[PATCH v4 00/17] ALSA: hda: cirrus: Add initial DSP support and firmware loading

[Takashi Iwai]

On Mon, 30 May 2022 11:36:39 +0200,
Charles Keepax wrote:
> 
> On Mon, May 30, 2022 at 11:18:43AM +0200, Takashi Iwai wrote:
> > On Mon, 30 May 2022 11:08:46 +0200,
> > Charles Keepax wrote:
> > > 
> > > On Fri, May 27, 2022 at 06:13:38PM +0200, Takashi Iwai wrote:
> > > > On Wed, 25 May 2022 15:16:21 +0200,
> > > > Vitaly Rodionov wrote:
> > > > The idea to add / delete controls by the control element change
> > > > doesn't sound good; as already mentioned in my reply to the previous
> > > > patch set, the change of the control elements can be triggered too
> > > > easily by any normal users who have the access to the sound devices.
> > > > It means thousands of additions and removals per second could be
> > > > attacked by any user.
> > > > 
> > > 
> > > This I am a little less sure how we handle. I mean arn't there
> > > already a few ways to do this? Both the existing ASoC wm_adsp
> > > stuff, and the topology stuff (used on all new Intel platforms,
> > > so very widely deployed) let you create controls by loading a
> > > firmware file. Also within ALSA itself can't user-space create
> > > user ALSA controls? Is there some rate limiting on that? How is
> > > this issue tackled there?
> > 
> > The creation of kctls via firmware loading would be OK, as the code
> > path can't be triggered so frequently.  Is it the case for this patch
> > set?  There was too little information about the implementation (and
> > more importantly, how to use the controls), so it's hard to judge...
> > 
> 
> Yeah that should be what is happening here. Although it looks
> like this code might be removing all the controls if the firmware
> is unloaded. I will discuss that with the guys, we normal just
> disable the controls on the wm_adsp stuff.

OK, that sounds good.  Basically my concern came up from the code
snippet doing asynchronous addition/removal via work.  This showed
some yellow signal, as such a pattern doesn't appear in the normal
implementation.  If this is (still) really necessary, it has to be
clarified as an exception.


thanks,

Takashi