Fix a regression from 4291c858
send_timer_rtcp() does its own locking of the SSRC. Release the lock
before calling it.
Change-Id: I185fb2fb4b47a343ab4be00d16629b5f330ee965
These are part of the extended tests which are normally disabled. Some
change in libopus results in different encodings in these cases, making
these tests fail with newer libopus versions. Accept an alternative test
result.
Change-Id: I846830a47651bf7a7f83fa0bb1f1220db77db71e
ps->out_lock must be held when sending packets, and the SSRC lock must
be held for sending RTCP
Change-Id: I19fe85762e330a9b8a03658a828f735db6eb6cb7
(cherry picked from commit 6c7277a84f)
Creating the codec handler can possibly fail. Catch that case.
Closes#1693
Change-Id: Iebd8d1c6aa3542cd9e1351756148c48ebada38ec
(cherry picked from commit 94ba3e0fa8)
... stream after a reconnect
if we're in the middle of a disconnect then ssrc_tls_state may have destroyed
the streambuf so we need to skip the rest of the function, otherwise the now
null pointer is passed into streambuf_write, which will return immediately and
the metadata will have been marked as sent when it hasn't been
closes#1694
Change-Id: I679fdd0ff9245e4413bc5a5ab40a70ea4f9ef7c0
(cherry picked from commit 88c0ffd49b)
Make sure we don't try to send on closed sockets or to endpoints which
haven't been initialised.
Fixes unexpected fallout from 83c7336e
Change-Id: If73d61e52edeb72257515adab7428ecef82c2797
(cherry picked from commit 8266b2da78)
Don't attempt to send packets to an endpoint address that is empty and
hasn't been initialised.
Change-Id: Iabeb034ebfc56fc49083c6a10a85b92c6983eebc
(cherry picked from commit 850a1a582e)
The default shell on i386 stretch has some sort of problem with wildcard
expansion. Force usage of bash.
Change-Id: I923375128b7861ec5f7a2c2d77a50850f888035e
(cherry picked from commit 1162f4a890)
Due to multiple threads polling all sockets for read events, it's
possible for one socket to receive a read event in one thread, then
immediately receive another read event in another thread, resulting in
two threads reading packets from the same socket at the same time.
While this is perfectly valid and correctly handled by mutex etc, it can
result in packets being processed out of order. In media passthrough
scenarios which don't do sequencing this can result in packets being
reordered.
Using a simple atomic counter we can ensure that only one thread is
reading from any one socket at a time.
Relevant to #1638
Change-Id: I406491d6ae5e13e618e153ba5463fd9169636016
(cherry picked from commit fdc9b14509)
Don't try to interpret or schedule DTMF packets if we're just doing RTP
passthrough.
Without tracking the seqencing and timestamping of the primary media
stream, we cannot actually schedule DTMF packets properly. Use the
presence of a defined output sequence number to make this determination.
The other case that uses a predefined output sequence number is when
media is coming from a pre-coded media player stream. The same thing
applies: These packets are already scheduled, so we can skip trying to
schedule them again.
closes#1638
Change-Id: I2d2c20e731f20a3136ca3eee37f0b521062cdb11
(cherry picked from commit 6f57d381e6)
Packet processing (decoding, encoding, file writing, etc) happens with
the main parent object (metafile) unlocked and only with the specific
child object (SSRC) locked.
During teardown we close all SSRC objects with the metafile locked,
without locking the SSRC objects. This opens up a race condition during
which SSRC closing can happen at the same time as packet processing.
Fix this by locking each SSRC object during teardown, and also clear out
the SSRC hash table while the metafile is locked, which serves as a
marker to a packet processing thread that processing should be skipped.
Closes#1663
Change-Id: I4e8a6a5abfd678a6cfd794a98470a49d60694207
(cherry picked from commit 237a3a6402)
Deprecate usage of the bit flags for return.
And instead do:
- only return true/false for the redis update
- check if we have to destroy the whole call using the
`call_monologues_associations_left()` only
Additionally remove unused variables.
Change-Id: Ie858c533d1aa4b74024d65c7064aa56437389882
Check globally left monologues (with associations left) after
the `monologue_delete_iter()` has already done its job.
This is needed to prevent cases, when a recursive call of the
`monologue_delete_iter()` leads to blocking of the whole call
destorying, meanwhile it's required, e.g.:
- A is in progress of destorying
- B gets recursive call of `monologue_delete_iter()` and sees
that A has still one association left with C, which A simply
hasn't manage to remove yet.
This commit introduces new function: `call_monologues_associations_left()`
Change-Id: I0941c18a76fa8c2a78d3864aee9e6433283bec88
(cherry picked from commit f4eb2d5a26)
Avoid trying to acquire a recursive lock by making sure the response is
always generated in a different thread.
Fixes#1656
Change-Id: I6c4c5bb52cb95a204823848bb427ab24f42dcccd
(cherry picked from commit 04ce204ef6)
This commit: 6f0ad0db0f
introduced an improved behavior of the call deconstruction
when dealing with multiple forked legs. But it also
introduced a regression related to TPCC related calls.
This is because it takes into consideration existing
`->viabranches`, which is not efficient in certain cases.
Instead, we just have to ensure, whether there are
other monologues of the call object, which are not directly
"visible" for this branch being cancelled (so they are not directly
associated to that), but they might have own associations
present, which means the call mustn't be globally destoroyed.
Change-Id: I630f2d88ef3b557af3a95816fc2703daccaff374
(cherry picked from commit 0994ffbe75)
The `encoder` is actually persistent and only freed when the output is
freed, so testing for its existence is not enough to determine whether
the output is open. The `fmtctx` can be used for that.
Change-Id: I2cc3f00a0e983fafd5915736089a0421385b6ce3
(cherry picked from commit 833c30d69e)
There's a mixture of RTPengine and RTPEngine, though we lowercase it
entirely most of the time, so let's stick to this and unify its style.
While at it remove the "the" article for rtpengine as well,
where applicable.
Change-Id: Idbad9313499942ad12ade2c9ce76fe3ec4cd2762
If a subscribe answer says that it doesn't want to receive any media
(i.e. `inactive` - `recv` flag unset), clear our corresponding `send`
flag so that we actually don't send media there.
Change-Id: I8862cc6f4121a82c723a8303ba70345d064cd243
... which is the more direct way to copy a bit flag in a mirror fashion,
instead of doing an if/else.
Change-Id: I7283d31f7a44e6bf1ee15ea1c14cb3b8fe0293b9
We're supposed to use the remote password for sending STUN error
responses (same as for sending STUN success responses), not the local
one.
Fixes a bug from 2015.
Closes#1626
Change-Id: I975178405dcd41661bdc5e0c1208295f9db70006
While LWS explicitly allows usage of lws_callback_on_writable() from
other threads, for some reason there is no internal locking in place,
and so a concurrently running lws_service() can interfere with internal
structures, in particular if lws_service() is closing connections at the
same time as lws_callback_on_writable() is invoked.
The suggested approach of using lws_cancel_service() in combination with
the LWS_CALLBACK_EVENT_WAIT_CANCELLED callback and a user-kept queue is
not feasible, as we need to support LWS 2.x, which doesn't have
LWS_CALLBACK_EVENT_WAIT_CANCELLED.
Closes#1624
Change-Id: Ia3ddeda66fd553c87f99404e0816d97ecbd4cdfe
Avoid calling lws_write() from threads other than the service thread, as
this might not be thread-safe. Instead store the values used for the
HTTP response headers in the websocket_output, then trigger a "writable"
callback, and finally do all the lws_write() calls from the service
thread.
Reported in #1624
Change-Id: Ifcb050193044e5543f750a12fb44f5e16d4c0a08
Richard Fuchs
Donat Zenichev
Alex Hermann
Tom Briden
Dennis Yurasov
Stefan Mititelu
Sipwise Jenkins Builder
Michael Prokop