DiG: use the same retry and fail-over logic for different failure types

DiG implements different logic in the `recv_done()` callback function
when processing a failure:

1. For a timed-out query it applies the "retries" logic first, then,
   when it fails, fail-overs to the next server.

2. For an EOF (end-of-file, or unexpected disconnect) error it tries to
   make a single retry attempt (even if the user has requested more
   retries), then, when it fails, fail-overs to the next server.

3. For other types of failures, DiG does not apply the "retries" logic,
   and tries to fail-over to the next servers (again, even if the user
   has requested to make retries).

Simplify the logic and apply the same logic (1) of first retries, and
then fail-over, for different types of failures in `recv_done()`.

(cherry picked from commit abfd0d363f4f495de00dc4bd41f74726370369ba)

diff --git a/bin/dig/dighost.c b/bin/dig/dighost.c
index 93b17598a3f2226e34d78585e4441dc542644e01..b1b2b605e6109104ab15b452a5c06bca11ca701d 100644 (file)
--- a/bin/dig/dighost.c
+++ b/bin/dig/dighost.c
@@ -811,7 +811,6 @@ clone_lookup(dig_lookup_t *lookold, bool servers) {
        looknew->done_as_is = lookold->done_as_is;
        looknew->dscp = lookold->dscp;
        looknew->rrcomments = lookold->rrcomments;
-       looknew->eoferr = lookold->eoferr;
 
        if (lookold->ecs_addr != NULL) {
                size_t len = sizeof(isc_sockaddr_t);
@@ -3367,8 +3366,7 @@ force_next(dig_query_t *query) {
                dighost_error("no response from %s\n", buf);
        } else {
                printf("%s", l->cmdline);
-               dighost_error("connection timed out; "
-                             "no servers could be reached\n");
+               dighost_error("no servers could be reached\n");
        }
 
        if (exitcode < 9) {
@@ -3382,28 +3380,6 @@ force_next(dig_query_t *query) {
        UNLOCK_LOOKUP;
 }
 
-/*%
- * Called when a peer closes a TCP socket prematurely.
- */
-static void
-requeue_or_update_exitcode(dig_lookup_t *lookup) {
-       if (lookup->eoferr == 0U && lookup->retries > 1) {
-               --lookup->retries;
-               /*
-                * Peer closed the connection prematurely for the first time
-                * for this lookup.  Try again, keeping track of this failure.
-                */
-               dig_lookup_t *requeued_lookup = requeue_lookup(lookup, true);
-               requeued_lookup->eoferr++;
-       } else {
-               /*
-                * Peer closed the connection prematurely and it happened
-                * previously for this lookup.  Indicate an error.
-                */
-               exitcode = 9;
-       }
-}
-
 /*%
  * For transfers that involve multiple recvs (XFR's in particular),
  * launch the next recv.
@@ -3968,7 +3944,13 @@ recv_done(isc_nmhandle_t *handle, isc_result_t eresult, isc_region_t *region,
                }
        }
 
-       if (eresult == ISC_R_TIMEDOUT) {
+       if (eresult != ISC_R_SUCCESS) {
+               char sockstr[ISC_SOCKADDR_FORMATSIZE];
+
+               isc_sockaddr_format(&query->sockaddr, sockstr, sizeof(sockstr));
+               dighost_warning("communications error to %s: %s", sockstr,
+                               isc_result_totext(eresult));
+
                if (l->retries > 1 && !l->tcp_mode) {
                        dig_query_t *newq = NULL;
 
@@ -4039,8 +4021,8 @@ recv_done(isc_nmhandle_t *handle, isc_result_t eresult, isc_region_t *region,
                         * and cancel the lookup.
                         */
                        printf("%s", l->cmdline);
-                       dighost_error("connection timed out; "
-                                     "no servers could be reached\n");
+                       dighost_error("no servers could be reached\n");
+
                        if (exitcode < 9) {
                                exitcode = 9;
                        }
@@ -4051,52 +4033,6 @@ recv_done(isc_nmhandle_t *handle, isc_result_t eresult, isc_region_t *region,
 
                        goto cancel_lookup;
                }
-       } else if (eresult != ISC_R_SUCCESS) {
-               dig_query_t *next = ISC_LIST_NEXT(query, link);
-               char sockstr[ISC_SOCKADDR_FORMATSIZE];
-               isc_sockaddr_format(&query->sockaddr, sockstr, sizeof(sockstr));
-
-               /*
-                * There was a communication error with the current query,
-                * go to the next query, if there is one.
-                */
-               if (next != NULL) {
-                       if (l->current_query == query) {
-                               query_detach(&l->current_query);
-                       }
-                       if (l->current_query == NULL) {
-                               debug("starting next query %p", next);
-                               if (l->tcp_mode) {
-                                       start_tcp(next);
-                               } else {
-                                       start_udp(next);
-                               }
-                       }
-                       if (check_if_queries_done(l, query)) {
-                               goto cancel_lookup;
-                       }
-
-                       goto detach_query;
-               }
-
-               /*
-                * Otherwise, print an error message and cancel the
-                * lookup.
-                */
-               dighost_error("communications error to %s: %s\n", sockstr,
-                             isc_result_totext(eresult));
-
-               if (keep != NULL) {
-                       isc_nmhandle_detach(&keep);
-               }
-
-               if (eresult == ISC_R_EOF) {
-                       requeue_or_update_exitcode(l);
-               } else if (exitcode < 9) {
-                       exitcode = 9;
-               }
-
-               goto cancel_lookup;
        }
 
        isc_buffer_init(&b, region->base, region->length);

diff --git a/bin/dig/dighost.h b/bin/dig/dighost.h
index d7246333e09189eb728caf9e0e4b7bf25be7943a..e9da8f618e6714025113bc2fcbc97269792941f5 100644 (file)
--- a/bin/dig/dighost.h
+++ b/bin/dig/dighost.h
@@ -169,7 +169,6 @@ struct dig_lookup {
        unsigned int ednsflags;
        dns_opcode_t opcode;
        int rrcomments;
-       unsigned int eoferr;
        uint16_t qid;
        struct {
                bool http_plain;

diff --git a/bin/tests/system/acl/tests.sh b/bin/tests/system/acl/tests.sh
index 9ee39843ae160a8e237b6621650e398502356f3b..ad98fa1f053f595cd2d0c9edb2190a4bb995f4ad 100644 (file)
--- a/bin/tests/system/acl/tests.sh
+++ b/bin/tests/system/acl/tests.sh
@@ -161,7 +161,8 @@ grep "status: NOERROR" dig.out.3.${t} > /dev/null 2>&1 || ret=1
 $DIG -p ${PORT} soa example. \
        @10.53.0.2 -b 10.53.0.8 > dig.out.4.${t}
 grep "status: NOERROR" dig.out.4.${t} > /dev/null 2>&1 && ret=1
-grep "connection timed out" dig.out.4.${t} > /dev/null 2>&1 || ret=1
+grep "timed out" dig.out.4.${t} > /dev/null 2>&1 || ret=1
+grep ";; no servers could be reached" dig.out.4.${t} > /dev/null 2>&1 || ret=1
 [ $ret -eq 0 ] || echo_i "failed"
 status=`expr $status + $ret`
 

diff --git a/bin/tests/system/legacy/tests.sh b/bin/tests/system/legacy/tests.sh
index 80d70899fb452a7464465be3ad1f7b9dde93f340..5adeead291c9a6e02a56e736c53cd84e13a47356 100755 (executable)
--- a/bin/tests/system/legacy/tests.sh
+++ b/bin/tests/system/legacy/tests.sh
@@ -106,7 +106,8 @@ n=`expr $n + 1`
 echo_i "checking drop edns server setup ($n)"
 ret=0
 $DIG $DIGOPTS +edns @10.53.0.2 dropedns soa > dig.out.1.test$n && ret=1
-grep "connection timed out; no servers could be reached" dig.out.1.test$n > /dev/null || ret=1
+grep "timed out" dig.out.1.test$n > /dev/null || ret=1
+grep ";; no servers could be reached" dig.out.1.test$n > /dev/null || ret=1
 $DIG $DIGOPTS +noedns @10.53.0.2 dropedns soa > dig.out.2.test$n || ret=1
 grep "status: NOERROR" dig.out.2.test$n > /dev/null || ret=1
 grep "EDNS: version:" dig.out.2.test$n > /dev/null && ret=1
@@ -114,7 +115,8 @@ $DIG $DIGOPTS +noedns +tcp @10.53.0.2 dropedns soa > dig.out.3.test$n || ret=1
 grep "status: NOERROR" dig.out.3.test$n > /dev/null || ret=1
 grep "EDNS: version:" dig.out.3.test$n > /dev/null && ret=1
 $DIG $DIGOPTS +edns +tcp @10.53.0.2 dropedns soa > dig.out.4.test$n && ret=1
-grep "connection timed out; no servers could be reached" dig.out.4.test$n > /dev/null || ret=1
+grep "timed out" dig.out.4.test$n > /dev/null || ret=1
+grep ";; no servers could be reached" dig.out.4.test$n > /dev/null || ret=1
 if [ $ret != 0 ]; then echo_i "failed"; fi
 status=`expr $status + $ret`
 
@@ -129,7 +131,8 @@ n=`expr $n + 1`
 echo_i "checking drop edns + no tcp server setup ($n)"
 ret=0
 $DIG $DIGOPTS +edns @10.53.0.3 dropedns-notcp soa > dig.out.1.test$n && ret=1
-grep "connection timed out; no servers could be reached" dig.out.1.test$n > /dev/null || ret=1
+grep "timed out" dig.out.1.test$n > /dev/null || ret=1
+grep ";; no servers could be reached" dig.out.1.test$n > /dev/null || ret=1
 $DIG $DIGOPTS +noedns +tcp @10.53.0.3 dropedns-notcp soa > dig.out.2.test$n && ret=1
 grep "connection refused" dig.out.2.test$n > /dev/null || ret=1
 $DIG $DIGOPTS +noedns @10.53.0.3 dropedns-notcp soa > dig.out.3.test$n || ret=1
@@ -192,7 +195,8 @@ $DIG $DIGOPTS +edns +tcp @10.53.0.6 edns512 txt > dig.out.2.test$n || ret=1
 grep "status: NOERROR" dig.out.2.test$n > /dev/null || ret=1
 grep "EDNS: version:" dig.out.2.test$n > /dev/null || ret=1
 $DIG $DIGOPTS +edns +dnssec @10.53.0.6 edns512 txt > dig.out.3.test$n && ret=1
-grep "connection timed out; no servers could be reached" dig.out.3.test$n > /dev/null || ret=1
+grep "timed out" dig.out.3.test$n > /dev/null || ret=1
+grep ";; no servers could be reached" dig.out.3.test$n > /dev/null || ret=1
 $DIG $DIGOPTS +edns +dnssec +bufsize=512 +ignore @10.53.0.6 edns512 soa > dig.out.4.test$n || ret=1
 grep "status: NOERROR" dig.out.4.test$n > /dev/null || ret=1
 grep "EDNS: version:" dig.out.4.test$n > /dev/null || ret=1
@@ -216,7 +220,8 @@ grep "EDNS: version:" dig.out.1.test$n > /dev/null || ret=1
 $DIG $DIGOPTS +edns +tcp @10.53.0.7 edns512-notcp soa > dig.out.2.test$n && ret=1
 grep "connection refused" dig.out.2.test$n > /dev/null || ret=1
 $DIG $DIGOPTS +edns +dnssec @10.53.0.7 edns512-notcp soa > dig.out.3.test$n && ret=1
-grep "connection timed out; no servers could be reached" dig.out.3.test$n > /dev/null || ret=1
+grep "timed out" dig.out.3.test$n > /dev/null || ret=1
+grep ";; no servers could be reached" dig.out.3.test$n > /dev/null || ret=1
 $DIG $DIGOPTS +edns +dnssec +bufsize=512 +ignore @10.53.0.7 edns512-notcp soa > dig.out.4.test$n || ret=1
 grep "status: NOERROR" dig.out.4.test$n > /dev/null || ret=1
 grep "EDNS: version:" dig.out.4.test$n > /dev/null || ret=1

diff --git a/bin/tests/system/rpz/tests.sh b/bin/tests/system/rpz/tests.sh
index 49a8316b08120b8cae44aabbe5a15466e4b14c89..9975b686cfd6858ef08a89a36a94b526a6820756 100644 (file)
--- a/bin/tests/system/rpz/tests.sh
+++ b/bin/tests/system/rpz/tests.sh
@@ -423,7 +423,7 @@ here () {
 }
 
 # check dropped response
-DROPPED='^;; connection timed out; no servers could be reached'
+DROPPED='^;; no servers could be reached'
 drop () {
     make_dignm
     digcmd $* >$DIGNM

diff --git a/bin/tests/system/serve-stale/tests.sh b/bin/tests/system/serve-stale/tests.sh
index ceea322a662aa837928fc979eb6c3e2c062e36a0..cced57ff021e41dea37110e68a37a30b5fb687a4 100755 (executable)
--- a/bin/tests/system/serve-stale/tests.sh
+++ b/bin/tests/system/serve-stale/tests.sh
@@ -1146,7 +1146,8 @@ n=$((n+1))
 echo_i "check notincache.example times out (max-stale-ttl default) ($n)"
 ret=0
 $DIG -p ${PORT} +tries=1 +timeout=3  @10.53.0.3 notfound.example TXT > dig.out.test$n 2>&1
-grep "connection timed out" dig.out.test$n > /dev/null || ret=1
+grep "timed out" dig.out.test$n > /dev/null || ret=1
+grep ";; no servers could be reached" dig.out.test$n > /dev/null || ret=1
 if [ $ret != 0 ]; then echo_i "failed"; fi
 status=$((status+ret))
 
@@ -1781,7 +1782,8 @@ n=$((n+1))
 echo_i "check not in cache longttl.example times out (stale-answer-client-timeout 1.8) ($n)"
 ret=0
 wait_for_log 4 "longttl.example client timeout, stale answer unavailable" ns3/named.run || ret=1
-grep "connection timed out" dig.out.test$n > /dev/null || ret=1
+grep "timed out" dig.out.test$n > /dev/null || ret=1
+grep ";; no servers could be reached" dig.out.test$n > /dev/null || ret=1
 if [ $ret != 0 ]; then echo_i "failed"; fi
 status=$((status+ret))