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 char uuid_s[UUID_LEN + 1];
313 sprintf(uuid_s, UUID_FMT, UUID_ARGS(&od->nb->header_.uuid));
314 const struct smap id = SMAP_CONST1(&id, "logical-switch", uuid_s);
315 sbrec_datapath_binding_set_external_ids(od->sb, &id);
317 sbrec_datapath_binding_set_tunnel_key(od->sb, tunnel_key);
319 destroy_tnlids(&dp_tnlids);
322 /* Delete southbound records without northbound matches. */
323 struct ovn_datapath *od, *next;
324 LIST_FOR_EACH_SAFE (od, next, list, &sb_only) {
325 list_remove(&od->list);
326 sbrec_datapath_binding_delete(od->sb);
327 ovn_datapath_destroy(datapaths, od);
332 struct hmap_node key_node; /* Index on 'key'. */
333 const char *key; /* nb->name and sb->logical_port */
335 const struct nbrec_logical_port *nb; /* May be NULL. */
336 const struct sbrec_port_binding *sb; /* May be NULL. */
338 struct ovn_datapath *od;
340 struct ovs_list list; /* In list of similar records. */
343 static struct ovn_port *
344 ovn_port_create(struct hmap *ports, const char *key,
345 const struct nbrec_logical_port *nb,
346 const struct sbrec_port_binding *sb)
348 struct ovn_port *op = xzalloc(sizeof *op);
352 hmap_insert(ports, &op->key_node, hash_string(op->key, 0));
357 ovn_port_destroy(struct hmap *ports, struct ovn_port *port)
360 /* Don't remove port->list. It is used within build_ports() as a
361 * private list and once we've exited that function it is not safe to
363 hmap_remove(ports, &port->key_node);
368 static struct ovn_port *
369 ovn_port_find(struct hmap *ports, const char *name)
373 HMAP_FOR_EACH_WITH_HASH (op, key_node, hash_string(name, 0), ports) {
374 if (!strcmp(op->key, name)) {
382 ovn_port_allocate_key(struct ovn_datapath *od)
384 return allocate_tnlid(&od->port_tnlids, "port",
385 (1u << 15) - 1, &od->port_key_hint);
389 join_logical_ports(struct northd_context *ctx,
390 struct hmap *datapaths, struct hmap *ports,
391 struct ovs_list *sb_only, struct ovs_list *nb_only,
392 struct ovs_list *both)
399 const struct sbrec_port_binding *sb;
400 SBREC_PORT_BINDING_FOR_EACH (sb, ctx->ovnsb_idl) {
401 struct ovn_port *op = ovn_port_create(ports, sb->logical_port,
403 list_push_back(sb_only, &op->list);
406 struct ovn_datapath *od;
407 HMAP_FOR_EACH (od, key_node, datapaths) {
408 for (size_t i = 0; i < od->nb->n_ports; i++) {
409 const struct nbrec_logical_port *nb = od->nb->ports[i];
410 struct ovn_port *op = ovn_port_find(ports, nb->name);
413 list_remove(&op->list);
414 list_push_back(both, &op->list);
416 op = ovn_port_create(ports, nb->name, nb, NULL);
417 list_push_back(nb_only, &op->list);
425 ovn_port_update_sbrec(const struct ovn_port *op)
427 sbrec_port_binding_set_type(op->sb, op->nb->type);
428 sbrec_port_binding_set_options(op->sb, &op->nb->options);
429 sbrec_port_binding_set_datapath(op->sb, op->od->sb);
430 sbrec_port_binding_set_parent_port(op->sb, op->nb->parent_name);
431 sbrec_port_binding_set_tag(op->sb, op->nb->tag, op->nb->n_tag);
432 sbrec_port_binding_set_mac(op->sb, (const char **) op->nb->macs,
437 build_ports(struct northd_context *ctx, struct hmap *datapaths,
440 struct ovs_list sb_only, nb_only, both;
442 join_logical_ports(ctx, datapaths, ports, &sb_only, &nb_only, &both);
444 /* For logical ports that are in both databases, update the southbound
445 * record based on northbound data. Also index the in-use tunnel_keys. */
446 struct ovn_port *op, *next;
447 LIST_FOR_EACH_SAFE (op, next, list, &both) {
448 ovn_port_update_sbrec(op);
450 add_tnlid(&op->od->port_tnlids, op->sb->tunnel_key);
451 if (op->sb->tunnel_key > op->od->port_key_hint) {
452 op->od->port_key_hint = op->sb->tunnel_key;
456 /* Add southbound record for each unmatched northbound record. */
457 LIST_FOR_EACH_SAFE (op, next, list, &nb_only) {
458 uint16_t tunnel_key = ovn_port_allocate_key(op->od);
463 op->sb = sbrec_port_binding_insert(ctx->ovnsb_txn);
464 ovn_port_update_sbrec(op);
466 sbrec_port_binding_set_logical_port(op->sb, op->key);
467 sbrec_port_binding_set_tunnel_key(op->sb, tunnel_key);
470 /* Delete southbound records without northbound matches. */
471 LIST_FOR_EACH_SAFE(op, next, list, &sb_only) {
472 list_remove(&op->list);
473 sbrec_port_binding_delete(op->sb);
474 ovn_port_destroy(ports, op);
478 #define OVN_MIN_MULTICAST 32768
479 #define OVN_MAX_MULTICAST 65535
481 struct multicast_group {
483 uint16_t key; /* OVN_MIN_MULTICAST...OVN_MAX_MULTICAST. */
486 #define MC_FLOOD "_MC_flood"
487 static const struct multicast_group mc_flood = { MC_FLOOD, 65535 };
489 #define MC_UNKNOWN "_MC_unknown"
490 static const struct multicast_group mc_unknown = { MC_UNKNOWN, 65534 };
493 multicast_group_equal(const struct multicast_group *a,
494 const struct multicast_group *b)
496 return !strcmp(a->name, b->name) && a->key == b->key;
499 /* Multicast group entry. */
500 struct ovn_multicast {
501 struct hmap_node hmap_node; /* Index on 'datapath' and 'key'. */
502 struct ovn_datapath *datapath;
503 const struct multicast_group *group;
505 struct ovn_port **ports;
506 size_t n_ports, allocated_ports;
510 ovn_multicast_hash(const struct ovn_datapath *datapath,
511 const struct multicast_group *group)
513 return hash_pointer(datapath, group->key);
516 static struct ovn_multicast *
517 ovn_multicast_find(struct hmap *mcgroups, struct ovn_datapath *datapath,
518 const struct multicast_group *group)
520 struct ovn_multicast *mc;
522 HMAP_FOR_EACH_WITH_HASH (mc, hmap_node,
523 ovn_multicast_hash(datapath, group), mcgroups) {
524 if (mc->datapath == datapath
525 && multicast_group_equal(mc->group, group)) {
533 ovn_multicast_add(struct hmap *mcgroups, const struct multicast_group *group,
534 struct ovn_port *port)
536 struct ovn_datapath *od = port->od;
537 struct ovn_multicast *mc = ovn_multicast_find(mcgroups, od, group);
539 mc = xmalloc(sizeof *mc);
540 hmap_insert(mcgroups, &mc->hmap_node, ovn_multicast_hash(od, group));
544 mc->allocated_ports = 4;
545 mc->ports = xmalloc(mc->allocated_ports * sizeof *mc->ports);
547 if (mc->n_ports >= mc->allocated_ports) {
548 mc->ports = x2nrealloc(mc->ports, &mc->allocated_ports,
551 mc->ports[mc->n_ports++] = port;
555 ovn_multicast_destroy(struct hmap *mcgroups, struct ovn_multicast *mc)
558 hmap_remove(mcgroups, &mc->hmap_node);
565 ovn_multicast_update_sbrec(const struct ovn_multicast *mc,
566 const struct sbrec_multicast_group *sb)
568 struct sbrec_port_binding **ports = xmalloc(mc->n_ports * sizeof *ports);
569 for (size_t i = 0; i < mc->n_ports; i++) {
570 ports[i] = CONST_CAST(struct sbrec_port_binding *, mc->ports[i]->sb);
572 sbrec_multicast_group_set_ports(sb, ports, mc->n_ports);
576 /* Logical flow generation.
578 * This code generates the Logical_Flow table in the southbound database, as a
579 * function of most of the northbound database.
583 struct hmap_node hmap_node;
585 struct ovn_datapath *od;
586 enum ovn_pipeline { P_IN, P_OUT } pipeline;
594 ovn_lflow_hash(const struct ovn_lflow *lflow)
596 size_t hash = uuid_hash(&lflow->od->key);
597 hash = hash_2words((lflow->table_id << 16) | lflow->priority, hash);
598 hash = hash_string(lflow->match, hash);
599 return hash_string(lflow->actions, hash);
603 ovn_lflow_equal(const struct ovn_lflow *a, const struct ovn_lflow *b)
605 return (a->od == b->od
606 && a->pipeline == b->pipeline
607 && a->table_id == b->table_id
608 && a->priority == b->priority
609 && !strcmp(a->match, b->match)
610 && !strcmp(a->actions, b->actions));
614 ovn_lflow_init(struct ovn_lflow *lflow, struct ovn_datapath *od,
615 enum ovn_pipeline pipeline, uint8_t table_id, uint16_t priority,
616 char *match, char *actions)
619 lflow->pipeline = pipeline;
620 lflow->table_id = table_id;
621 lflow->priority = priority;
622 lflow->match = match;
623 lflow->actions = actions;
627 ingress_stage_to_str(int stage) {
629 #define INGRESS_STAGE(NAME, STR) case S_IN_##NAME: return #STR;
632 default: return "<unknown>";
637 egress_stage_to_str(int stage) {
639 #define EGRESS_STAGE(NAME, STR) case S_OUT_##NAME: return #STR;
642 default: return "<unknown>";
646 /* Adds a row with the specified contents to the Logical_Flow table. */
648 ovn_lflow_add(struct hmap *lflow_map, struct ovn_datapath *od,
649 enum ovn_pipeline pipeline, uint8_t table_id, uint16_t priority,
650 const char *match, const char *actions)
652 struct ovn_lflow *lflow = xmalloc(sizeof *lflow);
653 ovn_lflow_init(lflow, od, pipeline, table_id, priority,
654 xstrdup(match), xstrdup(actions));
655 hmap_insert(lflow_map, &lflow->hmap_node, ovn_lflow_hash(lflow));
658 static struct ovn_lflow *
659 ovn_lflow_find(struct hmap *lflows, struct ovn_datapath *od,
660 enum ovn_pipeline pipeline, uint8_t table_id, uint16_t priority,
661 const char *match, const char *actions)
663 struct ovn_lflow target;
664 ovn_lflow_init(&target, od, pipeline, table_id, priority,
665 CONST_CAST(char *, match), CONST_CAST(char *, actions));
667 struct ovn_lflow *lflow;
668 HMAP_FOR_EACH_WITH_HASH (lflow, hmap_node, ovn_lflow_hash(&target),
670 if (ovn_lflow_equal(lflow, &target)) {
678 ovn_lflow_destroy(struct hmap *lflows, struct ovn_lflow *lflow)
681 hmap_remove(lflows, &lflow->hmap_node);
683 free(lflow->actions);
688 /* Appends port security constraints on L2 address field 'eth_addr_field'
689 * (e.g. "eth.src" or "eth.dst") to 'match'. 'port_security', with
690 * 'n_port_security' elements, is the collection of port_security constraints
691 * from an OVN_NB Logical_Port row. */
693 build_port_security(const char *eth_addr_field,
694 char **port_security, size_t n_port_security,
697 size_t base_len = match->length;
698 ds_put_format(match, " && %s == {", eth_addr_field);
701 for (size_t i = 0; i < n_port_security; i++) {
704 if (eth_addr_from_string(port_security[i], &ea)) {
705 ds_put_format(match, ETH_ADDR_FMT, ETH_ADDR_ARGS(ea));
706 ds_put_char(match, ' ');
710 ds_chomp(match, ' ');
711 ds_put_cstr(match, "}");
714 match->length = base_len;
719 lport_is_enabled(const struct nbrec_logical_port *lport)
721 return !lport->enabled || *lport->enabled;
724 /* Updates the Logical_Flow and Multicast_Group tables in the OVN_SB database,
725 * constructing their contents based on the OVN_NB database. */
727 build_lflows(struct northd_context *ctx, struct hmap *datapaths,
730 struct hmap lflows = HMAP_INITIALIZER(&lflows);
731 struct hmap mcgroups = HMAP_INITIALIZER(&mcgroups);
733 /* Ingress table 0: Admission control framework (priorities 0 and 100). */
734 struct ovn_datapath *od;
735 HMAP_FOR_EACH (od, key_node, datapaths) {
736 /* Logical VLANs not supported. */
737 ovn_lflow_add(&lflows, od, P_IN, S_IN_PORT_SEC, 100, "vlan.present",
740 /* Broadcast/multicast source address is invalid. */
741 ovn_lflow_add(&lflows, od, P_IN, S_IN_PORT_SEC, 100, "eth.src[40]",
744 /* Port security flows have priority 50 (see below) and will continue
745 * to the next table if packet source is acceptable. */
747 /* Otherwise drop the packet. */
748 ovn_lflow_add(&lflows, od, P_IN, S_IN_PORT_SEC, 0, "1", "drop;");
751 /* Ingress table 0: Ingress port security (priority 50). */
753 HMAP_FOR_EACH (op, key_node, ports) {
754 struct ds match = DS_EMPTY_INITIALIZER;
755 ds_put_cstr(&match, "inport == ");
756 json_string_escape(op->key, &match);
757 build_port_security("eth.src",
758 op->nb->port_security, op->nb->n_port_security,
760 ovn_lflow_add(&lflows, op->od, P_IN, S_IN_PORT_SEC, 50, ds_cstr(&match),
761 lport_is_enabled(op->nb) ? "next;" : "drop;");
765 /* Ingress table 1: Destination lookup, broadcast and multicast handling
767 HMAP_FOR_EACH (op, key_node, ports) {
768 if (lport_is_enabled(op->nb)) {
769 ovn_multicast_add(&mcgroups, &mc_flood, op);
772 HMAP_FOR_EACH (od, key_node, datapaths) {
773 ovn_lflow_add(&lflows, od, P_IN, S_IN_L2_LKUP, 100, "eth.dst[40]",
774 "outport = \""MC_FLOOD"\"; output;");
777 /* Ingress table 1: Destination lookup, unicast handling (priority 50), */
778 HMAP_FOR_EACH (op, key_node, ports) {
779 for (size_t i = 0; i < op->nb->n_macs; i++) {
782 if (eth_addr_from_string(op->nb->macs[i], &mac)) {
783 struct ds match, actions;
786 ds_put_format(&match, "eth.dst == %s", op->nb->macs[i]);
789 ds_put_cstr(&actions, "outport = ");
790 json_string_escape(op->nb->name, &actions);
791 ds_put_cstr(&actions, "; output;");
792 ovn_lflow_add(&lflows, op->od, P_IN, S_IN_L2_LKUP, 50,
793 ds_cstr(&match), ds_cstr(&actions));
794 ds_destroy(&actions);
796 } else if (!strcmp(op->nb->macs[i], "unknown")) {
797 ovn_multicast_add(&mcgroups, &mc_unknown, op);
798 op->od->has_unknown = true;
800 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
802 VLOG_INFO_RL(&rl, "%s: invalid syntax '%s' in macs column",
803 op->nb->name, op->nb->macs[i]);
808 /* Ingress table 1: Destination lookup for unknown MACs (priority 0). */
809 HMAP_FOR_EACH (od, key_node, datapaths) {
810 if (od->has_unknown) {
811 ovn_lflow_add(&lflows, od, P_IN, S_IN_L2_LKUP, 0, "1",
812 "outport = \""MC_UNKNOWN"\"; output;");
816 /* Egress table 0: ACLs (any priority). */
817 HMAP_FOR_EACH (od, key_node, datapaths) {
818 for (size_t i = 0; i < od->nb->n_acls; i++) {
819 const struct nbrec_acl *acl = od->nb->acls[i];
822 action = (!strcmp(acl->action, "allow") ||
823 !strcmp(acl->action, "allow-related"))
825 ovn_lflow_add(&lflows, od, P_OUT, S_OUT_ACL, acl->priority,
829 HMAP_FOR_EACH (od, key_node, datapaths) {
830 ovn_lflow_add(&lflows, od, P_OUT, S_OUT_ACL, 0, "1", "next;");
833 /* Egress table 1: Egress port security multicast/broadcast (priority
835 HMAP_FOR_EACH (od, key_node, datapaths) {
836 ovn_lflow_add(&lflows, od, P_OUT, S_OUT_PORT_SEC, 100, "eth.dst[40]",
840 /* Egress table 1: Egress port security (priority 50). */
841 HMAP_FOR_EACH (op, key_node, ports) {
845 ds_put_cstr(&match, "outport == ");
846 json_string_escape(op->key, &match);
847 build_port_security("eth.dst",
848 op->nb->port_security, op->nb->n_port_security,
851 ovn_lflow_add(&lflows, op->od, P_OUT, S_OUT_PORT_SEC, 50,
853 lport_is_enabled(op->nb) ? "output;" : "drop;");
858 /* Push changes to the Logical_Flow table to database. */
859 const struct sbrec_logical_flow *sbflow, *next_sbflow;
860 SBREC_LOGICAL_FLOW_FOR_EACH_SAFE (sbflow, next_sbflow, ctx->ovnsb_idl) {
861 struct ovn_datapath *od
862 = ovn_datapath_from_sbrec(datapaths, sbflow->logical_datapath);
864 sbrec_logical_flow_delete(sbflow);
868 struct ovn_lflow *lflow = ovn_lflow_find(
869 &lflows, od, (!strcmp(sbflow->pipeline, "ingress") ? P_IN : P_OUT),
870 sbflow->table_id, sbflow->priority,
871 sbflow->match, sbflow->actions);
873 ovn_lflow_destroy(&lflows, lflow);
875 sbrec_logical_flow_delete(sbflow);
878 struct ovn_lflow *lflow, *next_lflow;
879 HMAP_FOR_EACH_SAFE (lflow, next_lflow, hmap_node, &lflows) {
880 sbflow = sbrec_logical_flow_insert(ctx->ovnsb_txn);
881 sbrec_logical_flow_set_logical_datapath(sbflow, lflow->od->sb);
882 sbrec_logical_flow_set_pipeline(
883 sbflow, lflow->pipeline == P_IN ? "ingress" : "egress");
884 sbrec_logical_flow_set_table_id(sbflow, lflow->table_id);
885 sbrec_logical_flow_set_priority(sbflow, lflow->priority);
886 sbrec_logical_flow_set_match(sbflow, lflow->match);
887 sbrec_logical_flow_set_actions(sbflow, lflow->actions);
889 const struct smap ids = SMAP_CONST1(
891 (lflow->pipeline == P_IN
892 ? ingress_stage_to_str(lflow->table_id)
893 : egress_stage_to_str(lflow->table_id)));
894 sbrec_logical_flow_set_external_ids(sbflow, &ids);
896 ovn_lflow_destroy(&lflows, lflow);
898 hmap_destroy(&lflows);
900 /* Push changes to the Multicast_Group table to database. */
901 const struct sbrec_multicast_group *sbmc, *next_sbmc;
902 SBREC_MULTICAST_GROUP_FOR_EACH_SAFE (sbmc, next_sbmc, ctx->ovnsb_idl) {
903 struct ovn_datapath *od = ovn_datapath_from_sbrec(datapaths,
906 sbrec_multicast_group_delete(sbmc);
910 struct multicast_group group = { .name = sbmc->name,
911 .key = sbmc->tunnel_key };
912 struct ovn_multicast *mc = ovn_multicast_find(&mcgroups, od, &group);
914 ovn_multicast_update_sbrec(mc, sbmc);
915 ovn_multicast_destroy(&mcgroups, mc);
917 sbrec_multicast_group_delete(sbmc);
920 struct ovn_multicast *mc, *next_mc;
921 HMAP_FOR_EACH_SAFE (mc, next_mc, hmap_node, &mcgroups) {
922 sbmc = sbrec_multicast_group_insert(ctx->ovnsb_txn);
923 sbrec_multicast_group_set_datapath(sbmc, mc->datapath->sb);
924 sbrec_multicast_group_set_name(sbmc, mc->group->name);
925 sbrec_multicast_group_set_tunnel_key(sbmc, mc->group->key);
926 ovn_multicast_update_sbrec(mc, sbmc);
927 ovn_multicast_destroy(&mcgroups, mc);
929 hmap_destroy(&mcgroups);
933 ovnnb_db_changed(struct northd_context *ctx)
935 VLOG_DBG("ovn-nb db contents have changed.");
937 struct hmap datapaths, ports;
938 build_datapaths(ctx, &datapaths);
939 build_ports(ctx, &datapaths, &ports);
940 build_lflows(ctx, &datapaths, &ports);
942 struct ovn_datapath *dp, *next_dp;
943 HMAP_FOR_EACH_SAFE (dp, next_dp, key_node, &datapaths) {
944 ovn_datapath_destroy(&datapaths, dp);
946 hmap_destroy(&datapaths);
948 struct ovn_port *port, *next_port;
949 HMAP_FOR_EACH_SAFE (port, next_port, key_node, &ports) {
950 ovn_port_destroy(&ports, port);
952 hmap_destroy(&ports);
956 * The only change we get notified about is if the 'chassis' column of the
957 * 'Port_Binding' table changes. When this column is not empty, it means we
958 * need to set the corresponding logical port as 'up' in the northbound DB.
961 ovnsb_db_changed(struct northd_context *ctx)
963 struct hmap lports_hmap;
964 const struct sbrec_port_binding *sb;
965 const struct nbrec_logical_port *nb;
967 struct lport_hash_node {
968 struct hmap_node node;
969 const struct nbrec_logical_port *nb;
970 } *hash_node, *hash_node_next;
972 VLOG_DBG("Recalculating port up states for ovn-nb db.");
974 hmap_init(&lports_hmap);
976 NBREC_LOGICAL_PORT_FOR_EACH(nb, ctx->ovnnb_idl) {
977 hash_node = xzalloc(sizeof *hash_node);
979 hmap_insert(&lports_hmap, &hash_node->node, hash_string(nb->name, 0));
982 SBREC_PORT_BINDING_FOR_EACH(sb, ctx->ovnsb_idl) {
984 HMAP_FOR_EACH_WITH_HASH(hash_node, node,
985 hash_string(sb->logical_port, 0),
987 if (!strcmp(sb->logical_port, hash_node->nb->name)) {
994 /* The logical port doesn't exist for this port binding. This can
995 * happen under normal circumstances when ovn-northd hasn't gotten
996 * around to pruning the Port_Binding yet. */
1000 if (sb->chassis && (!nb->up || !*nb->up)) {
1002 nbrec_logical_port_set_up(nb, &up, 1);
1003 } else if (!sb->chassis && (!nb->up || *nb->up)) {
1005 nbrec_logical_port_set_up(nb, &up, 1);
1009 HMAP_FOR_EACH_SAFE(hash_node, hash_node_next, node, &lports_hmap) {
1010 hmap_remove(&lports_hmap, &hash_node->node);
1013 hmap_destroy(&lports_hmap);
1017 static char *default_db_;
1023 default_db_ = xasprintf("unix:%s/db.sock", ovs_rundir());
1029 parse_options(int argc OVS_UNUSED, char *argv[] OVS_UNUSED)
1032 DAEMON_OPTION_ENUMS,
1035 static const struct option long_options[] = {
1036 {"ovnsb-db", required_argument, NULL, 'd'},
1037 {"ovnnb-db", required_argument, NULL, 'D'},
1038 {"help", no_argument, NULL, 'h'},
1039 {"options", no_argument, NULL, 'o'},
1040 {"version", no_argument, NULL, 'V'},
1041 DAEMON_LONG_OPTIONS,
1043 STREAM_SSL_LONG_OPTIONS,
1046 char *short_options = ovs_cmdl_long_options_to_short_options(long_options);
1051 c = getopt_long(argc, argv, short_options, long_options, NULL);
1057 DAEMON_OPTION_HANDLERS;
1058 VLOG_OPTION_HANDLERS;
1059 STREAM_SSL_OPTION_HANDLERS;
1074 ovs_cmdl_print_options(long_options);
1078 ovs_print_version(0, 0);
1087 ovnsb_db = default_db();
1091 ovnnb_db = default_db();
1094 free(short_options);
1098 add_column_noalert(struct ovsdb_idl *idl,
1099 const struct ovsdb_idl_column *column)
1101 ovsdb_idl_add_column(idl, column);
1102 ovsdb_idl_omit_alert(idl, column);
1106 main(int argc, char *argv[])
1108 extern struct vlog_module VLM_reconnect;
1109 struct ovsdb_idl *ovnnb_idl, *ovnsb_idl;
1110 unsigned int ovnnb_seqno, ovn_seqno;
1111 int res = EXIT_SUCCESS;
1112 struct northd_context ctx = {
1115 bool ovnnb_changes_pending = false;
1116 bool ovn_changes_pending = false;
1117 struct unixctl_server *unixctl;
1121 fatal_ignore_sigpipe();
1122 set_program_name(argv[0]);
1123 service_start(&argc, &argv);
1124 vlog_set_levels(NULL, VLF_CONSOLE, VLL_WARN);
1125 vlog_set_levels(&VLM_reconnect, VLF_ANY_DESTINATION, VLL_WARN);
1126 parse_options(argc, argv);
1130 retval = unixctl_server_create(NULL, &unixctl);
1134 unixctl_command_register("exit", "", 0, 0, ovn_northd_exit, &exiting);
1136 daemonize_complete();
1141 /* We want to detect all changes to the ovn-nb db. */
1142 ctx.ovnnb_idl = ovnnb_idl = ovsdb_idl_create(ovnnb_db,
1143 &nbrec_idl_class, true, true);
1145 ctx.ovnsb_idl = ovnsb_idl = ovsdb_idl_create(ovnsb_db,
1146 &sbrec_idl_class, false, true);
1148 ovsdb_idl_add_table(ovnsb_idl, &sbrec_table_logical_flow);
1149 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_logical_datapath);
1150 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_pipeline);
1151 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_table_id);
1152 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_priority);
1153 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_match);
1154 add_column_noalert(ovnsb_idl, &sbrec_logical_flow_col_actions);
1156 ovsdb_idl_add_table(ovnsb_idl, &sbrec_table_multicast_group);
1157 add_column_noalert(ovnsb_idl, &sbrec_multicast_group_col_datapath);
1158 add_column_noalert(ovnsb_idl, &sbrec_multicast_group_col_tunnel_key);
1159 add_column_noalert(ovnsb_idl, &sbrec_multicast_group_col_name);
1160 add_column_noalert(ovnsb_idl, &sbrec_multicast_group_col_ports);
1162 ovsdb_idl_add_table(ovnsb_idl, &sbrec_table_datapath_binding);
1163 add_column_noalert(ovnsb_idl, &sbrec_datapath_binding_col_tunnel_key);
1164 add_column_noalert(ovnsb_idl, &sbrec_datapath_binding_col_external_ids);
1166 ovsdb_idl_add_table(ovnsb_idl, &sbrec_table_port_binding);
1167 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_datapath);
1168 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_logical_port);
1169 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_tunnel_key);
1170 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_parent_port);
1171 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_tag);
1172 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_type);
1173 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_options);
1174 add_column_noalert(ovnsb_idl, &sbrec_port_binding_col_mac);
1175 ovsdb_idl_add_column(ovnsb_idl, &sbrec_port_binding_col_chassis);
1178 * The loop here just runs the IDL in a loop waiting for the seqno to
1179 * change, which indicates that the contents of the db have changed.
1181 * If the contents of the ovn-nb db change, the mappings to the ovn-sb
1182 * db must be recalculated.
1184 * If the contents of the ovn-sb db change, it means the 'up' state of
1185 * a port may have changed, as that's the only type of change ovn-northd is
1189 ovnnb_seqno = ovsdb_idl_get_seqno(ovnnb_idl);
1190 ovn_seqno = ovsdb_idl_get_seqno(ovnsb_idl);
1193 ovsdb_idl_run(ovnnb_idl);
1194 ovsdb_idl_run(ovnsb_idl);
1195 unixctl_server_run(unixctl);
1197 if (!ovsdb_idl_is_alive(ovnnb_idl)) {
1198 int retval = ovsdb_idl_get_last_error(ovnnb_idl);
1199 VLOG_ERR("%s: database connection failed (%s)",
1200 ovnnb_db, ovs_retval_to_string(retval));
1205 if (!ovsdb_idl_is_alive(ovnsb_idl)) {
1206 int retval = ovsdb_idl_get_last_error(ovnsb_idl);
1207 VLOG_ERR("%s: database connection failed (%s)",
1208 ovnsb_db, ovs_retval_to_string(retval));
1213 if (ovnnb_seqno != ovsdb_idl_get_seqno(ovnnb_idl)) {
1214 ovnnb_seqno = ovsdb_idl_get_seqno(ovnnb_idl);
1215 ovnnb_changes_pending = true;
1218 if (ovn_seqno != ovsdb_idl_get_seqno(ovnsb_idl)) {
1219 ovn_seqno = ovsdb_idl_get_seqno(ovnsb_idl);
1220 ovn_changes_pending = true;
1224 * If there are any pending changes, we delay recalculating the
1225 * necessary updates until after an existing transaction finishes.
1226 * This avoids the possibility of rapid updates causing ovn-northd to
1227 * never be able to successfully make the corresponding updates to the
1228 * other db. Instead, pending changes are batched up until the next
1229 * time we get a chance to calculate the new state and apply it.
1232 if (ovnnb_changes_pending && !ctx.ovnsb_txn) {
1234 * The OVN-nb db contents have changed, so create a transaction for
1235 * updating the OVN-sb DB.
1237 ctx.ovnsb_txn = ovsdb_idl_txn_create(ctx.ovnsb_idl);
1238 ovsdb_idl_txn_add_comment(ctx.ovnsb_txn,
1239 "ovn-northd: northbound db changed");
1240 ovnnb_db_changed(&ctx);
1241 ovnnb_changes_pending = false;
1244 if (ovn_changes_pending && !ctx.ovnnb_txn) {
1246 * The OVN-sb db contents have changed, so create a transaction for
1247 * updating the northbound DB.
1249 ctx.ovnnb_txn = ovsdb_idl_txn_create(ctx.ovnnb_idl);
1250 ovsdb_idl_txn_add_comment(ctx.ovnnb_txn,
1251 "ovn-northd: southbound db changed");
1252 ovnsb_db_changed(&ctx);
1253 ovn_changes_pending = false;
1256 if (ctx.ovnnb_txn) {
1257 enum ovsdb_idl_txn_status txn_status;
1258 txn_status = ovsdb_idl_txn_commit(ctx.ovnnb_txn);
1259 switch (txn_status) {
1260 case TXN_UNCOMMITTED:
1261 case TXN_INCOMPLETE:
1262 /* Come back around and try to commit this transaction again */
1266 case TXN_NOT_LOCKED:
1268 /* Something went wrong, so try creating a new transaction. */
1269 ovn_changes_pending = true;
1272 ovsdb_idl_txn_destroy(ctx.ovnnb_txn);
1273 ctx.ovnnb_txn = NULL;
1277 if (ctx.ovnsb_txn) {
1278 enum ovsdb_idl_txn_status txn_status;
1279 txn_status = ovsdb_idl_txn_commit(ctx.ovnsb_txn);
1280 switch (txn_status) {
1281 case TXN_UNCOMMITTED:
1282 case TXN_INCOMPLETE:
1283 /* Come back around and try to commit this transaction again */
1287 case TXN_NOT_LOCKED:
1289 /* Something went wrong, so try creating a new transaction. */
1290 ovnnb_changes_pending = true;
1293 ovsdb_idl_txn_destroy(ctx.ovnsb_txn);
1294 ctx.ovnsb_txn = NULL;
1298 if (ovnnb_seqno == ovsdb_idl_get_seqno(ovnnb_idl) &&
1299 ovn_seqno == ovsdb_idl_get_seqno(ovnsb_idl)) {
1300 ovsdb_idl_wait(ovnnb_idl);
1301 ovsdb_idl_wait(ovnsb_idl);
1302 if (ctx.ovnnb_txn) {
1303 ovsdb_idl_txn_wait(ctx.ovnnb_txn);
1305 if (ctx.ovnsb_txn) {
1306 ovsdb_idl_txn_wait(ctx.ovnsb_txn);
1308 unixctl_server_wait(unixctl);
1310 poll_immediate_wake();
1314 if (should_service_stop()) {
1319 unixctl_server_destroy(unixctl);
1320 ovsdb_idl_destroy(ovnsb_idl);
1321 ovsdb_idl_destroy(ovnnb_idl);
1330 ovn_northd_exit(struct unixctl_conn *conn, int argc OVS_UNUSED,
1331 const char *argv[] OVS_UNUSED, void *exiting_)
1333 bool *exiting = exiting_;
1336 unixctl_command_reply(conn, NULL);