From: Mel Gorman <mgorman@suse.de>
To: Tommi Rantala <tt.rantala@gmail.com>
Cc: Stable <stable@vger.kernel.org>, Andi Kleen <ak@linux.intel.com>,
Andrew Morton <akpm@linux-foundation.org>,
KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>,
Dave Jones <davej@redhat.com>, Christoph Lameter <cl@linux.com>,
Hugh Dickins <hughd@google.com>,
LKML <linux-kernel@vger.kernel.org>,
Linux-MM <linux-mm@kvack.org>
Subject: Re: [PATCH 5/5] mempolicy: fix a memory corruption by refcount imbalance in alloc_pages_vma()
Date: Tue, 4 Dec 2012 14:15:43 +0000 [thread overview]
Message-ID: <20121204141501.GA2797@suse.de> (raw)
In-Reply-To: <CA+ydwtqQ7iK_1E+7ctLxYe8JZY+SzMfuRagjyHJ12OYsxbMcaA@mail.gmail.com>
On Tue, Dec 04, 2012 at 02:54:08PM +0200, Tommi Rantala wrote:
> 2012/10/9 Mel Gorman <mgorman@suse.de>:
> > commit 00442ad04a5eac08a98255697c510e708f6082e2 upstream.
> >
> > Commit cc9a6c877661 ("cpuset: mm: reduce large amounts of memory barrier
> > related damage v3") introduced a potential memory corruption.
> > shmem_alloc_page() uses a pseudo vma and it has one significant unique
> > combination, vma->vm_ops=NULL and vma->policy->flags & MPOL_F_SHARED.
> >
> > get_vma_policy() does NOT increase a policy ref when vma->vm_ops=NULL
> > and mpol_cond_put() DOES decrease a policy ref when a policy has
> > MPOL_F_SHARED. Therefore, when a cpuset update race occurs,
> > alloc_pages_vma() falls in 'goto retry_cpuset' path, decrements the
> > reference count and frees the policy prematurely.
>
> Hello,
>
> kmemleak is complaining about memory leaks that point to the mbind()
> syscall. I've seen this only in v3.7-rcX, so I bisected this, and
> found that this patch is the first mainline commit where I'm able to
> reproduce it with Trinity.
>
Uncool.
I'm writing this from an airport so am not in the position to test properly
but at a glance I'm not seeing what drops the reference count taken by
mpol_shared_policy_lookup() in all cases. vm_ops->get_policy() probably
gets it right but what about shmem_alloc_page() and shmem_swapin()?
This patch is only compile tested. If the reference counts are dropped
somewhere I did not spot quickly then it'll cause a use-after-free bug
instead but is worth trying anyway.
diff --git a/mm/shmem.c b/mm/shmem.c
index 89341b6..6229a43 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -912,6 +912,7 @@ static struct page *shmem_swapin(swp_entry_t swap, gfp_t gfp,
{
struct mempolicy mpol, *spol;
struct vm_area_struct pvma;
+ struct page *page;
spol = mpol_cond_copy(&mpol,
mpol_shared_policy_lookup(&info->policy, index));
@@ -922,13 +923,19 @@ static struct page *shmem_swapin(swp_entry_t swap, gfp_t gfp,
pvma.vm_pgoff = index + info->vfs_inode.i_ino;
pvma.vm_ops = NULL;
pvma.vm_policy = spol;
- return swapin_readahead(swap, gfp, &pvma, 0);
+ page = swapin_readahead(swap, gfp, &pvma, 0);
+
+ /* Drop reference taken by mpol_shared_policy_lookup() */
+ mpol_cond_put(pvma.vm_policy);
+
+ return page;
}
static struct page *shmem_alloc_page(gfp_t gfp,
struct shmem_inode_info *info, pgoff_t index)
{
struct vm_area_struct pvma;
+ struct page *page;
/* Create a pseudo vma that just contains the policy */
pvma.vm_start = 0;
@@ -940,7 +947,12 @@ static struct page *shmem_alloc_page(gfp_t gfp,
/*
* alloc_page_vma() will drop the shared policy reference
*/
- return alloc_page_vma(gfp, &pvma, 0);
+ page = alloc_page_vma(gfp, &pvma, 0);
+
+ /* Drop reference taken by mpol_shared_policy_lookup() */
+ mpol_cond_put(pvma.vm_policy);
+
+ return page;
}
#else /* !CONFIG_NUMA */
#ifdef CONFIG_TMPFS
next prev parent reply other threads:[~2012-12-04 14:24 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-10-09 16:58 [PATCH 0/5] Memory policy corruption fixes -stable Mel Gorman
2012-10-09 16:58 ` [PATCH 1/5] revert "mm: mempolicy: Let vma_merge and vma_split handle vma->vm_policy linkages" Mel Gorman
2012-10-09 16:58 ` [PATCH 2/5] mempolicy: remove mempolicy sharing Mel Gorman
2012-10-09 16:58 ` [PATCH 3/5] mempolicy: fix a race in shared_policy_replace() Mel Gorman
2012-10-09 16:58 ` [PATCH 4/5] mempolicy: fix refcount leak in mpol_set_shared_policy() Mel Gorman
2012-10-09 16:58 ` [PATCH 5/5] mempolicy: fix a memory corruption by refcount imbalance in alloc_pages_vma() Mel Gorman
2012-12-04 12:54 ` Tommi Rantala
2012-12-04 14:15 ` Mel Gorman [this message]
2012-12-05 5:11 ` Hugh Dickins
2012-12-05 6:28 ` Hugh Dickins
2012-12-05 7:24 ` [PATCH] tmpfs: fix shared mempolicy leak Hugh Dickins
2012-12-05 9:52 ` Mel Gorman
2012-12-05 20:25 ` Tommi Rantala
2012-12-05 21:59 ` Hugh Dickins
2012-12-05 22:01 ` Hugh Dickins
2012-10-10 0:47 ` [PATCH 0/5] Memory policy corruption fixes -stable Greg KH
2012-10-14 9:13 ` Ben Hutchings
-- strict thread matches above, loose matches on Subject: below --
2012-08-20 16:36 [PATCH 0/5] Memory policy corruption fixes V2 Mel Gorman
2012-08-20 16:36 ` [PATCH 5/5] mempolicy: fix a memory corruption by refcount imbalance in alloc_pages_vma() Mel Gorman
2012-08-20 19:51 ` Christoph Lameter
2012-08-21 7:26 ` Mel Gorman
2012-08-21 15:37 ` Christoph Lameter
2012-09-07 23:06 ` KOSAKI Motohiro
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=20121204141501.GA2797@suse.de \
--to=mgorman@suse.de \
--cc=ak@linux.intel.com \
--cc=akpm@linux-foundation.org \
--cc=cl@linux.com \
--cc=davej@redhat.com \
--cc=hughd@google.com \
--cc=kosaki.motohiro@jp.fujitsu.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=stable@vger.kernel.org \
--cc=tt.rantala@gmail.com \
/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;
as well as URLs for NNTP newsgroup(s).