Re: [PATCH] at73c213: Fix DMA size at the end of DMA buffer

[Takashi Iwai <tiwai@suse.de>]

At Fri, 14 Mar 2008 22:29:32 +0900 (JST),
Atsushi Nemoto wrote:
> 
> On Fri, 14 Mar 2008 10:44:45 +0100, Haavard Skinnemoen <haavard.skinnemoen@atmel.com> wrote:
> > > +		size = runtime->period_size * runtime->channels;
> > > +		if (next_period == runtime->periods - 1)
> > > +			size += (runtime->buffer_size % runtime->period_size)
> > > +				* runtime->channels;
> > 
> > Ow. That looks expensive. Isn't there any way we can force the client
> > to select sane values of buffer_size and period_size?
> 
> Well, I suppose it is not _too_ expensive. :)
> 
> > It seems like a reasonable demand that buffer_size is a multiple of
> > period_size, doesn't it?
> 
> But actually it can happen.  And I gave up understanding how are these
> parameters determined...  If there were any way the driver can enforce
> that constraint, it would be better fix.
> 
> Iwai-san, any comments from alsa guru?

Add the following constraint in the open callback:

	err = snd_pcm_hw_constraint_integer(runtime,
					SNDRV_PCM_HW_PARAM_PERIODS);
	if (err < 0)
		return err;

This will guarantee that the period size fits with the buffer size.


Takashi