Re: [PATCH v2] mm: fix the memory leak after collapsing the huge page fails (fwd)

[Vlastimil Babka <vbabka@suse.cz>]

On 05/10/2017 08:25 AM, Julia Lawall wrote:
> 
> 
> On Wed, 10 May 2017, zhong jiang wrote:
> 
>> On 2017/5/9 23:43, Julia Lawall wrote:
>>> Hello,
>>>
>>> I don't know if there is a bug here, but it could e worth checking on.  If
>>> the loop on line 1481 is executed, page will not be NULL at the out label
>>> on line 1560.  Instead it will have a dummy value.  Perhaps the value of
>>> result keeps the if at the out label from being taken.
>>>
>>> julia
>>   Hi, Julia
>>
>>    it has no memory leak.  so my initial thought is not correct. but I do not know you mean.
>>    The page is local variable.  it aybe a  dummy value. but it should not cause any issue.
>>    is it right? or I miss something.
> 
> I had first been thinking that the if branch was referencing page.  In
> that case, if page were a dummy value, then there could be a problem.  But
> now I see that the branch does not refer to page.  So the question is
> just, if the loop on lines 1481-1491 is executed, is it correct to execute
> the code put_page(new_page)?  Or will result be SCAN_SUCCEEDED in that
> case?

That loop is under "if (result == SCAN_SUCCEED)", so yeah. It also ends
with "*hpage = NULL;", so the put_page(hpage) in khugepaged_do_scan()
won't apply. I see no problem besides the very non-obvious code :/

--
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>