[pulseaudio-discuss] [PATCH] bluetooth: Let PulseAudio always handle the transition to SINK_STATE_PLAYING.

[Luiz Augusto von Dentz]

Hi Jarkko,

On Thu, Apr 19, 2012 at 3:52 PM, Jarkko Suontausta
<jarkko.suontausta at digia.com> wrote:
> Some A2DP audio devices initiate audio sink state transitions over the
> Audio/Video Remote Control (AVRCP) profile. Specifically, trying to
> switch the A2DP sink into PLAYING state is problematic, because the
> Bluetooth device has no way of knowing when streaming is ready to start
> on the host side. If it is detected that the A2DP audio sink has already
> been set to PLAYING state by Bluez when PulseAudio is starting to
> transition the sink from IDLE to PLAYING, the transport will be released
> and re-acquired (in effect, making PulseAudio the initiator again.)
> This prevents audio loss at least with Jabra SP700.
> ---
>  src/modules/bluetooth/module-bluetooth-device.c |   27 +++++++++++++++++++++-
>  1 files changed, 25 insertions(+), 2 deletions(-)
>
> diff --git a/src/modules/bluetooth/module-bluetooth-device.c b/src/modules/bluetooth/module-bluetooth-device.c
> index b7945ed..e86c745 100644
> --- a/src/modules/bluetooth/module-bluetooth-device.c
> +++ b/src/modules/bluetooth/module-bluetooth-device.c
> @@ -2356,7 +2356,30 @@ static int bt_transport_config(struct userdata *u) {
>  }
>
>  /* Run from main thread */
> -static int bt_transport_open(struct userdata *u) {
> +/* This function is only called when a Bluetooth profile is initialized */
> +static int bt_transport_initialize(struct userdata *u, const pa_bluetooth_device *device) {
> +    pa_assert(u);
> +    pa_assert(device);
> +
> +    /* Some A2DP audio devices initiate audio sink state transitions over the
> +     * Audio/Video Remote Control (AVRCP) profile. Specifically, trying to
> +     * switch the A2DP sink into PLAYING state is problematic, because the
> +     * Bluetooth device has no way of knowing when streaming is ready to start
> +     * on the host side. If it is detected that the A2DP audio sink has already
> +     * been set to PLAYING state by Bluez when PulseAudio is starting to
> +     * transition the sink from IDLE to PLAYING, the transport will be released
> +     * and re-acquired (in effect, making PulseAudio the initiator again.)
> +     * This prevents audio loss at least with Jabra SP700. */
> +    if (u->profile == PROFILE_A2DP && device->audio_sink_state == PA_BT_AUDIO_STATE_PLAYING) {
> +        pa_log_debug("The audio sink of device %s is already in PLAYING state. Release and re-acquire transport.",
> +                     device->name);
> +
> +        if (bt_transport_acquire(u, FALSE) < 0)
> +            return -1;
> +        else
> +            bt_transport_release(u);
> +    }
> +
>     if (bt_transport_acquire(u, FALSE) < 0)
>         return -1;
>
> @@ -2406,7 +2429,7 @@ static int setup_bt(struct userdata *u) {
>     t = pa_bluetooth_device_get_transport(d, u->profile);
>     if (t) {
>         u->transport = pa_xstrdup(t->path);
> -        return bt_transport_open(u);
> +        return bt_transport_initialize(u, d);
>     }
>
>     if (get_caps(u, 0) < 0)
> --
> 1.7.0.4

I don't get what exactly is wrong here, if it is playing we should not
do anything, perhaps this is a problem with the frequency being
different when the headset is the initiator. Btw, do you have any logs
about this problem?


-- 
Luiz Augusto von Dentz