cfg80211: fix deadlock with rfkill/sched_scan by adding new mutex
authorLuciano Coelho <coelho@ti.com>
Thu, 30 Jun 2011 05:32:41 +0000 (08:32 +0300)
committerJohn W. Linville <linville@tuxdriver.com>
Tue, 5 Jul 2011 18:42:36 +0000 (14:42 -0400)
commitc10841ca722a0bc960dc541c51582773f9a24f98
treef1e5901db0a530cfd4008f718a908a86bddcb333
parent37000b305bff81bb1ee2f7f37b1319b670a08f76
cfg80211: fix deadlock with rfkill/sched_scan by adding new mutex

There was a deadlock when rfkill-blocking a wireless interface,
because we were locking the rdev mutex on NETDEV_GOING_DOWN to stop
sched_scans that were eventually running.  The rfkill block code was
already holding a mutex under rdev:

kernel: =======================================================
kernel: [ INFO: possible circular locking dependency detected ]
kernel: 3.0.0-rc1-00049-g1fa7b6a #57
kernel: -------------------------------------------------------
kernel: kworker/0:1/4525 is trying to acquire lock:
kernel: (&rdev->mtx){+.+.+.}, at: [<ffffffff8164c831>] cfg80211_netdev_notifier_call+0x131/0x5b0
kernel:
kernel: but task is already holding lock:
kernel: (&rdev->devlist_mtx){+.+.+.}, at: [<ffffffff8164dcef>] cfg80211_rfkill_set_block+0x4f/0xa0
kernel:
kernel: which lock already depends on the new lock.

To fix this, add a new mutex specifically for sched_scan, to protect
the sched_scan_req element in the rdev struct, instead of using the
global rdev mutex.

Reported-by: Duane Griffin <duaneg@dghda.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
net/wireless/core.c
net/wireless/core.h
net/wireless/nl80211.c
net/wireless/scan.c