Re: [PATCH v3 3/3] arm: omap_hwmod disable ick autoidling when a hwmod requires that

[Tony Lindgren]

* Tero Kristo <t-kristo@xxxxxx> [190121 07:13]:
> On 18/01/2019 21:45, Tony Lindgren wrote:
> > * Andreas Kemnade <andreas@xxxxxxxxxxxx> [190118 19:39]:
> > > Hi,
> > > 
> > > On Fri, 18 Jan 2019 10:36:30 -0800
> > > Tony Lindgren <tony@xxxxxxxxxxx> wrote:
> > > 
> > > [...]
> > > > til the next workaround.
> > > > 
> > > > > That flags also causes the iclk being enabled/disabled
> > > > > manually.
> > > > 
> > > > Yes but SWSUP_IDLE for the interface clock to me currently
> > > > just means:
> > > > 
> > > > "manually enable and disable ocp interface clock"
> > > > 
> > > well, if we want to manually disable it and not automatically,
> > > we have to disable autoidle or it will be automatically disabled.
> > > 
> > > Disabling it manually when it is already auto-disabled (by autoidle) is
> > > just practically a no-op towards the clock.
> > 
> > OK I buy that :) It should be probably added to the patch
> > description to make it clear what it changes.
> > 
> > Tero, any comments on this change?
> 
> Well, seeing the flag is pretty much only used for a handful of hwmods
> nowadays, it should be fine. OMAP3 PM should be tested with this change
> though, as there are couple of hwmods impacted on that platform. I wonder
> what happens to cpuidle when display is active...

OK so that's a good test case. AFAIK, we should have DSS idle
and have the SoC hit at least core retention with DSI command mode
displays. I don't know if this patch would block DSS autoidle then
or not.. I'm guessing 80% chance that we still need DSS hit runtime
suspend to enter SoC idle states meaning this patch would not
affect it :)

> > > > and with your changes it becomes:
> > > > 
> > > > "manually enable and disable ocp interface clock and block
> > > >   autoidle while in use".
> > > > 
> > > > So aren't we now changing the way things behave in general
> > > > for SWSUP_IDLE?
> > > > 
> > > Yes, we are, so proper testing is needed. But If I read those comments
> > > it was always intended this way but not fully implemented because it
> > > appeared to be more work like needing a usecounter (which my patchset
> > > also adds) for that autoidle flag.
> > 
> > OK yeah the use count seems necessary. I'll test here
> > with my PM use cases.

Regards,

Tony