[alsa-devel] How to use implicit feedback with full duplex?

[Eldad Zack]

On Tue, 26 Mar 2013, Daniel Mack wrote:

> On 24.03.2013 23:49, Eldad Zack wrote:
> > 
> > On Wed, 6 Feb 2013, Clemens Ladisch wrote:
> > 
> >>> Can you try using jackd -d alsa -d hw:x with the device and see if that
> >>> works for you?
> >>
> >> That works.  This means that there is a race condition in the driver, or
> >> that the different open/hw_params/prepare order trips it up.
> > 
> > Yes, and also on closing (close/hw_free).
> > I finally had some time to look into this - patch below adds some 
> > checks and with it:
> >     
> > * Starting playback, waiting, starting capture: capture doesn't start,
> >   playback continues without breaking.
> >     
> > * Starting capture, waiting, starting playback: playback doesn't start,
> >   capture breaks - but it is possible to restart the streams afterwards.
> >   [On my setup (with xhci), when the streams break I must restart my box
> >   to get them to work again with current code].
> > 
> > * Starting both - no change, works good.
> > 
> > This is the order the ops get called when both are start at once, if it 
> > helps anyone:
> > 
> > playback	capture
> > open
> > 		open
> > 		hw_params
> > 		 set_format
> >  		prepare
> >  		 set_format
> > hw_params
> >  set_format
> > prepare
> >  set_format
> > prepare
> >  set_format
> > 		prepare
> > 		 set_format
> >  trigger
> > 		trigger
> > 
> > I'll try to figure out why capture breaks next.
> > I still quite slow around the code and don't understand some parts of 
> > it so this might take me a while.
> 
> Thanks for looking into this!
> 
> > @Daniel, do you have any hints for this case (capture breaking
> > when starting playback)?
> 
> Give me some time to catch up and rebuild my FTU setup please. I hope
> I'll be able to get back to you by the end of this week.

Cool, thank you!

And it just remembered something. I tried to make the "in use"
check non-fatal (return 0) and with a couple of more changes
(very hacky) I got it to start in both orders.

But the problem was that when I stopped the playback* stream my
kernel "exploded" (hard lock up). But if I stopped the capture,
it was ok. Playback didn't even stop.
(It might have been the other way around, but I think it was 
playback)

I can find that branch when I find some time if you think it
might help.

Cheers,
Eldad