Re: [PATCH 14/14] PCI: j721e: add suspend and resume support

[Thomas Richard <thomas.richard@bootlin.com>]

On 1/22/24 22:36, Andy Shevchenko wrote:
> On Mon, Jan 22, 2024 at 5:30 PM Thomas Richard
> <thomas.richard@bootlin.com> wrote:
>> On 1/15/24 21:13, Andy Shevchenko wrote:
>>> On Mon, Jan 15, 2024 at 6:16 PM Thomas Richard
>>> <thomas.richard@bootlin.com> wrote:
> 
> ...
> 
>>>> +               if (pcie->reset_gpio)
>>>
>>> Dup, why?
>>
>> This pcie->reset_gpio corresponds to PERST# of PCIe endpoints.
>> I assert it during suspend, because I have to deassert it (with a delay)
>> during resume stage [1].
> 
> Ah, sorry for being unclear, I meant that gpiod_set_value*() already
> has that check, you don't need it here.

ok understood, check is useless.

> 
>>>> +                       gpiod_set_value_cansleep(pcie->reset_gpio, 0);
> 
> ...
> 
>>>> +               if (pcie->reset_gpio) {
>>>> +                       usleep_range(100, 200);
>>>
>>> fsleep() ?
>>> Btw, why is it needed here, perhaps a comment?
>>
>> The comment should be the same than in the probe [1].
>> Should I copy it? Or should I just add a reference to the probe?
>>
>> [1]
>> https://elixir.bootlin.com/linux/v6.8-rc1/source/drivers/pci/controller/cadence/pci-j721e.c#L535
> 
> Either way works for me.
> 
>>>> +                       gpiod_set_value_cansleep(pcie->reset_gpio, 1);
>>>> +               }
> 
> ...
> 
>>>> +#define cdns_pcie_to_rc(p) container_of(p, struct cdns_pcie_rc, pcie)
>>>
>>> Is container_of.h included in this file?
>>
>> linux/container_of.h is included in linux/kernel.h.
>> And linux/kernel.h is included in pcie-cadence.h
>> (https://elixir.bootlin.com/linux/v6.8-rc1/source/drivers/pci/controller/cadence/pcie-cadence.h#L9).
> 
> Okay, so, try to clean up pcie-cadence.h so it won't use "proxy" headers.
> There is an IWYU (include what you use) principle, please follow it.

In fact, as cdns_pcie_to_rc is only used in pci-j721e.c, no need to have
it in a header file.
I prefer to move cdns_pcie_to_rc definition in pci-j721e.c.
As I don't modify pcie-cadence.h, the cleanup to not use proxy headers
is something that it can be done outside this series.

-- 
Thomas Richard, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com