Tyler Hicks <tyhicks@xxxxxxxxxxxxxxxxxx> writes:
> On Mon May 09, 2011 at 12:07:07PM -0700, Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
>> I assume this is a post-2.6.38 regression.
>>
>> I can't begin to think what might cause this. Is it reproducible?
>
> I'd bet on e7c0a167860620bd2938366896964f729ddaeaaa
>
> eCryptfs uses lookup_one_len() to lookup lower files, which means that
> the lower filesystem's d_revalidate() can get a NULL nameidata pointer.
> That commit dropped the check on nd before dereferencing it.
Looks like you hit the nail right on the head.
Following patch should fix it.
Thanks,
Miklos
commit d24339059d640f108c08ba99ef30e3bafa10f8e4
Author: Miklos Szeredi <mszeredi@xxxxxxx>
Date: Tue May 10 17:35:58 2011 +0200
fuse: fix oops in revalidate when called with NULL nameidata
Some cases (e.g. ecryptfs) can call ->dentry_revalidate with NULL
nameidata.
https://bugzilla.kernel.org/show_bug.cgi?id=34732
Tyler Hicks pointed out that this bug was introduced by commit
e7c0a16786 "fuse: make fuse_dentry_revalidate() RCU aware"
Reported-by: Witold Baryluk <baryluk@xxxxxxxxxxxxxxxx>
Signed-off-by: Miklos Szeredi <mszeredi@xxxxxxx>
diff --git a/fs/fuse/dir.c b/fs/fuse/dir.c
index c6ba49b..b32eb29 100644
--- a/fs/fuse/dir.c
+++ b/fs/fuse/dir.c
@@ -174,7 +174,7 @@ static int fuse_dentry_revalidate(struct dentry *entry, struct nameidata *nd)
if (!inode)
return 0;
- if (nd->flags & LOOKUP_RCU)
+ if (nd && (nd->flags & LOOKUP_RCU))
return -ECHILD;
fc = get_fuse_conn(inode);
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
