[pulseaudio-discuss] Blueman "failed to initialize pulseaudio bluetooth module"

[Colin Beckingham]

On 10/16/2009 01:31 PM, Colin Beckingham wrote:
> On 16/10/09 10:20 AM, Colin Beckingham wrote:
>> On 16/10/09 09:33 AM, Lennart Poettering wrote:
>>> On Fri, 16.10.09 08:51, Colin Beckingham (colbec at start.ca) wrote:
>>>
>>>> Using latest svn 602 of blueman on Opensuse 11.1
>>>> 0.9.19 of pulseaudio
>>>> Headset Jabra BT2040
>>>> with pulseaudio -vvvv
>>>
>>> Please use gnome-bluetooth. That's the only interface the bluez folks
>>> and I can recommend.
>>>
>>>> D: module-bluetooth-device.c: Received BT_INDICATION<- BT_NEW_STREAM
>>>> D: module-bluetooth-device.c: Stream properly set up, we're ready to
>>>> roll!
>>>> I: module-bluetooth-device.c: FD error: POLLERR
>>>
>>> POLLERR is signalled via the stream socket from the kernel. Running
>>> "bluetoothd -nd" gives you more debug output and might tell you what
>>> triggered this.
>>>
>>> But generally, given that this error is actually "behind" the
>>> bluetooth socket, I'd ping the bluez folks about this.
>>>
>>
>> OK thanks I will do that. I will be sharing this output from bluetoothd
>> with them, FYI:
>>
>>
>> bluetoothd[26861]: Audio API: BT_REQUEST <- BT_OPEN
>> bluetoothd[26861]: open sco -
>> object=/org/bluez/26861/hci0/dev_00_1A_45_32_E7_16 source=ANY
>> destination=ANY lock=readwrite
>> bluetoothd[26861]: Unable to open seid 64
>> bluetoothd[26861]: sending error Invalid argument(22)
>> bluetoothd[26861]: Audio API: BT_ERROR -> BT_OPEN
>> bluetoothd[26861]: Unix client disconnected (fd=31)
>>
>> Colin
>
> Noted that there is an upgrade to bluez, 4.56 (I was using 4.54) so
> compiled new version. Above "seid 64" error has gone, but pulseaudio
> still fails with POLLERR.
>
> Oct 16 13:14:37 : Audio API: BT_REQUEST <- BT_START_STREAM
> Oct 16 13:14:37 : State changed
> /org/bluez/11455/hci0/dev_00_1A_45_32_E7_16: HEADSET_STATE_CONNECTED ->
> HEADSET_STATE_PLAY_IN_PROGRESS
>
> Oct 16 13:14:37 : module-bluetooth-device.c: Trying to receive message
> from audio service...
> Oct 16 13:14:38 : Received AT+BTRH?
> Oct 16 13:14:39 : Received AT+VGS=15
> Oct 16 13:14:39 : Received AT+VGM=15
> Oct 16 13:15:17 : HUP or ERR on socket
> ...
> Oct 16 13:15:17 : module-bluetooth-device.c: Received BT_INDICATION <-
> BT_NEW_STREAM
> Oct 16 13:15:17 : module-bluetooth-device.c: Stream properly set up,
> we're ready to roll!
> Oct 16 13:15:17 : module-bluetooth-device.c: FD error: POLLERR
> Oct 16 13:15:17 : module-bluetooth-device.c: IO thread failed
> Oct 16 13:15:17 : module-suspend-on-idle.c: Sink
> bluez_sink.00_1A_45_32_E7_16 becomes idle, timeout in 5 seconds.

I did ping the bluez folks on this, but no takers.

Updating for others who may trip on this problem: I am now using 
Opensuse 11.2 with tools as supplied in the standard package. PA 0.9.19, 
Bluez 4.56, KBluetooth4 (I have also used gnome-bluetooth and have 
exactly the same symptoms, issues and reliability) and supplied graphic 
tools. Kernel 2.6.31.5-0.1-default.

This problem is difficult because it is intermittent. Sometimes there is 
no problem, other times it fails. Success happens about one time in 50 
(fifty) so it is rare for PA to automatically set up the sink and 
source. However I cannot claim that it _never_ does. It appears to be a 
timing problem, or my hardware is not communicating easily. I have two 
headsets, a Jabra BT2040 and a VBT185Z. Both headsets have triggered the 
PA sink and source properly at one time or another and both seem to have 
an equal chance of becoming fully activated even if that is small. By 
fully activated I mean that the PA bluetooth sink and source are 
activated and

$ arecord -f S16_LE -D pulse -r 48000 | aplay -D pulse

works perfectly.

I now have a copy of both the system log when it fails and a copy when 
it succeeds. If these are of any interest I can send them along.

Colin