[U-Boot] [PATCH 2/2] watchdog: designware: Convert to DM and DT probing

[Marek Vasut]

On 10/3/19 8:11 PM, Simon Goldschmidt wrote:
> Marek Vasut schrieb am Do., 3. Okt. 2019, 20:09:
> 
>> On 10/3/19 8:06 PM, Simon Goldschmidt wrote:
>>> Marek Vasut schrieb am Do., 3. Okt. 2019, 09:22:
>>>
>>>> On 10/3/19 3:57 AM, Ley Foon Tan wrote:
>>>>
>>>> [...]
>>>>
>>>>>> diff --git a/drivers/watchdog/Kconfig b/drivers/watchdog/Kconfig
>>>>>> index 6fd9b0a177..ec34993664 100644
>>>>>> --- a/drivers/watchdog/Kconfig
>>>>>> +++ b/drivers/watchdog/Kconfig
>>>>>> @@ -36,13 +36,6 @@ config ULP_WATCHDOG
>>>>>>         help
>>>>>>           Say Y here to enable i.MX7ULP watchdog driver.
>>>>>>
>>>>>> -config DESIGNWARE_WATCHDOG
>>>>>> -       bool "Designware watchdog timer support"
>>>>>> -       select HW_WATCHDOG
>>>>> CONFIG_HW_WATCHDOG is disabled now. Few areas of code in
>>>>> arm/mach-socfpga/ still using this CONFIG and call to
>>>>> hw_watchdog_init(). They need to remove too.
>>>>> Do we need call to uclass_get_device(UCLASS_WDT, 0, &dev) in SPL to
>>>>> probe watchdog and call to wdt_start() to start watchdog? Can't find
>>>>> place that start watchdog.
>>>>
>>>> You're right, and the WDT is enabled early on, before the DM structures
>>>> are available.
>>>>
>>>> I wonder whether we should do what we did on iMX -- have a non-DM WDT
>>>> driver for SPL and DM-one for U-Boot proper.
>>>>
>>>
>>> Are you aware that DM WDT is enabled in SPL by default for gen5 now? I
>> get
>>> a message that WDT is not found. I haven't sent a patch to fix that yet,
>>> since the message is the only thing that happens, works normally
>> otherwise.
>>>
>>> But if you're working on that, you might check side effects of that
>> setting
>>> (which is new for this release).
>>
>> That doesn't sound right, the SoCFPGA core code uses HW_WATCHDOG in some
>> places as Ley pointed out, so if DM WDT is enabled in SPL, that code in
>> HW_WATCHDOG ifdefs isn't used. And I think that will trigger some weird
>> problems. So in the end, enabling DM watchdog in SoCFPGA SPL right now
>> is a bug ?
>>
> 
> Hmm, it probably is. I'm not using  a watchdog right now, so I couldn't
> really tell... It'll be a week or so until I'll find the time for u-boot
> again :-(

We probably should fix that before the release though.