Re: [PATCH v2 6/8] mm: prevent get_user_pages() from overflowing page refcount

[Vlastimil Babka]

On 11/21/19 9:38 PM, gregkh@xxxxxxxxxxxxxxxxxxx wrote:
> On Mon, Nov 11, 2019 at 05:00:29AM +0000, Ajay Kaher wrote:
>>
>> ïOn 06/11/19, 2:25 PM, "Vlastimil Babka" <vbabka@xxxxxxx> wrote:
>>
>>> On 10/25/19 8:18 AM, Ajay Kaher wrote:
>>>> On 17/10/19, 9:58 PM, "Ajay Kaher" <akaher@xxxxxxxxxx> wrote:
>>>>     
>>>>>    
>>>>> Could we handle arch-specific gup.c in different patch sets and 
>>>>> let these patches to merge to 4.4.y?
>>>>   
>>>> Vlastimil, please suggest if it's fine to merge these patches to 4.4.y
>>>
>>> I'm not sure if it makes much sense to merge them without the arch-specific gup
>>> support, when we're aware that it's missing.
>>>
>>>> and handle arch-specific gup.c in different patch sets starts from 4.19.y,
>>>
>>> Actually arch-specific gup.c were removed in 4.13, so it's enough to start from
>>> 4.9.y, which I'm going to finally look into.
>>
>> Yes x86 gup.c is removed. s390 gup.c is present till 4.19,
>> so if you are making changes in this file for 4.4.y and 4.9.y, 
>> then same should be done for 4.14.y and v4.19.y.
> 
> Ok, I have no idea what to do here.  I have two different series from
> both of you, yet they are different.
> 
> Can you both come up with a series you agree on, and send it to me, with
> both of your acks so that I know this is what should be applied?  I've
> deleted both of your current series from my todo mbox.

I started by fixing up [1] 4.9, 4.14 and 4.19 which already got the
backport, but without arch-specific gup.c variants. I'll wait for Ajay's
feedback for the 4.4 series.

[1] https://lore.kernel.org/linux-mm/20191129090351.3507-1-vbabka@xxxxxxx/

> thanks,
> 
> greg k-h
>