Modifiers currently keep their conflicting and supported devices's
names, and these names are resolved to devices every time we need to use
them. Instead, resolve these device names while creating the modifier
struct and keep track of the resulting device structs in idxsets, same
as how device structs keep track of their support relations.
Signed-off-by: Alper Nebi Yasak <alpernebiyasak@gmail.com>
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/596>
This is intended to make the current and upcoming code a bit clearer, as
we won't need to constantly check for the existence of these idxsets
before using or operating on them.
Signed-off-by: Alper Nebi Yasak <alpernebiyasak@gmail.com>
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/596>
It may be possible that the ALSA control element appears
again. Allow this combination by checking, if the pulseaudio
mixer element already exists. Do not create the duplicate
mixer element in this case.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/729>
PulseAudio v5.99 or later hits assertion at alsa-lib mixer API due to
wrong handling of removal event for mixer element.
pulseaudio: mixer.c:149: hctl_elem_event_handler: Assertion `bag_empty(bag)' failed.
The removal event is defined as '~0U', thus it's not distinguished from
the other type of event just by bitwise operator.
At the removal event, class implementator for mixer API should detach
mixer element from hcontrol element in callback handler since alsa-lib
has assertion to check the list of mixer elements for a hcontrol element
is empty or not after calling all of handlers. In detail, please refer to
MR to alsa-lib:
* https://github.com/alsa-project/alsa-lib/pull/244
This commit fixes the above two issues. The issue can be regenerated by
`samples/ctl` Python 3 script of alsa-gobject.
* https://github.com/alsa-project/alsa-gobject/
It adds some user-defined elements into sound card 0. When terminated by
SIGINT signal, it removes the elements. Then PulseAudio dies due to the
assertion.
Fixes: 1fd8848e64 ("alsa-util: Add functions for accessing mixer elements through mixer class")
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/728>
GST_TYPE_BITMASK is 64-bit bit mask while corresponding channel_mask in
pulseaudio is int therefore usually 32-bit. Switch to uint64_t instead
to match internal representation in gstreamer.
Fixes pulseaudio crash on ARM 32-bit when pulseaudio is compiled with
gstreamer and either LDAC or aptX support is available.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/723>
Default build configuration would fail to run on a system without systemd-logind
(or elogind) and without ConsoleKit daemon responding on dbus interface. Here,
module-console-kit would fail to initialize, preventing daemon from starting.
Make module-console-kit an optional build feature to allow opt-out.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/719>
On FreeBSD (and probably other BSDs as well), the FIONREAD ioctl
on UDP sockets does not return the size of the next datagram (like
it does on Linux), but returns the size of the output buffer: this
count contain multiple datagrams and also contains the headers.
We fixed this by taking the result of the FIONREAD as lower bound
for the size, adding an upper bound and then removing the check
that the sizes should be exactly the same.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/718>
Turned out that pa_sdp_info::enable_opus is never initialized, which seldom
makes module-rtp-recv believe it will be playing OPUS-encoded stream even though
discovered SDP record does not indicate OPUS codec in metadata.
Fix this by adding missing initializer.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/720>
RFC 4566 states that SDP record is terminated with CRLF, and parsers should be
able to accept records terminated with just LF. Pulseaudio only accepts LF here.
Fix this by accepting both CRLF and LF terminators.
The combine sink used the current time and counter when calculating
the latency if smoother_2 was enabled. This lead to wrong latency
reports. This patch fixes the problem by using the snapshot time
and counter instead.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/711>
Bluetooth transport layer already allows for packets larger than mSBC frame, and
there are up to 1 + MTU / (mSBC packet size) complete frames to be decoded from
each incoming SCO packet.
Now decoder fails when there is more than one complete frame available, which
could happen if MTU size is larger than 1.5 * (mSBC packet size) = 90
Fix this by adding a loop over avialable frames, and adjust decoded buffer size
to allow decoding up to 1 + MTU / (mSBC packet size) frames at once.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/706>
When the module is loaded and avahi_client_new() fails because the client cannot
connect, a shutdown of the module is scheduled. In parallel, the client_callback
is called with AVAHI_ERR_DISCONNECTED and another connection attempt is made
which also fails and triggers a second unload of the module. This crashes PA,
because there is already an unload in progress.
This patch fixes the problem by checking if an unload is already scheduled.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/710>
The timestamp used for updating the smoother was taken at the wrong time.
It may take some time until an async message is executed (measured up to
2ms), therefore the timestamp used to update the smoother must be taken
before the message is executed and not inside the message.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/705>
When the tunnel modules had no connection and a re-init was pending, the module
could be unloaded without cancelling the pending re-init. When the timer expired
in that situation, this lead to a crash. This patch fixes the problem by keeping
a reference when the module is scheduled to be re-initialized.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/705>
The AVRCP service is known to not be connected before the A2DP transport
is, resulting in PulseAudio asking BlueZ for an initial 'Volume' value
but not getting it because the property doesn't exist.
To prevent end-users from conjecturing this to be the source of whatever
issue they're observing, demote it to a warning.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/707>
When configured, reinitialize the module instead of exiting. This
allows a restart/reconnect, but the module to appear to always be alive
when the user does: "pactl list modules". (The sink will still not
exist until the tcp connection is established.)
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/688>
The io thread, after connection, sends a message asking for a sink to be
created. After the ctl thread is done with creation, it sends a message
back to the io thread so it can continue. This ensures that the sink
only exists when it's connected to something.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/688>
When monitor source becomes idle it may happen that monitored sink has no
uncorked inputs anymore and can now be suspended. To allow this, detect if state
is changed for monitor source and check state of monitored sink instead.
This change allows pulseaudio to suspend devices when pavucontrol volume meters
are disabled and corresponding peaks resampled streams are corked.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/697>
Turned out that SelectConfiguration is only used for outgoing connections, and
incoming connection from bluetooth headset using SBC codec ends up with a
bitpool as large as declared by headset. When resulting bitpool is so large that
SBC frame size plus RTP header size exceeds write MTU size, number of frames per
packet becomes zero causing crash dividing by zero in update_sink_buffer_size()
Fix this by limiting available bitpool value exposed for SBC endpoints.
Fixes: 89082cbfa ("bluetooth: a2dp dual channel SBC XQ codec configurations")
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/695>
Commit c6d6ca541 ("bluetooth/gst: Replace buffer accumulation in adapter
with direct pull") removed the `timestamp` parameter from GStreamer
transcoders due to being unused, but these should instead be propagated
to the GStreamer encoding buffers.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/494>
Bluetooth codecs should always have fixed in/output and are hence able
to have their results directly read from the codec, instead of
accumulating in a buffer asynchronously that is subsequently only read
in the transcode callback. The Bluetooth backends calling encode/decode
also expect these fixed buffer sizes.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/494>
Handling multiple threads does not come without overhead, especially
when the end-goal is to ping-pong them making the whole system run
serially. This patch rips out all that thread handling and instead
"chains" buffers to be encoded/decoded directly into the pipeline,
making them execute their work on the current thread. The resulting
buffer can be pulled out from appsink immediately without require extra
locking and signalling. While the overhead on modern systems is found
to be negligible or unnoticable, code complexity of such locking and
signalling systems is prevalent making it the main drive behind this
refactor.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/494>
Drop rtpldacpay and payload the LDAC encoded output manually in the
RTP header.
The RTP payload seems to be required as it carries the frame count
information. Right now, rtpldacpay does not add this so construct
the RTP header and payload manually.
Strangely some devices like Shanling MP4 and Sony XM3 would still
work without this while some like the Sony XM4 does not.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/689>
If UCM defines the private alsa-lib configuration, the ELD controls
are expected to use this device configuration too.
With this change:
I: [pulseaudio] alsa-util.c: Successfully attached to mixer '_ucm0009.hw:Loopback'
Without:
I: [pulseaudio] alsa-util.c: Successfully attached to mixer '_ucm0009.hw:Loopback'
I: [pulseaudio] alsa-util.c: Successfully attached to mixer 'hw:4'
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/673>
The hw: device can be addressed using the card index (hw:0)
or the card identifier (ASCII string - hw:Loopback). Both
mixers are equal.
The previous code was fine for the mixers without the UCM
private prefixes (_ucmXXXX). Make code more robust, create
two aliased mixer structures in the mixers array.
Signed-off-by: Jaroslav Kysela <perex@perex.cz>
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/673>
This makes it possible to define multiple sinks/sources on detection
of the jack server. This allows one to for example create a separate
sink for conferencing software and route that in jack to another
channel on their audio interface.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/669>
Even though the file name is currently behringer-umc22.conf, the USB ID
actually belongs to Texas Instruments PCM2902, which is a generic chip
used in multiple products. Some products have true mono input unlike
Behringer UMC22, which has two mono inputs combined into one stereo PCM
device.
This patch removes the "mono,mono" mapping from Behringer UMC22, which
hopefully won't be missed too much (there are still "mono,aux1" and
"aux1,mono" mappings available for mono recording).
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/667>
If the preferred ports are not set in this function, the
entrys_equal() always returns false in the card_put_hook_callback().
This will make the entry be written into the metadata and the
preferred ports will be cleaned by a mistake.
And we met a hdmi audio bug which has sth to do with this issue, on
the machines with the legacy HDA audio driver, the hdmi port has lower
priority than speaker, users need to manually select the hdmi to be
active output port, then the preferred output port is hdmi for this
sound card, after reboot, the card_put_hook_callback() in the
module-card-restore.c will be called and the preferred ports are
cleaned by a mistake, then the hdmi output port or hdmi sink couldn't
switch to be active after reboot or resume automatically. That is
because the preferred ports are cleaned and hdmi port has lower
priority than speaker, the profile_good_for_output() in the
module-switch-on-port-available.c always returns false.
Signed-off-by: Hui Wang <hui.wang@canonical.com>
Change d7f95170a1 added a dependency on device
adapter pointer being valid while checking if bluetooth profile is supported by
device.
When adapter object is released, each device holding pointer to adapter being
released is notified to reset that to NULL. Since adapter objects are released
first when discovery object is unreferenced, each device will have adapter
pointer reset before the time device objects are released.
Fix observed crash by examining device adapter pointer. If it is NULL report
that device does not support any bluetooth profile instead of looking at UUIDs
supported by adapter.
Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/646>
