Re: [PATCH RFC] watchdog: add a new driver for VIA chipsets

[Marc Vertes <marc.vertes@sigfox.com>]

Wolfram Sang <w.sang@pengutronix.de> wrote:

> On Tue, Nov 22, 2011 at 05:32:28PM +0000, Mark Brown wrote:
> > On Tue, Nov 22, 2011 at 06:05:48PM +0100, Marc Vertes wrote:
> > 
> > > +static int wdt_start(struct watchdog_device *wdev)
> > > +{
> > > +	/* Nothing to do. The watchdog can only be started by the BIOS. */
> > > +	return 0;
> > > +}
> > 
> > Shouldn't we just update the framework to cope with missing functions?
>
> For start(), I dunno. I don't have enough experience with watchdog drivers to
> judge if this is acceptable, although I'd think so. For stop(), a few drivers
> already activate a timer to keep the watchdog happy; such a mechanism should
> definately go into the core. It's in my todo-list, but I won't be angry if
> someone is faster ;)

A question regarding "nowayout", in the case of this VIA chipset where
a watchdog can not be stopped once started from BIOS. I think I should
set unconditionally WDOG_NO_WAY_OUT in status, shouldn't I ?

I also do not think it is correct to implement a timer and ping
proactively in case of magicclose because 1) it complicates a bit the
code, 2) there is no way to know the timeout value, and it set very low
in the BIOS (1 s).

Regards,
--
Marc