RE: [PATCH v2 1/1] watchdog: atmel: atmel-sama5d4-wdt: Disable watchdog on system suspend

[Ken Sloat]

> -----Original Message-----
> From: Guenter Roeck <groeck7@xxxxxxxxx> On Behalf Of Guenter Roeck
> Sent: Friday, June 14, 2019 12:46 PM
> To: Ken Sloat <KSloat@xxxxxxxxxxxxxx>
> Cc: nicolas.ferre@xxxxxxxxxxxxx; alexandre.belloni@xxxxxxxxxxx;
> ludovic.desroches@xxxxxxxxxxxxx; wim@xxxxxxxxxxxxxxxxxx; linux-arm-
> kernel@xxxxxxxxxxxxxxxxxxx; linux-watchdog@xxxxxxxxxxxxxxx; linux-
> kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH v2 1/1] watchdog: atmel: atmel-sama5d4-wdt: Disable
> watchdog on system suspend
> 
> [This is an EXTERNAL EMAIL]
> ________________________________
> 
> On Fri, Jun 14, 2019 at 12:53:22PM +0000, Ken Sloat wrote:
> > From: Ken Sloat <ksloat@xxxxxxxxxxxxxx>
> >
> > Currently, the atmel-sama5d4-wdt continues to run after system suspend.
> > Unless the system resumes within the watchdog timeout period so the
> > userspace can kick it, the system will be reset. This change disables
> > the watchdog on suspend if it is active and re-enables on resume.
> > These actions occur during the late and early phases of suspend and
> > resume respectively to minimize chances where a lock could occur while
> > the watchdog is disabled.
> >
> > Signed-off-by: Ken Sloat <ksloat@xxxxxxxxxxxxxx>
> > ---
> >  Changes in v2:
> >  -Consolidate resume and resume early statements.
> >
> >  drivers/watchdog/sama5d4_wdt.c | 21 ++++++++++++++++++---
> >  1 file changed, 18 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/watchdog/sama5d4_wdt.c
> > b/drivers/watchdog/sama5d4_wdt.c index 111695223aae..0d123f8cbcc6
> > 100644
> > --- a/drivers/watchdog/sama5d4_wdt.c
> > +++ b/drivers/watchdog/sama5d4_wdt.c
> > @@ -280,7 +280,17 @@ static const struct of_device_id
> > sama5d4_wdt_of_match[] = {  MODULE_DEVICE_TABLE(of,
> > sama5d4_wdt_of_match);
> >
> >  #ifdef CONFIG_PM_SLEEP
> > -static int sama5d4_wdt_resume(struct device *dev)
> > +static int sama5d4_wdt_suspend_late(struct device *dev) {
> > +     struct sama5d4_wdt *wdt = dev_get_drvdata(dev);
> > +
> > +     if (watchdog_active(&wdt->wdd))
> > +             sama5d4_wdt_stop(&wdt->wdd);
> > +
> > +     return 0;
> > +}
> > +
> > +static int sama5d4_wdt_resume_early(struct device *dev)
> >  {
> >       struct sama5d4_wdt *wdt = dev_get_drvdata(dev);
> >
> > @@ -291,12 +301,17 @@ static int sama5d4_wdt_resume(struct device
> *dev)
> >        */
> >       sama5d4_wdt_init(wdt);
> >
> > +     if (watchdog_active(&wdt->wdd))
> > +             sama5d4_wdt_start(&wdt->wdd);
> > +
> 
> The call to sama5d4_wdt_init() above now explicitly stops the watchdog
> even if we want to (re)start it. I think this would be better handled with an
> else case here
> 
>         else
>                 sama5d4_wdt_stop(&wdt->wdd);
> 

So we completely remove the sama5d4_wdt_init() call then correct?

To leave the code as it behaves today with the addition
of wdt stop/start, shouldn't we call init in the else instead?

	if (watchdog_active(&wdt->wdd))
		sama5d4_wdt_start(&wdt->wdd);
	else
		sama5d4_wdt_init();

I guess I don't really understand the purpose of having the init statement in resume
in the first place. I agree, calling this first does end up essentially resetting the wdt
it will start again if it was running before, but the count will be reset.

> Guenter
> 
> >       return 0;
> >  }
> >  #endif
> >
> > -static SIMPLE_DEV_PM_OPS(sama5d4_wdt_pm_ops, NULL,
> > -                      sama5d4_wdt_resume);
> > +static const struct dev_pm_ops sama5d4_wdt_pm_ops = {
> > +     SET_LATE_SYSTEM_SLEEP_PM_OPS(sama5d4_wdt_suspend_late,
> > +                     sama5d4_wdt_resume_early) };
> >
> >  static struct platform_driver sama5d4_wdt_driver = {
> >       .probe          = sama5d4_wdt_probe,
> > --
> > 2.17.1
> >

Thanks,
Ken Sloat