Date: Tue, 14 Jan 2014 17:44:27 +0800
From: Dave Young <dyoung@redhat.com>
To: Wim Van Sebroeck <wim@iguana.be>
Cc: dzickus@redhat.com, bhe@redhat.com, vgoyal@redhat.com,
        linux-watchdog@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] watchdog: add a parameter for stop wdt before register
Message-ID: <20140114094427.GG3969@dhcp-16-126.nay.redhat.com>
References: <20140114082323.GA4485@dhcp-16-126.nay.redhat.com>
 <20140114082632.GA28648@spo001.leaseweb.com>
 <20140114084154.GC3969@dhcp-16-126.nay.redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20140114084154.GC3969@dhcp-16-126.nay.redhat.com>
User-Agent: Mutt/1.5.21 (2010-09-15)
Sender: linux-kernel-owner@vger.kernel.org

On 01/14/14 at 04:41pm, Dave Young wrote:
> On 01/14/14 at 09:26am, Wim Van Sebroeck wrote:
> > Hi Dave,
> > 
> > > In kdump kernel watchdog could interrupt vmcore capturing because we
> > > have no way to disable/stop it while crashing happens.
> > > 
> > > Add a module parameter stop_before_register so watchdog can be stopped
> > > before register in driver loading path. Thus we can try to load the
> > > watchdog driver as early as possible in kdump kernel to ensure vmcore
> > > capturing.
> > > 
> > > Don Zickus mentioned that there's the case that bios start the watchdog
> > > and it is expected that the kernel keep the watchdog alive. To address
> > > this case I added the module parameter which is false by default so
> > > it will stop the watchdog only when user provice kernel cmdline
> > > "watchdog.stop_before_register=1".  
> > > 
> > > Signed-off-by: Dave Young <dyoung@redhat.com>
> > > ---
> > >  drivers/watchdog/watchdog_core.c |    6 ++++++
> > >  1 file changed, 6 insertions(+)
> > > 
> > > --- linux.orig/drivers/watchdog/watchdog_core.c
> > > +++ linux/drivers/watchdog/watchdog_core.c
> > > @@ -42,6 +42,7 @@
> > >  
> > >  static DEFINE_IDA(watchdog_ida);
> > >  static struct class *watchdog_class;
> > > +static bool stop_before_register;
> > >  
> > >  static void watchdog_check_min_max_timeout(struct watchdog_device *wdd)
> > >  {
> > > @@ -119,6 +120,9 @@ int watchdog_register_device(struct watc
> > >  	if (wdd->ops->start == NULL || wdd->ops->stop == NULL)
> > >  		return -EINVAL;
> > >  
> > > +	if (stop_before_register)
> > > +		wdd->ops->stop(wdd);
> > > +
> > >  	watchdog_check_min_max_timeout(wdd);
> > >  
> > >  	/*
> > > @@ -220,6 +224,8 @@ static void __exit watchdog_exit(void)
> > >  subsys_initcall(watchdog_init);
> > >  module_exit(watchdog_exit);
> > >  
> > > +module_param(stop_before_register, bool, 0644);
> > > +
> > >  MODULE_AUTHOR("Alan Cox <alan@lxorguk.ukuu.org.uk>");
> > >  MODULE_AUTHOR("Wim Van Sebroeck <wim@iguana.be>");
> > >  MODULE_DESCRIPTION("WatchDog Timer Driver Core");
> > 
> > Hmm, need to look closer to this, but my first thought is:
> > what about devices that cannot be stopped once started...
> > They should be able to override this module_parameter...
> 
> Hi, Wim
> 
> Thanks for quick feedback!
> 
> I'm not sure the meaning of "cannot be stopped", if it means that
> the policy that it should not be stopped, I think since the watchdog_core
> is always built-in so the param can only be provided via boot cmdline it would
> be fine?

Hmm, the wdt driver can be removed then insmod again.
I think you are talking about the nowayout, for this case probably should add below:

if (stop_before_register && !nowayout)
  stop it.

But is there a general way for checkout nowayout, could you give some
hints?

> 
> For device which really *cannot* stop, the stop() will fail silently?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/