From: Gustavo Padovan Date: Fri, 15 Jun 2012 05:30:20 +0000 (-0300) Subject: Bluetooth: Fix possible deadlock in SCO code X-Git-Url: https://git.stricted.de/?a=commitdiff_plain;h=269c4845d5b3627b95b1934107251bacbe99bb68;p=GitHub%2FLineageOS%2Fandroid_kernel_motorola_exynos9610.git Bluetooth: Fix possible deadlock in SCO code sco_chan_del() only has conn != NULL when called from sco_conn_del() so just move the code from it that deal with conn to sco_conn_del(). [ 120.765529] [ 120.765529] ====================================================== [ 120.766529] [ INFO: possible circular locking dependency detected ] [ 120.766529] 3.5.0-rc1-10292-g3701f94-dirty #70 Tainted: G W [ 120.766529] ------------------------------------------------------- [ 120.766529] kworker/u:3/1497 is trying to acquire lock: [ 120.766529] (&(&conn->lock)->rlock#2){+.+...}, at: [] sco_chan_del+0x4c/0x170 [bluetooth] [ 120.766529] [ 120.766529] but task is already holding lock: [ 120.766529] (slock-AF_BLUETOOTH-BTPROTO_SCO){+.+...}, at: [] sco_conn_del+0x61/0xe0 [bluetooth] [ 120.766529] [ 120.766529] which lock already depends on the new lock. [ 120.766529] [ 120.766529] [ 120.766529] the existing dependency chain (in reverse order) is: [ 120.766529] [ 120.766529] -> #1 (slock-AF_BLUETOOTH-BTPROTO_SCO){+.+...}: [ 120.766529] [] lock_acquire+0x8e/0xb0 [ 120.766529] [] _raw_spin_lock+0x40/0x80 [ 120.766529] [] sco_connect_cfm+0x79/0x300 [bluetooth] [ 120.766529] [] hci_sync_conn_complete_evt.isra.90+0x343/0x400 [bluetooth] [ 120.766529] [] hci_event_packet+0x317/0xfb0 [bluetooth] [ 120.766529] [] hci_rx_work+0x2c8/0x890 [bluetooth] [ 120.766529] [] process_one_work+0x197/0x460 [ 120.766529] [] worker_thread+0x126/0x2d0 [ 120.766529] [] kthread+0x9d/0xb0 [ 120.766529] [] kernel_thread_helper+0x4/0x10 [ 120.766529] [ 120.766529] -> #0 (&(&conn->lock)->rlock#2){+.+...}: [ 120.766529] [] __lock_acquire+0x154a/0x1d30 [ 120.766529] [] lock_acquire+0x8e/0xb0 [ 120.766529] [] _raw_spin_lock+0x40/0x80 [ 120.766529] [] sco_chan_del+0x4c/0x170 [bluetooth] [ 120.766529] [] sco_conn_del+0x74/0xe0 [bluetooth] [ 120.766529] [] sco_disconn_cfm+0x32/0x60 [bluetooth] [ 120.766529] [] hci_disconn_complete_evt.isra.53+0x242/0x390 [bluetooth] [ 120.766529] [] hci_event_packet+0x617/0xfb0 [bluetooth] [ 120.766529] [] hci_rx_work+0x2c8/0x890 [bluetooth] [ 120.766529] [] process_one_work+0x197/0x460 [ 120.766529] [] worker_thread+0x126/0x2d0 [ 120.766529] [] kthread+0x9d/0xb0 [ 120.766529] [] kernel_thread_helper+0x4/0x10 [ 120.766529] [ 120.766529] other info that might help us debug this: [ 120.766529] [ 120.766529] Possible unsafe locking scenario: [ 120.766529] [ 120.766529] CPU0 CPU1 [ 120.766529] ---- ---- [ 120.766529] lock(slock-AF_BLUETOOTH-BTPROTO_SCO); [ 120.766529] lock(&(&conn->lock)->rlock#2); [ 120.766529] lock(slock-AF_BLUETOOTH-BTPROTO_SCO); [ 120.766529] lock(&(&conn->lock)->rlock#2); [ 120.766529] [ 120.766529] *** DEADLOCK *** Signed-off-by: Gustavo Padovan --- diff --git a/net/bluetooth/sco.c b/net/bluetooth/sco.c index 40bbe25dcff7..3589e21edb09 100644 --- a/net/bluetooth/sco.c +++ b/net/bluetooth/sco.c @@ -131,6 +131,15 @@ static int sco_conn_del(struct hci_conn *hcon, int err) sco_sock_clear_timer(sk); sco_chan_del(sk, err); bh_unlock_sock(sk); + + sco_conn_lock(conn); + conn->sk = NULL; + sco_pi(sk)->conn = NULL; + sco_conn_unlock(conn); + + if (conn->hcon) + hci_conn_put(conn->hcon); + sco_sock_kill(sk); } @@ -821,16 +830,6 @@ static void sco_chan_del(struct sock *sk, int err) BT_DBG("sk %p, conn %p, err %d", sk, conn, err); - if (conn) { - sco_conn_lock(conn); - conn->sk = NULL; - sco_pi(sk)->conn = NULL; - sco_conn_unlock(conn); - - if (conn->hcon) - hci_conn_put(conn->hcon); - } - sk->sk_state = BT_CLOSED; sk->sk_err = err; sk->sk_state_change(sk);