lkml.org 
[lkml]   [2013]   [Feb]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH 1/3] mm: use long type for page counts in mm_populate() and get_user_pages()
From
On Wed, Feb 6, 2013 at 5:10 PM, Andrew Morton <akpm@linux-foundation.org> wrote:
> On Wed, 06 Feb 2013 19:39:11 -0500
> Sasha Levin <sasha.levin@oracle.com> wrote:
>
>> We're now hitting the VM_BUG_ON() which was added in the last hunk of the
>> patch:
>
> hm, why was that added.
>
> Michel, I seem to have confused myself over this series. I saw a
> report this morning which led me to drop
> mm-accelerate-munlock-treatment-of-thp-pages.patch but now I can't find
> that report and I'm wondering if I should have dropped
> mm-accelerate-mm_populate-treatment-of-thp-pages.patch instead.
>
> Given that and Sasha's new report I think I'll drop
>
> mm-use-long-type-for-page-counts-in-mm_populate-and-get_user_pages.patch
> mm-use-long-type-for-page-counts-in-mm_populate-and-get_user_pages-fix.patch
> mm-use-long-type-for-page-counts-in-mm_populate-and-get_user_pages-fix-fix.patch
> mm-accelerate-mm_populate-treatment-of-thp-pages.patch
> mm-accelerate-munlock-treatment-of-thp-pages.patch
>
> and let's start again?

All right. My bad, there were issues in the patch series. I think
there were two:

- The VM_BUG_ON(!reg) in "mm: use long type for page counts in
mm_populate() and get_user_pages()". The intention there was to test
for what happened in the original overflow case, which is that gup
would return 0 as the passed nr_pages argument (when passed as an int)
would be <= 0. As it turns out, this didn't account for the other case
where gup returns 0, which is when the first page is file-backed, not
found in page cache, and the mmap_sem gets dropped due to a non-NULL
"nonblocking" argument (as is the case with mm_populate())

- The issue in munlock, where the page mask wasn't correctly computed
due to an off by 1 issue.

I agree a clean resend seems the most appropriate course of action at
this point. Sorry for the trouble :/

--
Michel "Walken" Lespinasse
A program is never fully debugged until the last user dies.


\
 
 \ /
  Last update: 2013-02-07 23:21    [W:0.059 / U:0.296 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site