Re: [RFC] [PATCH 6/10] Generic Watchdog Timer Driver

[Wim Van Sebroeck <wim@iguana.be>]

Hi Jamie,

> I don't fully understand why the module refcounting is done in the open 
> and release though.  If you moved it into the registration and 
> unregistration then doesn't that remove the need for WDOG_ORPHAN?

Hmm, nice suggestion. Will look into that.

> > diff --git a/drivers/watchdog/core/watchdog_core.c b/drivers/watchdog/core/watchdog_core.c
> > index 52bc520..d1a824e 100644
> > --- a/drivers/watchdog/core/watchdog_core.c
> > +++ b/drivers/watchdog/core/watchdog_core.c
> > @@ -60,6 +60,10 @@ int register_watchdogdevice(struct watchdog_device *wdd)
> >  	if (wdd == NULL || wdd->info == NULL || wdd->ops == NULL)
> >  		return -ENODATA;
> >  
> > +	/* Make sure that the owner of the watchdog operations exists */
> > +	if (wdd->ops->owner == NULL)
> > +		return -ENODATA;
> 
> Won't this be effectively NULL if the module is builtin?  It looks like 
> if it is builtin then THIS_MODULE would be defined as (struct module 
> *)0.

Same for this: I will look into this.

Thanks,
Wim.