Re: [PATCH] mm/mlock: fix BUG_ON unlocked page for nolinear VMAs

[Vlastimil Babka <vbabka@suse.cz>]

On 01/03/2014 09:52 PM, Linus Torvalds wrote:
> On Fri, Jan 3, 2014 at 12:17 PM, Sasha Levin <sasha.levin@oracle.com> wrote:
>>
>> Ping? This BUG() is triggerable in 3.13-rc6 right now.
> 
> So Andrew suggested just removing the BUG_ON(), but it's been there
> for a *long* time.

Yes, Andrew also merged this patch for that:
 http://ozlabs.org/~akpm/mmots/broken-out/mm-remove-bug_on-from-mlock_vma_page.patch

But there wasn't enough confidence in the fix to sent it to you yet, I guess.

The related thread: http://www.spinics.net/lists/linux-mm/msg66972.html

> And I detest the patch that was sent out that said "Should I check?"
> 
> Maybe we should just remove that mlock_vma_page() thing instead in

You mean that it it's already undeterministic because it can be already skipped when
mmap_sem can't be acquired for read? I think the assumption for this case is that mmap_sem
is already held for write which means VM_LOCKED is unset anyway (per comments at
try_to_unmap_file(), which calls try_to_unmap_cluster()). I'm however not sure how it is
protected from somebody else holding the semaphore...

> try_to_unmap_cluster()? Or maybe actually lock the page around calling
> it?

check_page is already locked, see try_to_munlock() which calls try_to_unmap_file(). So
this might smell of potential deadlock?

I'm for going with the removal of BUG_ON. The TestSetPageMlocked should provide enough
race protection.

>              Linus
> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
> 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>