Instead of
MGW(MGW_99)
use a name of
msc-mgcp(MSISDN:2331_GERAN_A:00000017_trans99)
1. The FSM is communication towards an MGW, not the MGW itself. When reading
combined logging (gsmtap_log), it is confusing to read 'MGW' in a log coming
from the MSC. The API is also called msc_mgcp_*.
2. Calling the FSM instance 'MGW_' again doesn't make sense.
3. Indicate 'trans' before the trans_id (trans_id was already shown, but not
indicated what it was).
4. Also indicate the actual subscriber's identification.
5. Also indicate the RAN connection and conn_id.
This comes up while trying to understand a call coming in on an already
established call: parsing the log with a human brain is near torture without
this info, taking extremely long to get grips on.
Change-Id: Ie5fc1ffb7eba0209fee4666a075655cd24d27473
Rationale: reading pcaps becomes so much easier when each of osmo-bsc and
osmo-msc address their MGW with differing domain names. Otherwise, both will
have a '0@mgw' endpoint and it gets really confusing.
After this, with according configuration, there can be a '0@bsc' and a '0@msc'
endpoint.
osmo-mgw-for-msc.cfg:
mgcp
domain msc
osmo-msc.cfg:
msc
mgw endpoint-domain msc
Depends: Ia662016f29dd8727d9c4626d726729641e21e1f8 (osmo-mgw)
Change-Id: I87ac11847d1a6d165ee9a2b5d8a4978e7ac73433
Replace locally defined enum ran_type with libosmocore's new enum
osmo_rat_type, and value_string ran_type_names with osmo_rat_type_names.
The string representations change, which has cosmetic effects on the test suite
expectations.
Depends: I659687aef7a4d67ca372a39fef31dee07aed7631 (libosmocore)
Change-Id: I2c78c265dc99df581e1b00e563d6912c7ffdb36b
If Assignment fails in the BSC, trigger an EV_TEARDOWN_ERROR in the mgcp_ctx
FSM instance, so that the call gets torn down immediately. Before this, the
non-call would idle around without anything happening.
Related: OS#3236
Depends: I11b182a03f5ecb6df7cd8f260757d3626c8e945d (libosmocore)
Change-Id: I358cfbaf0f44f25148e8b9bafcb9257b1952b35a
If a call is already busy and another call is coming in, do not try to
immediately assign an lchan (before this patch, it fails because there already
is an mgcp_ctx for the conn). Leave the second CC transaction waiting.
When a call is hung up, as soon as the old mgcp_ctx is discarded, look for
other CC transactions that are waiting. If there is one, trigger assignment, so
a new mgcp_ctx is set up for the new call.
This fixes the following scenario:
- from A, call B.
- from C, call B; B rings during ongoing call.
- in B, pick up the call, choose to drop the old call.
After this patch, and with osmo-bsc patch with change-id
I0c00ec2c120e5008281755adcd4944a3ce4d8355
we are now able to talk to the new caller.
I currently haven't tested yet what happens if there is *three* peers trying to
talk to the same number, running out of lab phones (not really, just not
bothering now). Possibly we should be taking over the particular call indicated
by the CC TI; instead, the current patch version takes on whichever waiting
call it finds first. This is fine if *one* additional call comes in on an
ongoing call, and this is already a huge improvement to what we had before.
Related: OS#3735
Change-Id: I0ba216b737909e92080a722db26e3577726c63cb
The flag is set to true when an assignment has been started, and it is only
relevant for a CC transaction. So fix naming and place in cc struct.
Cosmetic preparation for I1f8746e7babfcd3028a4d2c0ba260c608c686c76 and
I0ba216b737909e92080a722db26e3577726c63cb/
Change-Id: I8dacf46141ba0b664e85b0867ade330c97d8495f
Various places in the code check a flag whether assignment was started and
launch it. To fix incoming-call-during-ongoing-call, I will tweak that logic.
To be able to do that only in one place, remove code dup.
Cosmetic preparation for I1f8746e7babfcd3028a4d2c0ba260c608c686c76 and
I0ba216b737909e92080a722db26e3577726c63cb/
Depends: I11b182a03f5ecb6df7cd8f260757d3626c8e945d (libosmocore: LOGPFSMSL)
Change-Id: I11c0b7dc3f1a747028629b48e522bb3b864884ba
Do not break the currently ongoing call when rejecting a second incoming
caller.
There may be multiple (up to seven) simultaneous CC transactions, and there is
one mgcp_ctx for the currently active RTP stream.
Release the MGCP context only when the active CC transaction is releasing.
Before this patch, any CC transaction release would destroy the single MGCP
context, possibly breaking the currently ongoing call (another CC trans).
This also fixes a possible use-after-free if there were pending MGCP message
responses for the MGCP context; they are canceled properly for a released
transaction, but since one transaction would free the other transaction's MGCP
state, the clean up did not take place and possibly caused an mgcp client
response handling to access a freed mgcp_ctx.
Related: OS#3735
Change-Id: I1f8746e7babfcd3028a4d2c0ba260c608c686c76
Use local variables instead of writing trans->conn-> all the time.
Cosmetic preparation for I1f8746e7babfcd3028a4d2c0ba260c608c686c76 and
I0ba216b737909e92080a722db26e3577726c63cb/
Change-Id: I99717b3b72a9d7cbc95455ea25b2018ec1755308
Two reasons:
- the caller of msc_mgcp_ass_complete() from Iu, iucs_rx_rab_assign(), failed
to be adjusted, breaking IuCS, as an --enable-iu --enable-werror build shows.
Unfortunately our gerrit verification doesn't --enable-werror for osmo-msc.
- the condition of requiring ST_MDCX_RAN is faulty, breaking GSM CS.
This reverts commit 212c0c9bda.
Change-Id: I8348675c2f7c8856ea1682d05ee54160d4cfeb96
BSSMAP Assignment Complete: sort MGCP handling upon Assignment Complete to the
proper locations. a_iface_bssap.c is not the right place to invoke the MGCP
related procedures.
- in a_iface_bssap.c only decode the IEs.
- call ran_conn_assign_compl() and pass decoded values.
- drop msc_assign_compl(), it was dead code; instead:
- add ran_conn_assign_compl()
- pass on all MGCP related info to msc_mgcp_ass_complete()
- move all MGCP ctx related handling from a_iface_bssap.c to msc_mgcp.c.
I'm dropping some comments to save some time, because if I adjust them IMHO
they would still anyway restate the obvious.
ran_conn_assign_compl() is now quite a thin shim, but it makes sense to have
it:
- This is the place that should tear down the ran_conn in case assignment
failed, left for a future patch.
- In the light of upcoming inter-MSC handover, ran_conn_assign_compl() will be
the place where the Assignment Complete message might be relayed to a remote
MSC.
Change-Id: I8137215c443239bddf3e69b5715839a365b73b6c
In preparation for inter-BSC and inter-MSC handover, we need to separate the
subscriber management logic from the actual RAN connections. What better time
to finally rename gsm_subscriber_connection.
* Name choice:
In 2G, this is a connection to the BSS, but even though 3GPP TS commonly talk
of "BSS-A" and "BSS-B" when explaining handover, it's not good to call it
"bss_conn": in 3G a BSS is called RNS, IIUC.
The overall term for 2G (GERAN) and 3G (UTRAN) is RAN: Radio Access Network.
* Rationale:
A subscriber in the MSC so far has only one RAN connection, but e.g. for
inter-BSC handover, a second one needs to be created to handover to. Most of
the items in the former gsm_subscriber_connection are actually related to the
RAN, with only a few MM and RTP related items. So, as a first step, just rename
it to ran_conn, to cosmetically prepare for moving the not strictly RAN related
items away later.
Also:
- Rename some functions from msc_subscr_conn_* to ran_conn_*
- Rename "Subscr_Conn" FSM instance name to "RAN_conn"
- Rename SUBSCR_CONN_* to RAN_CONN_*
Change-Id: Ic595f7a558d3553c067f77dc67543ab59659707a
The function _handle_error() initalizes a struct gsm_mncc variable
on startup. The initalization accesses mgcp_ctx->trans->callref. All
this is done before the assertion on mgcp_ctx. Later in the code one
finds an if which tests on mgcp_ctx->free_ctx. This is the only part of
the code that accesses the mncc struct variable. We should move the
initalization there as well.
- Move initalization of struct gsm_mncc mncc into the if body
that uses it.
Change-Id: I86983eabd999c4275dcc0e4a169ef2aa1e33c747
Related: OS#3635
Move code which needs to test the mgcp_ctx->free_ctx flag upwards
such that it runs before we're calling functions which will
potentially free mgcp_ctx. The code being moved up takes effect
only in case mgcp_ctx won't be freed, so there should be no
functional difference.
Change-Id: I5df17c19e2a68c019f7eaf582b14585caa54b32a
Related: OS#2885
When the assignment completes a choosen codec is returned. At the
moment we do not use this information.
- add struct members for codec info (both, RAN and CN)
- parse codec info in BSSMAP ASSIGNMENT COMPLETE
- use codec info on mgcp
Since the MNCC API is not complete yet, we currently only use the
codec info only on the internal MNCC yet.
Change-Id: I9d5b1cd016d9a058b22a367d0e5e9f2ef447931a
Related: OS#2728
Since change If9a81d057f73150e483286472e73c45e7a453a6d removes the
RTP loopback at the beginning. This also means that the Hack we
do to run the IuUP negotiation via looping back the first few
RTP packets will not work anymore. However, we should keep that
hack as long as we do not have IuUP support in the MGW.
- Start RTP connection in loopback mode for IuUP
Change-Id: I4c7d90de4dc87e8baf7cf4a0c69d0e9e8c92e27b
If an error on the MGW side occurs, the MSC may send out
a DLCX command that contains a wildcarded endpoint name.
Wildcarded DLCX commands are legal in principle but not
in the context of an error on a single endpoint. Apart
from that osmo-mgw is (not yet) capable to handle
wildcarded DLCX command.
The problem is caused by a wrong error handling. When the
first (RAN) CRCX fails the error handling logic tries to
perform a DLCX, but since we did not receive a specific
endpoint name yet, the buffer containing the endpoint name
is still initalized with the wildcarded enpoint name, but
the error handler and the code that generates the DLCX is
not aware of that.
- Perform a check in the error handler function that
checks if a DLCX can be made (a specific endpoint
name is set
- Correct the flags in the code that handles the first
CRCX so that no DLCX is requested in the case of error
Related OS#2882
Change-Id: I64c2a82016d854ad446fd49a5d76a28324e8bd4b
The logtext currently logs the pointer (address) of the string
variable that holds the endpoint name, rather then the endpoint
name itself.
- Use %s instead of %p in format string
Change-Id: I01b3d07aeedd72be60361249a5bf80fbb68b7bb8
Fixes several of these type of warnings:
include/osmocom/core/fsm.h:123:38: warning: format ‘%x’ expects argument of type ‘unsigned int’, but argument 9 has type ‘char *’ [-Wformat=]
LOGP((fi)->fsm->log_subsys, level, "%s{%s}: " fmt, \
^
src/libmsc/msc_mgcp.c:277:71: note: format string is defined here
"CRCX/RAN: creating connection for the RAN side on MGW endpoint:0x%x...\n", mgcp_ctx->rtp_endpoint);
~^
Change-Id: I17b7bed8fc39612286ba66f250b6b26da01d38c0
The higher layers (gsm_04_08.c) are informed errors occur. But it
is not checked if the call was already released. If an error occurs
after the call control stack calls msc_mgcp_call_release() then
the higher layers might already have cleaned up and the code
accesses memory that is already freed (trans)
- fix use after free by guarding the call to mncc_tx_to_cc()
Change-Id: I78f1b6a9149488a4ad3f120c1e190a83c07d4b89
Related OS#2881
Related OS#2882
osmo-msc still uses endpoints that are allocated locally by the
MGCP-Client. Since osmo-mgw now supports the more comfortable,
dynamic variant we should make use of it.
- Replace the endpoint numer allocation by the client with a
wildcarded CRCX. Use the endpoint that is assigned by the
MGW.
Related: OS#2710
Change-Id: Iee3e446b6689626516f01c521abe3d4603cd3e13
When a truncation check (osmo_strlcpy()) fails handle_error()
is called with MGCP_ERR_NOMEM as cause code. Which finally
results in an "out of memory" message. MGCP_ERR_NOMEM is only
used for failed truncation checks, so it makes sense to choose
a message that describes the cause of the problem better.
- rename MGCP_ERR_NOMEM to MGCP_ERR_TOOLONG
- replace error message string
Change-Id: Ifedb85c1933a30b2aa491b495119919f45782e2c
When the FSM reaches ST_HALT it frees itsself and all context
information but it is not ensured that there are still pending
MGW transactions that might hit late and eventually cause a use after
free situation.
- if an MGW transaction is still pending, cancel it.
Change-Id: I8ff55e48a95cc4c556a97ad2593bad1cc1aa69bd
When the MGW does not respond to an MGCP message then the mgcp FSM
terminates, but the CC handler (gsm_04_08.c) is not informed. This
lets the CC handler think that the MGCP connection would be successful,
so it also does not take any action to release the non functional
connection.
- make sure the CC handler is always informed on any kind of
error, especially on MGW timeouts
Change-Id: I3fcd0d71fad53274e82f6d87c80042d06283bc5d
Related OS#2881
Related OS#2882
The FSM (fsm_msc_mgcp) lacks a proper definition of the FSM event
names. This causes problems when inspecting the FSM using the VTY.
- Add proper FSM Event names
Closes: OS#2924
Change-Id: I6823756a63b08a71e5518130e49751aa073dbcd2
The MGCP FSM implements a timeout when waiting for the RAN to complete
the call (assignment complete, alerting, connect...). This timeout
is currently set to 10sec. This means if the other end did not pick
up after 10sec, the MGCP connection will be lost while the phone keeps
ringing. When the other end finally picks up, the call gets
disconnected.
This behavior is odd and requires a proper fix. For now increasing the
timeout to 120sec. will decrese the probability that he problem occurs.
- Increas RAN timeout to 120sec (2 min).
Change-Id: I5a11d53f9701d9b11b18d7026ff2241c7c0b57f5
in the current implementation we still use osmo-bsc_mgcp, which
has many problems and is also obsoleted by osmo-mgw.
integrate osmo-mgw and re-implement the current switching using
an osmo fsm.
Depends: osmo-mgw Iab6a6038e7610c62f34e642cd49c93d11151252c
Depends: osmo-iuh I3c1a0455c5f25cae41ee19229d6daf299e023062
Closes: OS#2605
Change-Id: Ieea9630358b3963261fa1993cf1f3b563ff23538
Neels Hofmeyr
Philipp Maier
Pau Espin
Daniel Willmann
Harald Welte