Re: [PATCH 1/2] mm: fix missing handler for __GFP_NOWARN

[Qi Zheng]

On 2022/5/11 2:59 AM, Andrew Morton wrote:
> On Tue, 10 May 2022 19:38:08 +0800 Qi Zheng <zhengqi.arch@xxxxxxxxxxxxx> wrote:
>
> > We expect no warnings to be issued when we specify __GFP_NOWARN, but
> > currently in paths like alloc_pages() and kmalloc(), there are still
> > some warnings printed, fix it.
>
> Looks sane to me.
>
> > --- a/mm/internal.h
> > +++ b/mm/internal.h
> > @@ -35,6 +35,17 @@ struct folio_batch;
> >   /* Do not use these with a slab allocator */
> >   #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
> >   
> > +#define WARN_ON_ONCE_GFP(cond, gfp)	({				\
> > +	static bool __section(".data.once") __warned;			\
> > +	int __ret_warn_once = !!(cond);					\
> > +									\
> > +	if (unlikely(!(gfp & __GFP_NOWARN) && __ret_warn_once && !__warned)) { \
> > +		__warned = true;					\
> > +		WARN_ON(1);						\
> > +	}								\
> > +	unlikely(__ret_warn_once);					\
> > +})
>
> I don't think WARN_ON_ONCE_GFP is a good name for this.  But
> WARN_ON_ONCE_IF_NOT_GFP_NOWARN is too long :(
>
> WARN_ON_ONCE_NOWARN might be better.  No strong opinion here, really.

I've thought about WARN_ON_ONCE_NOWARN, but I feel a little weird 
putting 'WARN' and 'NOWARN' together, how about WARN_ON_ONCE_IF_ALLOWED?

> > @@ -4902,8 +4906,8 @@ __alloc_pages_slowpath(gfp_t gfp_mask, unsigned int order,
> >   	 * We also sanity check to catch abuse of atomic reserves being used by
> >   	 * callers that are not in atomic context.
> >   	 */
> > -	if (WARN_ON_ONCE((gfp_mask & (__GFP_ATOMIC|__GFP_DIRECT_RECLAIM)) ==
> > -				(__GFP_ATOMIC|__GFP_DIRECT_RECLAIM)))
> > +	if (WARN_ON_ONCE_GFP((gfp_mask & (__GFP_ATOMIC|__GFP_DIRECT_RECLAIM)) ==
> > +				(__GFP_ATOMIC|__GFP_DIRECT_RECLAIM), gfp_mask))
> >   		gfp_mask &= ~__GFP_ATOMIC;
> >   
> >   retry_cpuset:
>
> I dropped this hunk - Neil's "mm: discard __GFP_ATOMIC"
> (https://lkml.kernel.org/r/163712397076.13692.4727608274002939094@xxxxxxxxxxxxxxxxxxxxx)
> deleted this code.

This series is based on v5.18-rc5, I will rebase it to the latest next
branch and check if there are any missing WARN_ON_ONCEs that are not
being handled.

Thanks,
Qi

--
Thanks,
Qi