Re: mm: kernel BUG at mm/mempolicy.c:1203!

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

On 12/17/2013 01:11 AM, Naoya Horiguchi wrote:
> Hello Bob,
> 
> On Tue, Dec 17, 2013 at 12:38:49PM +0800, Bob Liu wrote:
>> On 12/17/2013 09:10 AM, Sasha Levin wrote:
>>> On 12/16/2013 07:44 PM, Bob Liu wrote:
>>>>
>>>> On 12/16/2013 07:37 AM, Sasha Levin wrote:
>>>>> Hi all,
>>>>>
>>>>> While fuzzing with trinity inside a KVM tools guest running latest -next
>>>>> kernel, I've
>>>>> stumbled on the following spew.
>>>>>
>>>>> This seems to be due to commit 0bf598d863e "mbind: add BUG_ON(!vma) in
>>>>> new_vma_page()"
>>>>> which added that BUG_ON.
>>>>
>>>> Could you take a try with this patch from Wanpeng Li?
>>>>
>>>> Thanks,
>>>> -Bob
>>>>
>>>> Subject: [PATCH] mm/mempolicy: fix !vma in new_vma_page()
>>>> ....
>>>> Signed-off-by: Wanpeng Li <liwanp@linux.vnet.ibm.com>
>>>> index eca4a31..73b5a35 100644
>>>> --- a/mm/mempolicy.c
>>>> +++ b/mm/mempolicy.c
>>>> @@ -1197,14 +1197,16 @@ static struct page *new_vma_page(struct page
>>>> *page, unsigned long private, int *
>>>>                break;
>>>>            vma = vma->vm_next;
>>>>        }
>>>> +
>>>> +    if (PageHuge(page)) {
>>>> +        if (vma)
>>>> +            return alloc_huge_page_noerr(vma, address, 1);
>>>> +        else
>>>> +            return NULL;
>>>> +    }
>>>>        /*
>>>> -     * queue_pages_range() confirms that @page belongs to some vma,
>>>> -     * so vma shouldn't be NULL.
>>>> +     * if !vma, alloc_page_vma() will use task or system default policy
>>>>         */
>>>> -    BUG_ON(!vma);
>>>> -
>>>> -    if (PageHuge(page))
>>>> -        return alloc_huge_page_noerr(vma, address, 1);
>>>>        return alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, address);
>>>>    }
>>>>    #else
>>>>
>>>
>>> Hmm... So in essence it's mostly a revert of Naoya's patch, who seemed
>>> pretty certain that this
>>> situation shouldn't happen at all. What's the reasoning behind just
>>
>> I think this assumption may not correct.
>> Even if
>> address = __vma_address(page, vma);
>> and
>> vma->start < address < vma->end;
>> page_address_in_vma() may still return -EFAULT because of many other
>> conditions in it.
>> As a result the while loop in new_vma_page() may end with vma=NULL.
>>
>> Naoya, any idea?
> 
> Yes, you totally make sense. So please apply Wanpeng's patch.

Shouldn't it just be a revert of Naoya's patch? Otherwise we're
changing code paths unnecessarily.


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>