- Fix #704: [FR] Statistics counter for number of outgoing UDP queries

sent; introduces 'num.query.udpout' to the 'unbound-control stats'
  command.

contrib/metrics.awk

@@ -28,6 +28,7 @@ END {
 	print "unbound_hits_queries{type=\"total.num.prefetch\"} " val["total.num.prefetch"];
 	print "unbound_hits_queries{type=\"num.query.tcp\"} " val["num.query.tcp"];
 	print "unbound_hits_queries{type=\"num.query.tcpout\"} " val["num.query.tcpout"];
+	print "unbound_hits_queries{type=\"num.query.udpout\"} " val["num.query.udpout"];
 	print "unbound_hits_queries{type=\"num.query.tls\"} " val["num.query.tls"];
 	print "unbound_hits_queries{type=\"num.query.tls.resume\"} " val["num.query.tls.resume"];
 	print "unbound_hits_queries{type=\"num.query.ipv6\"} " val["num.query.ipv6"];

contrib/unbound_munin_

@@ -253,6 +253,7 @@ if test "$1" = "config" ; then
 		p_config "total.num.prefetch" "cache prefetch" "ABSOLUTE"
 		p_config "num.query.tcp" "TCP queries" "ABSOLUTE"
 		p_config "num.query.tcpout" "TCP out queries" "ABSOLUTE"
+		p_config "num.query.udpout" "UDP out queries" "ABSOLUTE"
 		p_config "num.query.tls" "TLS queries" "ABSOLUTE"
 		p_config "num.query.tls.resume" "TLS resumes" "ABSOLUTE"
 		p_config "num.query.ipv6" "IPv6 queries" "ABSOLUTE"
@@ -452,7 +453,7 @@ hits)
 	for x in `grep "^thread[0-9][0-9]*\.num\.queries=" $state |
 		sed -e 's/=.*//'` total.num.queries \
 		total.num.cachehits total.num.prefetch num.query.tcp \
-		num.query.tcpout num.query.tls num.query.tls.resume \
+		num.query.tcpout num.query.udpout num.query.tls num.query.tls.resume \
 		num.query.ipv6 unwanted.queries \
 		unwanted.replies; do
 		if grep "^"$x"=" $state >/dev/null 2>&1; then

daemon/remote.c

@@ -988,6 +988,8 @@ print_ext(RES* ssl, struct ub_stats_info* s)
 		(unsigned long)s->svr.qtcp)) return 0;
 	if(!ssl_printf(ssl, "num.query.tcpout"SQ"%lu\n", 
 		(unsigned long)s->svr.qtcp_outgoing)) return 0;
+	if(!ssl_printf(ssl, "num.query.udpout"SQ"%lu\n",
+		(unsigned long)s->svr.qudp_outgoing)) return 0;
 	if(!ssl_printf(ssl, "num.query.tls"SQ"%lu\n", 
 		(unsigned long)s->svr.qtls)) return 0;
 	if(!ssl_printf(ssl, "num.query.tls.resume"SQ"%lu\n", 

daemon/stats.c

@@ -281,6 +281,7 @@ server_stats_compile(struct worker* worker, struct ub_stats_info* s, int reset)
 	/* values from outside network */
 	s->svr.unwanted_replies = (long long)worker->back->unwanted_replies;
 	s->svr.qtcp_outgoing = (long long)worker->back->num_tcp_outgoing;
+	s->svr.qudp_outgoing = (long long)worker->back->num_udp_outgoing;
 
 	/* get and reset validator rrset bogus number */
 	s->svr.rrset_bogus = (long long)get_rrset_bogus(worker, reset);
@@ -424,6 +425,7 @@ void server_stats_add(struct ub_stats_info* total, struct ub_stats_info* a)
 		total->svr.qclass_big += a->svr.qclass_big;
 		total->svr.qtcp += a->svr.qtcp;
 		total->svr.qtcp_outgoing += a->svr.qtcp_outgoing;
+		total->svr.qudp_outgoing += a->svr.qudp_outgoing;
 		total->svr.qtls += a->svr.qtls;
 		total->svr.qtls_resume += a->svr.qtls_resume;
 		total->svr.qhttps += a->svr.qhttps;

daemon/worker.c

@@ -2217,6 +2217,7 @@ void worker_stats_clear(struct worker* worker)
 	mesh_stats_clear(worker->env.mesh);
 	worker->back->unwanted_replies = 0;
 	worker->back->num_tcp_outgoing = 0;
+	worker->back->num_udp_outgoing = 0;
 }
 
 void worker_start_accept(void* arg)

doc/Changelog

@@ -3,6 +3,9 @@
 
 29 June 2022: George
 	- Add debug option to the mini_tdir.sh test code.
+	- Fix #704: [FR] Statistics counter for number of outgoing UDP queries
+	  sent; introduces 'num.query.udpout' to the 'unbound-control stats'
+	  command.
 
 28 June 2022: George
 	- Show the output of the exact .rpl run that failed with 'make test'.

doc/unbound-control.8.in

@@ -552,6 +552,10 @@ Number of queries that were made using TCP towards the Unbound server.
 Number of queries that the Unbound server made using TCP outgoing towards
 other servers.
 .TP
+.I num.query.udpout
+Number of queries that the Unbound server made using UDP outgoing towards
+other servers.
+.TP
 .I num.query.tls
 Number of queries that were made using TLS towards the Unbound server.
 These are also counted in num.query.tcp, because TLS uses TCP.

libunbound/unbound.h

@@ -725,6 +725,8 @@ struct ub_server_stats {
 	long long qtcp;
 	/** number of outgoing queries over TCP */
 	long long qtcp_outgoing;
+	/** number of outgoing queries over UDP */
+	long long qudp_outgoing;
 	/** number of queries over (DNS over) TLS */
 	long long qtls;
 	/** number of queries over (DNS over) HTTPS */

services/outside_network.c

@@ -1608,6 +1608,7 @@ outside_network_create(struct comm_base *base, size_t bufsize,
 	outnet->tcp_reuse_timeout= tcp_reuse_timeout;
 	outnet->tcp_auth_query_timeout = tcp_auth_query_timeout;
 	outnet->num_tcp_outgoing = 0;
+	outnet->num_udp_outgoing = 0;
 	outnet->infra = infra;
 	outnet->rnd = rnd;
 	outnet->sslctx = sslctx;
@@ -2142,6 +2143,7 @@ randomize_and_send_udp(struct pending* pend, sldns_buffer* packet, int timeout)
 		portcomm_loweruse(outnet, pend->pc);
 		return 0;
 	}
+	outnet->num_udp_outgoing++;
 
 	/* system calls to set timeout after sending UDP to make roundtrip
 	   smaller. */

services/outside_network.h

@@ -113,6 +113,8 @@ struct outside_network {
 	/** if we perform udp-connect, connect() for UDP socket to mitigate
 	 * ICMP side channel leakage */
 	int udp_connect;
+	/** number of udp packets sent. */
+	size_t num_udp_outgoing;
 
 	/** array of outgoing IP4 interfaces */
 	struct port_if* ip4_ifs;

smallapp/unbound-control.c

@@ -354,6 +354,7 @@ static void print_extended(struct ub_stats_info* s)
 	/* transport */
 	PR_UL("num.query.tcp", s->svr.qtcp);
 	PR_UL("num.query.tcpout", s->svr.qtcp_outgoing);
+	PR_UL("num.query.udpout", s->svr.qudp_outgoing);
 	PR_UL("num.query.tls", s->svr.qtls);
 	PR_UL("num.query.tls_resume", s->svr.qtls_resume);
 	PR_UL("num.query.ipv6", s->svr.qipv6);

testdata/stat_values.tdir/stat_values.conf

@@ -1,5 +1,5 @@
 server:
-	verbosity: 2
+	verbosity: 5
 	num-threads: 1
 	interface: 127.0.0.1
 	port: @PORT@
@@ -11,6 +11,9 @@ server:
 	do-not-query-localhost: no
 	extended-statistics: yes
 	identity: "stat_values"
+	outbound-msg-retry: 0
+	root-key-sentinel: no
+	trust-anchor-signaling: no
 
 	local-zone: local.zone static
 	local-data: "www.local.zone A 192.0.2.1"

testdata/stat_values.tdir/stat_values.test

@@ -186,6 +186,7 @@ num.query.opcode.QUERY=1
 num.query.flags.RD=1
 num.query.flags.AD=1
 num.query.edns.present=1
+num.query.udpout=1
 msg.cache.count=1
 rrset.cache.count=1
 infra.cache.count=1
@@ -223,6 +224,7 @@ num.query.flags.AD=1
 num.query.flags.RD=1
 num.query.opcode.QUERY=1
 num.query.type.A=1
+num.query.udpout=1
 msg.cache.count=1
 rrset.cache.count=1
 infra.cache.count=1"
@@ -249,6 +251,7 @@ num.query.opcode.QUERY=1
 num.query.flags.RD=1
 num.query.flags.AD=1
 num.query.edns.present=1
+num.query.udpout=1
 msg.cache.count=2
 rrset.cache.count=2
 infra.cache.count=2
@@ -263,6 +266,7 @@ if grep "192.0.2.1" outfile; then
 else
 	end 1
 fi
+sleep 1  # make sure the outgoing UDP (and the edns1xx0 retry) are accounted for.
 check_stats "\
 total.num.queries=1
 total.num.expired=1
@@ -274,6 +278,7 @@ num.query.flags.AD=1
 num.query.flags.RD=1
 num.query.opcode.QUERY=1
 num.query.type.A=1
+num.query.udpout=2
 total.num.cachemiss=1
 msg.cache.count=2
 rrset.cache.count=2
@@ -327,6 +332,7 @@ num.query.edns.present=1
 num.query.flags.RD=1
 num.query.opcode.QUERY=1
 num.query.type.A=1
+num.query.udpout=1
 total.num.queries=1
 total.num.recursivereplies=1
 msg.cache.count=3