Re: [PATCH 3/3] watchdog: xilinx: Add WDIOC_SETTIMEOUT ioctl function

From: Guenter Roeck <linux@roeck-us.net>
To: Michal Simek <monstr@monstr.eu>
Cc: Michal Simek <michal.simek@xilinx.com>,
	linux-kernel@vger.kernel.org, Wim Van Sebroeck <wim@iguana.be>,
	linux-watchdog@vger.kernel.org
Subject: Re: [PATCH 3/3] watchdog: xilinx: Add WDIOC_SETTIMEOUT ioctl function
Date: Thu, 30 May 2013 08:25:45 -0700	[thread overview]
Message-ID: <20130530152545.GA29874@roeck-us.net> (raw)
In-Reply-To: <51A76C58.5060107@monstr.eu>

On Thu, May 30, 2013 at 05:12:24PM +0200, Michal Simek wrote:
> On 05/30/2013 05:03 PM, Guenter Roeck wrote:
> > On Thu, May 30, 2013 at 04:34:02PM +0200, Michal Simek wrote:
> >> On 05/30/2013 04:21 PM, Guenter Roeck wrote:
> >>> On Thu, May 30, 2013 at 04:15:45PM +0200, Michal Simek wrote:
> >>>> On 05/30/2013 04:07 PM, Guenter Roeck wrote:
> >>>>> On Thu, May 30, 2013 at 02:26:04PM +0200, Michal Simek wrote:
> >>>>>> Standard watchdog programs try to setup timeout
> >>>>>> via ioctl and this functionality should be implemented.
> >>>>>> Timeout value is hardcoded in the hardware but
> >>>>>> based on Documentation/watchdog/watchdog-api.txt
> >>>>>> can return the real timeout used in the same variable.
> >>>>>>
> >>>>>> Signed-off-by: Michal Simek <michal.simek@xilinx.com>
> >>>>>> ---
> >>>>>>  drivers/watchdog/of_xilinx_wdt.c | 1 +
> >>>>>>  1 file changed, 1 insertion(+)
> >>>>>>
> >>>>>> diff --git a/drivers/watchdog/of_xilinx_wdt.c b/drivers/watchdog/of_xilinx_wdt.c
> >>>>>> index 79f358c..a3bbe72 100644
> >>>>>> --- a/drivers/watchdog/of_xilinx_wdt.c
> >>>>>> +++ b/drivers/watchdog/of_xilinx_wdt.c
> >>>>>> @@ -253,6 +253,7 @@ static long xwdt_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
> >>>>>>  		xwdt_keepalive();
> >>>>>>  		return 0;
> >>>>>>
> >>>>>> +	case WDIOC_SETTIMEOUT:
> >>>>>>  	case WDIOC_GETTIMEOUT:
> >>>>>>  		if (no_timeout)
> >>>>>>  			return -ENOTTY;
> >>>>>
> >>>>> Watchdog programs should check ident.options before trying to set the timeout.
> >>>>> If they don't, there is an application bug. I don't think it is a good idea
> >>>>> to start hacking the kernel to work around application bugs.
> >>>>
> >>>> Based on Documentation/watchdog/watchdog-api.txt
> >>>>
> >>>> "For some drivers it is possible to modify the watchdog timeout on the
> >>>> fly with the SETTIMEOUT ioctl, those drivers have the WDIOF_SETTIMEOUT
> >>>> flag set in their option field.  The argument is an integer
> >>>
> >>> Yes, but WDIOF_SETTIMEOUT is not set in the driver's option field.
> >>
> >> ok. It means I should probably enable it.
> >>
> > I am missing your point. Applications should not try to write the timeout
> > since WDIOF_SETTIMEOUT is not set. Any application doing it anyway is buggy
> > and should be fixed.
> 
> I fully understand your points and 100% agree with you
> 1. the application is broken and should be fixed
> 2. also the kernel shouldn't fix any problem with stupid application
> 
> But based on documentation the driver can support setup timeout
> and based on description "the driver returns the real timeout used
> in the same variable and this timeout might differ from the requested one
> due to limitation of the hardware"
> Based on this I still think that the driver can support set timeout
> feature and if the driver supports this option then WDIOF_SETTIMEOUT
> should be set in driver's option field. And I would add this to v2.
> 
> Can you see my point now?

No. The driver doesn't support setting the timeout. You just want it
to falsely claim that it does to work around an application problem.
With your logic, _every_ watchdog driver would "support" setting
the timeout.

I'll defer to a higher authority, ie Wim.

Guenter