Re: [PATCH v2] mm, slab: clean up slab->obj_exts always

[Andy Shevchenko]

On Thu, Apr 24, 2025 at 06:48:46PM +0200, Vlastimil Babka wrote:
> On 4/24/25 18:34, Andy Shevchenko wrote:
> > On Mon, Apr 21, 2025 at 03:52:32PM +0800, Zhenhua Huang wrote:
> >> When memory allocation profiling is disabled at runtime or due to an
> >> error, shutdown_mem_profiling() is called: slab->obj_exts which
> >> previously allocated remains.
> >> It won't be cleared by unaccount_slab() because of
> >> mem_alloc_profiling_enabled() not true. It's incorrect, slab->obj_exts
> >> should always be cleaned up in unaccount_slab() to avoid following error:
> >> 
> >> [...]BUG: Bad page state in process...
> >> ..
> >> [...]page dumped because: page still charged to cgroup
> > 
> > Please, always compile test with `make W=1`. Since CONFIG_WERROR=y this
> > effectively breaks the build with Clang.
> 
> I don't see why, nor observe any W=1 warnings, can you be more specific? Thanks.

Specifics are in the fix I sent. Just a relatively new Clang and
relatively recent enabling of warning for unused static inline functions
in the C code. If you are insisting in seeing the exact kernel
configuration I have, tell me where to send, I'll send it privately
to avoid noise here.

-- 
With Best Regards,
Andy Shevchenko