Re: [PATCH] MIPS: Drop virt_to_phys define to self

["Arnd Bergmann" <arnd@arndb.de>]

On Tue, Aug 8, 2023, at 15:12, Linus Walleij wrote:
> On Tue, Aug 8, 2023 at 11:29 AM Linus Walleij <linus.walleij@linaro.org> wrote:
>
>> The function virt_to_phys was defined to virt_to_phys and then
>> implemented right below.
>>
>> I can't understand why, just drop it and let the actual function
>> slot in.
>>
>> Fixes: dfad83cb7193 ("MIPS: Add support for CONFIG_DEBUG_VIRTUAL")
>> Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
>> ---
>>  arch/mips/include/asm/io.h | 1 -
>>  1 file changed, 1 deletion(-)
>>
>> diff --git a/arch/mips/include/asm/io.h b/arch/mips/include/asm/io.h
>> index affd21e9c20b..18ed44843541 100644
>> --- a/arch/mips/include/asm/io.h
>> +++ b/arch/mips/include/asm/io.h
>> @@ -111,7 +111,6 @@ extern phys_addr_t __virt_to_phys(volatile const void *x);
>>  #define __virt_to_phys(x)      __virt_to_phys_nodebug(x)
>>  #endif
>>
>> -#define virt_to_phys virt_to_phys
>>  static inline phys_addr_t virt_to_phys(const volatile void *x)
>>  {
>>         return __virt_to_phys(x);
>
> Just have to CC Arnd on this because I never feel I understand this
> properly.
>
> Normally you would do this so as to override the default virt_to_phys()
> from include/asm-generic/io.h, but we do not seem to be using it here?

Correct. In linux-next, we have converted arch/sh to use
include/asm-generic/io.h, so arch/mips is now the last one
to not use it. 

I see that Jiaxun Yang posted a patch for this in May and it looked
like it should finally work [1], but there seems to still be something
missing.

> Further right below we are implementing phys_to_virt() with no
> corresponding define, so in any case this needs to go or that one
> needs a define.
>
> That said they seem like a good candidate to replace with the
> generic variant because the content is the same sans some debug
> hacks that I doubt are still needed.

I think the ARCH_HAS_DEBUG_VIRTUAL option is useful in
general, I'd rather move that into the common code if
we were to unify it.

For the moment, I'd suggest we leave the #define in place
here in order to finish the patch that starts using the
asm-generic header, and then we can look into using more
of the generic code.

       Arnd

[1] https://lore.kernel.org/linux-mips/20230519195135.79600-1-jiaxun.yang@flygoat.com/