Re: [PATCH] mm/mempool: fix poisoning order>0 pages with HIGHMEM

[Vlastimil Babka <vbabka@suse.cz>]

On 11/14/25 06:04, Christoph Hellwig wrote:
> On Thu, Nov 13, 2025 at 07:54:35PM +0100, Vlastimil Babka wrote:
>> Christoph found out this is due to the poisoning code not dealing
>> properly with CONFIG_HIGHMEM because only the first page is mapped but
>> then the whole potentially high-order page is accessed.
>> 
>> This went unnoticed for years probably because nobody has yet used a
>> mempool for order>0 pages before the new block code in -next.
> 
> I did a quick audit: and bcache, dm-integrity (config dependent) and the
> KASAN unit tests create page based mempools with order > 0.  It looks
> like none of those ever got much testing on highmem systems.

Thanks,

KASAN unit tests sounds like something that would have been flagged by
automated testing long ago, however the mempool-specific poisoning isn't
compatible with it so poison_element() and check_element() both return
immediately with kasan_enabled().

bcache and dm-integrity are perhaps too complex setups for the test robots.
But I'll add cc: stable then.

> The fix looks good:
> 
> Reviewed-by: Christoph Hellwig <hch@lst.de>