[PATCH v3] Bluetooth: SCO: give the socket its own sco_conn reference

[Aldo Ariel Panzardo]

sco_conn_del() drops a reference it does not own. It takes one transient
reference via sco_conn_hold_unless_zero() and releases it with the
sco_conn_put() that follows sco_sock_hold(); the additional put in the
!sk branch releases a second one:

    conn = sco_conn_hold_unless_zero(conn);
    ...
    sk = sco_sock_hold(conn);
    sco_conn_unlock(conn);
    sco_conn_put(conn);

    if (!sk) {
            sco_conn_put(conn);
            return;
    }

When close() races the controller's Disconnection Complete, sco_chan_del()
clears conn->sk and drops the socket's reference while sco_conn_del() is
running. sco_conn_del() then sees sk == NULL, its own put drops the count
to zero and frees the conn, and the second put writes to the freed kref:

    BUG: KASAN: slab-use-after-free in sco_conn_put.part.0+0x1a/0x190
    Write of size 4 at addr ffff8881099dec74 by task kworker/u17:3/413
    Workqueue: hci1 hci_rx_work
    Call Trace:
     sco_conn_put.part.0+0x1a/0x190
     hci_disconn_complete_evt+0x1ee/0x3e0
     hci_event_packet+0x54a/0x650
     hci_rx_work+0x321/0x3d0
    Allocated by task 413:
     sco_conn_add+0x72/0x1a0
     sco_connect_cfm+0x88/0x670
    Freed by task 413:
     sco_conn_del.isra.0+0x3f/0xf0
     hci_disconn_complete_evt+0x1ee/0x3e0
    refcount_t: underflow; use-after-free.

The root cause is that the socket stores the connection without holding a
reference of its own. __sco_chan_add() does:

    sco_pi(sk)->conn = conn;

so the socket borrows whatever reference its caller happened to hold, and
the callers paper over that with ad-hoc holds and puts. Give the socket a
counted reference instead: __sco_chan_add() takes one and it is released
together with the channel (sco_chan_del()) and in sco_sock_destruct().
With the socket holding its own reference, sco_conn_del() no longer needs
the extra put and the redundant hold in sco_conn_ready() goes away.

Making the socket own its reference means the connection is now actually
freed on the error paths of sco_connect() where it used to leak, which in
turn runs sco_conn_free() and its hci_conn_drop(conn->hcon). To keep the
hci_conn accounting balanced, make that ownership explicit as well:
sco_conn_add() consumes one hci_conn reference and the sco_conn owns it for
its lifetime. sco_connect() hands over the reference returned by
hci_connect_sco() and no longer drops it on the error paths;
sco_connect_cfm(), which is not given a reference, takes one with
hci_conn_hold() before handing it to sco_conn_add() (and drops it again if
the allocation fails); and the explicit hci_conn_hold() in sco_conn_ready()
is removed. Every reference then has a single, clear owner.

Fixes: e6720779ae61 ("Bluetooth: SCO: Use kref to track lifetime of sco_conn")
Cc: stable@xxxxxxxxxxxxxxx
Suggested-by: Pauli Virtanen <pav@xxxxxx>
Signed-off-by: Aldo Ariel Panzardo <qwe.aldo@xxxxxxxxx>
---
v3:
 - Incorporate Pauli Virtanen's review: make sco_conn own one hci_conn
   reference for its whole lifetime -- sco_conn_add() consumes an hci_conn
   reference, sco_connect() no longer drops hcon on its error paths,
   sco_connect_cfm() holds one before sco_conn_add(), and the hci_conn_hold()
   in sco_conn_ready() is removed. This avoids the double hci_conn_drop() on
   sco_connect()'s error paths that v2 would otherwise introduce.
 - Keep the sco_pi(conn->sk)->conn = NULL clearing in sco_conn_free(). v2
   removed it as unreachable, but KASAN testing of close() racing the
   Disconnection Complete showed that dropping it reintroduces a
   use-after-free on the sco_sock_release() path, so it is retained.
 - Drop the hcon reference in sco_connect_cfm() when sco_conn_add() fails,
   so the allocation-failure path does not leak it.
v2:
 - Make the socket own its sco_conn reference rather than only deleting the
   extra put, per Pauli Virtanen's review.

Testing: on v7.2-rc4 with KASAN and a /dev/vhci reproducer that races
close() of an SCO socket against an injected Disconnection Complete, the
unpatched kernel hits the refcount_t underflow / use-after-free above
within a few thousand iterations; with this patch the sco_conn_del()
over-put on the Disconnection Complete path no longer reproduces.

 net/bluetooth/sco.c | 22 +++++++++++++---------
 1 file changed, 13 insertions(+), 9 deletions(-)

diff --git a/net/bluetooth/sco.c b/net/bluetooth/sco.c
index fcc597be5bbd..aa9f61a748ab 100644
--- a/net/bluetooth/sco.c
+++ b/net/bluetooth/sco.c
@@ -188,6 +188,9 @@ static void sco_sock_clear_timer(struct sock *sk)
 }
 
 /* ---- SCO connections ---- */
+/* Consumes a reference on @hcon, which the returned sco_conn owns until it is
+ * freed. On failure (NULL return) the reference is left for the caller to drop.
+ */
 static struct sco_conn *sco_conn_add(struct hci_conn *hcon)
 {
 	struct sco_conn *conn = hcon->sco_data;
@@ -198,6 +201,9 @@ static struct sco_conn *sco_conn_add(struct hci_conn *hcon)
 			sco_conn_lock(conn);
 			conn->hcon = hcon;
 			sco_conn_unlock(conn);
+		} else {
+			/* conn already owns a reference on hcon */
+			hci_conn_drop(hcon);
 		}
 		return conn;
 	}
@@ -265,10 +271,8 @@ static void sco_conn_del(struct hci_conn *hcon, int err)
 	sco_conn_unlock(conn);
 	sco_conn_put(conn);
 
-	if (!sk) {
-		sco_conn_put(conn);
+	if (!sk)
 		return;
-	}
 
 	/* Kill socket */
 	lock_sock(sk);
@@ -283,7 +287,7 @@ static void __sco_chan_add(struct sco_conn *conn, struct sock *sk,
 {
 	BT_DBG("conn %p", conn);
 
-	sco_pi(sk)->conn = conn;
+	sco_pi(sk)->conn = sco_conn_hold(conn);
 	conn->sk = sk;
 
 	if (parent)
@@ -366,15 +370,15 @@ static int sco_connect(struct sock *sk)
 	 */
 	if (sk->sk_state != BT_OPEN && sk->sk_state != BT_BOUND) {
 		release_sock(sk);
-		hci_conn_drop(hcon);
+		sco_conn_put(conn);
 		err = -EBADFD;
 		goto unlock;
 	}
 
 	err = sco_chan_add(conn, sk, NULL);
+	sco_conn_put(conn);
 	if (err) {
 		release_sock(sk);
-		hci_conn_drop(hcon);
 		goto unlock;
 	}
 
@@ -1439,8 +1443,6 @@ static void sco_conn_ready(struct sco_conn *conn)
 		bacpy(&sco_pi(sk)->src, &conn->hcon->src);
 		bacpy(&sco_pi(sk)->dst, &conn->hcon->dst);
 
-		sco_conn_hold(conn);
-		hci_conn_hold(conn->hcon);
 		__sco_chan_add(conn, sk, parent);
 
 		if (test_bit(BT_SK_DEFER_SETUP, &bt_sk(parent)->flags))
@@ -1496,10 +1498,12 @@ static void sco_connect_cfm(struct hci_conn *hcon, __u8 status)
 	if (!status) {
 		struct sco_conn *conn;
 
-		conn = sco_conn_add(hcon);
+		conn = sco_conn_add(hci_conn_hold(hcon));
 		if (conn) {
 			sco_conn_ready(conn);
 			sco_conn_put(conn);
+		} else {
+			hci_conn_drop(hcon);
 		}
 	} else
 		sco_conn_del(hcon, bt_to_errno(status));
-- 
2.43.0