Hi Vlad,
On Fri, Oct 4, 2024 at 8:35 AM Vlad Pruteanu <vlad.pruteanu@xxxxxxx> wrote:
>
> In order to sync to multiple BISes from the same BIG, the existing
> sync must be destroyed and a new one created. This is accomplished
> by prompting the audio server to release the existing, active,
> transports (by moving them to the IDLE state). They will later be
> identified by the RELEASING state of their streams and the process
> for reacquirement (along with the new transport) will begin.
> ---
> profiles/audio/transport.c | 41 ++++++++++++++++++++++++++++++++++++++
> 1 file changed, 41 insertions(+)
>
> diff --git a/profiles/audio/transport.c b/profiles/audio/transport.c
> index caa7287db..e68695c39 100644
> --- a/profiles/audio/transport.c
> +++ b/profiles/audio/transport.c
> @@ -1366,6 +1366,7 @@ static DBusMessage *select_transport(DBusConnection *conn, DBusMessage *msg,
> void *data)
> {
> struct media_transport *transport = data;
> + GSList *l;
>
> if (transport->owner != NULL)
> return btd_error_not_authorized(msg);
> @@ -1375,6 +1376,29 @@ static DBusMessage *select_transport(DBusConnection *conn, DBusMessage *msg,
>
> if (!strcmp(media_endpoint_get_uuid(transport->endpoint),
> BAA_SERVICE_UUID)) {
> + /* Check if there are any ACTIVE transports, from the same
> + * device. If there are, it means that this is a request to add
> + * a new BIS to the active BIG sync. This is done by releasing
> + * the ACTIVE transports, and then reaquiring them along with
> + * the new transport that needs to be added to the sync. To
> + * release the transports, bt_bap_stream_release is called,
> + * which will set the stream's state to
> + * BT_BAP_STREAM_STATE_RELEASING. On bap_state_changed, this
> + * will be detected and transport_update_playing will be called,
> + * with playing set to FALSE. This will move the transport to
> + * IDLE, prompting the audio server to release it.
> + */
> + for (l = transports; l; l = g_slist_next(l)) {
> + struct media_transport *tr = l->data;
> + struct bap_transport *bap_temp = tr->data;
> +
> + if (tr->device == transport->device &&
> + tr->state == TRANSPORT_STATE_ACTIVE) {
> + bt_bap_stream_release(bap_temp->stream,
> + NULL, NULL);
> + }
> + }
This seems a little counter intuitive, I don't think we should allow,
at least not by default, to sync BIS(s) one by one otherwise on every
select we would then need to do the release/reacquire logic that you
are proposing, instead what we should probably be doing is to list the
BIS(s) of the same BIG as Links:
https://github.com/bluez/bluez/blob/master/doc/org.bluez.MediaTransport.rst#arrayobject-links-readonly-optional-iso-only-experimental
> transport_update_playing(transport, TRUE);
> }
>
> @@ -1385,9 +1409,22 @@ static DBusMessage *unselect_transport(DBusConnection *conn, DBusMessage *msg,
> void *data)
> {
> struct media_transport *transport = data;
> + GSList *l;
>
> if (!strcmp(media_endpoint_get_uuid(transport->endpoint),
> BAA_SERVICE_UUID)) {
> + for (l = transports; l; l = g_slist_next(l)) {
> + struct media_transport *tr = l->data;
> + struct bap_transport *bap_temp = tr->data;
> +
> + if (tr->device == transport->device &&
> + tr->state == TRANSPORT_STATE_ACTIVE &&
> + tr != transport) {
> + bt_bap_stream_release(bap_temp->stream,
> + NULL, NULL);
> + }
> + }
> +
> transport_update_playing(transport, FALSE);
> }
>
> @@ -1768,6 +1805,10 @@ static void bap_state_changed(struct bt_bap_stream *stream, uint8_t old_state,
> bap_update_bcast_qos(transport);
> break;
> case BT_BAP_STREAM_STATE_RELEASING:
> + if (bt_bap_stream_io_dir(stream) == BT_BAP_BCAST_SOURCE) {
> + transport_update_playing(transport, FALSE);
> + return;
> + }
> if (bt_bap_stream_io_dir(stream) == BT_BAP_BCAST_SINK)
> return;
> break;
> --
> 2.40.1
>
--
Luiz Augusto von Dentz
