2 * Licensed under the Apache License, Version 2.0 (the "License");
3 * you may not use this file except in compliance with the License.
4 * You may obtain a copy of the License at:
6 * http://www.apache.org/licenses/LICENSE-2.0
8 * Unless required by applicable law or agreed to in writing, software
9 * distributed under the License is distributed on an "AS IS" BASIS,
10 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
11 * See the License for the specific language governing permissions and
12 * limitations under the License.
21 #include "command-line.h"
24 #include "dynamic-string.h"
25 #include "fatal-signal.h"
29 #include "ovn/lib/lex.h"
30 #include "ovn/lib/ovn-nb-idl.h"
31 #include "ovn/lib/ovn-sb-idl.h"
32 #include "poll-loop.h"
35 #include "stream-ssl.h"
39 #include "openvswitch/vlog.h"
41 VLOG_DEFINE_THIS_MODULE(ovn_northd);
43 static unixctl_cb_func ovn_northd_exit;
45 struct northd_context {
46 struct ovsdb_idl *ovnnb_idl;
47 struct ovsdb_idl *ovnsb_idl;
48 struct ovsdb_idl_txn *ovnnb_txn;
49 struct ovsdb_idl_txn *ovnsb_txn;
52 static const char *ovnnb_db;
53 static const char *ovnsb_db;
55 static const char *default_db(void);
58 /* Ingress pipeline stages.
60 * These must be listed in the order that the stages will be executed. */
61 #define INGRESS_STAGES \
62 INGRESS_STAGE(PORT_SEC, port_sec) \
63 INGRESS_STAGE(L2_LKUP, l2_lkup)
66 #define INGRESS_STAGE(NAME, STR) S_IN_##NAME,
72 /* Egress pipeline stages.
74 * These must be listed in the order that the stages will be executed. */
75 #define EGRESS_STAGES \
76 EGRESS_STAGE(ACL, acl) \
77 EGRESS_STAGE(PORT_SEC, port_sec)
80 #define EGRESS_STAGE(NAME, STR) S_OUT_##NAME,
90 %s: OVN northbound management daemon\n\
91 usage: %s [OPTIONS]\n\
94 --ovnnb-db=DATABASE connect to ovn-nb database at DATABASE\n\
96 --ovnsb-db=DATABASE connect to ovn-sb database at DATABASE\n\
98 -h, --help display this help message\n\
99 -o, --options list available options\n\
100 -V, --version display version information\n\
101 ", program_name, program_name, default_db(), default_db());
104 stream_usage("database", true, true, false);
108 struct hmap_node hmap_node;
113 destroy_tnlids(struct hmap *tnlids)
115 struct tnlid_node *node, *next;
116 HMAP_FOR_EACH_SAFE (node, next, hmap_node, tnlids) {
117 hmap_remove(tnlids, &node->hmap_node);
120 hmap_destroy(tnlids);
124 add_tnlid(struct hmap *set, uint32_t tnlid)
126 struct tnlid_node *node = xmalloc(sizeof *node);
127 hmap_insert(set, &node->hmap_node, hash_int(tnlid, 0));
132 tnlid_in_use(const struct hmap *set, uint32_t tnlid)
134 const struct tnlid_node *node;
135 HMAP_FOR_EACH_IN_BUCKET (node, hmap_node, hash_int(tnlid, 0), set) {
136 if (node->tnlid == tnlid) {
144 allocate_tnlid(struct hmap *set, const char *name, uint32_t max,
147 for (uint32_t tnlid = *hint + 1; tnlid != *hint;
148 tnlid = tnlid + 1 <= max ? tnlid + 1 : 1) {
149 if (!tnlid_in_use(set, tnlid)) {
150 add_tnlid(set, tnlid);
156 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
157 VLOG_WARN_RL(&rl, "all %s tunnel ids exhausted", name);
161 /* The 'key' comes from nb->header_.uuid or sb->external_ids:logical-switch. */
162 struct ovn_datapath {
163 struct hmap_node key_node; /* Index on 'key'. */
164 struct uuid key; /* nb->header_.uuid. */
166 const struct nbrec_logical_switch *nb; /* May be NULL. */
167 const struct sbrec_datapath_binding *sb; /* May be NULL. */
169 struct ovs_list list; /* In list of similar records. */
171 struct hmap port_tnlids;
172 uint32_t port_key_hint;
177 static struct ovn_datapath *
178 ovn_datapath_create(struct hmap *datapaths, const struct uuid *key,
179 const struct nbrec_logical_switch *nb,
180 const struct sbrec_datapath_binding *sb)
182 struct ovn_datapath *od = xzalloc(sizeof *od);
186 hmap_init(&od->port_tnlids);
187 od->port_key_hint = 0;
188 hmap_insert(datapaths, &od->key_node, uuid_hash(&od->key));
193 ovn_datapath_destroy(struct hmap *datapaths, struct ovn_datapath *od)
196 /* Don't remove od->list. It is used within build_datapaths() as a
197 * private list and once we've exited that function it is not safe to
199 hmap_remove(datapaths, &od->key_node);
200 destroy_tnlids(&od->port_tnlids);
205 static struct ovn_datapath *
206 ovn_datapath_find(struct hmap *datapaths, const struct uuid *uuid)
208 struct ovn_datapath *od;
210 HMAP_FOR_EACH_WITH_HASH (od, key_node, uuid_hash(uuid), datapaths) {
211 if (uuid_equals(uuid, &od->key)) {
218 static struct ovn_datapath *
219 ovn_datapath_from_sbrec(struct hmap *datapaths,
220 const struct sbrec_datapath_binding *sb)
224 if (!smap_get_uuid(&sb->external_ids, "logical-switch", &key)) {
227 return ovn_datapath_find(datapaths, &key);
231 join_datapaths(struct northd_context *ctx, struct hmap *datapaths,
232 struct ovs_list *sb_only, struct ovs_list *nb_only,
233 struct ovs_list *both)
235 hmap_init(datapaths);
240 const struct sbrec_datapath_binding *sb, *sb_next;
241 SBREC_DATAPATH_BINDING_FOR_EACH_SAFE (sb, sb_next, ctx->ovnsb_idl) {
243 if (!smap_get_uuid(&sb->external_ids, "logical-switch", &key)) {
244 ovsdb_idl_txn_add_comment(ctx->ovnsb_txn,
245 "deleting Datapath_Binding "UUID_FMT" that "
246 "lacks external-ids:logical-switch",
247 UUID_ARGS(&sb->header_.uuid));
248 sbrec_datapath_binding_delete(sb);
252 if (ovn_datapath_find(datapaths, &key)) {
253 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 1);
254 VLOG_INFO_RL(&rl, "deleting Datapath_Binding "UUID_FMT" with "
255 "duplicate external-ids:logical-switch "UUID_FMT,
256 UUID_ARGS(&sb->header_.uuid), UUID_ARGS(&key));
257 sbrec_datapath_binding_delete(sb);
261 struct ovn_datapath *od = ovn_datapath_create(datapaths, &key,
263 list_push_back(sb_only, &od->list);
266 const struct nbrec_logical_switch *nb;
267 NBREC_LOGICAL_SWITCH_FOR_EACH (nb, ctx->ovnnb_idl) {
268 struct ovn_datapath *od = ovn_datapath_find(datapaths,
272 list_remove(&od->list);
273 list_push_back(both, &od->list);
275 od = ovn_datapath_create(datapaths, &nb->header_.uuid, nb, NULL);
276 list_push_back(nb_only, &od->list);
282 ovn_datapath_allocate_key(struct hmap *dp_tnlids)
284 static uint32_t hint;
285 return allocate_tnlid(dp_tnlids, "datapath", (1u << 24) - 1, &hint);
289 build_datapaths(struct northd_context *ctx, struct hmap *datapaths)
291 struct ovs_list sb_only, nb_only, both;
293 join_datapaths(ctx, datapaths, &sb_only, &nb_only, &both);
295 if (!list_is_empty(&nb_only)) {
296 /* First index the in-use datapath tunnel IDs. */
297 struct hmap dp_tnlids = HMAP_INITIALIZER(&dp_tnlids);
298 struct ovn_datapath *od;
299 LIST_FOR_EACH (od, list, &both) {
300 add_tnlid(&dp_tnlids, od->sb->tunnel_key);
303 /* Add southbound record for each unmatched northbound record. */
304 LIST_FOR_EACH (od, list, &nb_only) {
305 uint16_t tunnel_key = ovn_datapath_allocate_key(&dp_tnlids);
310 od->sb = sbrec_datapath_binding_insert(ctx->ovnsb_txn);
312 struct smap external_ids = SMAP_INITIALIZER(&external_ids);
313 char uuid_s[UUID_LEN + 1];
314 sprintf(uuid_s, UUID_FMT, UUID_ARGS(&od->nb->header_.uuid));
315 smap_add(&external_ids, "logical-switch", uuid_s);
316 sbrec_datapath_binding_set_external_ids(od->sb, &external_ids);
317 smap_destroy(&external_ids);
319 sbrec_datapath_binding_set_tunnel_key(od->sb, tunnel_key);
321 destroy_tnlids(&dp_tnlids);
324 /* Delete southbound records without northbound matches. */
325 struct ovn_datapath *od, *next;
326 LIST_FOR_EACH_SAFE (od, next, list, &sb_only) {
327 list_remove(&od->list);
328 sbrec_datapath_binding_delete(od->sb);
329 ovn_datapath_destroy(datapaths, od);
334 struct hmap_node key_node; /* Index on 'key'. */
335 const char *key; /* nb->name and sb->logical_port */
337 const struct nbrec_logical_port *nb; /* May be NULL. */
338 const struct sbrec_port_binding *sb; /* May be NULL. */
340 struct ovn_datapath *od;
342 struct ovs_list list; /* In list of similar records. */
345 static struct ovn_port *
346 ovn_port_create(struct hmap *ports, const char *key,
347 const struct nbrec_logical_port *nb,
348 const struct sbrec_port_binding *sb)
350 struct ovn_port *op = xzalloc(sizeof *op);
354 hmap_insert(ports, &op->key_node, hash_string(op->key, 0));
359 ovn_port_destroy(struct hmap *ports, struct ovn_port *port)
362 /* Don't remove port->list. It is used within build_ports() as a
363 * private list and once we've exited that function it is not safe to
365 hmap_remove(ports, &port->key_node);
370 static struct ovn_port *
371 ovn_port_find(struct hmap *ports, const char *name)
375 HMAP_FOR_EACH_WITH_HASH (op, key_node, hash_string(name, 0), ports) {
376 if (!strcmp(op->key, name)) {
384 ovn_port_allocate_key(struct ovn_datapath *od)
386 return allocate_tnlid(&od->port_tnlids, "port",
387 (1u << 15) - 1, &od->port_key_hint);
391 join_logical_ports(struct northd_context *ctx,
392 struct hmap *datapaths, struct hmap *ports,
393 struct ovs_list *sb_only, struct ovs_list *nb_only,
394 struct ovs_list *both)
401 const struct sbrec_port_binding *sb;
402 SBREC_PORT_BINDING_FOR_EACH (sb, ctx->ovnsb_idl) {
403 struct ovn_port *op = ovn_port_create(ports, sb->logical_port,
405 list_push_back(sb_only, &op->list);
408 struct ovn_datapath *od;
409 HMAP_FOR_EACH (od, key_node, datapaths) {
410 for (size_t i = 0; i < od->nb->n_ports; i++) {
411 const struct nbrec_logical_port *nb = od->nb->ports[i];
412 struct ovn_port *op = ovn_port_find(ports, nb->name);
415 list_remove(&op->list);
416 list_push_back(both, &op->list);
418 op = ovn_port_create(ports, nb->name, nb, NULL);
419 list_push_back(nb_only, &op->list);
427 ovn_port_update_sbrec(const struct ovn_port *op)
429 sbrec_port_binding_set_type(op->sb, op->nb->type);
430 sbrec_port_binding_set_options(op->sb, &op->nb->options);
431 sbrec_port_binding_set_datapath(op->sb, op->od->sb);
432 sbrec_port_binding_set_parent_port(op->sb, op->nb->parent_name);
433 sbrec_port_binding_set_tag(op->sb, op->nb->tag, op->nb->n_tag);
434 sbrec_port_binding_set_mac(op->sb, (const char **) op->nb->macs,
439 build_ports(struct northd_context *ctx, struct hmap *datapaths,
442 struct ovs_list sb_only, nb_only, both;
444 join_logical_ports(ctx, datapaths, ports, &sb_only, &nb_only, &both);
446 /* For logical ports that are in both databases, update the southbound
447 * record based on northbound data. Also index the in-use tunnel_keys. */
448 struct ovn_port *op, *next;
449 LIST_FOR_EACH_SAFE (op, next, list, &both) {
450 ovn_port_update_sbrec(op);
452 add_tnlid(&op->od->port_tnlids, op->sb->tunnel_key);
453 if (op->sb->tunnel_key > op->od->port_key_hint) {
454 op->od->port_key_hint = op->sb->tunnel_key;
458 /* Add southbound record for each unmatched northbound record. */
459 LIST_FOR_EACH_SAFE (op, next, list, &nb_only) {
460 uint16_t tunnel_key = ovn_port_allocate_key(op->od);
465 op->sb = sbrec_port_binding_insert(ctx->ovnsb_txn);
466 ovn_port_update_sbrec(op);
468 sbrec_port_binding_set_logical_port(op->sb, op->key);
469 sbrec_port_binding_set_tunnel_key(op->sb, tunnel_key);
472 /* Delete southbound records without northbound matches. */
473 LIST_FOR_EACH_SAFE(op, next, list, &sb_only) {
474 list_remove(&op->list);
475 sbrec_port_binding_delete(op->sb);
476 ovn_port_destroy(ports, op);
480 #define OVN_MIN_MULTICAST 32768
481 #define OVN_MAX_MULTICAST 65535
483 struct multicast_group {
485 uint16_t key; /* OVN_MIN_MULTICAST...OVN_MAX_MULTICAST. */
488 #define MC_FLOOD "_MC_flood"
489 static const struct multicast_group mc_flood = { MC_FLOOD, 65535 };
491 #define MC_UNKNOWN "_MC_unknown"
492 static const struct multicast_group mc_unknown = { MC_UNKNOWN, 65534 };
495 multicast_group_equal(const struct multicast_group *a,
496 const struct multicast_group *b)
498 return !strcmp(a->name, b->name) && a->key == b->key;
501 /* Multicast group entry. */
502 struct ovn_multicast {
503 struct hmap_node hmap_node; /* Index on 'datapath' and 'key'. */
504 struct ovn_datapath *datapath;
505 const struct multicast_group *group;
507 struct ovn_port **ports;
508 size_t n_ports, allocated_ports;
512 ovn_multicast_hash(const struct ovn_datapath *datapath,
513 const struct multicast_group *group)
515 return hash_pointer(datapath, group->key);
518 static struct ovn_multicast *
519 ovn_multicast_find(struct hmap *mcgroups, struct ovn_datapath *datapath,
520 const struct multicast_group *group)
522 struct ovn_multicast *mc;
524 HMAP_FOR_EACH_WITH_HASH (mc, hmap_node,
525 ovn_multicast_hash(datapath, group), mcgroups) {
526 if (mc->datapath == datapath
527 && multicast_group_equal(mc->group, group)) {
535 ovn_multicast_add(struct hmap *mcgroups, const struct multicast_group *group,
536 struct ovn_port *port)
538 struct ovn_datapath *od = port->od;
539 struct ovn_multicast *mc = ovn_multicast_find(mcgroups, od, group);
541 mc = xmalloc(sizeof *mc);
542 hmap_insert(mcgroups, &mc->hmap_node, ovn_multicast_hash(od, group));
546 mc->allocated_ports = 4;
547 mc->ports = xmalloc(mc->allocated_ports * sizeof *mc->ports);
549 if (mc->n_ports >= mc->allocated_ports) {
550 mc->ports = x2nrealloc(mc->ports, &mc->allocated_ports,
553 mc->ports[mc->n_ports++] = port;
557 ovn_multicast_destroy(struct hmap *mcgroups, struct ovn_multicast *mc)
560 hmap_remove(mcgroups, &mc->hmap_node);
567 ovn_multicast_update_sbrec(const struct ovn_multicast *mc,
568 const struct sbrec_multicast_group *sb)
570 struct sbrec_port_binding **ports = xmalloc(mc->n_ports * sizeof *ports);
571 for (size_t i = 0; i < mc->n_ports; i++) {
572 ports[i] = CONST_CAST(struct sbrec_port_binding *, mc->ports[i]->sb);
574 sbrec_multicast_group_set_ports(sb, ports, mc->n_ports);
578 /* Logical flow generation.
580 * This code generates the Logical_Flow table in the southbound database, as a
581 * function of most of the northbound database.
585 struct hmap_node hmap_node;
587 struct ovn_datapath *od;
588 enum ovn_pipeline { P_IN, P_OUT } pipeline;
596 ovn_lflow_hash(const struct ovn_lflow *lflow)
598 size_t hash = uuid_hash(&lflow->od->key);
599 hash = hash_2words((lflow->table_id << 16) | lflow->priority, hash);
600 hash = hash_string(lflow->match, hash);
601 return hash_string(lflow->actions, hash);
605 ovn_lflow_equal(const struct ovn_lflow *a, const struct ovn_lflow *b)
607 return (a->od == b->od
608 && a->pipeline == b->pipeline
609 && a->table_id == b->table_id
610 && a->priority == b->priority
611 && !strcmp(a->match, b->match)
612 && !strcmp(a->actions, b->actions));
616 ovn_lflow_init(struct ovn_lflow *lflow, struct ovn_datapath *od,
617 enum ovn_pipeline pipeline, uint8_t table_id, uint16_t priority,
618 char *match, char *actions)
621 lflow->pipeline = pipeline;
622 lflow->table_id = table_id;
623 lflow->priority = priority;
624 lflow->match = match;
625 lflow->actions = actions;
629 ingress_stage_to_str(int stage) {
631 #define INGRESS_STAGE(NAME, STR) case S_IN_##NAME: return #STR;
634 default: return "<unknown>";
639 egress_stage_to_str(int stage) {
641 #define EGRESS_STAGE(NAME, STR) case S_OUT_##NAME: return #STR;
644 default: return "<unknown>";
648 /* Adds a row with the specified contents to the Logical_Flow table. */
650 ovn_lflow_add(struct hmap *lflow_map, struct ovn_datapath *od,
651 enum ovn_pipeline pipeline, uint8_t table_id, uint16_t priority,
652 const char *match, const char *actions)
654 struct ovn_lflow *lflow = xmalloc(sizeof *lflow);
655 ovn_lflow_init(lflow, od, pipeline, table_id, priority,
656 xstrdup(match), xstrdup(actions));
657 hmap_insert(lflow_map, &lflow->hmap_node, ovn_lflow_hash(lflow));
660 static struct ovn_lflow *
661 ovn_lflow_find(struct hmap *lflows, struct ovn_datapath *od,
662 enum ovn_pipeline pipeline, uint8_t table_id, uint16_t priority,
663 const char *match, const char *actions)
665 struct ovn_lflow target;
666 ovn_lflow_init(&target, od, pipeline, table_id, priority,
667 CONST_CAST(char *, match), CONST_CAST(char *, actions));
669 struct ovn_lflow *lflow;
670 HMAP_FOR_EACH_WITH_HASH (lflow, hmap_node, ovn_lflow_hash(&target),
672 if (ovn_lflow_equal(lflow, &target)) {
680 ovn_lflow_destroy(struct hmap *lflows, struct ovn_lflow *lflow)
683 hmap_remove(lflows, &lflow->hmap_node);
685 free(lflow->actions);
690 /* Appends port security constraints on L2 address field 'eth_addr_field'
691 * (e.g. "eth.src" or "eth.dst") to 'match'. 'port_security', with
692 * 'n_port_security' elements, is the collection of port_security constraints
693 * from an OVN_NB Logical_Port row. */
695 build_port_security(const char *eth_addr_field,
696 char **port_security, size_t n_port_security,
699 size_t base_len = match->length;
700 ds_put_format(match, " && %s == {", eth_addr_field);
703 for (size_t i = 0; i < n_port_security; i++) {
706 if (eth_addr_from_string(port_security[i], &ea)) {
707 ds_put_format(match, ETH_ADDR_FMT, ETH_ADDR_ARGS(ea));
708 ds_put_char(match, ' ');
712 ds_chomp(match, ' ');
713 ds_put_cstr(match, "}");
716 match->length = base_len;
721 lport_is_enabled(const struct nbrec_logical_port *lport)
723 return !lport->enabled || *lport->enabled;
726 /* Updates the Logical_Flow and Multicast_Group tables in the OVN_SB database,
727 * constructing their contents based on the OVN_NB database. */
729 build_lflows(struct northd_context *ctx, struct hmap *datapaths,
732 struct hmap lflows = HMAP_INITIALIZER(&lflows);
733 struct hmap mcgroups = HMAP_INITIALIZER(&mcgroups);
735 /* Ingress table 0: Admission control framework (priorities 0 and 100). */
736 struct ovn_datapath *od;
737 HMAP_FOR_EACH (od, key_node, datapaths) {
738 /* Logical VLANs not supported. */
739 ovn_lflow_add(&lflows, od, P_IN, S_IN_PORT_SEC, 100, "vlan.present",
742 /* Broadcast/multicast source address is invalid. */
743 ovn_lflow_add(&lflows, od, P_IN, S_IN_PORT_SEC, 100, "eth.src[40]",
746 /* Port security flows have priority 50 (see below) and will continue
747 * to the next table if packet source is acceptable. */
749 /* Otherwise drop the packet. */
750 ovn_lflow_add(&lflows, od, P_IN, S_IN_PORT_SEC, 0, "1", "drop;");
753 /* Ingress table 0: Ingress port security (priority 50). */
755 HMAP_FOR_EACH (op, key_node, ports) {
756 struct ds match = DS_EMPTY_INITIALIZER;
757 ds_put_cstr(&match, "inport == ");
758 json_string_escape(op->key, &match);
759 build_port_security("eth.src",
760 op->nb->port_security, op->nb->n_port_security,
762 ovn_lflow_add(&lflows, op->od, P_IN, S_IN_PORT_SEC, 50, ds_cstr(&match),
763 lport_is_enabled(op->nb) ? "next;" : "drop;");
767 /* Ingress table 1: Destination lookup, broadcast and multicast handling
769 HMAP_FOR_EACH (op, key_node, ports) {
770 if (lport_is_enabled(op->nb)) {
771 ovn_multicast_add(&mcgroups, &mc_flood, op);
774 HMAP_FOR_EACH (od, key_node, datapaths) {
775 ovn_lflow_add(&lflows, od, P_IN, S_IN_L2_LKUP, 100, "eth.dst[40]",
776 "outport = \""MC_FLOOD"\"; output;");
779 /* Ingress table 1: Destination lookup, unicast handling (priority 50), */
780 HMAP_FOR_EACH (op, key_node, ports) {
781 for (size_t i = 0; i < op->nb->n_macs; i++) {
784 if (eth_addr_from_string(op->nb->macs[i], &mac)) {
785 struct ds match, actions;
788 ds_put_format(&match, "eth.dst == %s", op->nb->macs[i]);
791 ds_put_cstr(&actions, "outport = ");
792 json_string_escape(op->nb->name, &actions);
793 ds_put_cstr(&actions, "; output;");
794 ovn_lflow_add(&lflows, op->od, P_IN, S_IN_L2_LKUP, 50,
795 ds_cstr(&match), ds_cstr(&actions));
796 ds_destroy(&actions);
798 } else if (!strcmp(op->nb->macs[i], "unknown")) {
799 ovn_multicast_add(&mcgroups, &mc_unknown, op);
800 op->od->has_unknown = true;
802 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
804 VLOG_INFO_RL(&rl, "%s: invalid syntax '%s' in macs column",
805 op->nb->name, op->nb->macs[i]);
810 /* Ingress table 1: Destination lookup for unknown MACs (priority 0). */
811 HMAP_FOR_EACH (od, key_node, datapaths) {
812 if (od->has_unknown) {
813 ovn_lflow_add(&lflows, od, P_IN, S_IN_L2_LKUP, 0, "1",
814 "outport = \""MC_UNKNOWN"\"; output;");
818 /* Egress table 0: ACLs (any priority). */
819 HMAP_FOR_EACH (od, key_node, datapaths) {
820 for (size_t i = 0; i < od->nb->n_acls; i++) {
821 const struct nbrec_acl *acl = od->nb->acls[i];
824 action = (!strcmp(acl->action, "allow") ||
825 !strcmp(acl->action, "allow-related"))
827 ovn_lflow_add(&lflows, od, P_OUT, S_OUT_ACL, acl->priority,
831 HMAP_FOR_EACH (od, key_node, datapaths) {
832 ovn_lflow_add(&lflows, od, P_OUT, S_OUT_ACL, 0, "1", "next;");
835 /* Egress table 1: Egress port security multicast/broadcast (priority
837 HMAP_FOR_EACH (od, key_node, datapaths) {
838 ovn_lflow_add(&lflows, od, P_OUT, S_OUT_PORT_SEC, 100, "eth.dst[40]",
842 /* Egress table 1: Egress port security (priority 50). */
843 HMAP_FOR_EACH (op, key_node, ports) {
847 ds_put_cstr(&match, "outport == ");
848 json_string_escape(op->key, &match);
849 build_port_security("eth.dst",
850 op->nb->port_security, op->nb->n_port_security,
853 ovn_lflow_add(&lflows, op->od, P_OUT, S_OUT_PORT_SEC, 50,
855 lport_is_enabled(op->nb) ? "output;" : "drop;");
860 /* Push changes to the Logical_Flow table to database. */
861 const struct sbrec_logical_flow *sbflow, *next_sbflow;
862 SBREC_LOGICAL_FLOW_FOR_EACH_SAFE (sbflow, next_sbflow, ctx->ovnsb_idl) {
863 struct ovn_datapath *od
864 = ovn_datapath_from_sbrec(datapaths, sbflow->logical_datapath);
866 sbrec_logical_flow_delete(sbflow);
870 struct ovn_lflow *lflow = ovn_lflow_find(
871 &lflows, od, (!strcmp(sbflow->pipeline, "ingress") ? P_IN : P_OUT),
872 sbflow->table_id, sbflow->priority,
873 sbflow->match, sbflow->actions);
875 ovn_lflow_destroy(&lflows, lflow);
877 sbrec_logical_flow_delete(sbflow);
880 struct ovn_lflow *lflow, *next_lflow;
881 HMAP_FOR_EACH_SAFE (lflow, next_lflow, hmap_node, &lflows) {
882 sbflow = sbrec_logical_flow_insert(ctx->ovnsb_txn);
883 sbrec_logical_flow_set_logical_datapath(sbflow, lflow->od->sb);
884 sbrec_logical_flow_set_pipeline(
885 sbflow, lflow->pipeline == P_IN ? "ingress" : "egress");
886 sbrec_logical_flow_set_table_id(sbflow, lflow->table_id);
887 sbrec_logical_flow_set_priority(sbflow, lflow->priority);
888 sbrec_logical_flow_set_match(sbflow, lflow->match);
889 sbrec_logical_flow_set_actions(sbflow, lflow->actions);
891 struct smap external_ids = SMAP_INITIALIZER(&external_ids);
892 smap_add(&external_ids, "stage-name",
893 lflow->pipeline == P_IN ?
894 ingress_stage_to_str(lflow->table_id) :
895 egress_stage_to_str(lflow->table_id));
896 sbrec_logical_flow_set_external_ids(sbflow, &external_ids);
897 smap_destroy(&external_ids);
899 ovn_lflow_destroy(&lflows, lflow);
901 hmap_destroy(&lflows);
903 /* Push changes to the Multicast_Group table to database. */
904 const struct sbrec_multicast_group *sbmc, *next_sbmc;
905 SBREC_MULTICAST_GROUP_FOR_EACH_SAFE (sbmc, next_sbmc, ctx->ovnsb_idl) {
906 struct ovn_datapath *od = ovn_datapath_from_sbrec(datapaths,
909 sbrec_multicast_group_delete(sbmc);
913 struct multicast_group group = { .name = sbmc->name,
914 .key = sbmc->tunnel_key };
915 struct ovn_multicast *mc = ovn_multicast_find(&mcgroups, od, &group);
917 ovn_multicast_update_sbrec(mc, sbmc);
918 ovn_multicast_destroy(&mcgroups, mc);
920 sbrec_multicast_group_delete(sbmc);
923 struct ovn_multicast *mc, *next_mc;
924 HMAP_FOR_EACH_SAFE (mc, next_mc, hmap_node, &mcgroups) {
925 sbmc = sbrec_multicast_group_insert(ctx->ovnsb_txn);
926 sbrec_multicast_group_set_datapath(sbmc, mc->datapath->sb);
927 sbrec_multicast_group_set_name(sbmc, mc->group->name);
928 sbrec_multicast_group_set_tunnel_key(sbmc, mc->group->key);
929 ovn_multicast_update_sbrec(mc, sbmc);
930 ovn_multicast_destroy(&mcgroups, mc);
932 hmap_destroy(&mcgroups);
936 ovnnb_db_changed(struct northd_context *ctx)
938 VLOG_DBG("ovn-nb db contents have changed.");
940 struct hmap datapaths, ports;
941 build_datapaths(ctx, &datapaths);
942 build_ports(ctx, &datapaths, &ports);
943 build_lflows(ctx, &datapaths, &ports);
945 struct ovn_datapath *dp, *next_dp;
946 HMAP_FOR_EACH_SAFE (dp, next_dp, key_node, &datapaths) {
947 ovn_datapath_destroy(&datapaths, dp);
949 hmap_destroy(&datapaths);
951 struct ovn_port *port, *next_port;
952 HMAP_FOR_EACH_SAFE (port, next_port, key_node, &ports) {
953 ovn_port_destroy(&ports, port);
955 hmap_destroy(&ports);
959 * The only change we get notified about is if the 'chassis' column of the
960 * 'Port_Binding' table changes. When this column is not empty, it means we
961 * need to set the corresponding logical port as 'up' in the northbound DB.
964 ovnsb_db_changed(struct northd_context *ctx)
966 struct hmap lports_hmap;
967 const struct sbrec_port_binding *sb;
968 const struct nbrec_logical_port *nb;
970 struct lport_hash_node {
971 struct hmap_node node;
972 const struct nbrec_logical_port *nb;
973 } *hash_node, *hash_node_next;
975 VLOG_DBG("Recalculating port up states for ovn-nb db.");
977 hmap_init(&lports_hmap);
979 NBREC_LOGICAL_PORT_FOR_EACH(nb, ctx->ovnnb_idl) {
980 hash_node = xzalloc(sizeof *hash_node);
982 hmap_insert(&lports_hmap, &hash_node->node, hash_string(nb->name, 0));
985 SBREC_PORT_BINDING_FOR_EACH(sb, ctx->ovnsb_idl) {
987 HMAP_FOR_EACH_WITH_HASH(hash_node, node,
988 hash_string(sb->logical_port, 0),
990 if (!strcmp(sb->logical_port, hash_node->nb->name)) {
997 /* The logical port doesn't exist for this port binding. This can
998 * happen under normal circumstances when ovn-northd hasn't gotten
999 * around to pruning the Port_Binding yet. */
1003 if (sb->chassis && (!nb->up || !*nb->up)) {
1005 nbrec_logical_port_set_up(nb, &up, 1);
1006 } else if (!sb->chassis && (!nb->up || *nb->up)) {
1008 nbrec_logical_port_set_up(nb, &up, 1);
1012 HMAP_FOR_EACH_SAFE(hash_node, hash_node_next, node, &lports_hmap) {
1013 hmap_remove(&lports_hmap, &hash_node->node);
1016 hmap_destroy(&lports_hmap);
1020 static char *default_db_;
1026 default_db_ = xasprintf("unix:%s/db.sock", ovs_rundir());
1032 parse_options(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
1035 DAEMON_OPTION_ENUMS,
1038 static const struct option long_options[] = {
1039 {"ovnsb-db", required_argument, NULL, 'd'},
1040 {"ovnnb-db", required_argument, NULL, 'D'},
1041 {"help", no_argument, NULL, 'h'},
1042 {"options", no_argument, NULL, 'o'},
1043 {"version", no_argument, NULL, 'V'},
1044 DAEMON_LONG_OPTIONS,
1046 STREAM_SSL_LONG_OPTIONS,
1049 char *short_options = ovs_cmdl_long_options_to_short_options(long_options);
1054 c = getopt_long(argc, argv, short_options, long_options, NULL);
1060 DAEMON_OPTION_HANDLERS;
1061 VLOG_OPTION_HANDLERS;
1062 STREAM_SSL_OPTION_HANDLERS;
1077 ovs_cmdl_print_options(long_options);
1081 ovs_print_version(0, 0);
1090 ovnsb_db = default_db();
1094 ovnnb_db = default_db();
1097 free(short_options);
1101 add_column_noalert(struct ovsdb_idl *idl,
1102 const struct ovsdb_idl_column *column)
1104 ovsdb_idl_add_column(idl, column);
1105 ovsdb_idl_omit_alert(idl, column);
1109 main(int argc, char *argv[])
1111 extern struct vlog_module VLM_reconnect;
1112 struct ovsdb_idl *ovnnb_idl, *ovnsb_idl;
1113 unsigned int ovnnb_seqno, ovn_seqno;
1114 int res = EXIT_SUCCESS;
1115 struct northd_context ctx = {
1118 bool ovnnb_changes_pending = false;
1119 bool ovn_changes_pending = false;
1120 struct unixctl_server *unixctl;
1124 fatal_ignore_sigpipe();
1125 set_program_name(argv[0]);
1126 service_start(&argc, &argv);
1127 vlog_set_levels(NULL, VLF_CONSOLE, VLL_WARN);
1128 vlog_set_levels(&VLM_reconnect, VLF_ANY_DESTINATION, VLL_WARN);
1129 parse_options(argc, argv);
1133 retval = unixctl_server_create(NULL, &unixctl);
1137 unixctl_command_register("exit", "", 0, 0, ovn_northd_exit, &exiting);
1139 daemonize_complete();
1144 /* We want to detect all changes to the ovn-nb db. */
1145 ctx.ovnnb_idl = ovnnb_idl = ovsdb_idl_create(ovnnb_db,
1146 &nbrec_idl_class, true, true);
1148 ctx.ovnsb_idl = ovnsb_idl = ovsdb_idl_create(ovnsb_db,
1149 &sbrec_idl_class, false, true);
1151 ovsdb_idl_add_table(ovnsb_idl, &sbrec_table_logical_flow);
1152 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_logical_datapath);
1153 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_pipeline);
1154 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_table_id);
1155 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_priority);
1156 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_match);
1157 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_actions);
1159 ovsdb_idl_add_table(ovnsb_idl, &sbrec_table_multicast_group);
1160 add_column_noalert(ovnsb_idl, &sbrec_multicast_group_col_datapath);
1161 add_column_noalert(ovnsb_idl, &sbrec_multicast_group_col_tunnel_key);
1162 add_column_noalert(ovnsb_idl, &sbrec_multicast_group_col_name);
1163 add_column_noalert(ovnsb_idl, &sbrec_multicast_group_col_ports);
1165 ovsdb_idl_add_table(ovnsb_idl, &sbrec_table_datapath_binding);
1166 add_column_noalert(ovnsb_idl, &sbrec_datapath_binding_col_tunnel_key);
1167 add_column_noalert(ovnsb_idl, &sbrec_datapath_binding_col_external_ids);
1169 ovsdb_idl_add_table(ovnsb_idl, &sbrec_table_port_binding);
1170 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_datapath);
1171 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_logical_port);
1172 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_tunnel_key);
1173 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_parent_port);
1174 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_tag);
1175 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_type);
1176 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_options);
1177 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_mac);
1178 ovsdb_idl_add_column(ovnsb_idl, &sbrec_port_binding_col_chassis);
1181 * The loop here just runs the IDL in a loop waiting for the seqno to
1182 * change, which indicates that the contents of the db have changed.
1184 * If the contents of the ovn-nb db change, the mappings to the ovn-sb
1185 * db must be recalculated.
1187 * If the contents of the ovn-sb db change, it means the 'up' state of
1188 * a port may have changed, as that's the only type of change ovn-northd is
1192 ovnnb_seqno = ovsdb_idl_get_seqno(ovnnb_idl);
1193 ovn_seqno = ovsdb_idl_get_seqno(ovnsb_idl);
1196 ovsdb_idl_run(ovnnb_idl);
1197 ovsdb_idl_run(ovnsb_idl);
1198 unixctl_server_run(unixctl);
1200 if (!ovsdb_idl_is_alive(ovnnb_idl)) {
1201 int retval = ovsdb_idl_get_last_error(ovnnb_idl);
1202 VLOG_ERR("%s: database connection failed (%s)",
1203 ovnnb_db, ovs_retval_to_string(retval));
1208 if (!ovsdb_idl_is_alive(ovnsb_idl)) {
1209 int retval = ovsdb_idl_get_last_error(ovnsb_idl);
1210 VLOG_ERR("%s: database connection failed (%s)",
1211 ovnsb_db, ovs_retval_to_string(retval));
1216 if (ovnnb_seqno != ovsdb_idl_get_seqno(ovnnb_idl)) {
1217 ovnnb_seqno = ovsdb_idl_get_seqno(ovnnb_idl);
1218 ovnnb_changes_pending = true;
1221 if (ovn_seqno != ovsdb_idl_get_seqno(ovnsb_idl)) {
1222 ovn_seqno = ovsdb_idl_get_seqno(ovnsb_idl);
1223 ovn_changes_pending = true;
1227 * If there are any pending changes, we delay recalculating the
1228 * necessary updates until after an existing transaction finishes.
1229 * This avoids the possibility of rapid updates causing ovn-northd to
1230 * never be able to successfully make the corresponding updates to the
1231 * other db. Instead, pending changes are batched up until the next
1232 * time we get a chance to calculate the new state and apply it.
1235 if (ovnnb_changes_pending && !ctx.ovnsb_txn) {
1237 * The OVN-nb db contents have changed, so create a transaction for
1238 * updating the OVN-sb DB.
1240 ctx.ovnsb_txn = ovsdb_idl_txn_create(ctx.ovnsb_idl);
1241 ovsdb_idl_txn_add_comment(ctx.ovnsb_txn,
1242 "ovn-northd: northbound db changed");
1243 ovnnb_db_changed(&ctx);
1244 ovnnb_changes_pending = false;
1247 if (ovn_changes_pending && !ctx.ovnnb_txn) {
1249 * The OVN-sb db contents have changed, so create a transaction for
1250 * updating the northbound DB.
1252 ctx.ovnnb_txn = ovsdb_idl_txn_create(ctx.ovnnb_idl);
1253 ovsdb_idl_txn_add_comment(ctx.ovnnb_txn,
1254 "ovn-northd: southbound db changed");
1255 ovnsb_db_changed(&ctx);
1256 ovn_changes_pending = false;
1259 if (ctx.ovnnb_txn) {
1260 enum ovsdb_idl_txn_status txn_status;
1261 txn_status = ovsdb_idl_txn_commit(ctx.ovnnb_txn);
1262 switch (txn_status) {
1263 case TXN_UNCOMMITTED:
1264 case TXN_INCOMPLETE:
1265 /* Come back around and try to commit this transaction again */
1269 case TXN_NOT_LOCKED:
1271 /* Something went wrong, so try creating a new transaction. */
1272 ovn_changes_pending = true;
1275 ovsdb_idl_txn_destroy(ctx.ovnnb_txn);
1276 ctx.ovnnb_txn = NULL;
1280 if (ctx.ovnsb_txn) {
1281 enum ovsdb_idl_txn_status txn_status;
1282 txn_status = ovsdb_idl_txn_commit(ctx.ovnsb_txn);
1283 switch (txn_status) {
1284 case TXN_UNCOMMITTED:
1285 case TXN_INCOMPLETE:
1286 /* Come back around and try to commit this transaction again */
1290 case TXN_NOT_LOCKED:
1292 /* Something went wrong, so try creating a new transaction. */
1293 ovnnb_changes_pending = true;
1296 ovsdb_idl_txn_destroy(ctx.ovnsb_txn);
1297 ctx.ovnsb_txn = NULL;
1301 if (ovnnb_seqno == ovsdb_idl_get_seqno(ovnnb_idl) &&
1302 ovn_seqno == ovsdb_idl_get_seqno(ovnsb_idl)) {
1303 ovsdb_idl_wait(ovnnb_idl);
1304 ovsdb_idl_wait(ovnsb_idl);
1305 if (ctx.ovnnb_txn) {
1306 ovsdb_idl_txn_wait(ctx.ovnnb_txn);
1308 if (ctx.ovnsb_txn) {
1309 ovsdb_idl_txn_wait(ctx.ovnsb_txn);
1311 unixctl_server_wait(unixctl);
1313 poll_immediate_wake();
1317 if (should_service_stop()) {
1322 unixctl_server_destroy(unixctl);
1323 ovsdb_idl_destroy(ovnsb_idl);
1324 ovsdb_idl_destroy(ovnnb_idl);
1333 ovn_northd_exit(struct unixctl_conn *conn, int argc OVS_UNUSED,
1334 const char *argv[] OVS_UNUSED, void *exiting_)
1336 bool *exiting = exiting_;
1339 unixctl_command_reply(conn, NULL);