sysfs: fix use-after-free in sysfs_kill_sb()
While restructuring the [u]mount path, 4b93dc9b1c68 ("sysfs, kernfs:
prepare mount path for kernfs") incorrectly updated sysfs_kill_sb() so
that it first kills super_block and then tries to dereference its
namespace tag to drop it. Fix it by caching namespace tag before
killing the superblock and then drop the cached namespace tag.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Tested-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Tested-by: Vlastimil Babka <vbabka@suse.cz>
Link: http://lkml.kernel.org/g/20131205031051.GC5135@yliu-dev.sh.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
diff --git a/fs/sysfs/mount.c b/fs/sysfs/mount.c
index e7e3aa8..8d07527 100644
--- a/fs/sysfs/mount.c
+++ b/fs/sysfs/mount.c
@@ -45,8 +45,10 @@
static void sysfs_kill_sb(struct super_block *sb)
{
+ void *ns = (void *)kernfs_super_ns(sb);
+
kernfs_kill_sb(sb);
- kobj_ns_drop(KOBJ_NS_TYPE_NET, (void *)kernfs_super_ns(sb));
+ kobj_ns_drop(KOBJ_NS_TYPE_NET, ns);
}
static struct file_system_type sysfs_fs_type = {