This issue was found by enabling ubsan. For me it consistently triggered
after about 28 seconds running a simple example that plays a sine wave
via the mainloop api.
I added a log and confirmed that before the ubsan is triggered the
index variable j is indeed 32 which is out-of-bounds.
Co-authored-by: Arun Raghavan <arun@asymptotic.io>
`pa_pstream_send_memblock()` would split incoming memblock into parts not
exceeding maximum pool block size.
To make sure split parts of memblock are still frame-aligned add new `align` arg
to `pa_pstream_send_memblock`, find out required alignment from stream sample
format and pass it there. Bump default alignment to 256 which is good up to
32bit 64ch frames.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/780>
Executing below command will not produce any audio:
pacat --channels=3 /dev/urandom
Turns out that pa_stream_write() breaks large audio buffers into
segments of the maximum memblock size available -- a value which
is not necessarily frame aligned.
Meanwhile the server discards any non-aligned client audio, as a
security measure, due to some earlier reported daemon crashes.
Thus divide sent audio to the expected aligned form.
CommitReference-1: 22827a5e1e
CommitReference-2: 150ace90f3
BugLink: https://bugs.freedesktop.org/show_bug.cgi?id=98475
BugLink: https://bugs.freedesktop.org/show_bug.cgi?id=77595
Signed-off-by: Ahmed S. Darwish <darwish.07@gmail.com>
In case the sample spec is not known, as can be the case when
pa_stream_new_extended is used, we cannot satisfy the PULSE_LATENCY_MSEC
request.
As a workaround disable being able to use PULSE_LATENCY_MSEC in this case.
Reported-by: Fritsch <fritsch@xbmc.org>
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
FSF addresses used in PA sources are no longer valid and rpmlint
generates numerous warnings during packaging because of this.
This patch changes all FSF addresses to FSF web page according to
the GPL how-to: https://www.gnu.org/licenses/gpl-howto.en.html
Done automatically by sed-ing through sources.
If somebody tries to push a non-frame-aligned memblock onto the
memblockq, then we should fail the write. Otherwise the daemon will
crash, see https://bugs.freedesktop.org/show_bug.cgi?id=77595
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
New function allows to pass data pointer that is a member
of the outer structure that need to be freed too when data
is not needed anymore.
Signed-off-by: Lukasz Marek <lukasz.m.luki2@gmail.com>
The check is done for clients that use pa_stream_new() but not for
clients that use pa_stream_new_extended(). This is inconsistent. We
could check that the volume channels match the channels set in the
format info struct that is passed to pa_stream_new_extended(), but
that doesn't work if the format info doesn't contain the channel
information (that can happen if the client wants the server to choose
the channel count for the stream). And it should also be possible to
pass a mono volume for a multi-channel stream. The check could be
extended to handle all these cases, but I don't see much point in
wasting time on that. The server will anyway validate the stream
parameters, it's not particularly important to fail already when the
stream is being created at the client side.
I think this makes the code a bit nicer to read and write. This also
reduces the chances of off-by-one errors when checking the bounds of
sample rate values.
This patch removes all occurrences of double and triple
newlines.
Command used for this:
find . -type d \( -name ffmpeg \) -prune -o \
-regex '\(.*\.[hc]\|.*\.cc\)' \
-a -not -name 'adrian-aec.*' -a -not \
-name reserve.c -a -not -name 'rtkit.*' \
-exec sed -i -e '/^$/{N;s/^\n$//}' {} \;
Two passes were needed to remove triple newlines.
The excluded files are mirrored files from external sources.
This patch replaces every occurrence of ')\n{' with ') {'.
Command used for this:
find . -type d \( -name ffmpeg \) -prune -o \
-regex '\(.*\.[hc]\|.*\.cc\)' \
-a -not -name core-util.c -a -not \
-name adrian-aec.c -a -not -name g711.c \
-exec sed -i -e '/)$/{N;s/)\n{$/) {/}' {} \;
The excluded files are mirrored files from external sources.
Previously, if there was a hole in a recording stream,
pa_stream_peek() would crash. Holes could be handled silently inside
pa_stream_peek() by generating silence (wouldn't work for compressed
streams, though) or by skipping any holes. However, I think it's
better to let the caller decide how the holes should be handled, so
in case of holes, pa_stream_peek() will return NULL data pointer and
the length of the hole in the nbytes argument.
This change is technically an interface break, because previously the
documentation didn't mention the possibility of holes that need
special handling. However, since holes caused crashing anyway in the
past, it's not a regression if applications keep misbehaving due to
not handing holes properly.
Some words about when holes can appear in recording streams: I think
it would be reasonable behavior if overruns due to the application
reading data too slowly would cause holes. Currently that's not the
case - overruns will just cause audio to be skipped. But the point is
that this might change some day. I'm not sure how holes can occur
with the current code, but as the linked bug shows, they can happen.
It's most likely due to recording from a monitor source where the
thing being monitored has holes in its playback stream.
BugLink: http://bugs.launchpad.net/bugs/1058200
This check was valid before we introduced per-source-output volumes, so
dropping it now. Thanks to Alban Browaeys <prahal@yahoo.com> for
catching this.
This fixes pa_sample_spec init to use the correct API. Not doing so
triggers a valgrind warning as we call pa_sample_spec_valid() on this
later on, which checks the rate and channels fields. Thanks to Rémi
Denis-Courmont for reporting this.
In pa_create_stream_callback, a stream is inserted into
s->context->record_streams only if it's a record stream. Otherwise it's
inserted into s->context->playback_streams. However, in stream_unlink the
stream is removed from s->context->playback_streams only if it's a playback
stream and otherwise it's removed from s->context->record_streams.
Thus, if the stream is an upload stream, we first insert it into
s->context->playback_streams in pa_create_stream_callback and then try to
remove it unsuccessfully from s->context->record_streams in stream_unlink. This
means that we are leaking hashmap entries until the context is freed,
constantly consuming more memory with applications that upload and unload a
large number of samples through one context.
Of course, this begs the question whether upload streams even belong in either
of those hashmaps. I don't want to mess around with the code too much at this
point though, so this patch should be a sufficient improvement.
These are not used for anything at this point, but this
makes it easy to add ad-hoc debug prints that show the
memblockq name and to convert between bytes and usecs.
This patch introduces some extra protocol information, so protocol
version is bumped. This functionality is primarily needed to solve
a long standing issue in alsa-plugins, which should ignore underruns
if and only if it is obsolete, i e, if more data has been written to
the pipe in the meantime (which will automatically end the underrun).
BugLink: http://bugs.launchpad.net/bugs/805940
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
This piggy backs onto the previous changes for protocol 22 and
thus does not bump the version. This and the previous commits should be
seen as mostly atomic. Apologies for any bisecting issues this causes
(although I would expect these to be minimal)
Passing a NULL-terminated array of pa_format_info pointers is a bit
unwieldy for clients. Instead of this, let's pass in an array of
pointers and the number of elements in the array.
This quite is an old patch. It was added to N900 to avoid unnecessary
wake-ups when the phone is in power save mode (= blank screen and
no user interaction). In this situation if the user had a browser
window with flash animation open pulseaudio kept waking up every
10 seconds, causing a severe hit to use times.
Anyway I do not see any reason to send timing updates if the sink or
source where the stream is connected to is suspended.
When the sink format changes and we kill the stream, clients need a way
to know (a) what device they should reconnect to, and (b) what the
stream running time was when the stream got killed (pa_stream_get_time()
won't work after the stream has been killed). This adds these two bits
of information in the event callback's proplist parameter.
This is the beginning of work to support compressed formats natively in
PulseAudio. This adds a pa_stream_new_extended() that takes a format
structure, sends it to the server (=> protocol extension) and has the
server negotiate with the appropropriate sink to figure out what format
it should use.
This is work in progress, and works only with PCM streams. Actual
compressed format support in some sink needs to be implemented, and
extensive testing is required.
More details on how this is supposed to work is available at:
http://pulseaudio.org/wiki/PassthroughSupport
