Re: [PATCH v2] kasan: fix cache shrink race with CPU hotplug

On Sat,  8 Aug 2026 11:14:59 +0800 Hui Su <[email protected]> wrote:

> kasan_quarantine_remove_cache() first invokes per_cpu_remove_cache() on
> all online CPUs. Each callback moves objects belonging to the cache from
> cpu_quarantine to the CPU's shrink_qlist, where they can later be freed
> from task context.
> 
> kmem_cache_destroy() invokes the quarantine removal path while holding
> cpus_read_lock(), but kmem_cache_shrink() does not. The latter can
> therefore race with CPU offlining as follows:
> 
>   kmem_cache_shrink()             CPU hotplug
>   -------------------             -----------
>   on_each_cpu()
>     CPU1 moves objects to
>     CPU1's shrink_qlist
>   on_each_cpu() returns
>                                   CPU1 goes offline
>                                   kasan_cpu_offline()
>                                     drains cpu_quarantine
>                                     leaves shrink_qlist untouched
>   for_each_online_cpu()
>     skips CPU1
> 
> The objects left on CPU1's shrink_qlist are not returned to the slab
> allocator. This may prevent kmem_cache_shrink() from releasing slabs
> that would otherwise become empty. If CPU1 remains offline, a later
> kmem_cache_destroy() also skips the list and can report that the cache
> still contains objects.
> 
> An intermittent occurrence was observed with a virtio-9p filesystem.
> The mount and umount commands both returned 0, but the kernel logged
> the following during the userspace-triggered teardown:
>
> ...
>

Thanks, I'll queue this for testing while we await maintainer review.

AI review suggests that there's a pre-existing quarantine_size
accounting flaw later in this function:

	https://sashiko.dev/#/patchset/[email protected]

If true, I'm surprised this hasn't yet been reported.


Also, I'd like to see a need_resched() wrapping that expensive

		/* Scanning whole quarantine can take a while. */
		raw_spin_unlock_irqrestore(&quarantine_lock, flags);
		cond_resched();
		raw_spin_lock_irqsave(&quarantine_lock, flags);