Hi Pauli,
[ EXTERNAL EMAIL ]
Hi,
ke, 2025-06-25 kello 13:24 +0800, Yang Li kirjoitti:
[clip]
When music is paused on the pixel 9 phone, the CIS link gets
disconnected. As the transport state changes from ACTIVE to IDLE, the
stream state transitions from config to qos.
> HCI Event: Disconnect Complete (0x05) plen 4 #1425 [hci0]
49.572089
Status: Success (0x00)
Handle: 512 Address: 6A:AB:51:47:3B:80 (Resolvable)
Identity type: Random (0x01)
Identity: E8:D5:2B:59:57:A6 (Static)
Reason: Remote User Terminated Connection (0x13)
bluetoothd[2313]: src/shared/bap.c:stream_io_disconnected() stream
0x1f9fc20 io disconnected
bluetoothd[2313]: src/shared/bap.c:bap_ucast_set_state() stream
0x1f9fc20 dir 0x01: releasing -> config
bluetoothd[2313]: src/shared/bap.c:stream_notify() stream 0x1f9fc20
state 1
bluetoothd[2313]: profiles/audio/transport.c:bap_state_changed()
stream 0x1f9fc20: releasing(6) -> config(1)
bluetoothd[2313]:
profiles/audio/transport.c:transport_update_playing()
/org/bluez/hci0/dev_6A_AB_51_47_3B_80/fd1 State=TRANSPORT_STATE_ACTIVE
Playing=0
bluetoothd[2313]:
profiles/audio/transport.c:media_transport_remove_owner() Transport
/org/bluez/hci0/dev_6A_AB_51_47_3B_80/fd1 Owner :1.1
bluetoothd[2313]: profiles/audio/transport.c:media_owner_free() Owner
:1.1
bluetoothd[2313]:
profiles/audio/transport.c:media_transport_suspend() Transport
/org/bluez/hci0/dev_6A_AB_51_47_3B_80/fd1 Owner
bluetoothd[2313]: profiles/audio/transport.c:transport_set_state()
State changed /org/bluez/hci0/dev_6A_AB_51_47_3B_80/fd1:
TRANSPORT_STATE_ACTIVE -> TRANSPORT_STATE_IDLE
bluetoothd[2313]: src/shared/bap.c:stream_disable() stream 0x1f9fc20
AFAICS the bug appears to be:
- bap.c:stream_disable() should do nothing if stream is
not ENABLING or STREAMING
since it's called from bt_bap_stream_disable() which is called on
transport suspend which should be noop for BAP server if stream is
already gone.
Following your suggestion, I updated the |stream_disable()| function,
which successfully resolved the issue.
--- a/src/shared/bap.c
+++ b/src/shared/bap.c
@@ -2059,7 +2059,9 @@ static uint8_t stream_disable(struct bt_bap_stream
*stream, struct iovec *rsp)
/* Sink can autonomously transit to QOS while source needs to go to
* Disabling until BT_ASCS_STOP is received.
*/
- if (stream->ep->dir == BT_BAP_SINK)
+ if (stream->ep->dir == BT_BAP_SINK &&
+ (stream->ep->state == BT_ASCS_ASE_STATE_ENABLING ||
+ stream->ep->state == BT_ASCS_ASE_STATE_STREAMING))
stream_set_state(stream, BT_BAP_STREAM_STATE_QOS);
bluetoothd[2313]: src/shared/bap.c:bap_ucast_set_state() stream
0x1f9fc20 dir 0x01: config -> qos
ATTbluetoothd[2313]: < ACL Data TX: H.. flags 0x00 dlen 51 #1426
[hci0] 49.585656
ATT: Handle Value Notification (0x1b) len 46
Handle: 0x007b Type: Sink ASE (0x2bc4)
Data[44]:
01010002050a00204e00409c00204e00409c0006000000001302010302020105030300000003042800020501
ASE ID: 1
State: Codec Configured (0x01)
Framing: Unframed PDUs supported (0x00)
PHY: 0x02
LE 2M PHY preffered (0x02)
RTN: 5
Max Transport Latency: 10
Presentation Delay Min: 20000 us
...
bluetoothd[2313]: < ACL Data TX: H.. flags 0x00 dlen 24 #1427 [hci0]
49.585725
ATT: Handle Value Notification (0x1b) len 19
Handle: 0x007b Type: Sink ASE (0x2bc4)
Data[17]: 0102010010270000025000050a00204e00
ASE ID: 1
State: QoS Configured (0x02)
CIG ID: 0x01
CIS ID: 0x00
...
when playback resumes on the phone, it attempts to set the ASE state to
Codec. However, since the stream has already transitioned from config to
qos, the phone ends up disconnecting the connection.
bluetoothd[2313]: < ACL Data TX: H.. flags 0x00 dlen 12 #1433 [hci0]
60.216004
ATT: Handle Value Notification (0x1b) len 7
Handle: 0x0087 Type: ASE Control Point (0x2bc6)
Data[5]: 0101010000
Opcode: Codec Configuration (0x01)
Number of ASE(s): 1
ASE: #0
ASE ID: 0x01
ASE Response Code: Success (0x00)
ASE Response Reason: None (0x00)
bluetoothd[2313]: < ACL Data TX: H.. flags 0x00 dlen 51 #1434 [hci0]
60.226086
ATT: Handle Value Notification (0x1b) len 46
Handle: 0x007b Type: Sink ASE (0x2bc4)
Data[44]:
01010002050a00204e00409c00204e00409c0006000000001302010302020105030300000003042800020501
ASE ID: 1
State: Codec Configured (0x01)
Framing: Unframed PDUs supported (0x00)
PHY: 0x02
LE 2M PHY preffered (0x02)
RTN: 5
Max Transport Latency: 10
Presentation Delay Min: 20000 us
Presentation Delay Max: 40000 us
Preferred Presentation Delay Min: 20000 us
Preferred Presentation Delay Max: 40000 us
Codec: LC3 (0x06)
Codec Specific Configuration: #0: len 0x02 type 0x01
Sampling Frequency: 16 Khz (0x03)
Codec Specific Configuration: #1: len 0x02 type 0x02
Frame Duration: 10 ms (0x01)
Codec Specific Configuration: #2: len 0x05 type 0x03
Location: 0x00000003
Front Left (0x00000001)
Front Right (0x00000002)
Codec Specific Configuration: #3: len 0x03 type 0x04
Frame Length: 40 (0x0028)
Codec Specific Configuration: #4: len 0x02 type 0x05
Frame Blocks per SDU: 1 (0x01)
...
> HCI Event: Disconnect Complete (0x05) plen 4 #1445 [hci0]
63.651497
Status: Success (0x00)
Handle: 16 Address: 6A:AB:51:47:3B:80 (Resolvable)
Identity type: Random (0x01)
Identity: E8:D5:2B:59:57:A6 (Static)
Reason: Remote User Terminated Connection (0x13)
Introducing a check in the stream logic to distinguish between Unicast
and Broadcast would indeed make it easier to handle different stream
types cleanly.
However, if we temporarily ignore Unicast caching, a simpler and cleaner
approach would be to transition the stream directly to IDLE when I/O is
disconnected.
This disconnect callback is used for all the roles: unicast client,
unicast server, broadcast. All of those require different handling, so
it's probably most clear to split it.
For unicast server:
The behavior has to follow BAP v1.0.2 Sec. 5.6.8 and ASCS Table 3.2.
Transition to IDLE is only allowed from RELEASING --- but one can as
well go to CONFIG like it is in current master.
CIS loss from STREAMING should go to QOS, and I think it currently does
so, via bap_stream_set_io.
From a brief look, the current version in master is maybe OK, although
one could test the above case again with stream_disable() fix.
For unicast client:
The current version in master is probably OK, although one could double
check it again.
I've added type checking and made the following modifications. Please
check if this meets the requirements. I believe that unicast source and
broadcast source are active behaviors, so there's no need to change the
stream state through I/O status.
@@ -6584,10 +6584,15 @@ static bool stream_io_disconnected(struct io
*io, void *user_data)
DBG(stream->bap, "stream %p io disconnected", stream);
- if (stream->ep->state == BT_ASCS_ASE_STATE_RELEASING)
+ if (stream->lpac->type == BT_BAP_SINK &&
+ bt_bap_stream_get_state(stream) ==
BT_ASCS_ASE_STATE_RELEASING) {
stream_set_state(stream, BT_BAP_STREAM_STATE_CONFIG);
+ bt_bap_stream_set_io(stream, -1);
+ }
+
+ if (stream->lpac->type == BT_BAP_BCAST_SINK)
+ stream_set_state(stream, BT_BAP_STREAM_STATE_IDLE);
- bt_bap_stream_set_io(stream, -1);
return false;
}
Once the Unicast caching issue is properly resolved, we can revisit and
introduce stream-type-based handling accordingly.
+ stream_set_state(stream, BT_BAP_STREAM_STATE_IDLE);
- bt_bap_stream_set_io(stream, -1);
return false;
}
---
base-commit: ae1b7f6ba805f82742bbc32ff275e268248ef9f8
change-id: 20250624-bap_for_big_sync_lost-63476c679dbb
Best regards,
--
Yang Li <yang.li@xxxxxxxxxxx>
--
Luiz Augusto von Dentz
--
Pauli Virtanen
