linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Linus Torvalds <torvalds@linux-foundation.org>
To: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Nick Piggin <npiggin@suse.de>, Hugh Dickins <hugh@veritas.com>,
	Linux Memory Management List <linux-mm@kvack.org>
Subject: Re: [patch] mm: fix anon_vma races
Date: Fri, 17 Oct 2008 16:53:39 -0700 (PDT)	[thread overview]
Message-ID: <alpine.LFD.2.00.0810171621180.3438@nehalem.linux-foundation.org> (raw)
In-Reply-To: <1224285222.10548.22.camel@lappy.programming.kicks-ass.net>


On Sat, 18 Oct 2008, Peter Zijlstra wrote:
> On Thu, 2008-10-16 at 06:10 +0200, Nick Piggin wrote:
> 
> > Signed-off-by: Nick Piggin <npiggin@suse.de>
> > ---
> > Index: linux-2.6/mm/rmap.c
> > ===================================================================
> > --- linux-2.6.orig/mm/rmap.c
> > +++ linux-2.6/mm/rmap.c
> > @@ -81,8 +81,15 @@ int anon_vma_prepare(struct vm_area_stru
> >  		/* page_table_lock to protect against threads */
> >  		spin_lock(&mm->page_table_lock);
> >  		if (likely(!vma->anon_vma)) {
> > -			vma->anon_vma = anon_vma;
> >  			list_add_tail(&vma->anon_vma_node, &anon_vma->head);
> > +			/*
> > +			 * This smp_wmb() is required to order all previous
> > +			 * stores to initialize the anon_vma (by the slab
> > +			 * ctor) and add this vma, with the store to make it
> > +			 * visible to other CPUs via vma->anon_vma.
> > +			 */
> > +			smp_wmb();
> > +			vma->anon_vma = anon_vma;
> 
> I'm not getting why you explicitly move the list_add_tail() before the
> wmb, doesn't the list also expose the anon_vma to other cpus?

I do think the anon_vma locking might be good to look over. It is very 
non-obvious. Especially the initial create is really really quite suspect. 
I suspect we should start out with the anon-vma locked, and do the

	spin_unlock(&anon_vma->lock);

unconditionally in anon_vma_prepare(), and just simplify it. As it is, 
newly allocated anon_vma's get exposed in unlocked state while we're still 
working on them.

But I think that what Nick did is correct - we always start traversal 
through anon_vma->head, so no, the "list_add_tail()" won't expose it to 
anybody else, because nobody else has seen the anon_vma().

That said, that's really too damn subtle. We shouldn't rely on memory 
ordering for the list handling, when the list handling is _supposed_ to be 
using that anon_vma->lock thing.

			Linus

--
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>

  reply	other threads:[~2008-10-17 23:53 UTC|newest]

Thread overview: 52+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-10-16  4:10 Nick Piggin
2008-10-17 22:14 ` Hugh Dickins
2008-10-17 23:05   ` Linus Torvalds
2008-10-18  0:13     ` Hugh Dickins
2008-10-18  0:25       ` Linus Torvalds
2008-10-18  1:53       ` Nick Piggin
2008-10-18  2:50         ` Paul Mackerras
2008-10-18  2:57           ` Linus Torvalds
2008-10-18  5:49           ` Nick Piggin
2008-10-18 10:49             ` Paul Mackerras
2008-10-18 17:00             ` Linus Torvalds
2008-10-18 18:44               ` Matthew Wilcox
2008-10-19  2:54                 ` Nick Piggin
2008-10-19  2:53               ` Nick Piggin
2008-10-17 23:13 ` Peter Zijlstra
2008-10-17 23:53   ` Linus Torvalds [this message]
2008-10-18  0:42     ` Linus Torvalds
2008-10-18  1:08       ` Linus Torvalds
2008-10-18  1:32         ` Nick Piggin
2008-10-18  2:11           ` Linus Torvalds
2008-10-18  2:25             ` Nick Piggin
2008-10-18  2:35               ` Nick Piggin
2008-10-18  2:53               ` Linus Torvalds
2008-10-18  5:20                 ` Nick Piggin
2008-10-18 10:38                   ` Peter Zijlstra
2008-10-19  9:52                     ` Hugh Dickins
2008-10-19 10:51                       ` Peter Zijlstra
2008-10-19 12:39                         ` Hugh Dickins
2008-10-19 18:25                         ` Linus Torvalds
2008-10-19 18:45                           ` Peter Zijlstra
2008-10-19 19:00                           ` Hugh Dickins
2008-10-20  4:03                           ` Hugh Dickins
2008-10-20 15:17                             ` Linus Torvalds
2008-10-20 18:21                               ` Hugh Dickins
2008-10-21  2:56                               ` Nick Piggin
2008-10-21  3:25                                 ` Linus Torvalds
2008-10-21  4:33                                   ` Nick Piggin
2008-10-21 12:58                                     ` Hugh Dickins
2008-10-21 15:59                                     ` Christoph Lameter
2008-10-22  9:29                                       ` Nick Piggin
2008-10-21  4:34                                   ` Nick Piggin
2008-10-21 13:55                                     ` Hugh Dickins
2008-10-21  2:44                           ` Nick Piggin
2008-10-18 19:14               ` Hugh Dickins
2008-10-19  3:03                 ` Nick Piggin
2008-10-19  7:07                   ` Hugh Dickins
2008-10-20  3:26                     ` Hugh Dickins
2008-10-21  2:45                       ` Nick Piggin
2008-10-19  1:13       ` Hugh Dickins
2008-10-19  2:41         ` Nick Piggin
2008-10-19  9:45           ` Hugh Dickins
2008-10-21  3:59             ` Nick Piggin

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=alpine.LFD.2.00.0810171621180.3438@nehalem.linux-foundation.org \
    --to=torvalds@linux-foundation.org \
    --cc=a.p.zijlstra@chello.nl \
    --cc=hugh@veritas.com \
    --cc=linux-mm@kvack.org \
    --cc=npiggin@suse.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox