lkml.org 
[lkml]   [2008]   [Sep]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: 2.6.27-rc7-sha1: EIP at proc_sys_compare+0x36/0x50


On Sun, 28 Sep 2008, Al Viro wrote:
>
> > Al, did I miss something?
>
> The real underlying bug, whatever it is. If this sucker ever becomes
> negative, we have a big problem. Where _could_ that happen? Remember,
> we do not allow ->rmdir() and ->unlink() to succeed there.

What about pure memory pressure? We're holding only the RCU read-side lock
when looking up dentries, and if there is any memory pressure, the
dentries may be unhashed and the inodes removed in parallel. Yes, yes, we
end up not actually _releasing_ the dentry, since it's all RCU, but it
will set D_UNHASHED and be scheduled for releasing later under RCU.

And d_compare() is called before we have done any validation that the name
is still active, including checking whether it even got released already!

I dunno. Do we want to move the D_UNHASHED check up earlier? Or am I still
missing something?

Linus

----
fs/dcache.c | 10 ++++++----
1 files changed, 6 insertions(+), 4 deletions(-)
diff --git a/fs/dcache.c b/fs/dcache.c
index 80e9395..e7a1a99 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1395,6 +1395,10 @@ struct dentry * __d_lookup(struct dentry * parent, struct qstr * name)
if (dentry->d_parent != parent)
goto next;

+ /* non-existing due to RCU? */
+ if (d_unhashed(dentry))
+ goto next;
+
/*
* It is safe to compare names since d_move() cannot
* change the qstr (protected by d_lock).
@@ -1410,10 +1414,8 @@ struct dentry * __d_lookup(struct dentry * parent, struct qstr * name)
goto next;
}

- if (!d_unhashed(dentry)) {
- atomic_inc(&dentry->d_count);
- found = dentry;
- }
+ atomic_inc(&dentry->d_count);
+ found = dentry;
spin_unlock(&dentry->d_lock);
break;
next:

\
 
 \ /
  Last update: 2008-09-28 22:49    [from the cache]
©2003-2011 Jasper Spaans. Advertise on this site