On Fri, Aug 02, 2019 at 03:51:09PM +0100, Thomas Preston wrote:
> On 02/08/2019 12:10, Mark Brown wrote:

> > You can use a read only control for the readback, or just have it be
> > triggered by overwriting the readback value.  You can cache the result.

> Keeping the trigger and result together like that would be better I think,
> although the routine isn't supposed to run mid way through playback. If
> we're mid playback the debugfs routine has to turn off AMP_ON, take the
> device back to a known state, run diagnostics, then restore. Which causes
> a gap in the audible sound.

Whatever method is used to do the triggering can always return -EBUSY
when you someone tries to do so during playback.

> >> Kirill Marinushkin mentioned this in the first review [0], it just didn't
> >> really sink in until now!

> > You could do that too, yeah.  Depends on what this is diagnosing and if
> > that'd be useful.

> The diagnostic status bits describe situations such as:
> - open load (no speaker connected)
> - short to GND
> - short to VCC
> - etc

> The intention is to test if all the speakers are connected. So, one might 
> have a self test which runs the diagnostic and verifies it outputs:

...

> I think the module parameter method is more appropriate for a
> "Turn-on diagnostic", even though I don't really like grepping dmesg
> for the result. I'll go ahead and implement that unless anyone has a
> particular preference for the kcontrol-trigger.

Right.  It's not ideal for use in production systems for example but
perhaps fine for support techs or whoever.  Up to you anyway.