Bluez does not connect to A2DP sink

Bluez does not connect to A2DP sink

[Thomas Ruprecht]

[-- Attachment #1: Type: text/plain, Size: 1322 bytes --]

Hi,

I got a problem with bluez since 5.51 that was recently discussed in the 
thread https://marc.info/?l=linux-bluetooth&m=157081619423081&w=2: My 
headset auto-connects but is not registered as a2dp sink. Unfortunately, 
the proposed patch did not solve this issue for my setup.

I also could track down this regression to commit 477ecca (policy: Add 
logic to connect a Sink); applying it to 5.50 yields the connection 
issues. After that, I applied your patch in commit 52faac4 (audio: Fix 
cancelling disconnect timeout), but it did not change. I used only these 
two commits on top of version 5.50 to isolate the problem (it is the 
same with version 5.51 and 5.52) and gathered the logs of running 
`bluetoothd -dn` and `btmon -w` for the three setups:
* version 5.50 (patch-0),
* version 5.50 + 477ecca (patch-1), and
* version 5.50 + 477ecca + 52faac4 (patch-2).
For all three logs I powered the headset on to connect it, waited a few 
moments, and then powered off to disconnect. You can find them at 
https://gist.github.com/6d733ba9ac6cf04c71625a08e8365178.

@Luiz Augusto von Dentz: Sorry if I sent you this mail multiple times, 
it probably got filtered from the mailing list as I attached the logs in 
previous attempts.

Please let me know if you need anything else. Thanks.



[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 5204 bytes --]

Re: Bluez does not connect to A2DP sink

[Luiz Augusto von Dentz]

Hi Thomas,

On Tue, Jan 7, 2020 at 12:51 AM Thomas Ruprecht
<thomas.ruprecht@tu-dresden.de> wrote:
>
> Hi,
>
> I got a problem with bluez since 5.51 that was recently discussed in the
> thread https://marc.info/?l=linux-bluetooth&m=157081619423081&w=2: My
> headset auto-connects but is not registered as a2dp sink. Unfortunately,
> the proposed patch did not solve this issue for my setup.
>
> I also could track down this regression to commit 477ecca (policy: Add
> logic to connect a Sink); applying it to 5.50 yields the connection
> issues. After that, I applied your patch in commit 52faac4 (audio: Fix
> cancelling disconnect timeout), but it did not change. I used only these
> two commits on top of version 5.50 to isolate the problem (it is the
> same with version 5.51 and 5.52) and gathered the logs of running
> `bluetoothd -dn` and `btmon -w` for the three setups:
> * version 5.50 (patch-0),
> * version 5.50 + 477ecca (patch-1), and
> * version 5.50 + 477ecca + 52faac4 (patch-2).
> For all three logs I powered the headset on to connect it, waited a few
> moments, and then powered off to disconnect. You can find them at
> https://gist.github.com/6d733ba9ac6cf04c71625a08e8365178.

From the logs it looks like it does connects but it aborts after the
suspend, so it might not be related to reconnection policy but
something else that you device do not like, perhaps the suspend being
initiated or something like it? What headset model is this btw?

> @Luiz Augusto von Dentz: Sorry if I sent you this mail multiple times,
> it probably got filtered from the mailing list as I attached the logs in
> previous attempts.
>
> Please let me know if you need anything else. Thanks.
>
>


-- 
Luiz Augusto von Dentz

Re: Bluez does not connect to A2DP sink

[Thomas Ruprecht]

[-- Attachment #1: Type: text/plain, Size: 925 bytes --]

Hey,

On 1/7/20 10:51 PM, Luiz Augusto von Dentz wrote:
>  From the logs it looks like it does connects but it aborts after the
> suspend, so it might not be related to reconnection policy but
> something else that you device do not like, perhaps the suspend being
> initiated or something like it? What headset model is this btw?

sorry, I am not familiar with the bluetooth stack, so I cannot make much 
of these logs. The most obvious things I noticed when I took a look into 
the diffs between 5.50 and the later versions are:
* in version 5.50 there are no logged calls to 
`plugins/policy.c:policy_connect()`, just `plugins/policy.c:service_cb()`
* only in version 5.50 there's a logged call to 
`profiles/audio/a2dp.c:confirm_cb()`.
Tomorrow I'll try adding some debug messages into the code that was 
added by the patch and see if I can make something of that.

The headset model is JBL E65BTNC.


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 5204 bytes --]