missing sound on kernel-5.15

[Takashi Iwai]

On Mon, 29 Aug 2022 09:56:42 +0200,
chihhao chen wrote:
> 
> Hi Takashi,
> 
> Because of this problem, is it okay to remove configure_endpoints
> function call in snd_usb_hw_params?
>
> Then configure_endpoints would be called in snd_usb_pcm_prepare and USB
> control transfers become the same as which in kernel-5.10. There will
> be only one SAMPLING_FREQ_CONTROL request and no sound-missing
> phenomenon.

Are you sure that removing the configure_endpoints() fixes the
problem?


Takashi

> 
> Thanks
> 
> 
> On Mon, 2022-08-22 at 08:57 -0300, Geraldo Nascimento wrote:
> > On Mon, Aug 22, 2022 at 04:06:58PM +0800, chihhao chen wrote:
> > > I am testing many headsets on Android platform with different
> > > kernel
> > > versions.
> > > 
> > > On kernel-5.10, calling sequence for playback is
> > > 1. snd_usb_hw_params -> set highest sampling rate, no
> > > SAMPLING_FREQ_CONTROL USB request sent in this stage
> > > 2. snd_pcm_release
> > > 3. snd_usb_hw_params -> set proper sampling rate, still no
> > > SAMPLING_FREQ_CONTROL USB request 
> > > 4. snd_usb_pcm_parepare -> in configure_endpoint
> > > SAMPLING_FREQ_CONTROL
> > > USB request was sent on USB bus
> > > 
> > > With the same calling sequence, a slightly behavior change on
> > > kernel-
> > > 5.15 is
> > > 1. snd_usb_hw_params -> set highest sampling rate,
> > > SAMPLING_FREQ_CONTROL USB request  was sent on USB bus
> > > 2. snd_pcm_release
> > > 3. snd_usb_hw_params -> set proper sampling rate,
> > > SAMPLING_FREQ_CONTROL
> > > USB request  was sent too
> > > (because configure_endpoint was moved to snd_usb_hw_params
> > > function)
> > > 4. snd_usb_pcm_parepare -> no SAMPLING_FREQ_CONTROL USB
> > > request  because of USB EP flag check
> > > 
> > > I checked all USB packets and confirmed audio data output was
> > > correct.
> > > But there may be missing sound problem for the first sound with
> > > many
> > > headsets (for example Samsung headset).
> > > I found this issue is related to two-times sampling rate set
> > > request.
> > > (I tried to forcibly skip first USB request, everything became
> > > okay.)
> > > So do you guys know why configure_endpoint was moved to
> > > snd_usb_hw_params?
> > 
> > Hi Chihhao,
> > 
> > the commit you're looking for is
> > bf6313a0ff766925462e97b4e733d5952de02367
> > ("ALSA: usb-audio: Refactor endpoint management")
> > 
> > Good luck,
> > Geraldo Nascimento
> > 
> > > Or could you provide patch for this problem?
> > > 
> > > Many thanks
> > > Chihhao
> > > 
>