Re: [PATCH v4 1/2] watchdog: New watchdog driver for MEN A21 watchdogs

[Guenter Roeck <linux@roeck-us.net>]

On Fri, May 31, 2013 at 03:05:50PM +0530, anish singh wrote:
> Hi Guenter,
> 
> On Fri, May 31, 2013 at 3:32 AM, Guenter Roeck <linux@roeck-us.net> wrote:
> > On Thu, May 30, 2013 at 11:59:28PM +0200, Wim Van Sebroeck wrote:
> >> Hi Guenter,
> >>
> >> > On Tue, May 28, 2013 at 10:10:53AM +0200, Johannes Thumshirn wrote:
> >> > > On Mon, May 27, 2013 at 08:25:54PM +0200, Wim Van Sebroeck wrote:
> >> > > [...]
> >> > > > > +     watchdog_set_drvdata(&a21_wdt, drv);
> >> > > >
> >> > > > I am missing the initialisation of the watchdog's timeout value here...
> >> > >
> >> > > This watchdog only knows two timeout values, 1s and 30s with the constraint
> >> > > that you can't go back to 30s once your in a 1s timeout without a reset of the
> >> > > CPLD. I could initially set it to 30s but that would be redundant.
> >> > >
> >> > I wonder - why bother with supporting one-second timeouts ?
> >> >
> >> > Is this realistic, ie can you guarantee that the watchdog will be pinged fast
> >> > enough to keep the system alive under all load conditions ? As far as I know
> >> > you can not even configure the watchdog application for less than 1 second
> >> > ping intervals.
> >>
> >> that's why certain drivers use a timer to keep kicking the watchdog at
> >> a rate of 0.5s (do a grep on HZ/2) untill userspace times out.
> >> (example: drivers/watchdog/pika_wdt.c or drivers/watchdog/pcwd.c).
> >>
> > Yes, I have used the same trick in a couple of my drivers. That isn't done here,
> > though.
> >
> > I even thought about adding this capability to the infrastructure.
> 
> If you don't mind can I give it a try?

I don't mind, but you might want to wait for Wim's feedback.

Thanks,
Guenter