[ovs-discuss] deadlock in 2.0.0 release

Xu (Simon) Chen xchenum at gmail.com
Thu Mar 20 16:19:35 UTC 2014


I experienced a deadlock while using openvswitch 2.0.0 release.

[2019033.190243] INFO: task kworker/u97:1:28426 blocked for more than 120
seconds.
[2019033.256661]  [<ffffffff814b685a>] mutex_lock+0x2a/0x50
[2019033.262653]  [<ffffffffa06ee855>] ovs_lock+0x15/0x20 [openvswitch]
[2019033.269795]  [<ffffffffa06f00e5>] ovs_exit_net+0x35/0x90 [openvswitch]
[2019033.277335]  [<ffffffff813ac529>] ops_exit_list.isra.3+0x39/0x60
[2019033.284291]  [<ffffffff813acea0>] cleanup_net+0x100/0x1a0
[2019033.290566]  [<ffffffff8105f065>] process_one_work+0x185/0x430
[2019033.297328]  [<ffffffff8105f7a2>] worker_thread+0x122/0x370

** cleanup_net() holds net_mutex and ovs_lock() is trying to grab ovs_mutex.

[2019153.562260] INFO: task ovs-vswitchd:41085 blocked for more than 120
seconds.
[2019153.636753]  [<ffffffff814b685a>] mutex_lock+0x2a/0x50
[2019153.642726]  [<ffffffff813accb4>] register_pernet_device+0x24/0x70
[2019153.649872]  [<ffffffffa06fa0e4>] vxlan_sock_add+0x244/0x310
[openvswitch]
[2019033.021548]  [<ffffffffa06f8599>] vxlan_tnl_create+0xc9/0x100
[openvswitch]
[2019033.029585]  [<ffffffffa06f62a2>] ovs_vport_add+0x52/0xb0 [openvswitch]
[2019033.037231]  [<ffffffffa06ee016>] new_vport+0x16/0x60 [openvswitch]
[2019033.044483]  [<ffffffffa06f089a>] ovs_vport_cmd_new+0x10a/0x1c0
[openvswitch]

** ovs_vport_cmd_new() holds ovs_mutex, and register_pernet_device tries to
grab net_mutex.


I noticed that in the newer 2.1.0 OVS release, vxlan_sock_add is done
differently. I am just wondering if this bug would be eliminated.

Thanks.
-Simon
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20140320/0e85a827/attachment-0002.html>


More information about the discuss mailing list