]> git.openfabrics.org - ~shefty/rdma-dev.git/commit
tipc: Prevent broadcast link stalling when another node fails
authorAllan Stephens <allan.stephens@windriver.com>
Thu, 7 Apr 2011 17:05:25 +0000 (13:05 -0400)
committerPaul Gortmaker <paul.gortmaker@windriver.com>
Thu, 1 Sep 2011 15:16:36 +0000 (11:16 -0400)
commit169073db442cb9e5aa2b70a2e4158d4f35a3b810
tree7ba297dce61f7374d0199e193d577903b0798771
parentc5bd4d85d356199ebdbc2c8bbfff86a292c65a9f
tipc: Prevent broadcast link stalling when another node fails

Ensure that broadcast link messages that have not been acknowledged
by a newly failed node do not get an implied acknowledgement until the
failed node is removed from the broadcast link's map of reachable nodes.

Previously, a race condition allowed a new broadcast link message to be
sent after the implicit acknowledgement processing was completed, but
before the map of reachable nodes was updated, resulting in the message
having an expected acknowledgement count that required the failed node
to explicitly acknowledge the message. Since this would never occur
the new message would remain in the broadcast link's transmit queue
forever, eventually causing the link to become congested and "stall".
Delaying the implicit acknowledgement processing until after the update
of the map of reachable nodes eliminates this race condition and prevents
stalling.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
net/tipc/node.c