On Tue, Apr 4, 2017 at 7:51 PM, Guenter Roeck <[email protected]> wrote:
> On 04/04/2017 08:47 AM, Andy Shevchenko wrote:
>> On Tue, Apr 4, 2017 at 5:10 PM, Andrey Smirnov <[email protected]>
>> wrote:

>>> +               ret = devm_register_reboot_notifier(wdd->parent,
>>> +                                                   &wdd->reboot_nb);
>>
>>
>> I'm not sure it's logically correct. Perhaps binding to the actual
>> watchdog device would be better.
>> Though, Guenter can correct me.
>>
>
> As mentioned in my other reply, it doesn't work since wdd->parent can be
> NULL.
> It might be possible to move the code into watchdog_dev_register() and tie
> it
> to the watchdog device, but I have not explored what that would require.

I would rather add device pointer to watchdoig struct, though it's your call :-)
So, what you decide I'll agree with.

-- 
With Best Regards,
Andy Shevchenko

Reply via email to