avoid retrying a server if its address is unusable

Previously, if there's no more server to try in force_next,
it attempted to retry the current server similar to a query
timeout case. But, force_next() is called only when the
current server's address is unusable and the query was not
even sent, so recvcount was not incremented, causing an
assertion failure.

We might be able to fix it so the retry doesn't cause a crash,
but it doesn't make sense to retry the server when its address
is known to be unusable. So, simply avoiding the retry would be
the easiest and safest way to prevent the crash.

(cherry picked from commit 022bf31fabe70e0555bba3aac98f752c79e3b3cc)

diff --git a/bin/dig/dighost.c b/bin/dig/dighost.c
index d496f72ad585473b79ab50ca01c06b5858442bda..0bb64fc87b405131aefe4f5c02531c419ee0163d 100644 (file)
--- a/bin/dig/dighost.c
+++ b/bin/dig/dighost.c
@@ -3438,19 +3438,11 @@ force_next(dig_query_t *query) {
                return;
        }
 
-       if (l->retries > 1) {
-               l->retries--;
-               debug("making new %s request, %d tries left",
-                     l->tcp_mode ? "TCP" : "UDP", l->retries);
-               requeue_lookup(l, true);
-               lookup_detach(&l);
-               isc_refcount_decrement0(&recvcount);
-               debug("recvcount=%" PRIuFAST32,
-                     isc_refcount_current(&recvcount));
-               query_detach(&query);
-               clear_current_lookup();
-               return;
-       }
+       /*
+        * We don't retry this server unlike other similar cases such as
+        * recv_done() because we're here due to get_address() failure,
+        * which shouldn't be transient.
+        */
 
        if (query->readhandle != NULL) {
                isc_refcount_decrement0(&recvcount);