Commit graph

354 commits

Author SHA1 Message Date
Nicolas Dufresne
bbe0053f8a rtpjitterbuffer: Add a faststart-min-packets property
When set this property will allow the jitterbuffer to start delivering
packets as soon as N most recent packets have consecutive seqnum. A
faststart-min-packets of zero disables this feature. This heuristic is
also used in rtpsource which implements the probation mechanism and a
similar heuristic is used to handle long gaps.

https://bugzilla.gnome.org/show_bug.cgi?id=769536
2017-06-28 11:51:10 -04:00
Andrew
76792a5c20 rtpjitterbuffer: Don't always reset PTS to 0 after a gap
In function rtp_jitter_buffer_calculate_pts: If gap in incoming RTP
timestamps is more than (3 * jbuf->clock_rate) we call
rtp_jitter_buffer_reset_skew which resets pts to 0. So components down
the pipeline (playes, mixers) just skip frames/samples until pts becomes
equal to pts before gap.

In version 1.10.2 and before this checking was bypassed for packets with
"estimated dts", and gaps were handled correctly.

https://bugzilla.gnome.org/show_bug.cgi?id=778341
2017-02-26 12:41:19 +02:00
Edward Hervey
e5158ca496 jitterbuffer: Don't leak duplicate items
When providing items with a seqnum, there is a (very small) probability
that an element with the same seqnum already exists. Don't forget
to free that item if it wasn't inserted.

And avoid returning undefined values when dealing with duplicate items
2016-12-02 09:01:57 +01:00
Havard Graff
1a4393fb4d rtpjitterbuffer: fix timer-reuse bug
When doing rtx, the jitterbuffer will always add an rtx-timer for the next
sequence number.

In the case of the packet corresponding to that sequence number arriving,
that same timer will be reused, and simply moved on to wait for the
following sequence number etc.

Once an rtx-timer expires (after all retries), it will be rescheduled as
a lost-timer instead for the same sequence number.

Now, if this particular sequence-number now arrives (after the timer has
become a lost-timer), the reuse mechanism *should* now set a new
rtx-timer for the next sequence number, but the bug is that it does
not change the timer-type, and hence schedules a lost-timer for that
following sequence number, with the result that you will have a very
early lost-event for a packet that might still arrive, and you will
never be able to send any rtx for this packet.

Found by Erlend Graff - erlend@pexip.com

https://bugzilla.gnome.org/show_bug.cgi?id=773891
2016-11-04 16:56:56 +02:00
Havard Graff
fb9c75db36 rtpjitterbuffer: fix lost-event using dts instead of pts
The lost-event was using a different time-domain (dts) than the outgoing
buffers (pts). Given certain network-conditions these two would become
sufficiently different and the lost-event contained timestamp/duration
that was really wrong. As an example GstAudioDecoder could produce
a stream that jumps back and forth in time after receiving a lost-event.

The previous behavior calculated the pts (based on the rtptime) inside the
rtp_jitter_buffer_insert function, but now this functionality has been
refactored into a new function rtp_jitter_buffer_calculate_pts that is
called much earlier in the _chain function to make pts available to
various calculations that wrongly used dts previously
(like the lost-event).

There are however two calculations where using dts is the right thing to
do: calculating the receive-jitter and the rtx-round-trip-time, where the
arrival time of the buffer from the network is the right metric
(and is what dts in fact is today).

The patch also adds two tests regarding B-frames or the
“rtptime-going-backwards”-scenario, as there were some concerns that this
patch might break this behavior (which the tests shows it does not).
2016-11-04 16:51:20 +02:00
Havard Graff
bea35f97c8 rtpjitterbuffer: fix bug in reschedule_timer
The new timeout is always going to be (timeout + delay), however, the
old behavior compared the current timeout to just (timeout), basically
being (delay) off.

This would happen if rtx-delay == rtx-retry-timeout, with the result that
a second rtx attempt for any buffers would be scheduled immediately instead
of after rtx-delay ms.

Simply calculate (new_timeout = timeout + delay) and then use that instead.

https://bugzilla.gnome.org/show_bug.cgi?id=773905
2016-11-04 16:40:14 +02:00
Thomas Bluemel
567afdd4d3 rtpjitterbuffer: Fix calculating next_seqnum when dropping old buffers from a full queue.
Fixes calculating the next sequence number when a ITEM_TYPE_LOST with more than one
definitely lost packets is encountered.

https://bugzilla.gnome.org/show_bug.cgi?id=769757
2016-09-14 19:47:28 -04:00
Havard Graff
f440b074b1 rtpjitterbuffer: improved rtx-rtt averaging
The basic idea is this:
1. For *larger* rtx-rtt, weigh a new measurement as before
2. For *smaller* rtx-rtt, be a bit more conservative and weigh a bit less
3. For very large measurements, consider them "outliers"
   and count them a lot less

The idea being that reducing the rtx-rtt is much more harmful then
increasing it, since we don't want to be underestimating the rtt of the
network, and when using this number to estimate the latency you need for
you jitterbuffer, you would rather want it to be a bit larger then a bit
smaller, potentially losing rtx-packets. The "outlier-detector" is there
to prevent a single skewed measurement to affect the outcome too much.
On wireless networks, these are surprisingly common.

https://bugzilla.gnome.org/show_bug.cgi?id=769768
2016-09-14 19:37:50 -04:00
Stian Selnes
f8238f0a9f rtpjitterbuffer: Detect whether to assume equidistant spacing when loss
Assuming equidistant packet spacing when that's not true leads to more
loss than necessary in the case of reordering and jitter. Typically this
is true for video where one frame often consists of multiple packets
with the same rtp timestamp. In this case it's better to assume that the
missing packets have the same timestamp as the last received packet, so
that the scheduled lost timer does not time out too early causing the
packets to be considered lost even though they may arrive in time.

https://bugzilla.gnome.org/show_bug.cgi?id=769768
2016-09-14 19:37:50 -04:00
Stian Selnes
2eb7383816 rtpjitterbuffer: Don't request rtx if 'now' is past retry period
There is no need to schedule another EXPECTED timer if we're already
past the retry period. Under normal operation this won't happen, but if
there are more timers than the jitterbuffer is able to process in
real-time, scheduling more timers will just make the situation worse.
Instead, consider this packet as lost and move on. This scenario can
occur with high loss rate, low rtt and high configured latency.

https://bugzilla.gnome.org/show_bug.cgi?id=769768
2016-09-14 19:37:50 -04:00
Stian Selnes
ab49dfd0b2 rtpjitterbuffer: Fix lost duration when gap after lost timer
This patch fixes an issue with the estimated gap duration when there is
a gap immediately after a lost timer has been processed. Previously
there was a discrepancy beteen the gap in seqnum and gap in dts which
would cause wrong calculated duration. The issue would only be seen with
retranmission enabled since when it's disabled lost timers are only
created when a packet is received and the actual gap length and last dts
is known.

https://bugzilla.gnome.org/show_bug.cgi?id=769768
2016-09-14 19:37:50 -04:00
Havard Graff
dd020f5cc8 rtpjitterbuffer: Expose rtx-deadline as a property
The default -1 gives the old behavior.

https://bugzilla.gnome.org/show_bug.cgi?id=769768
2016-09-14 19:37:50 -04:00
Havard Graff
8087a8a31c rtpjitterbuffer: Improved expected-timer handling when gap > 0
https://bugzilla.gnome.org/show_bug.cgi?id=769768
2016-09-14 19:37:50 -04:00
Stian Selnes
38a7545003 rtpjitterbuffer: Major improvements for RTX stats
Stats should also be collected for unsuccessful packets.

rtx-rtt is very important for determining the necessary configured
latency on the jitterbuffer. It's especially important to be able to
increase the latency when retransmitted packets arrive too late and are
considered lost. This patch includes these late packets in the
calculation of the various rtx stats, making them more correct and
useful.

Also in the case where the original packet arrives after a NACK is sent,
the received RTX packet should update the stats since it provides useful
information about RTT.

The RTT is only updated if and only if all requested retranmissions are
received. That way the RTT is guaranteed to make sense. If not we don't
know which request the packet is a response to and the RTT may be bogus.
A consequence of this patch is that RTT is not updated for a request
when one of the RTX packets for that seqnum is lost, but that since
measured RTT will be more accurate.

The implementation store the RTX information from the timed out timers
and use this when the retransmitted packet arrives. For performance
these timers are stored separately from the "normal" timers in order to
not impact performance (see attached performance test).

https://bugzilla.gnome.org/show_bug.cgi?id=769768
2016-09-14 19:37:50 -04:00
Havard Graff
1b868cc9b1 rtpjitterbuffer: Add and expose more stats and increase testing of it
Add num-pushed and num-lost.
Expose num-late, num-duplicates and avg-jitter.

https://bugzilla.gnome.org/show_bug.cgi?id=769768
2016-09-14 19:37:50 -04:00
Havard Graff
1436fc01e9 rtpjitterbuffer: Option to disable rtx-delay-reorder
When disabled we can save some iterations over timers.

There is probably an argument for rtx-delay-reorder to exist, but
for normal operations, handling jitter (reordering) is something a
jitterbuffer should do, and this variable feels like functionality that
is not "in-sync" with what the jitterbuffer is trying to achieve.

Example: You have 50ms jitter on your network, and are receiving
audio packets with 10ms durations. An audio packet should not be
considered late until its rtx-timeout has expired (and hence a rtx-event
is sent), but with rtx-delay-reorder, events will be sent pretty much
all the time due to the jitter on the network.

Point being: The jitterbuffer should adapt its size to the measured network
jitter, and then rtx-delay-reorder needs to adapt as well, or simply
get out of the way and let the other (better) rtx-mechanisms do their job.

Also change find_timer to only use seqnum as an argument, since there
will only ever be one timer per seqnum at any given time. In the
one case where the type matters, the caller simply checks the type.

https://bugzilla.gnome.org/show_bug.cgi?id=769768
2016-09-14 19:37:50 -04:00
Thomas Bluemel
4dff74358e rtpjitterbuffer: Actually calculate the packet rate for max-dropout and max-misorder calculations.
https://bugzilla.gnome.org/show_bug.cgi?id=751311
2016-08-10 19:49:27 +02:00
Thiago Santos
7f0381fdd9 rtpjitterbuffer: avoid unref of null buffer
The current 'l' pointer will be NULL when the loop
is interrupted with a 'break' statement. Need to have
it advance to the next list item before interrupting.
2016-08-04 00:36:28 -03:00
Olivier Crête
5328378132 rtpjitterbuffer: Work with non-TIME segments
With non-time segments, it now assumes that the arrival time of packets
is not relevant and that only the RTP timestamp matter and it produces
an output segment start at running time 0.

https://bugzilla.gnome.org/show_bug.cgi?id=766438
2016-06-08 14:49:49 -04:00
Olivier Crête
0ebdb97797 jitterbuffer: Upgrade debug message to error
It causes the entire pipeline to fail, it should be easier to find.
2016-05-14 12:36:08 +02:00
Havard Graff
8f7962e1c3 rtpjitterbuffer: Fix stall when receiving already lost packet
When a packet arrives that has already been considered lost as part of a
large gap the "lost timer" for this will be cancelled. If the remaining
packets of this large gap never arrives, there will be missing entries
in the queue and the loop function will keep waiting for these packets
to arrive and never push another packet, effectively stalling the
pipeline.

The proposed fix conciders parts of a large gap definitely lost (since
they are calculated from latency) and ignores the late arrivals.

In practice the issue is rare since large gaps are scheduled immediately,
and for the stall to happen the late arrival needs to be processed
before this times out.

https://bugzilla.gnome.org/show_bug.cgi?id=765933
2016-05-06 14:32:42 +03:00
Sebastian Dröge
608b4ee53c rtpjitterbuffer: Ensure to not take caps with the wrong pt for getting the clock-rate
Especially the caps on the pad might be out of date, and the new caps would be
provided for the current pt via the request-pt-map signal.

https://bugzilla.gnome.org/show_bug.cgi?id=765689
2016-04-27 20:52:27 +03:00
Edward Hervey
5fa1c2ba59 jiterbuffer: Move assertion to the right location
We shouldn't have "late" lost timers at that point
2016-04-07 13:01:52 +02:00
Edward Hervey
b82da62922 jitterbuffer: Speed up lost timeout handling
When downstream blocks, "lost" timers are created to notify the
outgoing thread that packets are lost.

The problem is that for high packet-rate streams, we might end up with
a big list of lost timeouts (had a use-case with ~1000...).

The problem isn't so much the amount of lost timeouts to handle, but
rather the way they were handled. All timers would first be iterated,
then the one selected would be handled ... to re-iterate the list again.

All of this is being done while the jbuf lock is taken, which in some use-cases
would return in holding that lock for 10s... blocking any buffers from
being accepted in input... which would then arrive late ... which would
create plenty of lost timers ... which would cause the same issue.

In order to avoid that situation, handle the lost timers immediately when
iterating the list of pending timers. This modifies the complexity from
a quadratic to a linear complexity.

https://bugzilla.gnome.org/show_bug.cgi?id=762988
2016-04-07 10:14:24 +02:00
Edward Hervey
d656fe8d54 jitterbuffer: Don't create lost events if we don't need them
When "do-lost" is set to FALSE we don't use/send the lost events.
In that case, don't create them to start with :)

https://bugzilla.gnome.org/show_bug.cgi?id=762988
2016-04-07 10:13:56 +02:00
Edward Hervey
cf866a8469 jitterbuffer: Add tracing of lock usage
Helps with debugging lock usage

https://bugzilla.gnome.org/show_bug.cgi?id=762988
2016-04-07 10:06:18 +02:00
Sebastian Dröge
df247f091c rtpjitterbuffer: Add RFC7273 media clock handling
https://bugzilla.gnome.org/show_bug.cgi?id=762259
2016-04-03 11:24:34 +03:00
Vineeth TM
1071309870 good: use new gst_element_class_add_static_pad_template()
https://bugzilla.gnome.org/show_bug.cgi?id=763076
2016-03-24 14:32:20 +02:00
Sebastian Dröge
b6e10be278 Revert "rtpjitterbuffer: don't forget to unlock mutex in error code path in two cases"
This reverts commit a7fb7b5359.

The mutex is taken by the caller, we should keep it locked when returning so
the caller can unlock it again.
2016-03-02 13:13:24 +02:00
Tim-Philipp Müller
a7fb7b5359 rtpjitterbuffer: don't forget to unlock mutex in error code path in two cases 2016-03-01 14:14:36 +00:00
Stian Selnes
5a2cc41398 rtpmanager: Don't warn for duplicate/reordered packets
This is a normal scenario and should not be a warning.

https://bugzilla.gnome.org/show_bug.cgi?id=762208
2016-02-21 22:37:57 +00:00
Sebastian Dröge
366bbffcd8 Revert "WIP: rtpjitterbuffer: Add RFC7273 media clock handling"
This reverts commit 271501f657.

It wasn't meant to be pushed yet as the commit message indicates.
2016-01-18 11:30:45 +02:00
Sebastian Dröge
271501f657 WIP: rtpjitterbuffer: Add RFC7273 media clock handling 2016-01-18 08:58:59 +02:00
Sebastian Dröge
e4b2360e6e rtpjitterbuffer: Fix packet dropping after a big discont
We would queue 5 consective packets before considering a reset and a proper
discont here. Instead of expecting the next output packet to have the current
seqnum (i.e. the fifth), expect it to have the first seqnum. Otherwise we're
going to drop all queued up packets.
2015-12-09 12:24:09 +02:00
Luis de Bethencourt
9fee2c7c9f rtpmanager: switch G_GINT64_FORMAT for GST_STIME_ARGS
No need to use G_GINT64_FORMAT for potentially negative values of
GstClockTimeDiff. Since 1.6 these can be handled with GST_STIME_ARGS.
Plus it creates more readable values in the logs.

https://bugzilla.gnome.org/show_bug.cgi?id=757480
2015-11-03 14:47:00 +00:00
Miguel París Díaz
f321bfeaf4 rtpmanager: Take into account packet rate for max-dropout and max-misorder calculations
https://bugzilla.gnome.org/show_bug.cgi?id=751311
2015-10-07 12:07:18 +01:00
Miguel París Díaz
4c96094fbb rtpmanager: add "max-dropout-time" and "max-misorder-time" props
https://bugzilla.gnome.org/show_bug.cgi?id=751311
2015-10-07 12:06:47 +01:00
Sebastian Dröge
7046852e7d gst: Don't use deprecated gst_segment_to_position() 2015-09-26 00:12:46 +02:00
Sebastian Dröge
01c0f8723f rtpbin/rtpjitterbuffer/rtspsrc: Add property to set maximum ms between RTCP SR RTP time and last observed RTP time
https://bugzilla.gnome.org/show_bug.cgi?id=755125
2015-09-25 23:55:05 +02:00
Mark Nauwelaerts
b7b244f356 rtpjitterbuffer: reset just a bit more upon flush_stop 2015-09-13 15:42:06 +02:00
Mark Nauwelaerts
1e7a3473fd rtpjitterbuffer: remove dead struct member 2015-09-13 15:41:03 +02:00
Sebastian Dröge
68a9209408 rtpjitterbuffer: Keep the DTS estimate if we got no DTS after a jitterbuffer reset
Otherwise we will just output buffers without timestamps after a reset if no
timestamps are provided by upstream, e.g. when using RTSP over TCP.

https://bugzilla.gnome.org/show_bug.cgi?id=749536
2015-08-13 16:45:16 +02:00
Sebastian Dröge
582ade2c42 rtpjitterbuffer: Fix indention 2015-07-10 00:13:32 +03:00
Sebastian Dröge
ae8acc0973 rtpjitterbuffer: Always estimate DTS from the current clock time
Estimating it from the RTP time will give us the PTS, so in cases of PTS!=DTS
we would produce wrong DTS. As now the estimated DTS is based on the clock,
don't store it in the jitterbuffer items as it would otherwise be used in the
skew calculations and would influence the results. We only really need the DTS
for timer calculations.

https://bugzilla.gnome.org/show_bug.cgi?id=749536
2015-07-10 00:13:22 +03:00
Sebastian Dröge
6e7c724afa rtpjitterbuffer: Calculate DTS from the clock if we had none for the first packet after a reset
https://bugzilla.gnome.org/show_bug.cgi?id=749536
2015-07-08 23:19:52 +03:00
Havard Graff
ddd032f56b rtpjitterbuffer: fix gap-time calculation and remove "late"
The amount of time that is completely expired and not worth waiting for,
is the duration of the packets in the gap (gap * duration) - the
latency (size) of the jitterbuffer (priv->latency_ns). This is the duration
that we make a "multi-lost" packet for.

The "late" concept made some sense in 0.10 as it reflected that a buffer
coming in had not been waited for at all, but had a timestamp that was
outside the jitterbuffer to wait for. With the rewrite of the waiting
(timeout) mechanism in 1.0, this no longer makes any sense, and the
variable no longer reflects anything meaningful (num > 0 is useless,
the duration is what matters)

Fixed up the tests that had been slightly modified in 1.0 to allow faulty
behavior to sneak in, and port some of them to use GstHarness.

https://bugzilla.gnome.org/show_bug.cgi?id=738363
2015-07-08 23:18:48 +03:00
Stian Selnes
40524e5a49 Revert "rtpjitterbuffer: Fix expected_dts calc in calculate_expected"
This reverts commit 05bd708fc5.

The reverted patch is wrong and introduces a regression because there
may still be time to receive some of the packets included in the gap
if they are reordered.
2015-07-08 23:18:48 +03:00
Sebastian Dröge
4e23481d9f rtpjitterbuffer: Calculate receive time if we don't have any
This is required to properly schedule packet loss timers and make
sure all our calculations work properly.

https://bugzilla.gnome.org/show_bug.cgi?id=749536
2015-07-08 17:02:05 +03:00
Sebastian Dröge
243730ced4 rtpjitterbuffer: Handle seqnum gaps in TCP streams without erroring out or overflowing calculations
That is, handle DTS==GST_CLOCK_TIME_NONE correctly.

https://bugzilla.gnome.org/show_bug.cgi?id=749536
2015-07-08 15:15:00 +03:00
Miguel París Díaz
5ae672fd22 rtpjitterbuffer: Consider timers len to compare with RTP_MAX_DROPOUT
When there are a lot of small gaps, we can consider that there is
a big gap (too losses) to reset the buffer.

https://bugzilla.gnome.org/show_bug.cgi?id=751636
2015-07-02 18:38:46 +02:00