IPoIB: Fix deadlock between dev_change_flags() and __ipoib_dev_flush()
authorErez Shitrit <erezsh@mellanox.com>
Wed, 16 Oct 2013 14:37:49 +0000 (17:37 +0300)
committerRoland Dreier <roland@purestorage.com>
Fri, 8 Nov 2013 22:42:49 +0000 (14:42 -0800)
commitf47944cc2dba3c7e6f753b81e9f713f4d12bdd5a
tree4dd24c53d8041eff9f251c9debf63bd54be083d8
parent22252b4e09f19431f82dc26428e0edf697d19a04
IPoIB: Fix deadlock between dev_change_flags() and __ipoib_dev_flush()

When ipoib interface is going down it takes all of its children with
it, under mutex.

For each child, dev_change_flags() is called.  That function calls
ipoib_stop() via the ndo, and causes flush of the workqueue.
Sometimes in the workqueue an __ipoib_dev_flush work() is waiting and
when invoked tries to get the same mutex, which leads to a deadlock,
as seen below.

The solution is to switch to rw-sem instead of mutex.

The deadlock:
[11028.165303]  [<ffffffff812b0977>] ? vgacon_scroll+0x107/0x2e0
[11028.171844]  [<ffffffff814eaac5>] schedule_timeout+0x215/0x2e0
[11028.178465]  [<ffffffff8105a5c3>] ? perf_event_task_sched_out+0x33/0x80
[11028.185962]  [<ffffffff814ea743>] wait_for_common+0x123/0x180
[11028.192491]  [<ffffffff8105fa40>] ? default_wake_function+0x0/0x20
[11028.199504]  [<ffffffff814ea85d>] wait_for_completion+0x1d/0x20
[11028.206224]  [<ffffffff8108b4f1>] flush_cpu_workqueue+0x61/0x90
[11028.212948]  [<ffffffff8108b5a0>] ? wq_barrier_func+0x0/0x20
[11028.219375]  [<ffffffff8108bfc4>] flush_workqueue+0x54/0x80
[11028.225712]  [<ffffffffa05a0576>] ipoib_mcast_stop_thread+0x66/0x90 [ib_ipoib]
[11028.233988]  [<ffffffffa059ccea>] ipoib_ib_dev_down+0x6a/0x100 [ib_ipoib]
[11028.241678]  [<ffffffffa059849a>] ipoib_stop+0x8a/0x140 [ib_ipoib]
[11028.248692]  [<ffffffff8142adf1>] dev_close+0x71/0xc0
[11028.254447]  [<ffffffff8142a631>] dev_change_flags+0xa1/0x1d0
[11028.261062]  [<ffffffffa059851b>] ipoib_stop+0x10b/0x140 [ib_ipoib]
[11028.268172]  [<ffffffff8142adf1>] dev_close+0x71/0xc0
[11028.273922]  [<ffffffff8142a631>] dev_change_flags+0xa1/0x1d0
[11028.280452]  [<ffffffff8148f20b>] devinet_ioctl+0x5eb/0x6a0
[11028.286786]  [<ffffffff814903b8>] inet_ioctl+0x88/0xa0
[11028.292633]  [<ffffffff8141591a>] sock_ioctl+0x7a/0x280
[11028.298576]  [<ffffffff81189012>] vfs_ioctl+0x22/0xa0
[11028.304326]  [<ffffffff81140540>] ? unmap_region+0x110/0x130
[11028.310756]  [<ffffffff811891b4>] do_vfs_ioctl+0x84/0x580
[11028.316897]  [<ffffffff81189731>] sys_ioctl+0x81/0xa0

and

11028.017533]  [<ffffffff8105a5c3>] ? perf_event_task_sched_out+0x33/0x80
[11028.025030]  [<ffffffff8100bb8e>] ? apic_timer_interrupt+0xe/0x20
[11028.031945]  [<ffffffff814eb2ae>] __mutex_lock_slowpath+0x13e/0x180
[11028.039053]  [<ffffffff814eb14b>] mutex_lock+0x2b/0x50
[11028.044910]  [<ffffffffa059f7e7>] __ipoib_ib_dev_flush+0x37/0x210 [ib_ipoib]
[11028.052894]  [<ffffffffa059fa00>] ? ipoib_ib_dev_flush_light+0x0/0x20 [ib_ipoib]
[11028.061363]  [<ffffffffa059fa17>] ipoib_ib_dev_flush_light+0x17/0x20 [ib_ipoib]
[11028.069738]  [<ffffffff8108b120>] worker_thread+0x170/0x2a0
[11028.076068]  [<ffffffff81090990>] ? autoremove_wake_function+0x0/0x40
[11028.083374]  [<ffffffff8108afb0>] ? worker_thread+0x0/0x2a0
[11028.089709]  [<ffffffff81090626>] kthread+0x96/0xa0
[11028.095266]  [<ffffffff8100c0ca>] child_rip+0xa/0x20
[11028.100921]  [<ffffffff81090590>] ? kthread+0x0/0xa0
[11028.106573]  [<ffffffff8100c0c0>] ? child_rip+0x0/0x20
[11028.112423] INFO: task ifconfig:23640 blocked for more than 120 seconds.

Signed-off-by: Erez Shitrit <erezsh@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
drivers/infiniband/ulp/ipoib/ipoib.h
drivers/infiniband/ulp/ipoib/ipoib_ib.c
drivers/infiniband/ulp/ipoib/ipoib_main.c
drivers/infiniband/ulp/ipoib/ipoib_netlink.c
drivers/infiniband/ulp/ipoib/ipoib_vlan.c