Path: csiph.com!aioe.org!bofh.it!news.nic.it!robomod
From: Dmitry Vyukov <dvyukov@google.com>
Newsgroups: linux.kernel
Subject: Re: kernel BUG at mm/mempolicy.c:1699!
Date: Fri, 29 Jul 2016 21:20:02 +0200
Message-ID: <s0lua-qL-7@gated-at.bofh.it>
References: <s0keK-7S4-25@gated-at.bofh.it> <s0koq-8aN-25@gated-at.bofh.it> <s0lua-qL-9@gated-at.bofh.it>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Sender: robomod@news.nic.it
Approved: robomod@news.nic.it
Lines: 33
Organization: linux.* mail to news gateway
Xref: csiph.com linux.kernel:1452526

On Fri, Jul 29, 2016 at 9:10 PM, Vegard Nossum <vegard.nossum@oracle.com> wrote:
> On 07/29/2016 08:05 PM, Dmitry Vyukov wrote:
>>
>> On Fri, Jul 29, 2016 at 7:57 PM, Vegard Nossum <vegard.nossum@oracle.com>
>> wrote:
>>>
>>> ------------[ cut here ]------------
>>> kernel BUG at mm/mempolicy.c:1699!
>
> [...]
>>>
>>> In particular, it's interesting that the kernel/exit.c line is
>>>
>>>      mpol_put(tsk->mempolicy);
>>>
>>> and alloc_pages_current() does (potentially):
>>>
>>>      pol = get_task_policy(current);.
>>>
>>> The bug seems very new or very rare or both.
>>
>>
>> This is https://github.com/google/kasan/issues/35
>> It is introduced with stackdepot.
>
>
> Ah, cool.
>
> Would it be enough to set __GFP_THISNODE in depot_save_stack() so it
> uses &default_policy instead of current->mempolicy?

I don't have deep understanding of that code. But looks at the code,
using &default_policy should help.