Re: [PATCH v3 0/6] Add support to retrieve the bootstatus from watchdog for RZ/V2H(P) SoC

[Guenter Roeck <linux@roeck-us.net>]

On 1/13/25 04:40, Lad, Prabhakar wrote:
> Hi Biju,
> 
> On Mon, Jan 13, 2025 at 11:38 AM Biju Das <biju.das.jz@bp.renesas.com> wrote:
>>
>> Hi Prabhakar,
>>
>>> -----Original Message-----
>>> From: Prabhakar <prabhakar.csengg@gmail.com>
>>> Sent: 13 January 2025 11:24
>>> Subject: [PATCH v3 0/6] Add support to retrieve the bootstatus from watchdog for RZ/V2H(P) SoC
>>>
>>> From: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
>>>
>>> Hi All,
>>>
>>> This patch series adds SYSCON support to retrieve boot status information for RZ/V2H(P) SoC.
>>> Summary of Changes,
>>>
>>>      Clock:
>>>          Add syscon compatible support to the CPG block in bindings and
>>>          device trees.
>>>
>>>      Watchdog:
>>>          Document the renesas,r9a09g057-syscon-wdt-errorrst property.
>>>          Update the watchdog driver to fetch and report boot status via
>>>          Error Reset Registers (CPG_ERROR_RSTm).
>>>
>>>      Device Tree:
>>>          Add the syscon property to CPG and WDT nodes in R9A09G057 and
>>>          R9A09G047 SoC DTSI.
>>>
>>> These changes enable the watchdog driver to identify boot sources like Power-on Reset and Watchdog
>>> Reset, improving system diagnostics.
>>
>> This means that, we should assume U-boot/bootloader should not clear the WDT reset status bit.
>>
>> If they clear it, there should be a way to propagate it from u-boot/bootloader to linux,
>> otherwise, we get wrong bootstatus in linux.
>> But the clearing of watchdog status by one of the cases:
>>
>> 1) u-boot identify the boot source and clear the status bit
>>
> I agree, if the tf-a/u-boot clears the register, the bootstatus
> reported by Linux will be in-correct.
> 

I would rephrase that: Today, the boot status is _always_ incorrect if a reboot
was triggered by a watchdog reset. After this patch, it will be correct unless
the boot loader resets the status bit.

> Alternative solution, would be:
> 
> Let the TF-A create WDT nodes for us and add a property
> `renesas,r9a09g057-wdt-bootstatus` and propagate the WDT nodes to
> Linux.
> 
> renesas,r9a09g057-wdt-bootstatus = <0/1>;
> 

That would require both a property change as well as a change in the bootloader.
I suspect it would also be inappropriate as standard property and would have
to be passed as 'chosen' property (if that is even permitted) or as command line
argument.

Either case, if the boot loader has to be changed anyway, why not just tell it
to leave the boot status alone ?

Guenter