If an rconn peer fails to send a hello message, the version number doesn't get set. Later, if the peer delays long enough, the rconn attempts to send an echo request but assert-fails instead because it doesn't know what version to use. This fixes the problem.
To reproduce this problem: make sandbox ovs-vsctl add-br br0 ovs-vsctl set-controller br0 ptcp:12345 nc 127.0.0.1 12345 and wait 10 seconds for ovs-vswitchd to die. (Then exit the sandbox.) Reported-by: 张东亚 <fortitude.zh...@gmail.com> Signed-off-by: Ben Pfaff <b...@ovn.org> --- lib/rconn.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-) diff --git a/lib/rconn.c b/lib/rconn.c index 0c1812a..8a29864 100644 --- a/lib/rconn.c +++ b/lib/rconn.c @@ -558,19 +558,23 @@ run_ACTIVE(struct rconn *rc) { if (timed_out(rc)) { unsigned int base = MAX(rc->last_activity, rc->state_entered); - int version; - VLOG_DBG("%s: idle %u seconds, sending inactivity probe", rc->name, (unsigned int) (time_now() - base)); - version = rconn_get_version__(rc); - ovs_assert(version >= 0 && version <= 0xff); - /* Ordering is important here: rconn_send() can transition to BACKOFF, * and we don't want to transition back to IDLE if so, because then we * can end up queuing a packet with vconn == NULL and then *boom*. */ state_transition(rc, S_IDLE); - rconn_send__(rc, make_echo_request(version), NULL); + + /* Send an echo request if we can. (If version negotiation is not + * complete, that is, if we did not yet receive a "hello" message from + * the peer, we do not know the version to use, so we don't send + * anything.) */ + int version = rconn_get_version__(rc); + if (version >= 0 && version <= 0xff) { + rconn_send__(rc, make_echo_request(version), NULL); + } + return; } -- 2.10.2 _______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev