALG_ID_A5_1;
return 0;
case 2:
*dst = GSM0808_ALG_ID_A5_2;
return 0;
case 3:
*dst = GSM0808_ALG_ID_A5_3;
return 0;
case 4:
*dst = GSM0808_ALG_ID_A5_4;
return 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
default:
return -ENOTSUP;
}
}
static int make_encrypt_info_perm_algo(struct osmo_fsm_inst *fi, struct gsm0808_encrypt_info *ei,
uint8_t a5_encryption_mask, const struct osmo_gsm48_classmark *cm)
{
int i;
int j = 0;
for (i = 0; i < 8; i++) {
int supported;
/* A5/n permitted by osmo-msc.cfg? */
if (!(a5_encryption_mask & (1 << i)))
continue;
/* A5/n supported by MS? */
supported = osmo_gsm48_classmark_supports_a5(cm, i);
if (supported != 1)
continue;
if (a5_n_to_gsm0808_chosen_enc_alg(&ei->perm_algo[j], i)) {
LOG_RAN_A_ENC(fi, LOGL_ERROR, "Not supported: A5/%d algorithm\n", i);
return -1;
}
j++;
ei->perm_algo_len = j;
}
return 0;
}
/* For ran_a_make_cipher_mode_command(), for
* memcpy(ei.key, cm->vec->kc, sizeof(cm->vec->kc));
*/
osmo_static_assert(sizeof(((struct gsm0808_encrypt_info*)0)->key) >= sizeof(((struct osmo_auth_vector*)0)->kc),
gsm0808_encrypt_info_key_fits_osmo_auth_vec_kc);
static struct msgb *ran_a_make_cipher_mode_command(struct osmo_fsm_inst *fi, const struct ran_cipher_mode_command *cm)
{
struct gsm0808_cipher_mode_command cmc = {
.cipher_response_mode_present = true,
.cipher_response_mode = 1, /* 1: include IMEISV (3GPP TS 48.008 3.2.2.34) */
};
struct gsm0808_encrypt_info *ei = &cmc.ei;
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
char buf[16 * 2 + 1];
if (make_encrypt_info_perm_algo(fi, ei, cm->geran.a5_encryption_mask, cm->classmark))
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 NULL;
if (ei->perm_algo_len == 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
LOG_RAN_A_ENC(fi, LOGL_ERROR, "cannot start ciphering, no intersection between MSC-configured"
" and MS-supported A5 algorithms. MSC: 0x%02x MS: %s\n",
cm->geran.a5_encryption_mask, osmo_gsm48_classmark_a5_name(cm->classmark));
return NULL;
}
/* In case of UMTS AKA, the Kc for ciphering must be derived from the 3G auth
* tokens. vec->kc was calculated from the GSM algorithm and is not
* necessarily a match for the UMTS AKA tokens. */
if (cm->geran.umts_aka) {
int i;
osmo_auth_c3(ei->key, cm->vec->ck, cm->vec->ik);
for (i = 0; i < ei->perm_algo_len; i++) {
if (ei->perm_algo[i] != GSM0808_ALG_ID_A5_4)
continue;
/* A5/4 is included, so need to generate Kc128 */
osmo_kdf_kc128(cm->vec->ck, cm->vec->ik, cmc.kc128);
cmc.kc128_present = true;
break;
}
} else {
memcpy(ei->key, cm->vec->kc, sizeof(cm->vec->kc));
}
ei->key_len = sizeof(cm->vec->kc);
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
/* Store chosen GERAN key where the caller asked it to be stored.
* alg_id remains unknown until we receive a Cipher Mode Complete from the BSC */
if (cm->geran.chosen_key) {
*cm->geran.chosen_key = (struct geran_encr){0};
if (ei->key_len > sizeof(cm->geran.chosen_key->key)) {
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_RAN_A_ENC(fi, LOGL_ERROR, "Chosen key is larger than I can store\n");
return NULL;
}
memcpy(cm->geran.chosen_key->key, ei->key, ei->key_len);
cm->geran.chosen_key->key_len = ei->key_len;
if (cmc.kc128_present) {
memcpy(cm->geran.chosen_key->kc128, cmc.kc128, 16);
cm->geran.chosen_key->kc128_present = true;
}
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_RAN_A_ENC(fi, LOGL_DEBUG, "Tx BSSMAP CIPHER MODE COMMAND to BSC, %u ciphers (%s) key %s\n",
ei->perm_algo_len, osmo_hexdump_nospc(ei->perm_algo, ei->perm_algo_len),
osmo_hexdump_buf(buf, sizeof(buf), ei->key, ei->key_len, NULL, false));
return gsm0808_create_cipher2(&cmc);
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 msgb *ran_a_make_handover_request(struct osmo_fsm_inst *log_fi, const struct ran_handover_request *n)
{
struct sockaddr_storage ss;
struct gsm0808_handover_request r = {
.cell_identifier_serving = n->cell_id_serving,
.cell_identifier_target = n->cell_id_target,
.cause = n->bssap_cause,
.current_channel_type_1_present = n->current_channel_type_1_present,
.current_channel_type_1 = n->current_channel_type_1,
.speech_version_used = n->speech_version_used,
.old_bss_to_new_bss_info_raw = n->old_bss_to_new_bss_info_raw,
.old_bss_to_new_bss_info_raw_len = n->old_bss_to_new_bss_info_raw_len,
.imsi = n->imsi,
.codec_list_msc_preferred = n->codec_list_msc_preferred,
.call_id_present = n->call_id_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
.call_id = n->call_id,
.global_call_reference = n->global_call_reference,
.global_call_reference_len = n->global_call_reference_len,
};
if (!n->geran.channel_type) {
LOG_RAN_A_ENC(log_fi, LOGL_ERROR, "Channel Type required for encoding Handover Request in BSSAP\n");
return NULL;
}
r.channel_type = *n->geran.channel_type;
/* Encryption Information */
make_encrypt_info_perm_algo(log_fi, &r.encryption_information, n->geran.a5_encryption_mask, n->classmark);
if (n->geran.chosen_encryption && n->geran.chosen_encryption->key_len) {
/* Prevent both source / destination buffer overrun / overflow */
if (n->geran.chosen_encryption->key_len > sizeof(r.encryption_information.key)
|| n->geran.chosen_encryption->key_len > sizeof(n->geran.chosen_encryption->key)) {
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_RAN_A_ENC(log_fi, LOGL_ERROR, "Handover Request: invalid chosen encryption key size %u\n",
n->geran.chosen_encryption->key_len);
return NULL;
}
memcpy(r.encryption_information.key,
n->geran.chosen_encryption->key, n->geran.chosen_encryption->key_len);
r.encryption_information.key_len = n->geran.chosen_encryption->key_len;
r.chosen_encryption_algorithm_serving = n->geran.chosen_encryption->alg_id;
if (n->geran.chosen_encryption->kc128_present) {
r.more_items = true;
memcpy(r.kc128, n->geran.chosen_encryption->kc128, sizeof(r.kc128));
r.kc128_present = true;
}
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 (n->classmark)
r.classmark_information = *n->classmark;
if (osmo_sockaddr_str_is_nonzero(n->rtp_ran_local)) {
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 (osmo_sockaddr_str_to_sockaddr(n->rtp_ran_local, &ss)) {
LOG_RAN_A_ENC(log_fi, LOGL_ERROR,
"Handover Request: invalid AoIP Transport Layer address/port: "
OSMO_SOCKADDR_STR_FMT "\n", OSMO_SOCKADDR_STR_FMT_ARGS(n->rtp_ran_local));
return NULL;
}
r.aoip_transport_layer = &ss;
}
return gsm0808_create_handover_request(&r);
}
static struct msgb *ran_a_make_handover_request_ack(struct osmo_fsm_inst *caller_fi, const struct ran_handover_request_ack *r)
{
struct sockaddr_storage ss;
struct gsm0808_handover_request_ack params = {
.l3_info = r->rr_ho_command,
.l3_info_len = r->rr_ho_command_len,
.chosen_channel_present = r->chosen_channel_present,
.chosen_channel = r->chosen_channel,
.chosen_encr_alg = r->chosen_encr_alg,
.chosen_speech_version = r->chosen_speech_version,
};
if (osmo_sockaddr_str_is_nonzero(&r->remote_rtp)) {
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_sockaddr_str_to_sockaddr(&r->remote_rtp, &ss);
params.aoip_transport_layer = &ss;
}
return gsm0808_create_handover_request_ack2(&params);
}
struct msgb *ran_a_make_handover_command(struct osmo_fsm_inst *log_fi, const struct ran_handover_command *n)
{
struct gsm0808_handover_command c = {
.l3_info = n->rr_ho_command,
.l3_info_len = n->rr_ho_command_len,
};
return gsm0808_create_handover_command(&c);
}
struct msgb *ran_a_make_handover_failure(struct osmo_fsm_inst *log_fi, const struct ran_msg *msg)
{
struct gsm0808_handover_failure params = {
.cause = msg->handover_failure.cause,
};
return gsm0808_create_handover_failure(&params);
}
static struct msgb *_ran_a_encode(struct osmo_fsm_inst *caller_fi, const struct ran_msg *ran_enc_msg)
{
LOG_RAN_A_ENC(caller_fi, LOGL_DEBUG, "%s\n", ran_msg_type_name(ran_enc_msg->msg_type));
switch (ran_enc_msg->msg_type) {
case RAN_MSG_DTAP:
return ran_a_wrap_dtap(ran_enc_msg->dtap);
case RAN_MSG_CLASSMARK_REQUEST:
return gsm0808_create_classmark_request();
case RAN_MSG_CLEAR_COMMAND:
return gsm0808_create_clear_command2(ran_enc_msg->clear_command.gsm0808_cause,
ran_enc_msg->clear_command.csfb_ind);
case RAN_MSG_ASSIGNMENT_COMMAND:
return ran_a_make_assignment_command(caller_fi, &ran_enc_msg->assignment_command);
case RAN_MSG_COMMON_ID:
return gsm0808_create_common_id(ran_enc_msg->common_id.imsi, NULL,
ran_enc_msg->common_id.last_eutran_plmn_present ?
&ran_enc_msg->common_id.last_eutran_plmn :
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
case RAN_MSG_CIPHER_MODE_COMMAND:
return ran_a_make_cipher_mode_command(caller_fi, &ran_enc_msg->cipher_mode_command);
case RAN_MSG_HANDOVER_REQUIRED_REJECT:
return gsm0808_create_handover_required_reject(&ran_enc_msg->handover_required_reject);
case RAN_MSG_HANDOVER_REQUEST:
return ran_a_make_handover_request(caller_fi, &ran_enc_msg->handover_request);
case RAN_MSG_HANDOVER_REQUEST_ACK:
return ran_a_make_handover_request_ack(caller_fi, &ran_enc_msg->handover_request_ack);
case RAN_MSG_HANDOVER_COMMAND:
return ran_a_make_handover_command(caller_fi, &ran_enc_msg->handover_command);
case RAN_MSG_HANDOVER_SUCCEEDED:
return gsm0808_create_handover_succeeded();
case RAN_MSG_HANDOVER_FAILURE:
return ran_a_make_handover_failure(caller_fi, ran_enc_msg);
default:
LOG_RAN_A_ENC(caller_fi, LOGL_ERROR, "Unimplemented RAN-encode message type: %s\n",
ran_msg_type_name(ran_enc_msg->msg_type));
return NULL;
}
}
struct msgb *ran_a_encode(struct osmo_fsm_inst *caller_fi, const struct ran_msg *ran_enc_msg)
{
struct msgb *msg = _ran_a_encode(caller_fi, ran_enc_msg);
if (!msg)
return NULL;
msg->l2h = msg->data;
/* some consistency checks to ensure we don't send invalid length */
switch (msg->l2h[0]) {
case BSSAP_MSG_DTAP:
OSMO_ASSERT(msgb_l2len(msg) == msg->l2h[2] + 3);
break;
case BSSAP_MSG_BSS_MANAGEMENT:
OSMO_ASSERT(msgb_l2len(msg) == msg->l2h[1] + 2);
break;
default:
break;
}
return msg;
}
static void cl_parse_osmux(struct osmo_fsm_inst *log_fi, struct msgb *msg, int *supports_osmux)
{
struct tlv_parsed tp;
int rc;
if (supports_osmux == NULL)
return;
rc = tlv_parse(&tp, gsm0808_att_tlvdef(), msgb_l3(msg) + 1, msgb_l3len(msg) - 1, 0, 0);
if (rc < 0) {
LOGPFSMSL(log_fi, DBSSAP, LOGL_ERROR, "BSSMAP: Failed parsing TLV looking for Osmux support\n");
return;
}
if (TLVP_PRESENT(&tp, GSM0808_IE_OSMO_OSMUX_SUPPORT)) {
*supports_osmux = true;
} else {
*supports_osmux = false;
}
}
/* Return 1 for a RESET, 2 for a RESET ACK message, 0 otherwise.
* In supports_osmux, return 0 for no information, 1 for support detected, -1 for non-support detected. */
enum reset_msg_type bssmap_is_reset_msg(const struct sccp_ran_inst *sri, struct osmo_fsm_inst *log_fi,
struct msgb *l2, int *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
{
struct bssmap_header *bs = (struct bssmap_header *)msgb_l2(l2);
if (supports_osmux != NULL)
*supports_osmux = 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
if (!bs
|| msgb_l2len(l2) < (sizeof(*bs) + 1)
|| bs->type != BSSAP_MSG_BSS_MANAGEMENT)
return SCCP_RAN_MSG_NON_RESET;
l2->l3h = l2->l2h + sizeof(struct bssmap_header);
switch (l2->l3h[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
case BSS_MAP_MSG_RESET:
cl_parse_osmux(log_fi, l2, 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
return SCCP_RAN_MSG_RESET;
case BSS_MAP_MSG_RESET_ACKNOWLEDGE:
cl_parse_osmux(log_fi, l2, 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
return SCCP_RAN_MSG_RESET_ACK;
default:
return SCCP_RAN_MSG_NON_RESET;
}
}
/* Patch regular BSSMAP RESET to add extra T to announce Osmux support (osmocom extension) */
static void _gsm0808_extend_announce_osmux(struct msgb *msg)
{
OSMO_ASSERT(msg->l3h[1] == msgb_l3len(msg) - 2); /*TL not in len */
msgb_put_u8(msg, GSM0808_IE_OSMO_OSMUX_SUPPORT);
msg->l3h[1] = msgb_l3len(msg) - 2;
}
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 msgb *bssmap_make_reset_msg(const struct sccp_ran_inst *sri, enum reset_msg_type type)
{
struct gsm_network *net = sri->user_data;
struct msgb *msg;
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 (type) {
case SCCP_RAN_MSG_RESET:
msg = gsm0808_create_reset();
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
case SCCP_RAN_MSG_RESET_ACK:
msg = gsm0808_create_reset_ack();
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:
return NULL;
}
if (!msg)
return NULL;
if (net->use_osmux != OSMUX_USAGE_OFF)
_gsm0808_extend_announce_osmux(msg);
return msg;
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 msgb *bssmap_make_paging_msg(const struct sccp_ran_inst *sri, const struct gsm0808_cell_id *page_cell_id,
const char *imsi, uint32_t tmsi, enum paging_cause cause)
{
struct gsm0808_cell_id_list2 cil;
gsm0808_cell_id_to_list(&cil, page_cell_id);
return gsm0808_create_paging2(imsi, tmsi == GSM_RESERVED_TMSI ? NULL : &tmsi, &cil, NULL);
}
const char *bssmap_msg_name(const struct sccp_ran_inst *sri, const struct msgb *l2)
{
struct bssmap_header *bs;
if (!l2->l2h)
return "?";
bs = (struct bssmap_header *)msgb_l2(l2);
switch (bs->type) {
case BSSAP_MSG_BSS_MANAGEMENT:
return gsm0808_bssmap_name(l2->l2h[0]);
case BSSAP_MSG_DTAP:
return "DTAP";
default:
return "?";
}
}