- Fix mesh cycle detection for configuration with respip CNAME

loop and tagged clients. Thanks to Qifan Zhang, Palo Alto
  Networks, for the report.

daemon/remote.c

@@ -5026,6 +5026,74 @@ fr_check_changed_cfg_str2list(struct config_str2list* cmp1,
 	}
 }
 
+/** fast reload thread, check if config str3list has changed. */
+#define FR_CHECK_CHANGED_CFG_STR3LIST(desc, var, buff) do {		\
+	fr_check_changed_cfg_str3list(cfg->var, newcfg->var, desc, buff,\
+		sizeof(buff));						\
+	} while(0);
+static void
+fr_check_changed_cfg_str3list(struct config_str3list* cmp1,
+	struct config_str3list* cmp2, const char* desc, char* str, size_t len)
+{
+	struct config_str3list* p1 = cmp1, *p2 = cmp2;
+	while(p1 && p2) {
+		if((!p1->str && p2->str) ||
+			(p1->str && !p2->str) ||
+			(p1->str && p2->str && strcmp(p1->str, p2->str) != 0)) {
+			/* The str3list is different. */
+			fr_add_incompatible_option(desc, str, len);
+			return;
+		}
+		if((!p1->str2 && p2->str2) ||
+			(p1->str2 && !p2->str2) ||
+			(p1->str2 && p2->str2 &&
+			strcmp(p1->str2, p2->str2) != 0)) {
+			/* The str3list is different. */
+			fr_add_incompatible_option(desc, str, len);
+			return;
+		}
+		if((!p1->str3 && p2->str3) ||
+			(p1->str3 && !p2->str3) ||
+			(p1->str3 && p2->str3 &&
+			strcmp(p1->str3, p2->str3) != 0)) {
+			/* The str3list is different. */
+			fr_add_incompatible_option(desc, str, len);
+			return;
+		}
+		p1 = p1->next;
+		p2 = p2->next;
+	}
+	if((!p1 && p2) || (p1 && !p2)) {
+		fr_add_incompatible_option(desc, str, len);
+	}
+}
+
+/** fast reload thread, check tag datas. */
+static int
+fr_check_tag_datas(struct fast_reload_thread* fr, struct config_file* newcfg)
+{
+	char changed_str[1024];
+	struct config_file* cfg = fr->worker->env.cfg;
+	changed_str[0]=0;
+
+	/* Check for tag_datas in acl_addr. */
+	FR_CHECK_CHANGED_CFG_STR3LIST("interface-tag-data", interface_tag_datas, changed_str);
+	FR_CHECK_CHANGED_CFG_STR3LIST("access-control-tag-data", acl_tag_datas, changed_str);
+
+	if(changed_str[0] != 0) {
+		if(fr->fr_drop_mesh)
+			return 1; /* already dropping queries */
+		fr->fr_drop_mesh = 1;
+		fr->worker->daemon->fast_reload_drop_mesh = fr->fr_drop_mesh;
+		if(!fr_output_printf(fr, "recursion referenced data has changed, with: '%s"
+			"', and the queries have to be dropped"
+			", setting '+d'\n", changed_str))
+			return 0;
+		fr_send_notification(fr, fast_reload_notification_printout);
+	}
+	return 1;
+}
+
 /** fast reload thread, check compatible config items */
 static int
 fr_check_compat_cfg(struct fast_reload_thread* fr, struct config_file* newcfg)
@@ -6911,6 +6979,10 @@ fr_load_config(struct fast_reload_thread* fr, struct timeval* time_read,
 		config_delete(newcfg);
 		return 0;
 	}
+	if(!fr_check_tag_datas(fr, newcfg)) {
+		config_delete(newcfg);
+		return 0;
+	}
 	if(!fr_check_compat_cfg(fr, newcfg)) {
 		config_delete(newcfg);
 		return 0;

doc/Changelog

@@ -12,6 +12,9 @@
 	  not considered valid when an NSEC next owner name is not
 	  under the signer zone name. Thanks to Qifan Zhang, Palo
 	  Alto Networks, for the report.
+	- Fix mesh cycle detection for configuration with respip CNAME
+	  loop and tagged clients. Thanks to Qifan Zhang, Palo Alto
+	  Networks, for the report.
 
 22 July 2026: Wouter
 	- Release tag for 1.25.2, with the security commits:

respip/respip.c

@@ -1164,8 +1164,10 @@ respip_operate(struct module_qstate* qstate, enum module_ev event, int id,
 				 * clients. */
 				qstate->is_drop = 1;
 			} else if(alias_rrset) {
-				if(!generate_cname_request(qstate, alias_rrset))
+				if(!generate_cname_request(qstate, alias_rrset)) {
+					errinf(qstate, "Could not generate CNAME request");
 					goto servfail;
+				}
 				next_state = module_wait_subquery;
 			}
 			qstate->return_msg->rep = new_rep;
@@ -1179,6 +1181,7 @@ respip_operate(struct module_qstate* qstate, enum module_ev event, int id,
   servfail:
 	qstate->return_rcode = LDNS_RCODE_SERVFAIL;
 	qstate->return_msg = NULL;
+	qstate->ext_state[id] = module_finished;
 }
 
 int
@@ -1275,6 +1278,7 @@ respip_inform_super(struct module_qstate* qstate, int id,
 	return;
 
   fail:
+	errinf(super, "CNAME lookup failed");
 	super->return_rcode = LDNS_RCODE_SERVFAIL;
 	super->return_msg = NULL;
 	return;

services/mesh.c

@@ -965,32 +965,9 @@ void mesh_report_reply(struct mesh_area* mesh, struct outbound_entry* e,
 	mesh_run(mesh, e->qstate->mesh_info, event, e);
 }
 
-/** copy strlist to region */
-static struct config_strlist*
-cfg_region_strlist_copy(struct regional* region, struct config_strlist* list)
-{
-	struct config_strlist* result = NULL, *last = NULL, *s = list;
-	while(s) {
-		struct config_strlist* n = regional_alloc_zero(region,
-			sizeof(*n));
-		if(!n)
-			return NULL;
-		n->str = regional_strdup(region, s->str);
-		if(!n->str)
-			return NULL;
-		if(last)
-			last->next = n;
-		else	result = n;
-		last = n;
-		s = s->next;
-	}
-	return result;
-}
-
 struct respip_client_info*
 mesh_copy_client_info(struct regional* region, struct respip_client_info* cinfo)
 {
-	size_t i;
 	struct respip_client_info* client_info;
 	client_info = regional_alloc_init(region, cinfo, sizeof(*cinfo));
 	if(!client_info)
@@ -1009,20 +986,13 @@ mesh_copy_client_info(struct regional* region, struct respip_client_info* cinfo)
 		if(!client_info->tag_actions)
 			return NULL;
 	}
-	if(cinfo->tag_datas) {
+	/* tag_datas is owned by the matched acl_addr in config_file; its
-		client_info->tag_datas = regional_alloc_zero(region,
+	 * lifetime is until config reload, which tears down all mesh states
-			sizeof(struct config_strlist*)*cinfo->tag_datas_size);
+	 * first. Keep the original pointer so client_info_compare()
-		if(!client_info->tag_datas)
+	 * can recognise two states from the same ACL entry. */
-			return NULL;
+	/* fast reload insists on dropping the queries when interface-tag-data
-		for(i=0; i<cinfo->tag_datas_size; i++) {
+	 * or access-control-tag-data are changed. */
-			if(cinfo->tag_datas[i]) {
+	/* client_info->tag_datas already copied by regional_alloc_init above */
-				client_info->tag_datas[i] = cfg_region_strlist_copy(
-					region, cinfo->tag_datas[i]);
-				if(!client_info->tag_datas[i])
-					return NULL;
-			}
-		}
-	}
 	if(cinfo->view) {
 		/* Do not copy the view pointer but store a name instead.
 		 * The name is looked up later when done, this means that
@@ -2306,8 +2276,29 @@ void mesh_run(struct mesh_area* mesh, struct mesh_state* mstate,
 	enum module_ev ev, struct outbound_entry* e)
 {
 	enum module_ext_state s;
+	int numrun = 0;
 	verbose(VERB_ALGO, "mesh_run: start");
 	while(mstate) {
+		if(numrun++ > MESH_MAX_RUN_ITER) {
+			/* These modules are too much to activate, stop them.*/
+			log_err("Too many module run iterations, deleting");
+			while(mstate) {
+				/* notify supers */
+				if(mstate->super_set.count > 0) {
+					verbose(VERB_ALGO, "notify supers of failure");
+					mstate->s.return_msg = NULL;
+					mstate->s.return_rcode = LDNS_RCODE_SERVFAIL;
+					mesh_walk_supers(mesh, mstate);
+				}
+				mesh_state_delete(&mstate->s);
+				if(mesh->run.count > 0) {
+					/* pop random element off the runnable tree */
+					mstate = (struct mesh_state*)mesh->run.root->key;
+					(void)rbtree_delete(&mesh->run, mstate);
+				} else mstate = NULL;
+			}
+			break;
+		}
 		/* run the module */
 		fptr_ok(fptr_whitelist_mod_operate(
 			mesh->mods.mod[mstate->s.curmod]->operate));

services/mesh.h

@@ -69,6 +69,13 @@ struct respip_client_info;
  */
 #define MESH_MAX_ACTIVATION 10000
 
+/**
+ * Maximum number of mesh state run items. These are different modules
+ * activated during a mesh run. Any more is likely an infinite loop
+ * in the module. It is then terminated, and states are deleted.
+ */
+#define MESH_MAX_RUN_ITER 10000
+
 /**
  * Max number of references-to-references-to-references.. search size.
  * Any more is treated like 'too large', and the creation of a new

testdata/respip_cname_loop_tagged.rpl

@@ -0,0 +1,254 @@
+; config options
+; The island of trust is at test.
+server:
+	target-fetch-policy: "0 0 0 0 0"
+	qname-minimisation: "no"
+	fake-sha1: yes
+	trust-anchor-signaling: no
+	minimal-responses: no
+	iter-scrub-promiscuous: no
+	aggressive-nsec: no
+	local-zone: test. nodefault
+	log-servfail: yes
+	discard-timeout: 0
+	module-config: "respip iterator"
+	define-tag: "turqoise"
+	access-control-tag: 127.0.0.0/8 "turqoise"
+	access-control-tag-data: 127.0.0.0/8 "turqoise" "A 127.0.0.1"
+
+	; These two CNAMEs form a loop.
+	response-ip: 192.0.2.1/32 redirect
+	response-ip-data: 192.0.2.1/32 "CNAME loop2.far.test."
+	response-ip: 192.0.2.2/32 redirect
+	response-ip-data: 192.0.2.2/32 "CNAME loop1.far.test."
+
+stub-zone:
+	name: "."
+	stub-addr: 193.0.14.129 	# K.ROOT-SERVERS.NET.
+CONFIG_END
+
+SCENARIO_BEGIN Test respip CNAME loop that is tagged.
+
+; K.ROOT-SERVERS.NET.
+RANGE_BEGIN 0 100
+	ADDRESS 193.0.14.129 
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR NOERROR
+SECTION QUESTION
+. IN NS
+SECTION ANSWER
+. IN NS	K.ROOT-SERVERS.NET.
+SECTION ADDITIONAL
+K.ROOT-SERVERS.NET.	IN	A	193.0.14.129
+ENTRY_END
+
+ENTRY_BEGIN
+MATCH opcode subdomain
+ADJUST copy_id copy_query
+REPLY QR NOERROR
+SECTION QUESTION
+test. IN NS
+SECTION AUTHORITY
+test.	IN NS	ns.test.
+SECTION ADDITIONAL
+ns.test. IN A 1.2.3.5
+ENTRY_END
+RANGE_END
+
+; ns.test
+RANGE_BEGIN 0 100
+	ADDRESS 1.2.3.5
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR AA NOERROR
+SECTION QUESTION
+test. IN NS
+SECTION ANSWER
+test.    IN NS   ns.test
+SECTION ADDITIONAL
+ns.test. IN A 1.2.3.5
+ENTRY_END
+
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR AA NOERROR
+SECTION QUESTION
+ns.test. IN A
+SECTION ANSWER
+ns.test. IN A 1.2.3.5
+ENTRY_END
+
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR AA NOERROR
+SECTION QUESTION
+ns.test. IN AAAA
+SECTION AUTHORITY
+test. 3600 IN SOA ns.test. host.test. 20201 3600 1800 604800 3600
+ENTRY_END
+
+ENTRY_BEGIN
+MATCH opcode subdomain
+ADJUST copy_id copy_query
+REPLY QR NOERROR
+SECTION QUESTION
+example.test. IN NS
+SECTION AUTHORITY
+example.test.    IN NS   ns.example.test.
+SECTION ADDITIONAL
+ns.example.test. IN A 1.2.3.4
+ENTRY_END
+
+ENTRY_BEGIN
+MATCH opcode subdomain
+ADJUST copy_id copy_query
+REPLY QR NOERROR
+SECTION QUESTION
+far.test. IN NS
+SECTION AUTHORITY
+far.test.    IN NS   ns.far.test.
+SECTION ADDITIONAL
+ns.far.test. IN A 1.2.3.6
+ENTRY_END
+RANGE_END
+
+; ns.example.test.
+RANGE_BEGIN 0 20
+	ADDRESS 1.2.3.4
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR NOERROR
+SECTION QUESTION
+example.test. IN NS
+SECTION ANSWER
+example.test.    IN NS   ns.example.test.
+SECTION ADDITIONAL
+ns.example.test.         IN      A       1.2.3.4
+ENTRY_END
+
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR AA NOERROR
+SECTION QUESTION
+ns.example.test. IN A
+SECTION ANSWER
+ns.example.test.         IN      A       1.2.3.4
+ENTRY_END
+
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR AA NOERROR
+SECTION QUESTION
+ns.example.test. IN AAAA
+SECTION AUTHORITY
+example.test. 3600 IN SOA ns.example.test. host.example.test. 20301 3600 1800 604800 3600
+ENTRY_END
+
+; response to query of interest
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR AA NOERROR
+SECTION QUESTION
+www.example.test. IN A
+SECTION ANSWER
+www.example.test. 1 IN A 192.0.2.1
+ENTRY_END
+RANGE_END
+
+; ns.example.test.
+RANGE_BEGIN 45 100
+	ADDRESS 1.2.3.4
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR AA NOERROR
+SECTION QUESTION
+www.example.test. IN A
+SECTION ANSWER
+www.example.test. 1 IN A 192.0.2.1
+ENTRY_END
+RANGE_END
+
+; ns.far.test.
+RANGE_BEGIN 0 100
+	ADDRESS 1.2.3.6
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR NOERROR
+SECTION QUESTION
+far.test. IN NS
+SECTION ANSWER
+far.test.    IN NS   ns.far.test.
+SECTION ADDITIONAL
+ns.far.test.         IN      A       1.2.3.6
+ENTRY_END
+
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR AA NOERROR
+SECTION QUESTION
+ns.far.test. IN A
+SECTION ANSWER
+ns.far.test.         IN      A       1.2.3.6
+ENTRY_END
+
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR AA NOERROR
+SECTION QUESTION
+ns.far.test. IN AAAA
+SECTION AUTHORITY
+far.test. 3600 IN SOA ns.far.test. host.far.test. 20301 3600 1800 604800 3600
+ENTRY_END
+
+; response to query of interest
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR AA NOERROR
+SECTION QUESTION
+loop1.far.test. IN A
+SECTION ANSWER
+loop1.far.test. IN A 192.0.2.1
+ENTRY_END
+
+ENTRY_BEGIN
+MATCH opcode qtype qname
+ADJUST copy_id
+REPLY QR AA NOERROR
+SECTION QUESTION
+loop2.far.test. IN A
+SECTION ANSWER
+loop2.far.test. IN A 192.0.2.2
+ENTRY_END
+RANGE_END
+
+STEP 1 QUERY
+ENTRY_BEGIN
+REPLY RD DO
+SECTION QUESTION
+loop1.far.test. IN A
+ENTRY_END
+
+STEP 2 CHECK_ANSWER
+ENTRY_BEGIN
+MATCH all
+REPLY QR RD RA DO SERVFAIL
+SECTION QUESTION
+loop1.far.test. IN A
+SECTION ANSWER
+ENTRY_END
+
+SCENARIO_END