IB/ipoib: Consolidate rtnl_lock tasks in workqueue
authorDoug Ledford <dledford@redhat.com>
Sun, 22 Feb 2015 00:27:01 +0000 (19:27 -0500)
committerDoug Ledford <dledford@redhat.com>
Wed, 15 Apr 2015 20:06:17 +0000 (16:06 -0400)
commitc84ca6d2b1a1bfbdab2dd2bc243840e9589a5aaf
tree90da63e83ca554c96167dfcd65998de7fd1440a5
parentbe7aa663fc1d9156798f5af3c60e6df45e1fe5de
IB/ipoib: Consolidate rtnl_lock tasks in workqueue

The ipoib_mcast_flush_dev routine is called with the rtnl_lock held and
needs to keep it held.  It also needs to call flush_workqueue() to flush
out any outstanding work.  In the past, we've had to try and make sure
that we didn't flush out any outstanding join completions because they
also wanted to grab rtnl_lock() and that would deadlock.  It turns out
that the only thing in the join completion handler that needs this lock
can be safely moved to our carrier_on_task, thereby reducing the
potential for the join completion code and the flush code to deadlock
against each other.

Signed-off-by: Doug Ledford <dledford@redhat.com>
drivers/infiniband/ulp/ipoib/ipoib_multicast.c