STABLE12-rx-cleanup-deadlock-and-refcnt-leak-20020121

The first part of the patch fixes a likely deadlock in the
case when we're destroying a client connection with active
calls.

The second part fixes a minor connection refcount leak in
my recent modification to protect against half-reachable
clients.

diff --git a/src/rx/rx.c b/src/rx/rx.c
index 21d456ba3472fe1a9538ac7358f78b968af5ae6e..f4c255da7af42fd4c70f22916fa0aafbdc0cf0ce 100644 (file)
--- a/src/rx/rx.c
+++ b/src/rx/rx.c
@@ -899,7 +899,7 @@ static void rxi_DestroyConnectionNoLock(conn)
                                   RX_CALL_REFCOUNT_DELAY);
                    if (call->state == RX_STATE_PRECALL ||
                        call->state == RX_STATE_ACTIVE) {
-                       rxi_SendDelayedAck(call->delayedAckEvent, call, 0);
+                       rxi_SendAck(call, 0, 0, 0, 0, RX_ACK_DELAY, 0);
                    } else {
                        rxi_AckAll((struct rxevent *)0, call, 0);
                    }