David Rientjes wrote:
> Since exit_mmap() is done without the protection of mm->mmap_sem, it is
> possible for the oom reaper to concurrently operate on an mm until
> MMF_OOM_SKIP is set.
>
> This allows munlock_vma_pages_all() to concurrently run while the oom
> reaper is operating on a vma. Since munlock_vma_pages_range() depends on
> clearing VM_LOCKED from vm_flags before actually doing the munlock to
> determine if any other vmas are locking the same memory, the check for
> VM_LOCKED in the oom reaper is racy.
>
> This is especially noticeable on architectures such as powerpc where
> clearing a huge pmd requires kick_all_cpus_sync(). If the pmd is zapped
> by the oom reaper during follow_page_mask() after the check for pmd_none()
> is bypassed, this ends up deferencing a NULL ptl.
I don't know whether the explanation above is correct.
Did you actually see a crash caused by this race?
>
> Fix this by reusing MMF_UNSTABLE to specify that an mm should not be
> reaped. This prevents the concurrent munlock_vma_pages_range() and
> unmap_page_range(). The oom reaper will simply not operate on an mm that
> has the bit set and leave the unmapping to exit_mmap().
But this patch is setting MMF_OOM_SKIP without reaping any memory as soon as
MMF_UNSTABLE is set, which is the situation described in 212925802454:
At the same time if the OOM reaper doesn't wait at all for the memory of
the current OOM candidate to be freed by exit_mmap->unmap_vmas, it would
generate a spurious OOM kill.
If exit_mmap() does not wait for any pages and __oom_reap_task_mm() can not
handle mlock()ed pages, isn't it better to revert 212925802454 like I mentioned
at https://patchwork.kernel.org/patch/10095661/ and let the OOM reaper reclaim
as much as possible before setting MMF_OOM_SKIP?
