Re: mm: BUG in resv_map_release

[Mike Kravetz]

On 03/28/2017 04:38 PM, Andrew Morton wrote:
> On Thu, 23 Mar 2017 11:19:38 +0100 Dmitry Vyukov <dvyukov@xxxxxxxxxx> wrote:
> 
>> Hello,
>>
>> I've got the following BUG while running syzkaller fuzzer.
>> Note the injected kmalloc failure, most likely it's the root cause.
>>
> 
> Yes, probably the logic(?) in region_chg() leaked a
> resv->adds_in_progress++, although I'm not sure how.  And afaict that
> code can leak the memory at *nrg if the `trg' allocation attempt failed
> on the second or later pass around the retry loop.
> 
> Blah.  Does someone want to take a look at it?

I sent out a patch to address this and Hillf Acked.  Unfortunately,
there was a typo in your e-mail when I sent out the patch.  So, you
may not have noticed.

[PATCH] mm/hugetlb: Don't call region_abort if region_chg fails
http://marc.info/?l=linux-mm&m=149033588500724&w=2

If you need/want me to send again, let me know.
-- 
Mike Kravetz