Bluetooth: Reduce critical section in sco_conn_ready
authorAndre Guedes <andre.guedes@openbossa.org>
Tue, 29 Jan 2013 22:59:56 +0000 (19:59 -0300)
committerGustavo Padovan <gustavo.padovan@collabora.co.uk>
Fri, 1 Feb 2013 17:50:18 +0000 (15:50 -0200)
commit405280887f8fb4e168a1bbc865917bb2b881db95
treea940fda4a0684f2ee33c99d36aa77237b3b1e6b6
parent3810285cf8cef5c3f9c4334a317b71b876125269
Bluetooth: Reduce critical section in sco_conn_ready

This patch reduces the critical section protected by sco_conn_lock in
sco_conn_ready function. The lock is acquired only when it is really
needed.

This patch fixes the following lockdep warning which is generated
when the host terminates a SCO connection.

Today, this warning is a false positive. There is no way those
two threads reported by lockdep are running at the same time since
hdev->workqueue (where rx_work is queued) is single-thread. However,
if somehow this behavior is changed in future, we will have a
potential deadlock.

======================================================
[ INFO: possible circular locking dependency detected ]
3.8.0-rc1+ #7 Not tainted
-------------------------------------------------------
kworker/u:1H/1018 is trying to acquire lock:
 (&(&conn->lock)->rlock){+.+...}, at: [<ffffffffa0033ba6>] sco_chan_del+0x66/0x190 [bluetooth]

but task is already holding lock:
 (slock-AF_BLUETOOTH-BTPROTO_SCO){+.+...}, at: [<ffffffffa0033d5a>] sco_conn_del+0x8a/0xe0 [bluetooth]

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (slock-AF_BLUETOOTH-BTPROTO_SCO){+.+...}:
       [<ffffffff81083011>] lock_acquire+0xb1/0xe0
       [<ffffffff813efd01>] _raw_spin_lock+0x41/0x80
       [<ffffffffa003436e>] sco_connect_cfm+0xbe/0x350 [bluetooth]
       [<ffffffffa0015d6c>] hci_event_packet+0xd3c/0x29b0 [bluetooth]
       [<ffffffffa0004583>] hci_rx_work+0x133/0x870 [bluetooth]
       [<ffffffff8104d65f>] process_one_work+0x2bf/0x4f0
       [<ffffffff81050022>] worker_thread+0x2b2/0x3e0
       [<ffffffff81056021>] kthread+0xd1/0xe0
       [<ffffffff813f14bc>] ret_from_fork+0x7c/0xb0

-> #0 (&(&conn->lock)->rlock){+.+...}:
       [<ffffffff81082215>] __lock_acquire+0x1465/0x1c70
       [<ffffffff81083011>] lock_acquire+0xb1/0xe0
       [<ffffffff813efd01>] _raw_spin_lock+0x41/0x80
       [<ffffffffa0033ba6>] sco_chan_del+0x66/0x190 [bluetooth]
       [<ffffffffa0033d6d>] sco_conn_del+0x9d/0xe0 [bluetooth]
       [<ffffffffa0034653>] sco_disconn_cfm+0x53/0x60 [bluetooth]
       [<ffffffffa000fef3>] hci_disconn_complete_evt.isra.54+0x363/0x3c0 [bluetooth]
       [<ffffffffa00150f7>] hci_event_packet+0xc7/0x29b0 [bluetooth]
       [<ffffffffa0004583>] hci_rx_work+0x133/0x870 [bluetooth]
       [<ffffffff8104d65f>] process_one_work+0x2bf/0x4f0
       [<ffffffff81050022>] worker_thread+0x2b2/0x3e0
       [<ffffffff81056021>] kthread+0xd1/0xe0
       [<ffffffff813f14bc>] ret_from_fork+0x7c/0xb0

other info that might help us debug this:

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(slock-AF_BLUETOOTH-BTPROTO_SCO);
                               lock(&(&conn->lock)->rlock);
                               lock(slock-AF_BLUETOOTH-BTPROTO_SCO);
  lock(&(&conn->lock)->rlock);

 *** DEADLOCK ***

4 locks held by kworker/u:1H/1018:
 #0:  (hdev->name#2){.+.+.+}, at: [<ffffffff8104d5f8>] process_one_work+0x258/0x4f0
 #1:  ((&hdev->rx_work)){+.+.+.}, at: [<ffffffff8104d5f8>] process_one_work+0x258/0x4f0
 #2:  (&hdev->lock){+.+.+.}, at: [<ffffffffa000fbe9>] hci_disconn_complete_evt.isra.54+0x59/0x3c0 [bluetooth]
 #3:  (slock-AF_BLUETOOTH-BTPROTO_SCO){+.+...}, at: [<ffffffffa0033d5a>] sco_conn_del+0x8a/0xe0 [bluetooth]

stack backtrace:
Pid: 1018, comm: kworker/u:1H Not tainted 3.8.0-rc1+ #7
Call Trace:
 [<ffffffff813e92f9>] print_circular_bug+0x1fb/0x20c
 [<ffffffff81082215>] __lock_acquire+0x1465/0x1c70
 [<ffffffff81083011>] lock_acquire+0xb1/0xe0
 [<ffffffffa0033ba6>] ? sco_chan_del+0x66/0x190 [bluetooth]
 [<ffffffff813efd01>] _raw_spin_lock+0x41/0x80
 [<ffffffffa0033ba6>] ? sco_chan_del+0x66/0x190 [bluetooth]
 [<ffffffffa0033ba6>] sco_chan_del+0x66/0x190 [bluetooth]
 [<ffffffffa0033d6d>] sco_conn_del+0x9d/0xe0 [bluetooth]
 [<ffffffffa0034653>] sco_disconn_cfm+0x53/0x60 [bluetooth]
 [<ffffffffa000fef3>] hci_disconn_complete_evt.isra.54+0x363/0x3c0 [bluetooth]
 [<ffffffffa000fbd0>] ? hci_disconn_complete_evt.isra.54+0x40/0x3c0 [bluetooth]
 [<ffffffffa00150f7>] hci_event_packet+0xc7/0x29b0 [bluetooth]
 [<ffffffff81202e90>] ? __dynamic_pr_debug+0x80/0x90
 [<ffffffff8133ff7d>] ? kfree_skb+0x2d/0x40
 [<ffffffffa0021644>] ? hci_send_to_monitor+0x1a4/0x1c0 [bluetooth]
 [<ffffffffa0004583>] hci_rx_work+0x133/0x870 [bluetooth]
 [<ffffffff8104d5f8>] ? process_one_work+0x258/0x4f0
 [<ffffffff8104d65f>] process_one_work+0x2bf/0x4f0
 [<ffffffff8104d5f8>] ? process_one_work+0x258/0x4f0
 [<ffffffff8104fdc1>] ? worker_thread+0x51/0x3e0
 [<ffffffffa0004450>] ? hci_tx_work+0x800/0x800 [bluetooth]
 [<ffffffff81050022>] worker_thread+0x2b2/0x3e0
 [<ffffffff8104fd70>] ? busy_worker_rebind_fn+0x100/0x100
 [<ffffffff81056021>] kthread+0xd1/0xe0
 [<ffffffff81055f50>] ? flush_kthread_worker+0xc0/0xc0
 [<ffffffff813f14bc>] ret_from_fork+0x7c/0xb0
 [<ffffffff81055f50>] ? flush_kthread_worker+0xc0/0xc0

Signed-off-by: Andre Guedes <andre.guedes@openbossa.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
net/bluetooth/sco.c