User namespaces: Only put the userns when we unhash the uid
Serge E. Hallyn [Fri, 13 Feb 2009 14:04:21 +0000 (14:04 +0000)]
uids in namespaces other than init don't get a sysfs entry.

For those in the init namespace, while we're waiting to remove
the sysfs entry for the uid the uid is still hashed, and
alloc_uid() may re-grab that uid without getting a new
reference to the user_ns, which we've already put in free_user
before scheduling remove_user_sysfs_dir().

Reported-and-tested-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Serge E. Hallyn <serue@us.ibm.com>
Acked-by: David Howells <dhowells@redhat.com>
Tested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/user.c

index 477b666..3551ac7 100644 (file)
@@ -72,6 +72,7 @@ static void uid_hash_insert(struct user_struct *up, struct hlist_head *hashent)
 static void uid_hash_remove(struct user_struct *up)
 {
        hlist_del_init(&up->uidhash_node);
+       put_user_ns(up->user_ns);
 }
 
 static struct user_struct *uid_hash_find(uid_t uid, struct hlist_head *hashent)
@@ -334,7 +335,6 @@ static void free_user(struct user_struct *up, unsigned long flags)
        atomic_inc(&up->__count);
        spin_unlock_irqrestore(&uidhash_lock, flags);
 
-       put_user_ns(up->user_ns);
        INIT_WORK(&up->work, remove_user_sysfs_dir);
        schedule_work(&up->work);
 }
@@ -357,7 +357,6 @@ static void free_user(struct user_struct *up, unsigned long flags)
        sched_destroy_user(up);
        key_put(up->uid_keyring);
        key_put(up->session_keyring);
-       put_user_ns(up->user_ns);
        kmem_cache_free(uid_cachep, up);
 }