[RFC PATCH v2 0/6] ALSA: compress: add support to change codec profile in gapless playback

[Vinod Koul]

On 23-07-20, 22:33, Takashi Iwai wrote:
> On Thu, 23 Jul 2020 17:56:12 +0200,
> Vinod Koul wrote:
> > 
> > On 23-07-20, 15:17, Takashi Iwai wrote:
> > > On Thu, 23 Jul 2020 15:05:22 +0200,
> > > Vinod Koul wrote:
> > > > 
> > > > On 23-07-20, 14:38, Takashi Iwai wrote:
> > > > > On Tue, 21 Jul 2020 19:00:01 +0200,
> > > > > Srinivas Kandagatla wrote:
> > > > > > 
> > > > > > For gapless playback it is possible that each track can have different
> > > > > > codec profile with same decoder, for example we have WMA album,
> > > > > > we may have different tracks as WMA v9, WMA v10 and so on
> > > > > > Or if DSP's like QDSP have abililty to switch decoders on single stream
> > > > > > for each track, then this call could be used to set new codec parameters.
> > > > > > 
> > > > > > Existing code does not allow to change this profile while doing gapless
> > > > > > playback.
> > > > > > 
> > > > > > This patchset adds new SNDRV_COMPRESS_SET_CODEC_PARAMS IOCTL along with
> > > > > > flags in capablity structure to allow userspace to set this new
> > > > > > parameters required which switching codec profile, either for gapless
> > > > > > or cross fade usecase.
> > > > > 
> > > > > One idea that came up at the previous audio conference regarding this
> > > > > implementation was to just allow SET_PARAMS during the stream is
> > > > > running (only if the driver sets the capability) instead of
> > > > > introducing yet a new ioctl and an ops.
> > > > > Would it make sense?
> > > > 
> > > > That does sound good but only issue would be that we need to somehow
> > > > mark/document that buffer info is useless and would be discarded, how do
> > > > we do that?
> > > 
> > > Yes, the buffer and no_wake_mode can be ignored in the gapless
> > > re-setup.  Is your concern only about the documentation?  Or something
> > > else needs to be changed significantly?  It's a new scheme in anyway,
> > > so the documentation update is required...
> > 
> > My concern is potential abuse of API down the road, if you feel it is
> > okay, I am okay with the proposal
> 
> If this can be potentially dangerous, it shouldn't be used, of course.
> What kind of scenario could it be?

I can think of users trying to invoke this incorrectly, right now we
would reject this.

Maybe, we can add checks like, if next_track is set and then copy the
codec params only to prevent any misuse.

Do you think that would be okay?

-- 
~Vinod