RE: [PATCH net 4/4] hv_netvsc: avoid deadlocks between rtnl lock and netvsc_inject_disable()
From: Haiyang Zhang <haiyangz@microsoft.com>
Date: 2016-08-11 17:12:40
Also in:
lkml
-----Original Message-----
From: Vitaly Kuznetsov [mailto:vkuznets@redhat.com]
Sent: Thursday, August 11, 2016 6:59 AM
To: netdev@vger.kernel.org
Cc: devel@linuxdriverproject.org; linux-kernel@vger.kernel.org; Haiyang
Zhang [off-list ref]; KY Srinivasan [off-list ref]
Subject: [PATCH net 4/4] hv_netvsc: avoid deadlocks between rtnl lock and
netvsc_inject_disable()
Here is a deadlock scenario:
- netvsc_vf_up() schedules netvsc_notify_peers() work and quits.
- netvsc_vf_down() runs before netvsc_notify_peers() gets executed. As it
is being executed from netdev notifier chain we hold rtnl lock when we
get here.
- we enter netvsc_inject_disable() and loop and wait till
netvsc_notify_peers() drops vf_use_cnt.
- netvsc_notify_peers() starts on some other CPU but netdev_notify_peers()
will hang on rtnl_lock().
- deadlock!
Similar deadlocks are possible between netvsc_vf_{up,down}() and
netvsc_unregister_vf() as it also waits till vf_use_cnt drops to zero.
Instead of introducing additional synchronization I suggest we drop
gwrk.dwrk completely and call NETDEV_NOTIFY_PEERS directly. As we're
acting under rtnl lock this is legitimate.
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>Acked-by: Haiyang Zhang <haiyangz@microsoft.com>