Age | Commit message (Collapse) | Author | Files | Lines |
|
Recent tries on the minnow board have show that the setting of channels
on a PCM must be done -after- the setting of the rate.
When not doing that, there can be issues (really, complicated to find)
around the boundaries values (open intervals) on the period bytes.
Additionally, the number of channels can now be forced in the json file.
Change-Id: I2c20853daadf9ccab06be6e04d1a185440f57780
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
Reading sound frames relies on usage of poll before calling snd_pcm_writei
Recent seeks for bugs and CPU load issues have shown that the same thing
must be done on the write side.
Thus this commit uses the same mechanism, including the dedicated event
pipe for the thread termination.
Change-Id: Ifd5407c9e27da7801623583a6f7d4a2e845b43ea
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
- Replace controller binder functions definition with the binder ones.
and remove the last used submodules to use the separated libraries.
- Add liburcu as a project dependency in addition to the others ones.
Bug-AGL: SPEC-2139
Change-Id: If46eed24018e28e58d1e62397f4355d2aa46a58e
Signed-off-by: Romain Forlot <romain.forlot@iot.bzh>
|
|
Fixes most memory leaks in softmixer.
The concept of 'transaction' for dynamic streams has
been generalized to the objects created at startup.
The cleanup is done via a handle set through a atexit()
call.
Also added a missing strdup in alsa-api-loop, that fixes
a double free.
Warning, the bluez-alsa PCM are not closed in this
version. This is intentional due to a BUG in the
bluealsa ioplug PCM, that crashes upon close
(pthread_cancel is used to terminate the io_thread
and things get very bad. I have a pending fix for
that, relying on a cancellation pipe, but deeper
testing must be done).
As an effect, only one phone call can be made,
else 4a needs to be restarted
Change-Id: Idb84cafe15f17c0ef02fcc70296d541dc55a2dcf
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
Signed-off-by: Stephane Desneux <stephane.desneux@iot.bzh>
|
|
This adds support for bluetooth telephony.
A big rework in the softmixer internals has been
mandatory, in order to support dynamic streams creation
and deletions.
Bluetooth telephony relies on the recent evolutions
of bluez-alsa, the most important one being the
support of HFP over Ofono. The softmixer opens
PCM ioplugs provided by bluez-alsa.
Bluetooth SCO needs 2 streams, one for listening
and the other for talking. These streams are created
upon requests sent by the hal-manager.
The hal manager subscribes for bluez-alsa events
and request the list of availalble transports.
For each "attach" transaction verb, the softmixer
maintains a list of the all created objects
(sources, sinks, zones, ramps, streams, and more)
Additionnally, it creates a new verb when the attach
succeeds, that verb is typically something like
"sco_XX:XX:XX:XX:XX:XX", and the only supported action
at the present time is {"action":"remove"}, that performs
all the cleanup of the registered objects.
Change-Id: I1b119e6c079e60daf771e63c083a1ef33a39f379
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
Changed the ringbuffer "xxx_free" function to something
more explicit, because it does not free anything.
Changed "single letter" variables to nicer names
Change-Id: I000c57aa5cc684d387105441889e011a45a6ccf3
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
In the writing thread, handle the case when 'writing threshold'
was smaller than ALSA sound card's PCM buffer (defined in the driver).
This avoids waiting indefinitely for a buffer to contain a wanted number
of frames that it cannot contain.
Change-Id: Ie1aa69ea1a93471ed46d571c669ec08a1b827476
Signed-off-by: Jonathan Aillet <jonathan.aillet@iot.bzh>
|
|
Replace some 'printf' by application framework debug functions.
Improve 'ALSA_PCM_UID' macro.
Correct minor errors in debug prints.
Change-Id: I49899fd904d42b5b3bec46017eb319ed7d321dc5
Signed-off-by: Jonathan Aillet <jonathan.aillet@iot.bzh>
|
|
This fixes the mute/unmute logic. This was broken due to
a forgotten push for a commit
Change-Id: I6cbdedefc11ebd508d30ece71df364d9b69c74f0
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
There were too many things printed in notice mode.
Change-Id: Ia7841e219cb2e9ca71e0c4436a2048a812ef6b07
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
The null pcmplug case was not correctly handled in various places.
Also fixed some typos in the log
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
This adds sound playback for incoming sound from
connected bluetooth devices.
In this version, the softmixer relies on a modified
bluez-alsa version (https://github.com/iotbzh/bluez-alsa),
that provides an ioplug PCM bluezalsa connection proxy.
The softmixer api has a new verb to dynamically set the
device to listen to:
afb-client-demo ws://localhost:1234/api?token=\uuid=123 smixer bluezalsa_dev '{"interface":"hci0", "device":"F6:32:15:2A:80:70", "profile":"a2dp"}'
In this way it is possible to switch from a bluezalsa audio source to
another without any further configuration.
For now, only interface hci0 is supported.
This commit also migrates the softmixer binding to API v3
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
Moves the log level of the message from info to debug.
Change-Id: I348421a17875b9061255c3b353e27070b0068c3b
Signed-off-by: Stephane Desneux <stephane.desneux@iot.bzh>
|
|
Now uses two threads for in the playing loop
The first one reads from the capture device (ie, a phys. capture, or snd_aloop) and writes
data to a circular buffer. The second one gets data from the circular buffer and outputs
it to the playback. This model solves a lot of correlated timing bugs between read & write
tasks.
The read tasks only wakes up the write task when the buffer is 80% full.
The buffer size big enough to hold 2 seconds of sound.
The mute implementation has also been simplified, since it has been found out that it was possible
to recover from an interrupted read, by calling snd_pcm_start additionnally to snd_pcm_prepare.
Thus, the mute code consists in listening to an extra file descriptor in the read loop.
Reading from that descriptor gives the mute or unmute command sent at higher level (in the
PCM control event callback).
When a 'mute' order is get, the capture sound fd is simply backup and replaced by '-1' in the
set of the poll of the read task.
When a 'unmute' order is get, the fd is simply restored.
The start threshold is only computed for capture, and hardcoded to 1 for playback.
This removes most of the remaining EPIPE on playback.
The stop threshold has been removed. It had bad side effects on the amount of writeable
data returned by snd_pcm_avail_update (was returning too small chunks)
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
Uses the same logic as 'aplay' to set the start & stop
thresholds. With this fix, there are no more
periodic EPIPE errors when writing to playback
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
when the stream is muted (due to the configuration, or
due to a HAL request), the read/write loop is stopped.
The benefit is that muting will work with capture devices
that do not implement the mute in their driver.
The inconvenient of stopping the read loop is that it
has made appear an unexpected side-effect: the poll on
capture does not trig for further incoming frames.
The workaround is to completely close, then reopen
and configure the capture PCM.
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
- removed the systemd polling usage
- uses a while loop for reading
- improved the write loop, leveraging the available space
at each iteration
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
This fixes the numerous XRUN issues seen on some cards.
The trick is to set the buffer_size & period size in
hw parameters. These ones are calculated from an expected
maximum latency.
Also, the writei is done in a loop to be robust to overruns
and partial writes.
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
|
|
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
|
|
Fixes the UNDERRUN issue (seen while playing radio, for instance)
Signed-off-by: Thierry Bultel <thierry.bultel@iot.bzh>
|
|
|
|
|
|
|
|
Mixing with volume and mute per audio role works.
|
|
|
|
|