Cyril Chemparathy <[email protected]> writes:

> Sekhar,
>
>> > +struct platform_device davinci_wdt_device = {
>> > +  .name           = "watchdog",
>> > +  .id             = -1,
>> > +  .num_resources  = ARRAY_SIZE(wdt_resources),
>> > +  .resource       = wdt_resources,
>> > +};
>>
>> IMO, a better way to overcome the 'no watchdog timer'
>> limitation would be to make the watchdog reset code
>> not use the wdt platform data directly, but instead
>> 'search' for watchdog device using bus_for_each_dev()
>> iterator on platform bus.
>>
>> In your case, the watchdog device wouldn't be found
>> and the reset function should exist gracefully.
>
> I realize that I agreed to your viewpoint earlier, but I see a problem
> with this approach.  If an arch_reset() were to somehow get triggered
> in early boot (e.g. via panic()), the watchdog device may not have
> been registered yet.
>
> Any thoughts on how to approach this situation?

Then just fail with an 'unable to reset' message due to missing wdt.

Kevin
_______________________________________________
Davinci-linux-open-source mailing list
[email protected]
http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source

Reply via email to