Re: [PATCH 1/3] mm/mlock.c: convert put_page() to put_user_page*()

[John Hubbard]

On 8/7/19 4:01 AM, Michal Hocko wrote:
> On Mon 05-08-19 15:20:17, john.hubbard@xxxxxxxxx wrote:
>> From: John Hubbard <jhubbard@xxxxxxxxxx>
>>
>> For pages that were retained via get_user_pages*(), release those pages
>> via the new put_user_page*() routines, instead of via put_page() or
>> release_pages().
> 
> Hmm, this is an interesting code path. There seems to be a mix of pages
> in the game. We get one page via follow_page_mask but then other pages
> in the range are filled by __munlock_pagevec_fill and that does a direct
> pte walk. Is using put_user_page correct in this case? Could you explain
> why in the changelog?
> 

Actually, I think follow_page_mask() gets all the pages, right? And the
get_page() in __munlock_pagevec_fill() is there to allow a pagevec_release() 
later.

But I still think I mighthave missed an error case, because the pvec_putback
in __munlock_pagevec() is never doing put_user_page() on the put-backed pages.

Let me sort through this one more time and maybe I'll need to actually
change the code. And either way, comments and changelog will need some notes, 
agreed.

thanks,
-- 
John Hubbard
NVIDIA