Previously the wrapping of the 24-bit reference time was not handled
correctly when transforming it into GstClockTime. Given the unit of 64ms
the span that could be represented by 24 bits is 12 days and depending
on the start value we could get a wrapping problem anytime within this
time frame. This turned out to be particularly problematic for the GCC
algorithm in gst-plugins-rs which tried to evict old packages based on
the "oldest" timestamp, which due to wrapping problems could be in the
future. Thus, the container managing the packets could grow without
limits for a long time thereby creating both CPU and memory problems.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7527>
If a stream has an 'irregular' frame rate (e.g. metadata) RTCP SR
may be generated way too early, before the RTPSource has received
the first packet after Latency was configured in the pipeline.
We skip such RTPSources in the RTCP generation.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7740>
Move RB info from receiver reports into the internal source that the RR
are about, and deprecate (but retain) the old mapping where each
external source has only a single RB entry in the rtp statistics.
The old method is broken if a remote peer uses a single ssrc to send
receiver reports for more than one of our internal sources, other
as multiple RB in a single packet, or alternate RB in different reports.
In each case only the most recent entry was kept, overwriting data for
other internal sources.
In multicast scenarios each internal source may receive multiple
receiver reports from different peers. To support that, all received
RR's are now stored into a hash table indexed by the sender's SSRC,
and all RRs are placed into an array when generating statistics, so
that the information from all peers is retrievable.
The current deficient behaviour (adding RB info into non-internal RTPSources) is
deprecated but kept in order to be backward compatible, and retained
that way in the generated statistics structure.
Refs
[1] https://tools.ietf.org/html/rfc3550#section-6.4.1
Based on a patch by Fede Claramonte <fclaramonte@twilio.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7424>
If two (or more) rtpfunnel elements are cascaded, then only one will
realistically have information on the particular ssrc that is in use for a
particular input stream. As such, any key unit requests may never reach the
corresponding encoder.
This has been discovered by combining simulcast and BUNDLE with webrtcbin.
simulcast uses one rtpfunnel, and BUNDLE uses another rtpfunnel.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7405>
Even if no new synchronization information is available.
This is necessary because the timestamp offset logic in rtpbin depends
on the base RTP time that is determined by the jitterbuffer, but this
changes all the time (especially in mode=slave) and the timestamp
offsets have to be updated accordingly. Doing so is especially important
if they're only determined by the RTP-Info, which never changes from the
very beginning.
The interval can be configured via the new min-sync-interval property.
Synchronization happens at least that often, but at most as often as the
old sync-interval property allows.
Both intervals are now based on the monotonic system clock.
Additionally, clean up synchronization code a bit, only emit either
inband NTP or RTCP SR synchronization at the same time, based on which
one has the more recent time information, and only emit RTP-Info
synchronization if it wasn't provided previously at the same time as the
NTP-based synchronization information.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
There is generally no requirement to ignore RTCP SR if the RTP time of
the SR differs a lot from the last received RTP packet. The mapping
between RTP and NTP time stays valid until there was a stream reset, in
which case we wouldn't use that information anyway.
When using rtcp-sync-send-time=false the default of 1s difference can
easily be exceeded, e.g. if encoding of the stream after capture adds
more than 1s of latency.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
Never is useful for some RTSP servers that report plain garbage both via
RTCP SR and RTP-Info, for example.
NTP is useful if synchronization should only ever happen based on RTCP
SR or NTP-64 RTP header extension.
Also slightly change the behaviour of always/initial to take RTP-Info
based synchronization into account too. It's supposed to give the same
values as the RTCP SR and is available earlier, so will generally cause
fewer synchronization glitches if it's made use of.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
Instead of switching on the very first stream, require that all streams
have switched before switching to the different synchronization
mechanism.
Without this there will be a noticeable gap during the switch. E.g. when
going from RTP-Info to NTP-based association, first the first stream
only would get an offset, then the first two, ... then all of them.
Depending on the order of streams this will cause a lot of changes in
ts-offset during the transition.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
Previously these parameters were randomly changed in the body of the
function to avoid having to declare a new variable, which made the code
very hard to follow. By marking them as const this won't be possible
anymore in the future.
Also the RTP clock-base (RTP time from RTSP RTP-Info) is an unsigned
64 bit integer as it's an extended RTP timestamp.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
Both were entangled previously and very hard to follow what happens
under which conditions. Now as a very first step the code decides which
of the two cases it is going to apply, and then proceeds accordingly.
This also avoids calculating completely invalid values along the way and
even printing them int the debug output.
Also improve debug output in various places.
This shouldn't cause any behaviour changes.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
This simplifies the code as it's a much simpler case than the normal
inter-stream synchronization, and interleaving it with that only
reduces readability of the code.
Also improve some debug output in this code path.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6543>
During RTP-Info synchronization, clock_base was temporarily switched
from the actual clock-base to the base RTP time and then back some lines
later.
Instead directly work with the base RTP time. The comment about using a
signed variable for convenience doesn't make any sense because all
calculations done with the value are unsigned.
Similarly, rtp_clock_base was overridden with the rtp_delta when
calculating it, which was fine because it is not used anymore
afterwards. Instead, introduce a new variable `rtp_delta` to make this
calculation clearer.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6536>
It's not in the same period as the current RTP base time but always in
the very first period. This avoids using it again at a much later time.
The code in question is only triggered with rtcp-sync=rtp-info.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6536>
It is compared to other extended RTP timestamps all over rtpjitterbuffer
and since 4df3da3bab the initial extended RTP timestamp is not equal
anymore to the plain RTP time.
Continue passing a non-extended RTP timestamp via the `sync` signal for
backwards compatibility. It will always be a timestamp inside the first
extended timestamp period anyway.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6536>
When the buffer DTS is estimated based on arrival time at the
jitterbuffer (rather than provided on the incoming buffer itself),
it shouldn't be used for skew adjustment. The typical case is
packets being deinterleaved from a tunnelled TCP/HTTP RTSP stream,
and the arrival times at the jitter buffer are not well enough
correlated to usefully do skew adjustments.
This restores the original intended behaviour for the 'estimated dts'
path, that was broken years ago during other jitterbuffer refactoring.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6509>
The transport stream only returned the CAPS for the first matching PT entry
from the `ptmap`. Other SSRC with the same PT where not included. For a stream
which bundled multiple audio streams for instance, only the first SSRC was
knowed to the SSRC demux and downstream elements.
This commit adds all the `ssrc-` attributes from the matching PT entries.
The RTP jitter buffer can now find the CNAME corresponding its SSRC even if it
was not the first to be registered for a particular PT.
The RTP PT demux removes `ssrc-*` attributes cooresponding to other SSRCs
before pushing SSRC specific CAPS to downstream elements.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6119>
In rtpbin we already systematically check for all property names
except latency, correct that.
In webrtcbin we need to check before trying to use the do-retransmission
property.
This is useful for the case where an element like identity gets passed
to rtpbin's request-jitterbuffer property, when the application wants
to use webrtcbin in an SFU situation, with no reordering and no added
latency
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6112>
When level value is greater than 127, it was being clamped but this clamped
value was not the one being actually used. For level values greater than 127
this resulted in an incorrect value being used. As an example, a level value
of 187, after and'ed with 0x7F, it would result in 0x3B being reported as the
level value.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5893>
If this property is enabled then the jitterbuffer will do the normal PTS
calculations according to the configured mode instead of making use of
the RFC7273 media clock.
The timestamp calculated from the RFC7273 media clock will only be
stored in the reference timestamp meta, if addition of that meta is enabled.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5512>
When this property is used, it is assumed that the system clock is
synced close enough to the media clock used by an RFC7273 stream.
As long as both clocks are at most a few seconds from each other this
will give the correct results and avoids having to create an actual
network clock that has to sync first.
If the system clock is actually synchronized to the media clock then
everything will behave exactly the same, otherwise the reference
timestamp meta will be correct but the buffer timestamps will be off by
the difference between the two clocks.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5512>
Do more checks for clock equality than just checking pointers. The same
NTP/PTP clock might be used as pipeline clock but a new instance, so
instead also check what clock they are synced to.
Also handling setting / resetting of the media clock and pipeline clock
correctly by resetting the media clock's state accordingly.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5512>
If we end up with GST_CLOCK_TIME_NONE as running time for an RTP packet
then this can't be used for bitrate estimation, and also not for
constructing the next RTCP SR. Both would end up with completely wrong
values, and an RTCP SR with wrong values can easily break
synchronization in receivers.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5329>
The timestamp offset can be negative, and it can be a bigger negative
number than the latency introduced by the rtpjitterbuffer so the overall
timeout offset can be negative.
Using the negative offset for calculating how many packets can still
arrive in time when encountering a lost packet in an equidistant stream
would then overflow and instead of considering fewer packets lost a lot
more packets are considered lost.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5296>
It could indeed be used uninitialized, but only if one of the
g_return_val_if_fail() caused an early return.
../subprojects/gst-plugins-good/gst/rtpmanager/rtpjitterbuffer.c: In function ‘rtp_jitter_buffer_append_query’:
../subprojects/gst-plugins-good/gst/rtpmanager/rtpjitterbuffer.c🔢10: warning: ‘head’ may be used uninitialized
[-Wmaybe-uninitialized]
1234 | return head;
| ^~~~
../subprojects/gst-plugins-good/gst/rtpmanager/rtpjitterbuffer.c:1232:12: note: ‘head’ was declared here
1232 | gboolean head;
| ^~~~
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4616>
This is a fix for a data race leading to:
> GLib-CRITICAL: g_hash_table_foreach:
> assertion 'version == hash_table->version' failed
Identified sequence:
* `rtp_session_on_timeout` acquires the lock on `session` and proceeds with its
processing.
* `rtp_session_process_rtcp` is called (debug log : received RTCP packet) and
attempts to acquire the lock on `session`, which is still held by
`rtp_session_on_timeout`.
* as part of an hash table iterator, `rtp_session_on_timeout` transitively
invokes `source_caps` which releases the lock on `session` so as to call
`session->callbacks.caps`.
* Since `rtp_session_process_rtcp` was waiting for the lock to be released, it
succeeds in acquiring it and proceeds with `rtp_session_process_rr` which
transitively calls `g_hash_table_insert` via `add_source`.
* After `source_caps` re-acquires the lock and gives the control flow back to
`rtp_session_on_timeout`, the hash table iterator is changed, resulting in the
assertion failure.
This commits copies `sess->ssrcs[sess->mask_idx]` and iterates on the copy so
the iterator is not affected by a concurrent change due to the lock being
released in the `source_caps` callback.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4555>
While testing the [implementation for insertable streams] in `webrtcsink` &
`webrtcsrc`, I encountered critical warnings, which turned out to result from
two race conditions in `rtpsession`. Both race conditions produce:
> GLib-CRITICAL: g_hash_table_foreach:
> assertion 'version == hash_table->version' failed
This commit fixes one of the race conditions observed.
In its simplest form, the test consists in 2 pipelines and a Signalling server:
* pipelines_sink: audiotestsrc ! webrtcsink
* pipelines_src: webrtcsrc ! appsrc
1. Set `pipelines_sink` to `Playing`.
2. The Signalling server delivers the `producer_id`.
3. Initialize `pipelines_src` to establish a session with `producer_id`.
4. Set `pipelines_src` to `Playing`.
5. Wait for a buffer to be received by the `appsrc`.
6. Set `pipelines_src` to `Null`.
7. Set `pipelines_sink` to `Null`.
The race condition happens in the following sequence:
* `webrtcsink` runs a task to periodically retrieve statistics from `webrtcbin`.
This transitively ends up executing `rtp_session_create_stats`.
* `pipelines_sink` is set to `Null`.
* In `Paused` to `Ready`, `gst_rtp_session_change_state()` calls
`rtp_session_reset()`.
* The assertion failure occurs when `rtp_session_reset` is called while
`rtp_session_create_stats` is executing.
This is because `rtp_session_create_stats` acquires the lock on `session` prior
to calling `g_hash_table_foreach`, but `rtp_session_reset` doesn't acquire the
lock before calling `g_hash_table_remove_all`.
Acquiring the lock in `rtp_session_reset` fixes the issue.
[implementing insertable streams support]: https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1176
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4528>
The previous code would only check if two packets in a row were duplicates. If
not (i.e. a packet is a duplicate of a packet received slightly before) the code
would generate completely bogus FCI because it assumes there were no duplicates
present in the array.
In order to be efficient, just store all received packets and remove the
duplicates just before the FCI is generated once the array of observations have
been sorted by seqnum.
Fixes TWCC usage with moderate to high packet duplication.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4328>
This patch prevents a possible race condition from taking place between the EOS event handling and rtcp send
function/thread.
The condition starts by getting the GST_EVENT_EOS event on the send_rtp_sink pad, which causes two core things
to happen -- the event gets pushed down to the send_rtp_src pad and all sessions get marked "bye" prior to
completion of the event handler. In another thread the rtp_session_on_timeout function gets called after an
expiration of gst_clock_id_wait in the rtcp_thread function. This results in a call to the
ess->callbacks.send_rtcp(), which is configured as a function pointer to gst_rtp_session_send_rtcp via the
RTPSessionCallbacks structure passed to rtp_session_set_callbacks in the gst_rtp_session_init function.
In the race condition, the call to gst_rtp_session_send_rtcp can have the all_sources_bye boolean set to true
while GST_PAD_IS_EOS(rtpsession->send_rtp_sink) evaluates to false. This is the result of gst_rtp_session_send_rtcp
running before the send_rtp_sink's GST_EVENT_EOS handler completes. The exact point at which this condition occurs
is if there's a context switch to the rtcp_thread right after the call to rtp_session_mark_all_bye in the
GET_EVENT_EOS handler, but before the handler returns.
Normally, this would not be an issue because the rtcp_thread continues to run and indirectly call
gst_rtp_session_send_rtcp. However, the call to rtp_source_reset sets the sent_bye boolean to false, which ends up
causing rtp_session_are_all_sources_bye to return false. This gets passed to gst_rtp_session_send_rtcp and the EOS
event never gets sent.
The race condition results in the EOS event never getting passed to the rtcp_src pad, which prevents the bin and
pipeline from ever completing with EOS.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3798>
All the RTP src pads were sharing the same stream-id while each actually
carry a different stream.
This was causing problem for example when funneling the streams together
and then trying to split them using 'streamiddemux'.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3855>