lkml.org 
[lkml]   [2011]   [May]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH] mm: fix page_lock_anon_vma leaving mutex locked
On Sat, May 28, 2011 at 3:02 PM, Hugh Dickins <hughd@google.com> wrote:
>
> But I'm replying before I've given it enough thought,
> mainly to let you know that I am back on it now.

So I applied your other two patches as obvious, but not this one.

I'm wondering - wouldn't it be nicer to just re-check (after getting
the anon_vma lock) that page->mapping still matches anon_mapping?

That said, I do agree with the "anon_vma_root" part of your patch. I
just think you mixed up two independent issues with it: the fact that
we may be unlocking a new root, and the precise check used to
determine whether the anon_vma might have changed.

So my gut feeling is that we should do the "anon_vma" root thing
independently as a fix for the "maybe anon_vma->root changed" issue,
and then as a separate patch decide on how to check whether anon_vma
is still valid.

Hmm?

Linus


\
 
 \ /
  Last update: 2011-05-29 01:27    [from the cache]
©2003-2011 Jasper Spaans