Due to us not properly acknowleding the time when the last RTX was sent
when scheduling a new one, it can easily happen that due to the packet
you are requesting have a PTS that is slightly old (but not too old when
adding the latency of the jitterbuffer), both its calculated second and
third (etc.) timeout could already have passed. This would lead to a burst
of RTX requests, which acts completely against its purpose, potentially
spending a lot more bandwidth than needed.
This has been properly reproduced in the test:
test_rtx_not_bursting_requests
The good news is that slightly re-thinking the logic concerning
re-requesting RTX, made it a lot simpler to understand, and allows us
to remove two members of the RtpTimer which no longer serves any purpose
due to the refactoring. If desirable the whole "delay" concept can actually
be removed completely from the timers, and simply just added to the timeout
by the caller of the API. But that can be a change for a another time.
The only external change (other than the improved behavior around bursting
RTX) is that the "delay" field now stricly represents the delay between
the PTS of the RTX-requested packet and the time it is requested on,
whereas before this calculation was more about the theoretical calculated
delay. This is visible in three other RTX-tests where the delay had
to be adjusted slightly. I am confident however that this change is
correct.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/-/merge_requests/789>
In order to support the symbol g_enum_to_string in various
project using GStreamer ( gst-validate etc.), the glib minimum
version should be 2.56.0.
Remove compat code as glib requirement
is now > 2.56
Version used by Ubuntu 18.04 LTS
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/-/merge_requests/774>
Until now, do_expected_timeout() was shortly dropping the JBUF_LOCK in order
to push RTX event event without causing deadlock. As a side effect, some
CPU hung would happen as the timerqueue would get filled while looping over
the due timers. To mitigate this, we were processing the lost timer first and
placing into a queue the remainign to be processed later.
In the gap caused by an unlock, we could endup receiving one of the seqnum
present in the pending timers. In that case, the timer would not be found and
a new one was created. When we then update the expected timer, the seqnum
would already exist and the updated timer would be lost.
In this patch we remove the unlock from do_expected_timeout() and place all
pending RTX event into a queue (instead of pending timer). Then, as soon as
we have selected a timer to wait (or if there is no timer to wait for) we send
all the upstream RTX events. As we no longer unlock, we no longer need to pop
more then one timer from the queue, and we do so with the lock held, which
blocks any new colliding timers from being created.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/-/merge_requests/616>
When a GST_EVENT_FLUSH_START reaches the jitterbuffer, there is a chance that
our task is currently blocking waiting for a timer.
There was two problems:
* That wait wasn't checking for flushing situations
* The flushing handling wasn't waking up that conditional (to check whether it
should abort)
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-good/-/merge_requests/608>
The problem was this:
Due to the highly irregular arrival of RTX-packet the max-misorder variable
could be pushed very low. (-10).
If you then at some point get a big in the sequence-numbers (62 in the
test) you end up sending RTX-requests for some of those packets, and then
if the sender answers those requests, you are going to get a bunch of
RTX-packets arriving. (-13 and then 5 more packets in the test)
Now, if max-misorder is pushed very low at this point, these RTX-packets
will trigger the handle_big_gap_buffer() logic, and because they arriving
so neatly in order, (as they would, since they have been requested like
that), the gst_rtp_jitter_buffer_reset() will be called, and two things
will happen:
1. priv->next_seqnum will be set to the first RTX packet
2. the 5 RTX-packet will be pushed into the chain() function
However, at this point, these RTX-packets are no longer valid, the
jitterbuffer has already pushed lost-events for these, so they will now
be dropped on the floor, and never make it to the waiting loop-function.
And, since we now have a priv->next_seqnum that will never arrive
in the loop-function, the jitterbuffer is now stalled forever, and will
not push out another buffer.
The proposed fixes:
1. Don't use RTX in calculation of the packet-rate.
2. Don't use RTX in large-gap logic, as they are likely to be dropped.
Changing the types from boolean to guint due to the ++ operand used on
them, and only call JBUF_SIGNAL_QUEUE after settling down,
or else you end up signaling the waiting code in chain() for every buffer
pushed out.
This is a concept that only applies when a buffer arrives in the chain
function, and it has already been scheduled as part of a "multi"-lost
timer.
However, "multi"-lost timers are now a thing of the past, making this
whole concept superflous, and this buffer is now simply counted as "late",
having already been pushed out (albeit as a lost-event).
There is a problem with the code today, where a single timer will
be scheduled for a series of lost packets, and then if the first packet
in that series arrives, it will cause a rescheduling of that timer, going
from a "multi"-timer to a single-timer, causing a lot of the packets
in that timer to be unaccounted for, and creating a situation in where
the jitterbuffer will never again push out another packet.
This patch solves the problem by instead of scheduling those lost packets
as another timer, it instead asks to have that lost-event pushed straight
out.
This very much goes with the intent of the code here: These packets are
so desperately late that no cure exists, and we might as well get the
lost-event out of the way and get on with it.
This change has some interesting knock-on effect being presented in
later commits. It completely removes the concept of "already-lost", so
that is why that test has been disabled in this commit, to be
removed later.
When calling gst_rtp_jitter_buffer_reset you pass in a seqnum.
This is considered the starting-point for a new stream.
However, the old behavior would unref this buffer, basically lying to
the thread that is pushing out buffers saying that it can expect
this buffer, when it would never arrive. The resulting effect being no
more buffer pushed out of the jitterbuffer, and it would buffer
incoming data indefinitely.
By instead inserting the buffer in the gap_packets queue, the _reset()
function will take responsibility for using that as the first buffer
of the new stream.
Fixes#703
We do not have a way to know the format modifiers to use with string
functions provided by the system. G_GUINT64_FORMAT and other string
modifiers only work for glib string formatting functions. We cannot
use them for string functions provided by the stdlib. See:
https://developer.gnome.org/glib/stable/glib-Basic-Types.html#glib-Basic-Types.description
```
../gst/rtpmanager/gstrtpjitterbuffer.c: In function 'gst_jitter_buffer_sink_parse_caps':
../gst/rtpmanager/gstrtpjitterbuffer.c:1523:32: error: unknown conversion type character 'l' in format [-Werror=format=]
|| sscanf (mediaclk, "direct=%" G_GUINT64_FORMAT, &clock_offset) != 1)
^~~~~~~~~~
In file included from /home/nirbheek/cerbero/build/dist/windows_x86/include/glib-2.0/glib/gtypes.h:32,
from /home/nirbheek/cerbero/build/dist/windows_x86/include/glib-2.0/glib/galloca.h:32,
from /home/nirbheek/cerbero/build/dist/windows_x86/include/glib-2.0/glib.h:30,
from /home/nirbheek/cerbero/build/dist/windows_x86/include/gstreamer-1.0/gst/gst.h:27,
from /home/nirbheek/cerbero/build/dist/windows_x86/include/gstreamer-1.0/gst/rtp/gstrtpbuffer.h:27,
from ../gst/rtpmanager/gstrtpjitterbuffer.c:108:
/home/nirbheek/cerbero/build/dist/windows_x86/lib/glib-2.0/include/glibconfig.h:69:28: note: format string is defined here
#define G_GUINT64_FORMAT "llu"
^
../gst/rtpmanager/gstrtpjitterbuffer.c:1523:32: error: too many arguments for format [-Werror=format-extra-args]
|| sscanf (mediaclk, "direct=%" G_GUINT64_FORMAT, &clock_offset) != 1)
^~~~~~~~~~
```
See also: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/379
By passing `NULL` to `g_signal_new` instead of a marshaller, GLib will
actually internally optimize the signal (if the marshaller is available
in GLib itself) by also setting the valist marshaller. This makes the
signal emission a bit more performant than the regular marshalling,
which still needs to box into `GValue` and call libffi in case of a
generic marshaller.
Note that for custom marshallers, one would use
`g_signal_set_va_marshaller()` with the valist marshaller instead.
The do_expected_timeout() function may release the JBUF_LOCK, so we need
to check if nothing wanted the timer thread to exit after this call.
The side effect was that we may endup going back into waiting for a timer
which will cause arbitrary delay on tear down (or deadlock when test
clock is used).
Fixes#653
JBUF_WAIT_QUEUE drops the JBUF_LOCK, which means the stop condition
for the chain function may have changed (change_state to NULL). Check
this immediately after the wait so that we don't delay shutting down.
This commit adds custom element messages for when gstrtpjitterbuffer
drops an incoming rtp packets due to for example arriving too late.
Applications can listen to these messages on the bus which enables
actions to be taken when packets are dropped due to for example high
network jitter.
Two properties has been added, one to enable posting drop messages and
one to set a minimum time between each message to enable throttling the
posting of messages as high drop rates.
When the queue is full (and adding more packets would risk a seqnum
roll-over), the best approach is to just start pushing out packets
from the other side. Just pushing out the packets results in the
timers being left hanging with old seqnums, so it's safer to just
execute them immediately in this case. It does limit the timer space
to the time it takes to receiver about 32k packets, but without
extended sequence number, this is the best RTP can do.
This also results in the test no longer needed to have timeouts or
timers as pushing packets in drives everything.
Fixes#619
The timer passed to update_timers may be from the stats timer. At the
moment, we could endup rescheduling (reusing) that timer onto the normal
timer queue, unschedul it as if it was from the normal timer queue or
duplicate it into the stats timer queue again. This was protected before
as the with the fact the stats timer didn't have a valid idx.
As the offset is already applied now, we need to update and reschedule
all timers each time the offset is changed. I'm not sure who expect this
to be retro-actively applied, but there was a unit test for it.
If the jitterbuffer head change, there is no need to systematically
wakeup the timer thread. The timer thread will be waken up on if
an earlier timeout has been pushed. This prevent some more spurious
wakeup when the system is loaded. As a side effect, cranking the clock
may set the clock at an earlier position.
In this patch we now make use of the new RtpTimerQueue instead of the
old GArray. This required a lot of changes all over the place, some of
the important changes are that `timer->timeout` is no longer a PTS but
the actual timeout. This was required to get the RtpTimerQueue sorting
right. The applied offset is saved as `timer->offset`, this allow
retreiving back the PTS when needed.
The clockid updates only happens once per incoming packet. If the
currently schedule timer is before the earliest timer in the queue, we
no longer wakeup the thread. This way, if other timers get setup in the
meantime, this will reduce the number of wakup.
The timer loop code has been mostly rewritten, though the behaviour of
running the lost timers first has been kept (even though there is no
test to show what would be the side effect of doing this differently).
Fixes#608
Implement a single timer queue for all timers. The goal is to always use
ordered queues for storing timers. This way, extracting timers for
execution becomes O(1). This also allow separating the clock wait
scheduling from the timer itself and ensure that we only wake up the
timer thread when strictly needed.
The knew data structure is still O(n) on insertions and reschedule,
but we now use proximity optimization so that normal cases should be
really fast. The GList structure is also embeded intot he RtpTimer
structure to reduce the number of allocations.
This moves the RtpJitterBufferStructure type, alloc, free into
rtpjitterbuffer.c/h implementation. jitterbuffer.c strictly rely on
the fact this structure is compatible with GList, and so it make more
sense to keep encapsulate it. Also, anything that could possibly
reduce the amount of code in the element is a win.
In order to support that move, a function pointer to free the data
was added. This also allow making the free function option when
flushing the jitterbuffer.
This helps understanding which function modify the Timerdata
and which one does not. This is not always obvious from thelper
name considering recalculate_timer() does not.
If an rtx packet arrives that hasn't been requested (it might
have been requested from prior to a reset), ignore it so that
it doesn't inadvertently trigger a clock skew.
In the case of reordered packets, calculating skew would cause
pts values to be off. Only calculate skew when packets come
in as expected. Also, late RTX packets should not trigger
clock skew adjustments.
Fixes#612
If, say, a rtx-packet arrives really late, this can have a dramatic
effect on the jitterbuffer clock-skew logic, having it being reset
and losing track of the current dts-to-pts calculations, directly affecting
the packets that arrive later.
This is demonstrated in the test, where a RTX packet is pushed in really
late, and without this patch the last packet will have its PTS affected
by this, where as a late RTX packet should be redundant information, and
not affect anything.
This patch corrects the delay set on EXPECTED timers that are added when
processing gaps. Previously the delay could be too small so that
'timout + delay' was much less than 'now', causing the following retries
to be scheduled too early. (They were sent earlier than
rtx-retry-timeout after the previous timeout.)
