[pulseaudio-discuss] [PATCH v2] bluetooth: Add support for automatic switch between hsp and a2dp profiles

[Tanu Kaskinen]

On Fri, 2014-10-31 at 13:07 +0100, Pali Rohár wrote:
> On Friday 31 October 2014 12:49:08 Tanu Kaskinen wrote:
> > On Tue, 2014-10-28 at 12:33 +0100, Pali Rohár wrote:
> > > On Tuesday 28 October 2014 09:44:53 Tanu Kaskinen wrote:
> > > > On Sun, 2014-10-19 at 13:42 +0200, Pali Rohár wrote:
> > > > > I would like to remind this patch.
> > > > > 
> > > > > This patch for bluetooth headsets will automatically
> > > > > switch between hsp (phone quality input+output) and
> > > > > a2dp (full quality output) bluetooth profiles based on
> > > > > running pulseaudio clients.
> > > > > 
> > > > > I you want to use microphone of your bluetooth headset
> > > > > you need to switch to hsp profile (manually or somehow
> > > > > else... or with this patch).
> > > > > 
> > > > > This patch will primary use information from
> > > > > "media.role" and will switch only if "media.role" is
> > > > > phone.
> > > > > 
> > > > > If "media.role" information is not available (e.g
> > > > > pulseaudio client did not set it) then in patch is used
> > > > > some heuristic (ignore resample method peaks, ignore
> > > > > virtual streams, ignore recording from monitors).
> > > > > 
> > > > > In previous discussion Arun was against last part --
> > > > > using that heuristic and wanted to use only media.role.
> > > > > 
> > > > > I would like to hear what other developers thinks. There
> > > > > are more pulseaudio users who want to see this feature
> > > > > in pulseaudio and it is *bad* that user must open
> > > > > volume control settings, change bluetooth configuration
> > > > > (switch from a2dp to hsp) before he wants to make voice
> > > > > call or use blueooth microphone for other purpose.
> > > > 
> > > > I'd prefer if the patch would switch the profile to hsp
> > > > only if a stream has the "phone" role, and only if the
> > > > application hasn't explicitly configured any other device
> > > > for the stream.
> > > 
> > > Ok. And what to do when application does not specify any
> > > role?
> > 
> > Nothing.
> > 
> 
> Ok, what about adding new bool parameter to this module (which 
> will be off by default) to use also heuristic when application 
> does not specify media.role?

Do you have a real use case (not something that you are able to imagine,
but something that either you or someone else really needs)? I'm
hesitant to accept any more heuristics than really necessary, because
this is not how I think routing policy "should" be implemented, but the
way I think routing policy "should" be implemented takes so much work
that it doesn't make sense to ask you to do it.

So yes, adding another parameter to module-bluetooth-policy should be
fine if there's real need for it. In a separate patch, please, because
I'd like to merge the by-media-role-only heuristic first.

> > > > Btw, does this patch actually do what it's supposed to do?
> > > > It seems to switch the profile in the source output put
> > > > hook, but at that point the stream has already been
> > > > routed, so I'd expect the stream to not get routed to the
> > > > hsp source, since it didn't exist at the time when the
> > > > routing for the stream was decided.
> > > 
> > > Patch automatically switch between hsp and a2dp mode in
> > > source output hook function. Because switching procedure
> > > takes some time (and it is asynchronous) it is not finished
> > > before returning from hook function.
> > 
> > As far as I know, switching card profiles is synchronous.
> > 
> > > Once profile switch is done new source card output
> > > appears and can be used for recording audio. When I
> > > configured hsp source with high priority in KDE4 multimedia
> > > settings PA automatically switch to hsp bluetooth card when
> > > it is available for all exiting source streams.
> > 
> > Ok, so this patch relies on some other policy module to
> > automatically move streams when a new sink/source appears. I
> > think that's fine for now.
> 
> Ok. So only problematic part of my patch is using heuristic when 
> application does not provide media.role information?

I haven't done thorough review, so there may be smaller issues, but
overall the logic seems ok to me.

-- 
Tanu