Re: [PATCH] mm/mempolicy: Fix weighted interleave auto sysfs name

[Huang, Ying]

Joshua Hahn <joshua.hahnjy@xxxxxxxxx> writes:

> On Tue, 7 Apr 2026 13:09:04 -0400 Gregory Price <gourry@xxxxxxxxxx> wrote:
>
>> On Tue, Apr 07, 2026 at 07:14:14AM -0700, Joshua Hahn wrote:
>> > The __ATTR macro is a utility that makes defining kobj_attributes easier
>> > by stringfying the name, verifying the mode, and setting the show/store
>> > fields in a single initializer. It takes a raw token as the first value,
>> > rather than a string, so that __ATTR family macros like __ATTR_RW can
>> > token-paste it for inferring the _show / _store function names.
>> > 
>> > Commit e341f9c3c841 ("mm/mempolicy: Weighted Interleave Auto-tuning")
>> > used the __ATTR macro to define the "auto" sysfs for weighted
>> > interleave. A few months later, commit 2fb6915fa22d ("compiler_types.h:
>> > add "auto" as a macro for "__auto_type"") introduced a #define macro
>> > which expanded auto into __auto_type.
>> > 
>> 
>
> Hello Gregory, thanks for reviewing my patch!!
>
>> Well that's really unfortunately.
>> 
>> Is this really the only "auto" in the entire sysfs structure?
>
> Thankfully, it seems so:
>
> $ rg "__ATTR\(auto"
>
> mm/mempolicy.c
> 3786:   __ATTR(auto, 0664, weighted_interleave_auto_show,
>
> arch/powerpc/platforms/pseries/power.c
> 44:     __ATTR(auto_poweron, 0644, auto_poweron_show, auto_poweron_store);
>
> Although maybe it is more valuable to check any __ATTR family macro
> for a macro as the first parameter.  In hindsight, I should have been more
> wary of passing a keyword as a token.
>
> [...snip...]
>
>> > Signed-off-by: Joshua Hahn <joshua.hahnjy@xxxxxxxxx>
>> 
>> Maybe - 
>> 
>> Fixes: 2fb6915fa22d ("compiler_types.h: add "auto" as a macro for "__auto_type"")
>
> I thought about adding a Fixes tag, decided against it because it's
> purely cosmetic.
>
> ... but now that I think about it, I think this could break some workflows
> for users who have automated writing to the auto sysfs file.

Fixes tag helps people to decide whether to backport the commit.  So, I
think that we need it anyway.  Feel free to add

Acked-by: Huang Ying <ying.huang@xxxxxxxxxxxxxxxxx>

in the future versions.

[snip]

---
Best Regards,
Huang, Ying