]> www.infradead.org Git - users/jedix/linux-maple.git/commitdiff
ksmbd: fix use-after-free in SMB request handling
authorYunseong Kim <yskelg@gmail.com>
Mon, 25 Nov 2024 07:45:55 +0000 (16:45 +0900)
committerSteve French <stfrench@microsoft.com>
Tue, 26 Nov 2024 00:58:27 +0000 (18:58 -0600)
A race condition exists between SMB request handling in
`ksmbd_conn_handler_loop()` and the freeing of `ksmbd_conn` in the
workqueue handler `handle_ksmbd_work()`. This leads to a UAF.
- KASAN: slab-use-after-free Read in handle_ksmbd_work
- KASAN: slab-use-after-free in rtlock_slowlock_locked

This race condition arises as follows:
- `ksmbd_conn_handler_loop()` waits for `conn->r_count` to reach zero:
  `wait_event(conn->r_count_q, atomic_read(&conn->r_count) == 0);`
- Meanwhile, `handle_ksmbd_work()` decrements `conn->r_count` using
  `atomic_dec_return(&conn->r_count)`, and if it reaches zero, calls
  `ksmbd_conn_free()`, which frees `conn`.
- However, after `handle_ksmbd_work()` decrements `conn->r_count`,
  it may still access `conn->r_count_q` in the following line:
  `waitqueue_active(&conn->r_count_q)` or `wake_up(&conn->r_count_q)`
  This results in a UAF, as `conn` has already been freed.

The discovery of this UAF can be referenced in the following PR for
syzkaller's support for SMB requests.
Link: https://github.com/google/syzkaller/pull/5524
Fixes: ee426bfb9d09 ("ksmbd: add refcnt to ksmbd_conn struct")
Cc: linux-cifs@vger.kernel.org
Cc: stable@vger.kernel.org # v6.6.55+, v6.10.14+, v6.11.3+
Cc: syzkaller@googlegroups.com
Signed-off-by: Yunseong Kim <yskelg@gmail.com>
Acked-by: Namjae Jeon <linkinjeon@kernel.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
fs/smb/server/server.c

index b3dceefe6c5f7f2c49f6fcba6cd7b9f2cdf97631..930d7566b52efdd624051b1cf0a0cc06be0254ce 100644 (file)
@@ -276,8 +276,12 @@ static void handle_ksmbd_work(struct work_struct *wk)
         * disconnection. waitqueue_active is safe because it
         * uses atomic operation for condition.
         */
+       atomic_inc(&conn->refcnt);
        if (!atomic_dec_return(&conn->r_count) && waitqueue_active(&conn->r_count_q))
                wake_up(&conn->r_count_q);
+
+       if (atomic_dec_and_test(&conn->refcnt))
+               kfree(conn);
 }
 
 /**