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

[Sasha Levin <sasha.levin@oracle.com>]

On 12/18/2013 04:43 PM, Andrew Morton wrote:
> On Wed, 18 Dec 2013 17:23:03 +0800 Wanpeng Li <liwanp@linux.vnet.ibm.com> wrote:
>
>>>> diff --git a/mm/rmap.c b/mm/rmap.c
>>>> index 55c8b8d..1e24813 100644
>>>> --- a/mm/rmap.c
>>>> +++ b/mm/rmap.c
>>>> @@ -1347,6 +1347,7 @@ static int try_to_unmap_cluster(unsigned long cursor, unsigned int *mapcount,
>>>>   	unsigned long end;
>>>>   	int ret = SWAP_AGAIN;
>>>>   	int locked_vma = 0;
>>>> +	int we_locked = 0;
>>>>
>>>>   	address = (vma->vm_start + cursor) & CLUSTER_MASK;
>>>>   	end = address + CLUSTER_SIZE;
>>>> @@ -1385,9 +1386,15 @@ static int try_to_unmap_cluster(unsigned long cursor, unsigned int *mapcount,
>>>>   		BUG_ON(!page || PageAnon(page));
>>>>
>>>>   		if (locked_vma) {
>>>> -			mlock_vma_page(page);   /* no-op if already mlocked */
>>>> -			if (page == check_page)
>>>> +			if (page != check_page) {
>>>> +				we_locked = trylock_page(page);
>>>
>>> If it's not us who has the page already locked, but somebody else, he
>>> might unlock it at this point and then the BUG_ON in mlock_vma_page()
>>> will trigger again.
>
> yes, this patch is pretty weak.
>
>> Any better idea is appreciated. ;-)
>
> Remove the BUG_ON() from mlock_vma_page()?  Why was it added?
> isolate_lru_page() and putback_lru_page() and *might* require
> the page be locked, but I don't immediately see issues?

Ping? This BUG() is triggerable in 3.13-rc6 right now.


Thanks,
Sasha

--
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>