When hitting the hard rfkill in 3.0-rc4, lockdep spots some likely lock misuse:
=======================================================
[ INFO: possible circular locking dependency detected ]
3.0.0-rc4-340c #1
-------------------------------------------------------
kworker/0:0/4 is trying to acquire lock:
(&rdev->mtx){+.+.+.}, at: [<ffffffff816cefce>]
cfg80211_netdev_notifier_call+0x11e/0x650
but task is already holding lock:
(&rdev->devlist_mtx){+.+.+.}, at: [<ffffffff816cff16>]
cfg80211_rfkill_set_block+0x46/0xa0
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #1 (&rdev->devlist_mtx){+.+.+.}:
[<ffffffff81096feb>] validate_chain.clone.23+0x54b/0x630
[<ffffffff81098084>] __lock_acquire+0x474/0x960
[<ffffffff81098af5>] lock_acquire+0x55/0x70
[<ffffffff81716a5e>] mutex_lock_nested+0x5e/0x390
[<ffffffff816cf307>] cfg80211_netdev_notifier_call+0x457/0x650
[<ffffffff81084f3b>] notifier_call_chain+0x8b/0x100
[<ffffffff81084fd1>] raw_notifier_call_chain+0x11/0x20
[<ffffffff815a73b2>] call_netdevice_notifiers+0x32/0x60
[<ffffffff815ab844>] __dev_notify_flags+0x34/0x90
[<ffffffff815ab8e0>] dev_change_flags+0x40/0x70
[<ffffffff815b9e8e>] do_setlink+0x17e/0x890
[<ffffffff815ba687>] rtnl_setlink+0xe7/0x130
[<ffffffff815badbf>] rtnetlink_rcv_msg+0x22f/0x260
[<ffffffff815d35a9>] netlink_rcv_skb+0xa9/0xd0
[<ffffffff815baaa0>] rtnetlink_rcv+0x20/0x30
[<ffffffff815d2e3e>] netlink_unicast+0x1ee/0x240
[<ffffffff815d30d1>] netlink_sendmsg+0x241/0x3b0
[<ffffffff8159184c>] sock_sendmsg+0xdc/0x120
[<ffffffff81591b48>] __sys_sendmsg+0x1d8/0x340
[<ffffffff815937b4>] sys_sendmsg+0x44/0x80
[<ffffffff817195bb>] system_call_fastpath+0x16/0x1b
-> #0 (&rdev->mtx){+.+.+.}:
[<ffffffff81096a8b>] check_prev_add+0x70b/0x720
[<ffffffff81096feb>] validate_chain.clone.23+0x54b/0x630
[<ffffffff81098084>] __lock_acquire+0x474/0x960
[<ffffffff81098af5>] lock_acquire+0x55/0x70
[<ffffffff81716a5e>] mutex_lock_nested+0x5e/0x390
[<ffffffff816cefce>] cfg80211_netdev_notifier_call+0x11e/0x650
[<ffffffff81084f3b>] notifier_call_chain+0x8b/0x100
[<ffffffff81084fd1>] raw_notifier_call_chain+0x11/0x20
[<ffffffff815a73b2>] call_netdevice_notifiers+0x32/0x60
[<ffffffff815a742d>] __dev_close_many+0x4d/0xf0
[<ffffffff815a75a8>] dev_close_many+0x88/0x110
[<ffffffff815a7668>] dev_close+0x38/0x50
[<ffffffff816cff3a>] cfg80211_rfkill_set_block+0x6a/0xa0
[<ffffffff816cff94>] cfg80211_rfkill_sync_work+0x24/0x30
[<ffffffff81076aa7>] process_one_work+0x1b7/0x450
[<ffffffff81079f51>] worker_thread+0x161/0x350
[<ffffffff8107ec06>] kthread+0xb6/0xc0
[<ffffffff8171a714>] kernel_thread_helper+0x4/0x10
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&rdev->devlist_mtx);
lock(&rdev->mtx);
lock(&rdev->devlist_mtx);
lock(&rdev->mtx);
*** DEADLOCK ***
4 locks held by kworker/0:0/4:
#0: (events){.+.+.+}, at: [<ffffffff81076a49>] process_one_work+0x159/0x450
#1: ((&rdev->rfkill_sync)){+.+...}, at: [<ffffffff81076a49>]
process_one_work+0x159/0x450
#2: (rtnl_mutex){+.+.+.}, at: [<ffffffff815baa72>] rtnl_lock+0x12/0x20
#3: (&rdev->devlist_mtx){+.+.+.}, at: [<ffffffff816cff16>]
cfg80211_rfkill_set_block+0x46/0xa0
stack backtrace:
Pid: 4, comm: kworker/0:0 Tainted: G C 3.0.0-rc4-340c #1
Call Trace:
[<ffffffff81095579>] print_circular_bug+0x109/0x110
[<ffffffff81096a8b>] check_prev_add+0x70b/0x720
[<ffffffff81096feb>] validate_chain.clone.23+0x54b/0x630
[<ffffffff81098084>] __lock_acquire+0x474/0x960
[<ffffffff810936ce>] ? __bfs+0x11e/0x260
[<ffffffff8109632f>] ? check_irq_usage+0x9f/0xf0
[<ffffffff816cefce>] ? cfg80211_netdev_notifier_call+0x11e/0x650
[<ffffffff81098af5>] lock_acquire+0x55/0x70
[<ffffffff816cefce>] ? cfg80211_netdev_notifier_call+0x11e/0x650
[<ffffffff81056fed>] ? add_preempt_count+0x9d/0xd0
[<ffffffff81716a5e>] mutex_lock_nested+0x5e/0x390
[<ffffffff816cefce>] ? cfg80211_netdev_notifier_call+0x11e/0x650
[<ffffffff81096feb>] ? validate_chain.clone.23+0x54b/0x630
[<ffffffff816cefce>] cfg80211_netdev_notifier_call+0x11e/0x650
[<ffffffff81098084>] ? __lock_acquire+0x474/0x960
[<ffffffff81096feb>] ? validate_chain.clone.23+0x54b/0x630
[<ffffffff81084f3b>] notifier_call_chain+0x8b/0x100
[<ffffffff81084fd1>] raw_notifier_call_chain+0x11/0x20
[<ffffffff815a73b2>] call_netdevice_notifiers+0x32/0x60
[<ffffffff815a742d>] __dev_close_many+0x4d/0xf0
[<ffffffff815a75a8>] dev_close_many+0x88/0x110
[<ffffffff815a7668>] dev_close+0x38/0x50
[<ffffffff816cff3a>] cfg80211_rfkill_set_block+0x6a/0xa0
[<ffffffff816cff94>] cfg80211_rfkill_sync_work+0x24/0x30
[<ffffffff81076aa7>] process_one_work+0x1b7/0x450
[<ffffffff81076a49>] ? process_one_work+0x159/0x450
[<ffffffff816cff70>] ? cfg80211_rfkill_set_block+0xa0/0xa0
[<ffffffff81079f51>] worker_thread+0x161/0x350
[<ffffffff81079df0>] ? manage_workers.clone.23+0x120/0x120
[<ffffffff8107ec06>] kthread+0xb6/0xc0
[<ffffffff810994ad>] ? trace_hardirqs_on_caller+0x13d/0x180
[<ffffffff8171a714>] kernel_thread_helper+0x4/0x10
[<ffffffff810498a7>] ? finish_task_switch+0x77/0x100
[<ffffffff81718e44>] ? retint_restore_args+0xe/0xe
[<ffffffff8107eb50>] ? __init_kthread_worker+0x70/0x70
[<ffffffff8171a710>] ? gs_change+0xb/0xb
--
Daniel J Blueman
On Tue, 2011-06-21 at 21:10 +0800, Daniel J Blueman wrote:
> When hitting the hard rfkill in 3.0-rc4, lockdep spots some likely lock misuse:
>
> =======================================================
> [ INFO: possible circular locking dependency detected ]
> 3.0.0-rc4-340c #1
> -------------------------------------------------------
> kworker/0:0/4 is trying to acquire lock:
> (&rdev->mtx){+.+.+.}, at: [<ffffffff816cefce>]
> cfg80211_netdev_notifier_call+0x11e/0x650
>
> but task is already holding lock:
> (&rdev->devlist_mtx){+.+.+.}, at: [<ffffffff816cff16>]
> cfg80211_rfkill_set_block+0x46/0xa0
>
> which lock already depends on the new lock.
This should be fixed with the patch for 3.0 that I have just sent,
"cfg80211: fix deadlock with rfkill/sched_scan by adding new mutex".
--
Cheers,
Luca.