Re: [PATCH 2/2] mm/hwpoison: fix refcount of THP head page in no-injection case

[Wanpeng Li <wanpeng.li@hotmail.com>]

On 8/10/15 5:20 PM, Naoya Horiguchi wrote:
> On Mon, Aug 10, 2015 at 05:06:25PM +0800, Wanpeng Li wrote:
> ...
>>>>> diff --git a/mm/hwpoison-inject.c b/mm/hwpoison-inject.c
>>>>> index 5015679..c343a45 100644
>>>>> --- a/mm/hwpoison-inject.c
>>>>> +++ b/mm/hwpoison-inject.c
>>>>> @@ -56,6 +56,8 @@ inject:
>>>>>      return memory_failure(pfn, 18, MF_COUNT_INCREASED);
>>>>>  put_out:
>>>>>      put_page(p);
>>>>> +    if (p != hpage)
>>>>> +        put_page(hpage);
>>>> Yes, we need this when we inject to a thp tail page and "goto put_out"
>>>> is
>>>> called. But it seems that this code can be called also when injecting
>>>> error
>>>> to a hugetlb tail page and hwpoison_filter() returns non-zero, which is
>>>> not
>>>> expected. Unfortunately simply doing like below
>>>>
>>>> +    if (!PageHuge(p) && p != hpage)
>>>> +        put_page(hpage);
>>>>
>>>> doesn't work, because exisiting put_page(p) can release refcount of
>>>> hugetlb
>>>> tail page, while get_hwpoison_page() takes refcount of hugetlb head
>>>> page.
>>>>
>>>> So I feel that we need put_hwpoison_page() to properly release the
>>>> refcount
>>>> taken by memory error handlers.
>>> Good point. I think I will continue to do it and will post it out soon. :)
>> How about something like this:
>>
>> +void put_hwpoison_page(struct page *page)
>> +{
>> +       struct page *head = compound_head(page);
>> +
>> +       if (PageHuge(head))
>> +               goto put_out;
>> +
>> +       if (PageTransHuge(head))
>> +               if (page != head)
>> +                       put_page(head);
>> +
>> +put_out:
>> +       put_page(page);
>> +       return;
>> +}
>> +
> Looks good.
>
>> Any comments are welcome, I can update the patch by myself. :)
> Most of callsites of put_page() in memory_failure(), soft_offline_page(),
> and unpoison_page() can be replaced with put_hwpoison_page().

Cool, thanks for your pointing out. I will do it soon. :)

Regards,
Wanpeng Li

>
> Thanks,
> Naoya Horiguchi

--
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>