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.
22 #include "command-line.h"
25 #include "openvswitch/dynamic-string.h"
26 #include "fatal-signal.h"
28 #include "openvswitch/hmap.h"
29 #include "openvswitch/json.h"
31 #include "ovn/lib/chassis-index.h"
32 #include "ovn/lib/logical-fields.h"
33 #include "ovn/lib/ovn-l7.h"
34 #include "ovn/lib/ovn-nb-idl.h"
35 #include "ovn/lib/ovn-sb-idl.h"
36 #include "ovn/lib/ovn-util.h"
37 #include "ovn/actions.h"
39 #include "openvswitch/poll-loop.h"
43 #include "stream-ssl.h"
47 #include "openvswitch/vlog.h"
49 VLOG_DEFINE_THIS_MODULE(ovn_northd
);
51 static unixctl_cb_func ovn_northd_exit
;
53 struct northd_context
{
54 struct ovsdb_idl
*ovnnb_idl
;
55 struct ovsdb_idl
*ovnsb_idl
;
56 struct ovsdb_idl_txn
*ovnnb_txn
;
57 struct ovsdb_idl_txn
*ovnsb_txn
;
60 static const char *ovnnb_db
;
61 static const char *ovnsb_db
;
62 static const char *unixctl_path
;
64 #define MAC_ADDR_PREFIX 0x0A0000000000ULL
65 #define MAC_ADDR_SPACE 0xffffff
67 /* MAC address management (macam) table of "struct eth_addr"s, that holds the
68 * MAC addresses allocated by the OVN ipam module. */
69 static struct hmap macam
= HMAP_INITIALIZER(&macam
);
71 #define MAX_OVN_TAGS 4096
73 /* Pipeline stages. */
75 /* The two pipelines in an OVN logical flow table. */
77 P_IN
, /* Ingress pipeline. */
78 P_OUT
/* Egress pipeline. */
81 /* The two purposes for which ovn-northd uses OVN logical datapaths. */
82 enum ovn_datapath_type
{
83 DP_SWITCH
, /* OVN logical switch. */
84 DP_ROUTER
/* OVN logical router. */
87 /* Returns an "enum ovn_stage" built from the arguments.
89 * (It's better to use ovn_stage_build() for type-safety reasons, but inline
90 * functions can't be used in enums or switch cases.) */
91 #define OVN_STAGE_BUILD(DP_TYPE, PIPELINE, TABLE) \
92 (((DP_TYPE) << 9) | ((PIPELINE) << 8) | (TABLE))
94 /* A stage within an OVN logical switch or router.
96 * An "enum ovn_stage" indicates whether the stage is part of a logical switch
97 * or router, whether the stage is part of the ingress or egress pipeline, and
98 * the table within that pipeline. The first three components are combined to
99 * form the stage's full name, e.g. S_SWITCH_IN_PORT_SEC_L2,
100 * S_ROUTER_OUT_DELIVERY. */
102 #define PIPELINE_STAGES \
103 /* Logical switch ingress stages. */ \
104 PIPELINE_STAGE(SWITCH, IN, PORT_SEC_L2, 0, "ls_in_port_sec_l2") \
105 PIPELINE_STAGE(SWITCH, IN, PORT_SEC_IP, 1, "ls_in_port_sec_ip") \
106 PIPELINE_STAGE(SWITCH, IN, PORT_SEC_ND, 2, "ls_in_port_sec_nd") \
107 PIPELINE_STAGE(SWITCH, IN, PRE_ACL, 3, "ls_in_pre_acl") \
108 PIPELINE_STAGE(SWITCH, IN, PRE_LB, 4, "ls_in_pre_lb") \
109 PIPELINE_STAGE(SWITCH, IN, PRE_STATEFUL, 5, "ls_in_pre_stateful") \
110 PIPELINE_STAGE(SWITCH, IN, ACL, 6, "ls_in_acl") \
111 PIPELINE_STAGE(SWITCH, IN, QOS_MARK, 7, "ls_in_qos_mark") \
112 PIPELINE_STAGE(SWITCH, IN, QOS_METER, 8, "ls_in_qos_meter") \
113 PIPELINE_STAGE(SWITCH, IN, LB, 9, "ls_in_lb") \
114 PIPELINE_STAGE(SWITCH, IN, STATEFUL, 10, "ls_in_stateful") \
115 PIPELINE_STAGE(SWITCH, IN, ARP_ND_RSP, 11, "ls_in_arp_rsp") \
116 PIPELINE_STAGE(SWITCH, IN, DHCP_OPTIONS, 12, "ls_in_dhcp_options") \
117 PIPELINE_STAGE(SWITCH, IN, DHCP_RESPONSE, 13, "ls_in_dhcp_response") \
118 PIPELINE_STAGE(SWITCH, IN, DNS_LOOKUP, 14, "ls_in_dns_lookup") \
119 PIPELINE_STAGE(SWITCH, IN, DNS_RESPONSE, 15, "ls_in_dns_response") \
120 PIPELINE_STAGE(SWITCH, IN, L2_LKUP, 16, "ls_in_l2_lkup") \
122 /* Logical switch egress stages. */ \
123 PIPELINE_STAGE(SWITCH, OUT, PRE_LB, 0, "ls_out_pre_lb") \
124 PIPELINE_STAGE(SWITCH, OUT, PRE_ACL, 1, "ls_out_pre_acl") \
125 PIPELINE_STAGE(SWITCH, OUT, PRE_STATEFUL, 2, "ls_out_pre_stateful") \
126 PIPELINE_STAGE(SWITCH, OUT, LB, 3, "ls_out_lb") \
127 PIPELINE_STAGE(SWITCH, OUT, ACL, 4, "ls_out_acl") \
128 PIPELINE_STAGE(SWITCH, OUT, QOS_MARK, 5, "ls_out_qos_mark") \
129 PIPELINE_STAGE(SWITCH, OUT, QOS_METER, 6, "ls_out_qos_meter") \
130 PIPELINE_STAGE(SWITCH, OUT, STATEFUL, 7, "ls_out_stateful") \
131 PIPELINE_STAGE(SWITCH, OUT, PORT_SEC_IP, 8, "ls_out_port_sec_ip") \
132 PIPELINE_STAGE(SWITCH, OUT, PORT_SEC_L2, 9, "ls_out_port_sec_l2") \
134 /* Logical router ingress stages. */ \
135 PIPELINE_STAGE(ROUTER, IN, ADMISSION, 0, "lr_in_admission") \
136 PIPELINE_STAGE(ROUTER, IN, IP_INPUT, 1, "lr_in_ip_input") \
137 PIPELINE_STAGE(ROUTER, IN, DEFRAG, 2, "lr_in_defrag") \
138 PIPELINE_STAGE(ROUTER, IN, UNSNAT, 3, "lr_in_unsnat") \
139 PIPELINE_STAGE(ROUTER, IN, DNAT, 4, "lr_in_dnat") \
140 PIPELINE_STAGE(ROUTER, IN, ND_RA_OPTIONS, 5, "lr_in_nd_ra_options") \
141 PIPELINE_STAGE(ROUTER, IN, ND_RA_RESPONSE, 6, "lr_in_nd_ra_response") \
142 PIPELINE_STAGE(ROUTER, IN, IP_ROUTING, 7, "lr_in_ip_routing") \
143 PIPELINE_STAGE(ROUTER, IN, ARP_RESOLVE, 8, "lr_in_arp_resolve") \
144 PIPELINE_STAGE(ROUTER, IN, GW_REDIRECT, 9, "lr_in_gw_redirect") \
145 PIPELINE_STAGE(ROUTER, IN, ARP_REQUEST, 10, "lr_in_arp_request") \
147 /* Logical router egress stages. */ \
148 PIPELINE_STAGE(ROUTER, OUT, UNDNAT, 0, "lr_out_undnat") \
149 PIPELINE_STAGE(ROUTER, OUT, SNAT, 1, "lr_out_snat") \
150 PIPELINE_STAGE(ROUTER, OUT, EGR_LOOP, 2, "lr_out_egr_loop") \
151 PIPELINE_STAGE(ROUTER, OUT, DELIVERY, 3, "lr_out_delivery")
153 #define PIPELINE_STAGE(DP_TYPE, PIPELINE, STAGE, TABLE, NAME) \
154 S_##DP_TYPE##_##PIPELINE##_##STAGE \
155 = OVN_STAGE_BUILD(DP_##DP_TYPE, P_##PIPELINE, TABLE),
157 #undef PIPELINE_STAGE
160 /* Due to various hard-coded priorities need to implement ACLs, the
161 * northbound database supports a smaller range of ACL priorities than
162 * are available to logical flows. This value is added to an ACL
163 * priority to determine the ACL's logical flow priority. */
164 #define OVN_ACL_PRI_OFFSET 1000
166 /* Register definitions specific to switches. */
167 #define REGBIT_CONNTRACK_DEFRAG "reg0[0]"
168 #define REGBIT_CONNTRACK_COMMIT "reg0[1]"
169 #define REGBIT_CONNTRACK_NAT "reg0[2]"
170 #define REGBIT_DHCP_OPTS_RESULT "reg0[3]"
171 #define REGBIT_DNS_LOOKUP_RESULT "reg0[4]"
172 #define REGBIT_ND_RA_OPTS_RESULT "reg0[5]"
174 /* Register definitions for switches and routers. */
175 #define REGBIT_NAT_REDIRECT "reg9[0]"
176 /* Indicate that this packet has been recirculated using egress
177 * loopback. This allows certain checks to be bypassed, such as a
178 * logical router dropping packets with source IP address equals
179 * one of the logical router's own IP addresses. */
180 #define REGBIT_EGRESS_LOOPBACK "reg9[1]"
182 /* Returns an "enum ovn_stage" built from the arguments. */
183 static enum ovn_stage
184 ovn_stage_build(enum ovn_datapath_type dp_type
, enum ovn_pipeline pipeline
,
187 return OVN_STAGE_BUILD(dp_type
, pipeline
, table
);
190 /* Returns the pipeline to which 'stage' belongs. */
191 static enum ovn_pipeline
192 ovn_stage_get_pipeline(enum ovn_stage stage
)
194 return (stage
>> 8) & 1;
197 /* Returns the pipeline name to which 'stage' belongs. */
199 ovn_stage_get_pipeline_name(enum ovn_stage stage
)
201 return ovn_stage_get_pipeline(stage
) == P_IN
? "ingress" : "egress";
204 /* Returns the table to which 'stage' belongs. */
206 ovn_stage_get_table(enum ovn_stage stage
)
211 /* Returns a string name for 'stage'. */
213 ovn_stage_to_str(enum ovn_stage stage
)
216 #define PIPELINE_STAGE(DP_TYPE, PIPELINE, STAGE, TABLE, NAME) \
217 case S_##DP_TYPE##_##PIPELINE##_##STAGE: return NAME;
219 #undef PIPELINE_STAGE
220 default: return "<unknown>";
224 /* Returns the type of the datapath to which a flow with the given 'stage' may
226 static enum ovn_datapath_type
227 ovn_stage_to_datapath_type(enum ovn_stage stage
)
230 #define PIPELINE_STAGE(DP_TYPE, PIPELINE, STAGE, TABLE, NAME) \
231 case S_##DP_TYPE##_##PIPELINE##_##STAGE: return DP_##DP_TYPE;
233 #undef PIPELINE_STAGE
234 default: OVS_NOT_REACHED();
242 %s: OVN northbound management daemon\n\
243 usage: %s [OPTIONS]\n\
246 --ovnnb-db=DATABASE connect to ovn-nb database at DATABASE\n\
248 --ovnsb-db=DATABASE connect to ovn-sb database at DATABASE\n\
250 --unixctl=SOCKET override default control socket name\n\
251 -h, --help display this help message\n\
252 -o, --options list available options\n\
253 -V, --version display version information\n\
254 ", program_name
, program_name
, default_nb_db(), default_sb_db());
257 stream_usage("database", true, true, false);
261 struct hmap_node hmap_node
;
266 destroy_tnlids(struct hmap
*tnlids
)
268 struct tnlid_node
*node
;
269 HMAP_FOR_EACH_POP (node
, hmap_node
, tnlids
) {
272 hmap_destroy(tnlids
);
276 add_tnlid(struct hmap
*set
, uint32_t tnlid
)
278 struct tnlid_node
*node
= xmalloc(sizeof *node
);
279 hmap_insert(set
, &node
->hmap_node
, hash_int(tnlid
, 0));
284 tnlid_in_use(const struct hmap
*set
, uint32_t tnlid
)
286 const struct tnlid_node
*node
;
287 HMAP_FOR_EACH_IN_BUCKET (node
, hmap_node
, hash_int(tnlid
, 0), set
) {
288 if (node
->tnlid
== tnlid
) {
296 next_tnlid(uint32_t tnlid
, uint32_t max
)
298 return tnlid
+ 1 <= max
? tnlid
+ 1 : 1;
302 allocate_tnlid(struct hmap
*set
, const char *name
, uint32_t max
,
305 for (uint32_t tnlid
= next_tnlid(*hint
, max
); tnlid
!= *hint
;
306 tnlid
= next_tnlid(tnlid
, max
)) {
307 if (!tnlid_in_use(set
, tnlid
)) {
308 add_tnlid(set
, tnlid
);
314 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
315 VLOG_WARN_RL(&rl
, "all %s tunnel ids exhausted", name
);
319 struct ovn_chassis_qdisc_queues
{
320 struct hmap_node key_node
;
322 struct uuid chassis_uuid
;
326 destroy_chassis_queues(struct hmap
*set
)
328 struct ovn_chassis_qdisc_queues
*node
;
329 HMAP_FOR_EACH_POP (node
, key_node
, set
) {
336 add_chassis_queue(struct hmap
*set
, struct uuid
*chassis_uuid
,
339 struct ovn_chassis_qdisc_queues
*node
= xmalloc(sizeof *node
);
340 node
->queue_id
= queue_id
;
341 memcpy(&node
->chassis_uuid
, chassis_uuid
, sizeof node
->chassis_uuid
);
342 hmap_insert(set
, &node
->key_node
, uuid_hash(chassis_uuid
));
346 chassis_queueid_in_use(const struct hmap
*set
, struct uuid
*chassis_uuid
,
349 const struct ovn_chassis_qdisc_queues
*node
;
350 HMAP_FOR_EACH_WITH_HASH (node
, key_node
, uuid_hash(chassis_uuid
), set
) {
351 if (uuid_equals(chassis_uuid
, &node
->chassis_uuid
)
352 && node
->queue_id
== queue_id
) {
360 allocate_chassis_queueid(struct hmap
*set
, struct sbrec_chassis
*chassis
)
362 for (uint32_t queue_id
= QDISC_MIN_QUEUE_ID
+ 1;
363 queue_id
<= QDISC_MAX_QUEUE_ID
;
365 if (!chassis_queueid_in_use(set
, &chassis
->header_
.uuid
, queue_id
)) {
366 add_chassis_queue(set
, &chassis
->header_
.uuid
, queue_id
);
371 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
372 VLOG_WARN_RL(&rl
, "all %s queue ids exhausted", chassis
->name
);
377 free_chassis_queueid(struct hmap
*set
, struct sbrec_chassis
*chassis
,
380 struct ovn_chassis_qdisc_queues
*node
;
381 HMAP_FOR_EACH_WITH_HASH (node
, key_node
,
382 uuid_hash(&chassis
->header_
.uuid
),
384 if (uuid_equals(&chassis
->header_
.uuid
, &node
->chassis_uuid
)
385 && node
->queue_id
== queue_id
) {
386 hmap_remove(set
, &node
->key_node
);
393 port_has_qos_params(const struct smap
*opts
)
395 return (smap_get(opts
, "qos_max_rate") ||
396 smap_get(opts
, "qos_burst"));
403 unsigned long *allocated_ipv4s
; /* A bitmap of allocated IPv4s */
404 bool ipv6_prefix_set
;
405 struct in6_addr ipv6_prefix
;
408 /* The 'key' comes from nbs->header_.uuid or nbr->header_.uuid or
409 * sb->external_ids:logical-switch. */
410 struct ovn_datapath
{
411 struct hmap_node key_node
; /* Index on 'key'. */
412 struct uuid key
; /* (nbs/nbr)->header_.uuid. */
414 const struct nbrec_logical_switch
*nbs
; /* May be NULL. */
415 const struct nbrec_logical_router
*nbr
; /* May be NULL. */
416 const struct sbrec_datapath_binding
*sb
; /* May be NULL. */
418 struct ovs_list list
; /* In list of similar records. */
420 /* Logical switch data. */
421 struct ovn_port
**router_ports
;
422 size_t n_router_ports
;
424 struct hmap port_tnlids
;
425 uint32_t port_key_hint
;
430 struct ipam_info
*ipam_info
;
432 /* OVN northd only needs to know about the logical router gateway port for
433 * NAT on a distributed router. This "distributed gateway port" is
434 * populated only when there is a "redirect-chassis" specified for one of
435 * the ports on the logical router. Otherwise this will be NULL. */
436 struct ovn_port
*l3dgw_port
;
437 /* The "derived" OVN port representing the instance of l3dgw_port on
438 * the "redirect-chassis". */
439 struct ovn_port
*l3redirect_port
;
440 struct ovn_port
*localnet_port
;
444 struct hmap_node hmap_node
;
445 struct eth_addr mac_addr
; /* Allocated MAC address. */
449 cleanup_macam(struct hmap
*macam_
)
451 struct macam_node
*node
;
452 HMAP_FOR_EACH_POP (node
, hmap_node
, macam_
) {
457 static struct ovn_datapath
*
458 ovn_datapath_create(struct hmap
*datapaths
, const struct uuid
*key
,
459 const struct nbrec_logical_switch
*nbs
,
460 const struct nbrec_logical_router
*nbr
,
461 const struct sbrec_datapath_binding
*sb
)
463 struct ovn_datapath
*od
= xzalloc(sizeof *od
);
468 hmap_init(&od
->port_tnlids
);
469 od
->port_key_hint
= 0;
470 hmap_insert(datapaths
, &od
->key_node
, uuid_hash(&od
->key
));
475 ovn_datapath_destroy(struct hmap
*datapaths
, struct ovn_datapath
*od
)
478 /* Don't remove od->list. It is used within build_datapaths() as a
479 * private list and once we've exited that function it is not safe to
481 hmap_remove(datapaths
, &od
->key_node
);
482 destroy_tnlids(&od
->port_tnlids
);
484 bitmap_free(od
->ipam_info
->allocated_ipv4s
);
487 free(od
->router_ports
);
492 /* Returns 'od''s datapath type. */
493 static enum ovn_datapath_type
494 ovn_datapath_get_type(const struct ovn_datapath
*od
)
496 return od
->nbs
? DP_SWITCH
: DP_ROUTER
;
499 static struct ovn_datapath
*
500 ovn_datapath_find(struct hmap
*datapaths
, const struct uuid
*uuid
)
502 struct ovn_datapath
*od
;
504 HMAP_FOR_EACH_WITH_HASH (od
, key_node
, uuid_hash(uuid
), datapaths
) {
505 if (uuid_equals(uuid
, &od
->key
)) {
512 static struct ovn_datapath
*
513 ovn_datapath_from_sbrec(struct hmap
*datapaths
,
514 const struct sbrec_datapath_binding
*sb
)
518 if (!smap_get_uuid(&sb
->external_ids
, "logical-switch", &key
) &&
519 !smap_get_uuid(&sb
->external_ids
, "logical-router", &key
)) {
522 return ovn_datapath_find(datapaths
, &key
);
526 lrouter_is_enabled(const struct nbrec_logical_router
*lrouter
)
528 return !lrouter
->enabled
|| *lrouter
->enabled
;
532 init_ipam_info_for_datapath(struct ovn_datapath
*od
)
538 const char *subnet_str
= smap_get(&od
->nbs
->other_config
, "subnet");
539 const char *ipv6_prefix
= smap_get(&od
->nbs
->other_config
, "ipv6_prefix");
542 od
->ipam_info
= xzalloc(sizeof *od
->ipam_info
);
543 od
->ipam_info
->ipv6_prefix_set
= ipv6_parse(
544 ipv6_prefix
, &od
->ipam_info
->ipv6_prefix
);
551 ovs_be32 subnet
, mask
;
552 char *error
= ip_parse_masked(subnet_str
, &subnet
, &mask
);
553 if (error
|| mask
== OVS_BE32_MAX
|| !ip_is_cidr(mask
)) {
554 static struct vlog_rate_limit rl
555 = VLOG_RATE_LIMIT_INIT(5, 1);
556 VLOG_WARN_RL(&rl
, "bad 'subnet' %s", subnet_str
);
561 if (!od
->ipam_info
) {
562 od
->ipam_info
= xzalloc(sizeof *od
->ipam_info
);
564 od
->ipam_info
->start_ipv4
= ntohl(subnet
) + 1;
565 od
->ipam_info
->total_ipv4s
= ~ntohl(mask
);
566 od
->ipam_info
->allocated_ipv4s
=
567 bitmap_allocate(od
->ipam_info
->total_ipv4s
);
569 /* Mark first IP as taken */
570 bitmap_set1(od
->ipam_info
->allocated_ipv4s
, 0);
572 /* Check if there are any reserver IPs (list) to be excluded from IPAM */
573 const char *exclude_ip_list
= smap_get(&od
->nbs
->other_config
,
575 if (!exclude_ip_list
) {
580 lexer_init(&lexer
, exclude_ip_list
);
581 /* exclude_ip_list could be in the format -
582 * "10.0.0.4 10.0.0.10 10.0.0.20..10.0.0.50 10.0.0.100..10.0.0.110".
585 while (lexer
.token
.type
!= LEX_T_END
) {
586 if (lexer
.token
.type
!= LEX_T_INTEGER
) {
587 lexer_syntax_error(&lexer
, "expecting address");
590 uint32_t start
= ntohl(lexer
.token
.value
.ipv4
);
593 uint32_t end
= start
+ 1;
594 if (lexer_match(&lexer
, LEX_T_ELLIPSIS
)) {
595 if (lexer
.token
.type
!= LEX_T_INTEGER
) {
596 lexer_syntax_error(&lexer
, "expecting address range");
599 end
= ntohl(lexer
.token
.value
.ipv4
) + 1;
603 /* Clamp start...end to fit the subnet. */
604 start
= MAX(od
->ipam_info
->start_ipv4
, start
);
605 end
= MIN(od
->ipam_info
->start_ipv4
+ od
->ipam_info
->total_ipv4s
, end
);
607 bitmap_set_multiple(od
->ipam_info
->allocated_ipv4s
,
608 start
- od
->ipam_info
->start_ipv4
,
611 lexer_error(&lexer
, "excluded addresses not in subnet");
615 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
616 VLOG_WARN_RL(&rl
, "logical switch "UUID_FMT
": bad exclude_ips (%s)",
617 UUID_ARGS(&od
->key
), lexer
.error
);
619 lexer_destroy(&lexer
);
623 ovn_datapath_update_external_ids(struct ovn_datapath
*od
)
625 /* Get the logical-switch or logical-router UUID to set in
627 char uuid_s
[UUID_LEN
+ 1];
628 sprintf(uuid_s
, UUID_FMT
, UUID_ARGS(&od
->key
));
629 const char *key
= od
->nbs
? "logical-switch" : "logical-router";
631 /* Get names to set in external-ids. */
632 const char *name
= od
->nbs
? od
->nbs
->name
: od
->nbr
->name
;
633 const char *name2
= (od
->nbs
634 ? smap_get(&od
->nbs
->external_ids
,
635 "neutron:network_name")
636 : smap_get(&od
->nbr
->external_ids
,
637 "neutron:router_name"));
639 /* Set external-ids. */
640 struct smap ids
= SMAP_INITIALIZER(&ids
);
641 smap_add(&ids
, key
, uuid_s
);
642 smap_add(&ids
, "name", name
);
643 if (name2
&& name2
[0]) {
644 smap_add(&ids
, "name2", name2
);
646 sbrec_datapath_binding_set_external_ids(od
->sb
, &ids
);
651 join_datapaths(struct northd_context
*ctx
, struct hmap
*datapaths
,
652 struct ovs_list
*sb_only
, struct ovs_list
*nb_only
,
653 struct ovs_list
*both
)
655 hmap_init(datapaths
);
656 ovs_list_init(sb_only
);
657 ovs_list_init(nb_only
);
660 const struct sbrec_datapath_binding
*sb
, *sb_next
;
661 SBREC_DATAPATH_BINDING_FOR_EACH_SAFE (sb
, sb_next
, ctx
->ovnsb_idl
) {
663 if (!smap_get_uuid(&sb
->external_ids
, "logical-switch", &key
) &&
664 !smap_get_uuid(&sb
->external_ids
, "logical-router", &key
)) {
665 ovsdb_idl_txn_add_comment(
667 "deleting Datapath_Binding "UUID_FMT
" that lacks "
668 "external-ids:logical-switch and "
669 "external-ids:logical-router",
670 UUID_ARGS(&sb
->header_
.uuid
));
671 sbrec_datapath_binding_delete(sb
);
675 if (ovn_datapath_find(datapaths
, &key
)) {
676 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
678 &rl
, "deleting Datapath_Binding "UUID_FMT
" with "
679 "duplicate external-ids:logical-switch/router "UUID_FMT
,
680 UUID_ARGS(&sb
->header_
.uuid
), UUID_ARGS(&key
));
681 sbrec_datapath_binding_delete(sb
);
685 struct ovn_datapath
*od
= ovn_datapath_create(datapaths
, &key
,
687 ovs_list_push_back(sb_only
, &od
->list
);
690 const struct nbrec_logical_switch
*nbs
;
691 NBREC_LOGICAL_SWITCH_FOR_EACH (nbs
, ctx
->ovnnb_idl
) {
692 struct ovn_datapath
*od
= ovn_datapath_find(datapaths
,
696 ovs_list_remove(&od
->list
);
697 ovs_list_push_back(both
, &od
->list
);
698 ovn_datapath_update_external_ids(od
);
700 od
= ovn_datapath_create(datapaths
, &nbs
->header_
.uuid
,
702 ovs_list_push_back(nb_only
, &od
->list
);
705 init_ipam_info_for_datapath(od
);
708 const struct nbrec_logical_router
*nbr
;
709 NBREC_LOGICAL_ROUTER_FOR_EACH (nbr
, ctx
->ovnnb_idl
) {
710 if (!lrouter_is_enabled(nbr
)) {
714 struct ovn_datapath
*od
= ovn_datapath_find(datapaths
,
719 ovs_list_remove(&od
->list
);
720 ovs_list_push_back(both
, &od
->list
);
721 ovn_datapath_update_external_ids(od
);
724 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
726 "duplicate UUID "UUID_FMT
" in OVN_Northbound",
727 UUID_ARGS(&nbr
->header_
.uuid
));
731 od
= ovn_datapath_create(datapaths
, &nbr
->header_
.uuid
,
733 ovs_list_push_back(nb_only
, &od
->list
);
739 ovn_datapath_allocate_key(struct hmap
*dp_tnlids
)
741 static uint32_t hint
;
742 return allocate_tnlid(dp_tnlids
, "datapath", (1u << 24) - 1, &hint
);
745 /* Updates the southbound Datapath_Binding table so that it contains the
746 * logical switches and routers specified by the northbound database.
748 * Initializes 'datapaths' to contain a "struct ovn_datapath" for every logical
749 * switch and router. */
751 build_datapaths(struct northd_context
*ctx
, struct hmap
*datapaths
)
753 struct ovs_list sb_only
, nb_only
, both
;
755 join_datapaths(ctx
, datapaths
, &sb_only
, &nb_only
, &both
);
757 if (!ovs_list_is_empty(&nb_only
)) {
758 /* First index the in-use datapath tunnel IDs. */
759 struct hmap dp_tnlids
= HMAP_INITIALIZER(&dp_tnlids
);
760 struct ovn_datapath
*od
;
761 LIST_FOR_EACH (od
, list
, &both
) {
762 add_tnlid(&dp_tnlids
, od
->sb
->tunnel_key
);
765 /* Add southbound record for each unmatched northbound record. */
766 LIST_FOR_EACH (od
, list
, &nb_only
) {
767 uint16_t tunnel_key
= ovn_datapath_allocate_key(&dp_tnlids
);
772 od
->sb
= sbrec_datapath_binding_insert(ctx
->ovnsb_txn
);
773 ovn_datapath_update_external_ids(od
);
774 sbrec_datapath_binding_set_tunnel_key(od
->sb
, tunnel_key
);
776 destroy_tnlids(&dp_tnlids
);
779 /* Delete southbound records without northbound matches. */
780 struct ovn_datapath
*od
, *next
;
781 LIST_FOR_EACH_SAFE (od
, next
, list
, &sb_only
) {
782 ovs_list_remove(&od
->list
);
783 sbrec_datapath_binding_delete(od
->sb
);
784 ovn_datapath_destroy(datapaths
, od
);
789 struct hmap_node key_node
; /* Index on 'key'. */
790 char *key
; /* nbs->name, nbr->name, sb->logical_port. */
791 char *json_key
; /* 'key', quoted for use in JSON. */
793 const struct sbrec_port_binding
*sb
; /* May be NULL. */
795 /* Logical switch port data. */
796 const struct nbrec_logical_switch_port
*nbsp
; /* May be NULL. */
798 struct lport_addresses
*lsp_addrs
; /* Logical switch port addresses. */
799 unsigned int n_lsp_addrs
;
801 struct lport_addresses
*ps_addrs
; /* Port security addresses. */
802 unsigned int n_ps_addrs
;
804 /* Logical router port data. */
805 const struct nbrec_logical_router_port
*nbrp
; /* May be NULL. */
807 struct lport_addresses lrp_networks
;
809 bool derived
; /* Indicates whether this is an additional port
810 * derived from nbsp or nbrp. */
814 * - A switch port S of type "router" has a router port R as a peer,
815 * and R in turn has S has its peer.
817 * - Two connected logical router ports have each other as peer. */
818 struct ovn_port
*peer
;
820 struct ovn_datapath
*od
;
822 struct ovs_list list
; /* In list of similar records. */
825 static struct ovn_port
*
826 ovn_port_create(struct hmap
*ports
, const char *key
,
827 const struct nbrec_logical_switch_port
*nbsp
,
828 const struct nbrec_logical_router_port
*nbrp
,
829 const struct sbrec_port_binding
*sb
)
831 struct ovn_port
*op
= xzalloc(sizeof *op
);
833 struct ds json_key
= DS_EMPTY_INITIALIZER
;
834 json_string_escape(key
, &json_key
);
835 op
->json_key
= ds_steal_cstr(&json_key
);
837 op
->key
= xstrdup(key
);
842 hmap_insert(ports
, &op
->key_node
, hash_string(op
->key
, 0));
847 ovn_port_destroy(struct hmap
*ports
, struct ovn_port
*port
)
850 /* Don't remove port->list. It is used within build_ports() as a
851 * private list and once we've exited that function it is not safe to
853 hmap_remove(ports
, &port
->key_node
);
855 for (int i
= 0; i
< port
->n_lsp_addrs
; i
++) {
856 destroy_lport_addresses(&port
->lsp_addrs
[i
]);
858 free(port
->lsp_addrs
);
860 for (int i
= 0; i
< port
->n_ps_addrs
; i
++) {
861 destroy_lport_addresses(&port
->ps_addrs
[i
]);
863 free(port
->ps_addrs
);
865 destroy_lport_addresses(&port
->lrp_networks
);
866 free(port
->json_key
);
872 static struct ovn_port
*
873 ovn_port_find(struct hmap
*ports
, const char *name
)
877 HMAP_FOR_EACH_WITH_HASH (op
, key_node
, hash_string(name
, 0), ports
) {
878 if (!strcmp(op
->key
, name
)) {
886 ovn_port_allocate_key(struct ovn_datapath
*od
)
888 return allocate_tnlid(&od
->port_tnlids
, "port",
889 (1u << 15) - 1, &od
->port_key_hint
);
893 chassis_redirect_name(const char *port_name
)
895 return xasprintf("cr-%s", port_name
);
899 ipam_is_duplicate_mac(struct eth_addr
*ea
, uint64_t mac64
, bool warn
)
901 struct macam_node
*macam_node
;
902 HMAP_FOR_EACH_WITH_HASH (macam_node
, hmap_node
, hash_uint64(mac64
),
904 if (eth_addr_equals(*ea
, macam_node
->mac_addr
)) {
906 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
907 VLOG_WARN_RL(&rl
, "Duplicate MAC set: "ETH_ADDR_FMT
,
908 ETH_ADDR_ARGS(macam_node
->mac_addr
));
917 ipam_insert_mac(struct eth_addr
*ea
, bool check
)
923 uint64_t mac64
= eth_addr_to_uint64(*ea
);
924 /* If the new MAC was not assigned by this address management system or
925 * check is true and the new MAC is a duplicate, do not insert it into the
927 if (((mac64
^ MAC_ADDR_PREFIX
) >> 24)
928 || (check
&& ipam_is_duplicate_mac(ea
, mac64
, true))) {
932 struct macam_node
*new_macam_node
= xmalloc(sizeof *new_macam_node
);
933 new_macam_node
->mac_addr
= *ea
;
934 hmap_insert(&macam
, &new_macam_node
->hmap_node
, hash_uint64(mac64
));
938 ipam_insert_ip(struct ovn_datapath
*od
, uint32_t ip
)
940 if (!od
|| !od
->ipam_info
|| !od
->ipam_info
->allocated_ipv4s
) {
944 if (ip
>= od
->ipam_info
->start_ipv4
&&
945 ip
< (od
->ipam_info
->start_ipv4
+ od
->ipam_info
->total_ipv4s
)) {
946 bitmap_set1(od
->ipam_info
->allocated_ipv4s
,
947 ip
- od
->ipam_info
->start_ipv4
);
952 ipam_insert_lsp_addresses(struct ovn_datapath
*od
, struct ovn_port
*op
,
955 if (!od
|| !op
|| !address
|| !strcmp(address
, "unknown")
956 || !strcmp(address
, "router") || is_dynamic_lsp_address(address
)) {
960 struct lport_addresses laddrs
;
961 if (!extract_lsp_addresses(address
, &laddrs
)) {
962 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
963 VLOG_WARN_RL(&rl
, "Extract addresses failed.");
966 ipam_insert_mac(&laddrs
.ea
, true);
968 /* IP is only added to IPAM if the switch's subnet option
969 * is set, whereas MAC is always added to MACAM. */
970 if (!od
->ipam_info
|| !od
->ipam_info
->allocated_ipv4s
) {
971 destroy_lport_addresses(&laddrs
);
975 for (size_t j
= 0; j
< laddrs
.n_ipv4_addrs
; j
++) {
976 uint32_t ip
= ntohl(laddrs
.ipv4_addrs
[j
].addr
);
977 ipam_insert_ip(od
, ip
);
980 destroy_lport_addresses(&laddrs
);
984 ipam_add_port_addresses(struct ovn_datapath
*od
, struct ovn_port
*op
)
991 /* Add all the port's addresses to address data structures. */
992 for (size_t i
= 0; i
< op
->nbsp
->n_addresses
; i
++) {
993 ipam_insert_lsp_addresses(od
, op
, op
->nbsp
->addresses
[i
]);
995 if (op
->nbsp
->dynamic_addresses
) {
996 ipam_insert_lsp_addresses(od
, op
, op
->nbsp
->dynamic_addresses
);
998 } else if (op
->nbrp
) {
999 struct lport_addresses lrp_networks
;
1000 if (!extract_lrp_networks(op
->nbrp
, &lrp_networks
)) {
1001 static struct vlog_rate_limit rl
1002 = VLOG_RATE_LIMIT_INIT(1, 1);
1003 VLOG_WARN_RL(&rl
, "Extract addresses failed.");
1006 ipam_insert_mac(&lrp_networks
.ea
, true);
1008 if (!op
->peer
|| !op
->peer
->nbsp
|| !op
->peer
->od
|| !op
->peer
->od
->nbs
1009 || !smap_get(&op
->peer
->od
->nbs
->other_config
, "subnet")) {
1010 destroy_lport_addresses(&lrp_networks
);
1014 for (size_t i
= 0; i
< lrp_networks
.n_ipv4_addrs
; i
++) {
1015 uint32_t ip
= ntohl(lrp_networks
.ipv4_addrs
[i
].addr
);
1016 ipam_insert_ip(op
->peer
->od
, ip
);
1019 destroy_lport_addresses(&lrp_networks
);
1024 ipam_get_unused_mac(void)
1026 /* Stores the suffix of the most recently ipam-allocated MAC address. */
1027 static uint32_t last_mac
;
1030 struct eth_addr mac
;
1031 uint32_t mac_addr_suffix
, i
;
1032 for (i
= 0; i
< MAC_ADDR_SPACE
- 1; i
++) {
1033 /* The tentative MAC's suffix will be in the interval (1, 0xfffffe). */
1034 mac_addr_suffix
= ((last_mac
+ i
) % (MAC_ADDR_SPACE
- 1)) + 1;
1035 mac64
= MAC_ADDR_PREFIX
| mac_addr_suffix
;
1036 eth_addr_from_uint64(mac64
, &mac
);
1037 if (!ipam_is_duplicate_mac(&mac
, mac64
, false)) {
1038 last_mac
= mac_addr_suffix
;
1043 if (i
== MAC_ADDR_SPACE
) {
1044 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
1045 VLOG_WARN_RL(&rl
, "MAC address space exhausted.");
1053 ipam_get_unused_ip(struct ovn_datapath
*od
)
1055 if (!od
|| !od
->ipam_info
|| !od
->ipam_info
->allocated_ipv4s
) {
1059 size_t new_ip_index
= bitmap_scan(od
->ipam_info
->allocated_ipv4s
, 0, 0,
1060 od
->ipam_info
->total_ipv4s
- 1);
1061 if (new_ip_index
== od
->ipam_info
->total_ipv4s
- 1) {
1062 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
1063 VLOG_WARN_RL( &rl
, "Subnet address space has been exhausted.");
1067 return od
->ipam_info
->start_ipv4
+ new_ip_index
;
1071 ipam_allocate_addresses(struct ovn_datapath
*od
, struct ovn_port
*op
,
1072 const char *addrspec
)
1074 if (!op
->nbsp
|| !od
->ipam_info
) {
1078 /* Get or generate MAC address. */
1079 struct eth_addr mac
;
1082 if (ovs_scan(addrspec
, ETH_ADDR_SCAN_FMT
" dynamic%n",
1083 ETH_ADDR_SCAN_ARGS(mac
), &n
)
1084 && addrspec
[n
] == '\0') {
1085 dynamic_mac
= false;
1087 uint64_t mac64
= ipam_get_unused_mac();
1091 eth_addr_from_uint64(mac64
, &mac
);
1095 /* Generate IPv4 address, if desirable. */
1096 bool dynamic_ip4
= od
->ipam_info
->allocated_ipv4s
!= NULL
;
1097 uint32_t ip4
= dynamic_ip4
? ipam_get_unused_ip(od
) : 0;
1099 /* Generate IPv6 address, if desirable. */
1100 bool dynamic_ip6
= od
->ipam_info
->ipv6_prefix_set
;
1101 struct in6_addr ip6
;
1103 in6_generate_eui64(mac
, &od
->ipam_info
->ipv6_prefix
, &ip6
);
1106 /* If we didn't generate anything, bail out. */
1107 if (!dynamic_ip4
&& !dynamic_ip6
) {
1111 /* Save the dynamic addresses. */
1112 struct ds new_addr
= DS_EMPTY_INITIALIZER
;
1113 ds_put_format(&new_addr
, ETH_ADDR_FMT
, ETH_ADDR_ARGS(mac
));
1114 if (dynamic_ip4
&& ip4
) {
1115 ipam_insert_ip(od
, ip4
);
1116 ds_put_format(&new_addr
, " "IP_FMT
, IP_ARGS(htonl(ip4
)));
1119 char ip6_s
[INET6_ADDRSTRLEN
+ 1];
1120 ipv6_string_mapped(ip6_s
, &ip6
);
1121 ds_put_format(&new_addr
, " %s", ip6_s
);
1123 ipam_insert_mac(&mac
, !dynamic_mac
);
1124 nbrec_logical_switch_port_set_dynamic_addresses(op
->nbsp
,
1125 ds_cstr(&new_addr
));
1126 ds_destroy(&new_addr
);
1131 build_ipam(struct hmap
*datapaths
, struct hmap
*ports
)
1133 /* IPAM generally stands for IP address management. In non-virtualized
1134 * world, MAC addresses come with the hardware. But, with virtualized
1135 * workloads, they need to be assigned and managed. This function
1136 * does both IP address management (ipam) and MAC address management
1139 /* If the switch's other_config:subnet is set, allocate new addresses for
1140 * ports that have the "dynamic" keyword in their addresses column. */
1141 struct ovn_datapath
*od
;
1142 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
1143 if (!od
->nbs
|| !od
->ipam_info
) {
1147 struct ovn_port
*op
;
1148 for (size_t i
= 0; i
< od
->nbs
->n_ports
; i
++) {
1149 const struct nbrec_logical_switch_port
*nbsp
=
1156 op
= ovn_port_find(ports
, nbsp
->name
);
1157 if (!op
|| (op
->nbsp
&& op
->peer
)) {
1158 /* Do not allocate addresses for logical switch ports that
1163 for (size_t j
= 0; j
< nbsp
->n_addresses
; j
++) {
1164 if (is_dynamic_lsp_address(nbsp
->addresses
[j
])
1165 && !nbsp
->dynamic_addresses
) {
1166 if (!ipam_allocate_addresses(od
, op
, nbsp
->addresses
[j
])
1167 || !extract_lsp_addresses(nbsp
->dynamic_addresses
,
1168 &op
->lsp_addrs
[op
->n_lsp_addrs
])) {
1169 static struct vlog_rate_limit rl
1170 = VLOG_RATE_LIMIT_INIT(1, 1);
1171 VLOG_INFO_RL(&rl
, "Failed to allocate address.");
1179 if (!nbsp
->n_addresses
&& nbsp
->dynamic_addresses
) {
1180 nbrec_logical_switch_port_set_dynamic_addresses(op
->nbsp
,
1187 /* Tag allocation for nested containers.
1189 * For a logical switch port with 'parent_name' and a request to allocate tags,
1190 * keeps a track of all allocated tags. */
1191 struct tag_alloc_node
{
1192 struct hmap_node hmap_node
;
1194 unsigned long *allocated_tags
; /* A bitmap to track allocated tags. */
1198 tag_alloc_destroy(struct hmap
*tag_alloc_table
)
1200 struct tag_alloc_node
*node
;
1201 HMAP_FOR_EACH_POP (node
, hmap_node
, tag_alloc_table
) {
1202 bitmap_free(node
->allocated_tags
);
1203 free(node
->parent_name
);
1206 hmap_destroy(tag_alloc_table
);
1209 static struct tag_alloc_node
*
1210 tag_alloc_get_node(struct hmap
*tag_alloc_table
, const char *parent_name
)
1212 /* If a node for the 'parent_name' exists, return it. */
1213 struct tag_alloc_node
*tag_alloc_node
;
1214 HMAP_FOR_EACH_WITH_HASH (tag_alloc_node
, hmap_node
,
1215 hash_string(parent_name
, 0),
1217 if (!strcmp(tag_alloc_node
->parent_name
, parent_name
)) {
1218 return tag_alloc_node
;
1222 /* Create a new node. */
1223 tag_alloc_node
= xmalloc(sizeof *tag_alloc_node
);
1224 tag_alloc_node
->parent_name
= xstrdup(parent_name
);
1225 tag_alloc_node
->allocated_tags
= bitmap_allocate(MAX_OVN_TAGS
);
1226 /* Tag 0 is invalid for nested containers. */
1227 bitmap_set1(tag_alloc_node
->allocated_tags
, 0);
1228 hmap_insert(tag_alloc_table
, &tag_alloc_node
->hmap_node
,
1229 hash_string(parent_name
, 0));
1231 return tag_alloc_node
;
1235 tag_alloc_add_existing_tags(struct hmap
*tag_alloc_table
,
1236 const struct nbrec_logical_switch_port
*nbsp
)
1238 /* Add the tags of already existing nested containers. If there is no
1239 * 'nbsp->parent_name' or no 'nbsp->tag' set, there is nothing to do. */
1240 if (!nbsp
->parent_name
|| !nbsp
->parent_name
[0] || !nbsp
->tag
) {
1244 struct tag_alloc_node
*tag_alloc_node
;
1245 tag_alloc_node
= tag_alloc_get_node(tag_alloc_table
, nbsp
->parent_name
);
1246 bitmap_set1(tag_alloc_node
->allocated_tags
, *nbsp
->tag
);
1250 tag_alloc_create_new_tag(struct hmap
*tag_alloc_table
,
1251 const struct nbrec_logical_switch_port
*nbsp
)
1253 if (!nbsp
->tag_request
) {
1257 if (nbsp
->parent_name
&& nbsp
->parent_name
[0]
1258 && *nbsp
->tag_request
== 0) {
1259 /* For nested containers that need allocation, do the allocation. */
1262 /* This has already been allocated. */
1266 struct tag_alloc_node
*tag_alloc_node
;
1268 tag_alloc_node
= tag_alloc_get_node(tag_alloc_table
,
1270 tag
= bitmap_scan(tag_alloc_node
->allocated_tags
, 0, 1, MAX_OVN_TAGS
);
1271 if (tag
== MAX_OVN_TAGS
) {
1272 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
1273 VLOG_ERR_RL(&rl
, "out of vlans for logical switch ports with "
1274 "parent %s", nbsp
->parent_name
);
1277 bitmap_set1(tag_alloc_node
->allocated_tags
, tag
);
1278 nbrec_logical_switch_port_set_tag(nbsp
, &tag
, 1);
1279 } else if (*nbsp
->tag_request
!= 0) {
1280 /* For everything else, copy the contents of 'tag_request' to 'tag'. */
1281 nbrec_logical_switch_port_set_tag(nbsp
, nbsp
->tag_request
, 1);
1287 * This function checks if the MAC in "address" parameter (if present) is
1288 * different from the one stored in Logical_Switch_Port.dynamic_addresses
1292 check_and_update_mac_in_dynamic_addresses(
1293 const char *address
,
1294 const struct nbrec_logical_switch_port
*nbsp
)
1296 if (!nbsp
->dynamic_addresses
) {
1301 if (!ovs_scan_len(address
, &buf_index
,
1302 ETH_ADDR_SCAN_FMT
, ETH_ADDR_SCAN_ARGS(ea
))) {
1306 struct eth_addr present_ea
;
1308 if (ovs_scan_len(nbsp
->dynamic_addresses
, &buf_index
,
1309 ETH_ADDR_SCAN_FMT
, ETH_ADDR_SCAN_ARGS(present_ea
))
1310 && !eth_addr_equals(ea
, present_ea
)) {
1311 /* MAC address has changed. Update it */
1312 char *new_addr
= xasprintf(
1313 ETH_ADDR_FMT
"%s", ETH_ADDR_ARGS(ea
),
1314  
->dynamic_addresses
[buf_index
]);
1315 nbrec_logical_switch_port_set_dynamic_addresses(
1322 join_logical_ports(struct northd_context
*ctx
,
1323 struct hmap
*datapaths
, struct hmap
*ports
,
1324 struct hmap
*chassis_qdisc_queues
,
1325 struct hmap
*tag_alloc_table
, struct ovs_list
*sb_only
,
1326 struct ovs_list
*nb_only
, struct ovs_list
*both
)
1329 ovs_list_init(sb_only
);
1330 ovs_list_init(nb_only
);
1331 ovs_list_init(both
);
1333 const struct sbrec_port_binding
*sb
;
1334 SBREC_PORT_BINDING_FOR_EACH (sb
, ctx
->ovnsb_idl
) {
1335 struct ovn_port
*op
= ovn_port_create(ports
, sb
->logical_port
,
1337 ovs_list_push_back(sb_only
, &op
->list
);
1340 struct ovn_datapath
*od
;
1341 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
1343 for (size_t i
= 0; i
< od
->nbs
->n_ports
; i
++) {
1344 const struct nbrec_logical_switch_port
*nbsp
1345 = od
->nbs
->ports
[i
];
1346 struct ovn_port
*op
= ovn_port_find(ports
, nbsp
->name
);
1348 if (op
->nbsp
|| op
->nbrp
) {
1349 static struct vlog_rate_limit rl
1350 = VLOG_RATE_LIMIT_INIT(5, 1);
1351 VLOG_WARN_RL(&rl
, "duplicate logical port %s",
1356 ovs_list_remove(&op
->list
);
1358 uint32_t queue_id
= smap_get_int(&op
->sb
->options
,
1359 "qdisc_queue_id", 0);
1360 if (queue_id
&& op
->sb
->chassis
) {
1362 chassis_qdisc_queues
, &op
->sb
->chassis
->header_
.uuid
,
1366 ovs_list_push_back(both
, &op
->list
);
1368 /* This port exists due to a SB binding, but should
1369 * not have been initialized fully. */
1370 ovs_assert(!op
->n_lsp_addrs
&& !op
->n_ps_addrs
);
1372 op
= ovn_port_create(ports
, nbsp
->name
, nbsp
, NULL
, NULL
);
1373 ovs_list_push_back(nb_only
, &op
->list
);
1376 if (!strcmp(nbsp
->type
, "localnet")) {
1377 od
->localnet_port
= op
;
1381 = xmalloc(sizeof *op
->lsp_addrs
* nbsp
->n_addresses
);
1382 for (size_t j
= 0; j
< nbsp
->n_addresses
; j
++) {
1383 if (!strcmp(nbsp
->addresses
[j
], "unknown")
1384 || !strcmp(nbsp
->addresses
[j
], "router")) {
1387 if (is_dynamic_lsp_address(nbsp
->addresses
[j
])) {
1388 if (nbsp
->dynamic_addresses
) {
1389 check_and_update_mac_in_dynamic_addresses(
1390 nbsp
->addresses
[j
], nbsp
);
1391 if (!extract_lsp_addresses(nbsp
->dynamic_addresses
,
1392 &op
->lsp_addrs
[op
->n_lsp_addrs
])) {
1393 static struct vlog_rate_limit rl
1394 = VLOG_RATE_LIMIT_INIT(1, 1);
1395 VLOG_INFO_RL(&rl
, "invalid syntax '%s' in "
1396 "logical switch port "
1397 "dynamic_addresses. No "
1398 "MAC address found",
1399 op
->nbsp
->dynamic_addresses
);
1405 } else if (!extract_lsp_addresses(nbsp
->addresses
[j
],
1406 &op
->lsp_addrs
[op
->n_lsp_addrs
])) {
1407 static struct vlog_rate_limit rl
1408 = VLOG_RATE_LIMIT_INIT(1, 1);
1409 VLOG_INFO_RL(&rl
, "invalid syntax '%s' in logical "
1410 "switch port addresses. No MAC "
1412 op
->nbsp
->addresses
[j
]);
1419 = xmalloc(sizeof *op
->ps_addrs
* nbsp
->n_port_security
);
1420 for (size_t j
= 0; j
< nbsp
->n_port_security
; j
++) {
1421 if (!extract_lsp_addresses(nbsp
->port_security
[j
],
1422 &op
->ps_addrs
[op
->n_ps_addrs
])) {
1423 static struct vlog_rate_limit rl
1424 = VLOG_RATE_LIMIT_INIT(1, 1);
1425 VLOG_INFO_RL(&rl
, "invalid syntax '%s' in port "
1426 "security. No MAC address found",
1427 op
->nbsp
->port_security
[j
]);
1434 ipam_add_port_addresses(od
, op
);
1435 tag_alloc_add_existing_tags(tag_alloc_table
, nbsp
);
1438 for (size_t i
= 0; i
< od
->nbr
->n_ports
; i
++) {
1439 const struct nbrec_logical_router_port
*nbrp
1440 = od
->nbr
->ports
[i
];
1442 struct lport_addresses lrp_networks
;
1443 if (!extract_lrp_networks(nbrp
, &lrp_networks
)) {
1444 static struct vlog_rate_limit rl
1445 = VLOG_RATE_LIMIT_INIT(5, 1);
1446 VLOG_WARN_RL(&rl
, "bad 'mac' %s", nbrp
->mac
);
1450 if (!lrp_networks
.n_ipv4_addrs
&& !lrp_networks
.n_ipv6_addrs
) {
1454 struct ovn_port
*op
= ovn_port_find(ports
, nbrp
->name
);
1456 if (op
->nbsp
|| op
->nbrp
) {
1457 static struct vlog_rate_limit rl
1458 = VLOG_RATE_LIMIT_INIT(5, 1);
1459 VLOG_WARN_RL(&rl
, "duplicate logical router port %s",
1464 ovs_list_remove(&op
->list
);
1465 ovs_list_push_back(both
, &op
->list
);
1467 /* This port exists but should not have been
1468 * initialized fully. */
1469 ovs_assert(!op
->lrp_networks
.n_ipv4_addrs
1470 && !op
->lrp_networks
.n_ipv6_addrs
);
1472 op
= ovn_port_create(ports
, nbrp
->name
, NULL
, nbrp
, NULL
);
1473 ovs_list_push_back(nb_only
, &op
->list
);
1476 op
->lrp_networks
= lrp_networks
;
1478 ipam_add_port_addresses(op
->od
, op
);
1480 const char *redirect_chassis
= smap_get(&op
->nbrp
->options
,
1481 "redirect-chassis");
1482 if (redirect_chassis
|| op
->nbrp
->n_gateway_chassis
) {
1483 /* Additional "derived" ovn_port crp represents the
1484 * instance of op on the "redirect-chassis". */
1485 const char *gw_chassis
= smap_get(&op
->od
->nbr
->options
,
1488 static struct vlog_rate_limit rl
1489 = VLOG_RATE_LIMIT_INIT(1, 1);
1490 VLOG_WARN_RL(&rl
, "Bad configuration: "
1491 "redirect-chassis configured on port %s "
1492 "on L3 gateway router", nbrp
->name
);
1495 if (od
->l3dgw_port
|| od
->l3redirect_port
) {
1496 static struct vlog_rate_limit rl
1497 = VLOG_RATE_LIMIT_INIT(1, 1);
1498 VLOG_WARN_RL(&rl
, "Bad configuration: multiple ports "
1499 "with redirect-chassis on same logical "
1500 "router %s", od
->nbr
->name
);
1504 char *redirect_name
= chassis_redirect_name(nbrp
->name
);
1505 struct ovn_port
*crp
= ovn_port_find(ports
, redirect_name
);
1507 crp
->derived
= true;
1509 ovs_list_remove(&crp
->list
);
1510 ovs_list_push_back(both
, &crp
->list
);
1512 crp
= ovn_port_create(ports
, redirect_name
,
1514 crp
->derived
= true;
1515 ovs_list_push_back(nb_only
, &crp
->list
);
1518 free(redirect_name
);
1520 /* Set l3dgw_port and l3redirect_port in od, for later
1521 * use during flow creation. */
1522 od
->l3dgw_port
= op
;
1523 od
->l3redirect_port
= crp
;
1529 /* Connect logical router ports, and logical switch ports of type "router",
1530 * to their peers. */
1531 struct ovn_port
*op
;
1532 HMAP_FOR_EACH (op
, key_node
, ports
) {
1533 if (op
->nbsp
&& !strcmp(op
->nbsp
->type
, "router") && !op
->derived
) {
1534 const char *peer_name
= smap_get(&op
->nbsp
->options
, "router-port");
1539 struct ovn_port
*peer
= ovn_port_find(ports
, peer_name
);
1540 if (!peer
|| !peer
->nbrp
) {
1546 op
->od
->router_ports
= xrealloc(
1547 op
->od
->router_ports
,
1548 sizeof *op
->od
->router_ports
* (op
->od
->n_router_ports
+ 1));
1549 op
->od
->router_ports
[op
->od
->n_router_ports
++] = op
;
1551 /* Fill op->lsp_addrs for op->nbsp->addresses[] with
1552 * contents "router", which was skipped in the loop above. */
1553 for (size_t j
= 0; j
< op
->nbsp
->n_addresses
; j
++) {
1554 if (!strcmp(op
->nbsp
->addresses
[j
], "router")) {
1555 if (extract_lrp_networks(peer
->nbrp
,
1556 &op
->lsp_addrs
[op
->n_lsp_addrs
])) {
1562 } else if (op
->nbrp
&& op
->nbrp
->peer
&& !op
->derived
) {
1563 struct ovn_port
*peer
= ovn_port_find(ports
, op
->nbrp
->peer
);
1567 } else if (peer
->nbsp
) {
1568 /* An ovn_port for a switch port of type "router" does have
1569 * a router port as its peer (see the case above for
1570 * "router" ports), but this is set via options:router-port
1571 * in Logical_Switch_Port and does not involve the
1572 * Logical_Router_Port's 'peer' column. */
1573 static struct vlog_rate_limit rl
=
1574 VLOG_RATE_LIMIT_INIT(5, 1);
1575 VLOG_WARN_RL(&rl
, "Bad configuration: The peer of router "
1576 "port %s is a switch port", op
->key
);
1584 ip_address_and_port_from_lb_key(const char *key
, char **ip_address
,
1585 uint16_t *port
, int *addr_family
);
1588 get_router_load_balancer_ips(const struct ovn_datapath
*od
,
1589 struct sset
*all_ips
, int *addr_family
)
1595 for (int i
= 0; i
< od
->nbr
->n_load_balancer
; i
++) {
1596 struct nbrec_load_balancer
*lb
= od
->nbr
->load_balancer
[i
];
1597 struct smap
*vips
= &lb
->vips
;
1598 struct smap_node
*node
;
1600 SMAP_FOR_EACH (node
, vips
) {
1601 /* node->key contains IP:port or just IP. */
1602 char *ip_address
= NULL
;
1605 ip_address_and_port_from_lb_key(node
->key
, &ip_address
, &port
,
1611 if (!sset_contains(all_ips
, ip_address
)) {
1612 sset_add(all_ips
, ip_address
);
1620 /* Returns an array of strings, each consisting of a MAC address followed
1621 * by one or more IP addresses, and if the port is a distributed gateway
1622 * port, followed by 'is_chassis_resident("LPORT_NAME")', where the
1623 * LPORT_NAME is the name of the L3 redirect port or the name of the
1624 * logical_port specified in a NAT rule. These strings include the
1625 * external IP addresses of all NAT rules defined on that router, and all
1626 * of the IP addresses used in load balancer VIPs defined on that router.
1628 * The caller must free each of the n returned strings with free(),
1629 * and must free the returned array when it is no longer needed. */
1631 get_nat_addresses(const struct ovn_port
*op
, size_t *n
)
1634 struct eth_addr mac
;
1635 if (!op
->nbrp
|| !op
->od
|| !op
->od
->nbr
1636 || (!op
->od
->nbr
->n_nat
&& !op
->od
->nbr
->n_load_balancer
)
1637 || !eth_addr_from_string(op
->nbrp
->mac
, &mac
)) {
1642 struct ds c_addresses
= DS_EMPTY_INITIALIZER
;
1643 ds_put_format(&c_addresses
, ETH_ADDR_FMT
, ETH_ADDR_ARGS(mac
));
1644 bool central_ip_address
= false;
1647 addresses
= xmalloc(sizeof *addresses
* (op
->od
->nbr
->n_nat
+ 1));
1649 /* Get NAT IP addresses. */
1650 for (size_t i
= 0; i
< op
->od
->nbr
->n_nat
; i
++) {
1651 const struct nbrec_nat
*nat
= op
->od
->nbr
->nat
[i
];
1654 char *error
= ip_parse_masked(nat
->external_ip
, &ip
, &mask
);
1655 if (error
|| mask
!= OVS_BE32_MAX
) {
1660 /* Determine whether this NAT rule satisfies the conditions for
1661 * distributed NAT processing. */
1662 if (op
->od
->l3redirect_port
&& !strcmp(nat
->type
, "dnat_and_snat")
1663 && nat
->logical_port
&& nat
->external_mac
) {
1664 /* Distributed NAT rule. */
1665 if (eth_addr_from_string(nat
->external_mac
, &mac
)) {
1666 struct ds address
= DS_EMPTY_INITIALIZER
;
1667 ds_put_format(&address
, ETH_ADDR_FMT
, ETH_ADDR_ARGS(mac
));
1668 ds_put_format(&address
, " %s", nat
->external_ip
);
1669 ds_put_format(&address
, " is_chassis_resident(\"%s\")",
1671 addresses
[n_nats
++] = ds_steal_cstr(&address
);
1674 /* Centralized NAT rule, either on gateway router or distributed
1676 ds_put_format(&c_addresses
, " %s", nat
->external_ip
);
1677 central_ip_address
= true;
1681 /* A set to hold all load-balancer vips. */
1682 struct sset all_ips
= SSET_INITIALIZER(&all_ips
);
1684 get_router_load_balancer_ips(op
->od
, &all_ips
, &addr_family
);
1686 const char *ip_address
;
1687 SSET_FOR_EACH (ip_address
, &all_ips
) {
1688 ds_put_format(&c_addresses
, " %s", ip_address
);
1689 central_ip_address
= true;
1691 sset_destroy(&all_ips
);
1693 if (central_ip_address
) {
1694 /* Gratuitous ARP for centralized NAT rules on distributed gateway
1695 * ports should be restricted to the "redirect-chassis". */
1696 if (op
->od
->l3redirect_port
) {
1697 ds_put_format(&c_addresses
, " is_chassis_resident(%s)",
1698 op
->od
->l3redirect_port
->json_key
);
1701 addresses
[n_nats
++] = ds_steal_cstr(&c_addresses
);
1710 gateway_chassis_equal(const struct nbrec_gateway_chassis
*nb_gwc
,
1711 const struct sbrec_chassis
*nb_gwc_c
,
1712 const struct sbrec_gateway_chassis
*sb_gwc
)
1714 bool equal
= !strcmp(nb_gwc
->name
, sb_gwc
->name
)
1715 && nb_gwc
->priority
== sb_gwc
->priority
1716 && smap_equal(&nb_gwc
->options
, &sb_gwc
->options
)
1717 && smap_equal(&nb_gwc
->external_ids
, &sb_gwc
->external_ids
);
1723 /* If everything else matched and we were unable to find the SBDB
1724 * Chassis entry at this time, assume a match and return true.
1725 * This happens when an ovn-controller is restarting and the Chassis
1726 * entry is gone away momentarily */
1728 || (sb_gwc
->chassis
&& !strcmp(nb_gwc_c
->name
,
1729 sb_gwc
->chassis
->name
));
1733 sbpb_gw_chassis_needs_update(
1734 const struct sbrec_port_binding
*port_binding
,
1735 const struct nbrec_logical_router_port
*lrp
,
1736 const struct chassis_index
*chassis_index
)
1738 if (!lrp
|| !port_binding
) {
1742 /* These arrays are used to collect valid Gateway_Chassis and valid
1743 * Chassis records from the Logical_Router_Port Gateway_Chassis list,
1744 * we ignore the ones we can't match on the SBDB */
1745 struct nbrec_gateway_chassis
**lrp_gwc
= xzalloc(lrp
->n_gateway_chassis
*
1747 const struct sbrec_chassis
**lrp_gwc_c
= xzalloc(lrp
->n_gateway_chassis
*
1750 /* Count the number of gateway chassis chassis names from the logical
1751 * router port that we are able to match on the southbound database */
1752 int lrp_n_gateway_chassis
= 0;
1754 for (n
= 0; n
< lrp
->n_gateway_chassis
; n
++) {
1756 if (!lrp
->gateway_chassis
[n
]->chassis_name
) {
1760 const struct sbrec_chassis
*chassis
=
1761 chassis_lookup_by_name(chassis_index
,
1762 lrp
->gateway_chassis
[n
]->chassis_name
);
1764 lrp_gwc_c
[lrp_n_gateway_chassis
] = chassis
;
1765 lrp_gwc
[lrp_n_gateway_chassis
] = lrp
->gateway_chassis
[n
];
1766 lrp_n_gateway_chassis
++;
1768 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
1770 &rl
, "Chassis name %s referenced in NBDB via Gateway_Chassis "
1771 "on logical router port %s does not exist in SBDB",
1772 lrp
->gateway_chassis
[n
]->chassis_name
, lrp
->name
);
1776 /* Basic check, different amount of Gateway_Chassis means that we
1777 * need to update southbound database Port_Binding */
1778 if (lrp_n_gateway_chassis
!= port_binding
->n_gateway_chassis
) {
1784 for (n
= 0; n
< lrp_n_gateway_chassis
; n
++) {
1786 /* For each of the valid gw chassis on the lrp, check if there's
1787 * a match on the Port_Binding list, we assume order is not
1789 for (i
= 0; i
< port_binding
->n_gateway_chassis
; i
++) {
1790 if (gateway_chassis_equal(lrp_gwc
[n
],
1792 port_binding
->gateway_chassis
[i
])) {
1793 break; /* we found a match */
1797 /* if no Port_Binding gateway chassis matched for the entry... */
1798 if (i
== port_binding
->n_gateway_chassis
) {
1801 return true; /* found no match for this gateway chassis on lrp */
1805 /* no need for update, all ports matched */
1811 /* This functions translates the gw chassis on the nb database
1812 * to sb database entries, the only difference is that SB database
1813 * Gateway_Chassis table references the chassis directly instead
1814 * of using the name */
1816 copy_gw_chassis_from_nbrp_to_sbpb(
1817 struct northd_context
*ctx
,
1818 const struct nbrec_logical_router_port
*lrp
,
1819 const struct chassis_index
*chassis_index
,
1820 const struct sbrec_port_binding
*port_binding
) {
1822 if (!lrp
|| !port_binding
|| !lrp
->n_gateway_chassis
) {
1826 struct sbrec_gateway_chassis
**gw_chassis
= NULL
;
1830 /* XXX: This can be improved. This code will generate a set of new
1831 * Gateway_Chassis and push them all in a single transaction, instead
1832 * this would be more optimal if we just add/update/remove the rows in
1833 * the southbound db that need to change. We don't expect lots of
1834 * changes to the Gateway_Chassis table, but if that proves to be wrong
1835 * we should optimize this. */
1836 for (n
= 0; n
< lrp
->n_gateway_chassis
; n
++) {
1837 struct nbrec_gateway_chassis
*lrp_gwc
= lrp
->gateway_chassis
[n
];
1838 if (!lrp_gwc
->chassis_name
) {
1842 const struct sbrec_chassis
*chassis
=
1843 chassis_lookup_by_name(chassis_index
, lrp_gwc
->chassis_name
);
1845 gw_chassis
= xrealloc(gw_chassis
, (n_gwc
+ 1) * sizeof *gw_chassis
);
1847 struct sbrec_gateway_chassis
*pb_gwc
=
1848 sbrec_gateway_chassis_insert(ctx
->ovnsb_txn
);
1850 sbrec_gateway_chassis_set_name(pb_gwc
, lrp_gwc
->name
);
1851 sbrec_gateway_chassis_set_priority(pb_gwc
, lrp_gwc
->priority
);
1852 sbrec_gateway_chassis_set_chassis(pb_gwc
, chassis
);
1853 sbrec_gateway_chassis_set_options(pb_gwc
, &lrp_gwc
->options
);
1854 sbrec_gateway_chassis_set_external_ids(pb_gwc
, &lrp_gwc
->external_ids
);
1856 gw_chassis
[n_gwc
++] = pb_gwc
;
1858 sbrec_port_binding_set_gateway_chassis(port_binding
, gw_chassis
, n_gwc
);
1863 ovn_port_update_sbrec(struct northd_context
*ctx
,
1864 const struct ovn_port
*op
,
1865 const struct chassis_index
*chassis_index
,
1866 struct hmap
*chassis_qdisc_queues
)
1868 sbrec_port_binding_set_datapath(op
->sb
, op
->od
->sb
);
1870 /* If the router is for l3 gateway, it resides on a chassis
1871 * and its port type is "l3gateway". */
1872 const char *chassis_name
= smap_get(&op
->od
->nbr
->options
, "chassis");
1874 sbrec_port_binding_set_type(op
->sb
, "chassisredirect");
1875 } else if (chassis_name
) {
1876 sbrec_port_binding_set_type(op
->sb
, "l3gateway");
1878 sbrec_port_binding_set_type(op
->sb
, "patch");
1884 const char *redirect_chassis
= smap_get(&op
->nbrp
->options
,
1885 "redirect-chassis");
1886 if (op
->nbrp
->n_gateway_chassis
&& redirect_chassis
) {
1887 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
1889 &rl
, "logical router port %s has both options:"
1890 "redirect-chassis and gateway_chassis populated "
1891 "redirect-chassis will be ignored in favour of "
1892 "gateway chassis", op
->nbrp
->name
);
1895 if (op
->nbrp
->n_gateway_chassis
) {
1896 if (sbpb_gw_chassis_needs_update(op
->sb
, op
->nbrp
,
1898 copy_gw_chassis_from_nbrp_to_sbpb(ctx
, op
->nbrp
,
1899 chassis_index
, op
->sb
);
1902 } else if (redirect_chassis
) {
1903 /* Handle ports that had redirect-chassis option attached
1904 * to them, and for backwards compatibility convert them
1905 * to a single Gateway_Chassis entry */
1906 const struct sbrec_chassis
*chassis
=
1907 chassis_lookup_by_name(chassis_index
, redirect_chassis
);
1909 /* If we found the chassis, and the gw chassis on record
1910 * differs from what we expect go ahead and update */
1911 if (op
->sb
->n_gateway_chassis
!= 1
1912 || !op
->sb
->gateway_chassis
[0]->chassis
1913 || strcmp(op
->sb
->gateway_chassis
[0]->chassis
->name
,
1915 || op
->sb
->gateway_chassis
[0]->priority
!= 0) {
1916 /* Construct a single Gateway_Chassis entry on the
1917 * Port_Binding attached to the redirect_chassis
1919 struct sbrec_gateway_chassis
*gw_chassis
=
1920 sbrec_gateway_chassis_insert(ctx
->ovnsb_txn
);
1922 char *gwc_name
= xasprintf("%s_%s", op
->nbrp
->name
,
1925 /* XXX: Again, here, we could just update an existing
1926 * Gateway_Chassis, instead of creating a new one
1927 * and replacing it */
1928 sbrec_gateway_chassis_set_name(gw_chassis
, gwc_name
);
1929 sbrec_gateway_chassis_set_priority(gw_chassis
, 0);
1930 sbrec_gateway_chassis_set_chassis(gw_chassis
, chassis
);
1931 sbrec_gateway_chassis_set_external_ids(gw_chassis
,
1932 &op
->nbrp
->external_ids
);
1933 sbrec_port_binding_set_gateway_chassis(op
->sb
,
1938 VLOG_WARN("chassis name '%s' from redirect from logical "
1939 " router port '%s' redirect-chassis not found",
1940 redirect_chassis
, op
->nbrp
->name
);
1941 if (op
->sb
->n_gateway_chassis
) {
1942 sbrec_port_binding_set_gateway_chassis(op
->sb
, NULL
,
1947 smap_add(&new, "distributed-port", op
->nbrp
->name
);
1950 smap_add(&new, "peer", op
->peer
->key
);
1953 smap_add(&new, "l3gateway-chassis", chassis_name
);
1956 sbrec_port_binding_set_options(op
->sb
, &new);
1959 sbrec_port_binding_set_parent_port(op
->sb
, NULL
);
1960 sbrec_port_binding_set_tag(op
->sb
, NULL
, 0);
1962 struct ds s
= DS_EMPTY_INITIALIZER
;
1963 ds_put_cstr(&s
, op
->nbrp
->mac
);
1964 for (int i
= 0; i
< op
->nbrp
->n_networks
; ++i
) {
1965 ds_put_format(&s
, " %s", op
->nbrp
->networks
[i
]);
1967 const char *addresses
= ds_cstr(&s
);
1968 sbrec_port_binding_set_mac(op
->sb
, &addresses
, 1);
1971 struct smap ids
= SMAP_INITIALIZER(&ids
);
1972 sbrec_port_binding_set_external_ids(op
->sb
, &ids
);
1974 if (strcmp(op
->nbsp
->type
, "router")) {
1975 uint32_t queue_id
= smap_get_int(
1976 &op
->sb
->options
, "qdisc_queue_id", 0);
1977 bool has_qos
= port_has_qos_params(&op
->nbsp
->options
);
1978 struct smap options
;
1980 if (op
->sb
->chassis
&& has_qos
&& !queue_id
) {
1981 queue_id
= allocate_chassis_queueid(chassis_qdisc_queues
,
1983 } else if (!has_qos
&& queue_id
) {
1984 free_chassis_queueid(chassis_qdisc_queues
,
1990 smap_clone(&options
, &op
->nbsp
->options
);
1992 smap_add_format(&options
,
1993 "qdisc_queue_id", "%d", queue_id
);
1995 sbrec_port_binding_set_options(op
->sb
, &options
);
1996 smap_destroy(&options
);
1997 if (ovn_is_known_nb_lsp_type(op
->nbsp
->type
)) {
1998 sbrec_port_binding_set_type(op
->sb
, op
->nbsp
->type
);
2000 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
2002 &rl
, "Unknown port type '%s' set on logical switch '%s'.",
2003 op
->nbsp
->type
, op
->nbsp
->name
);
2006 const char *chassis
= NULL
;
2007 if (op
->peer
&& op
->peer
->od
&& op
->peer
->od
->nbr
) {
2008 chassis
= smap_get(&op
->peer
->od
->nbr
->options
, "chassis");
2011 /* A switch port connected to a gateway router is also of
2012 * type "l3gateway". */
2014 sbrec_port_binding_set_type(op
->sb
, "l3gateway");
2016 sbrec_port_binding_set_type(op
->sb
, "patch");
2019 const char *router_port
= smap_get(&op
->nbsp
->options
,
2021 if (router_port
|| chassis
) {
2025 smap_add(&new, "peer", router_port
);
2028 smap_add(&new, "l3gateway-chassis", chassis
);
2030 sbrec_port_binding_set_options(op
->sb
, &new);
2034 const char *nat_addresses
= smap_get(&op
->nbsp
->options
,
2036 if (nat_addresses
&& !strcmp(nat_addresses
, "router")) {
2037 if (op
->peer
&& op
->peer
->od
2038 && (chassis
|| op
->peer
->od
->l3redirect_port
)) {
2040 char **nats
= get_nat_addresses(op
->peer
, &n_nats
);
2042 sbrec_port_binding_set_nat_addresses(op
->sb
,
2043 (const char **) nats
, n_nats
);
2044 for (size_t i
= 0; i
< n_nats
; i
++) {
2049 sbrec_port_binding_set_nat_addresses(op
->sb
, NULL
, 0);
2052 sbrec_port_binding_set_nat_addresses(op
->sb
, NULL
, 0);
2054 /* Only accept manual specification of ethernet address
2055 * followed by IPv4 addresses on type "l3gateway" ports. */
2056 } else if (nat_addresses
&& chassis
) {
2057 struct lport_addresses laddrs
;
2058 if (!extract_lsp_addresses(nat_addresses
, &laddrs
)) {
2059 static struct vlog_rate_limit rl
=
2060 VLOG_RATE_LIMIT_INIT(1, 1);
2061 VLOG_WARN_RL(&rl
, "Error extracting nat-addresses.");
2062 sbrec_port_binding_set_nat_addresses(op
->sb
, NULL
, 0);
2064 sbrec_port_binding_set_nat_addresses(op
->sb
,
2066 destroy_lport_addresses(&laddrs
);
2069 sbrec_port_binding_set_nat_addresses(op
->sb
, NULL
, 0);
2072 sbrec_port_binding_set_parent_port(op
->sb
, op
->nbsp
->parent_name
);
2073 sbrec_port_binding_set_tag(op
->sb
, op
->nbsp
->tag
, op
->nbsp
->n_tag
);
2074 sbrec_port_binding_set_mac(op
->sb
, (const char **) op
->nbsp
->addresses
,
2075 op
->nbsp
->n_addresses
);
2077 struct smap ids
= SMAP_INITIALIZER(&ids
);
2078 smap_clone(&ids
, &op
->nbsp
->external_ids
);
2079 const char *name
= smap_get(&ids
, "neutron:port_name");
2080 if (name
&& name
[0]) {
2081 smap_add(&ids
, "name", name
);
2083 sbrec_port_binding_set_external_ids(op
->sb
, &ids
);
2088 /* Remove mac_binding entries that refer to logical_ports which are
2091 cleanup_mac_bindings(struct northd_context
*ctx
, struct hmap
*ports
)
2093 const struct sbrec_mac_binding
*b
, *n
;
2094 SBREC_MAC_BINDING_FOR_EACH_SAFE (b
, n
, ctx
->ovnsb_idl
) {
2095 if (!ovn_port_find(ports
, b
->logical_port
)) {
2096 sbrec_mac_binding_delete(b
);
2101 /* Updates the southbound Port_Binding table so that it contains the logical
2102 * switch ports specified by the northbound database.
2104 * Initializes 'ports' to contain a "struct ovn_port" for every logical port,
2105 * using the "struct ovn_datapath"s in 'datapaths' to look up logical
2108 build_ports(struct northd_context
*ctx
, struct hmap
*datapaths
,
2109 const struct chassis_index
*chassis_index
, struct hmap
*ports
)
2111 struct ovs_list sb_only
, nb_only
, both
;
2112 struct hmap tag_alloc_table
= HMAP_INITIALIZER(&tag_alloc_table
);
2113 struct hmap chassis_qdisc_queues
= HMAP_INITIALIZER(&chassis_qdisc_queues
);
2115 join_logical_ports(ctx
, datapaths
, ports
, &chassis_qdisc_queues
,
2116 &tag_alloc_table
, &sb_only
, &nb_only
, &both
);
2118 struct ovn_port
*op
, *next
;
2119 /* For logical ports that are in both databases, update the southbound
2120 * record based on northbound data. Also index the in-use tunnel_keys.
2121 * For logical ports that are in NB database, do any tag allocation
2123 LIST_FOR_EACH_SAFE (op
, next
, list
, &both
) {
2125 tag_alloc_create_new_tag(&tag_alloc_table
, op
->nbsp
);
2127 ovn_port_update_sbrec(ctx
, op
, chassis_index
, &chassis_qdisc_queues
);
2129 add_tnlid(&op
->od
->port_tnlids
, op
->sb
->tunnel_key
);
2130 if (op
->sb
->tunnel_key
> op
->od
->port_key_hint
) {
2131 op
->od
->port_key_hint
= op
->sb
->tunnel_key
;
2135 /* Add southbound record for each unmatched northbound record. */
2136 LIST_FOR_EACH_SAFE (op
, next
, list
, &nb_only
) {
2137 uint16_t tunnel_key
= ovn_port_allocate_key(op
->od
);
2142 op
->sb
= sbrec_port_binding_insert(ctx
->ovnsb_txn
);
2143 ovn_port_update_sbrec(ctx
, op
, chassis_index
, &chassis_qdisc_queues
);
2145 sbrec_port_binding_set_logical_port(op
->sb
, op
->key
);
2146 sbrec_port_binding_set_tunnel_key(op
->sb
, tunnel_key
);
2149 bool remove_mac_bindings
= false;
2150 if (!ovs_list_is_empty(&sb_only
)) {
2151 remove_mac_bindings
= true;
2154 /* Delete southbound records without northbound matches. */
2155 LIST_FOR_EACH_SAFE(op
, next
, list
, &sb_only
) {
2156 ovs_list_remove(&op
->list
);
2157 sbrec_port_binding_delete(op
->sb
);
2158 ovn_port_destroy(ports
, op
);
2160 if (remove_mac_bindings
) {
2161 cleanup_mac_bindings(ctx
, ports
);
2164 tag_alloc_destroy(&tag_alloc_table
);
2165 destroy_chassis_queues(&chassis_qdisc_queues
);
2168 #define OVN_MIN_MULTICAST 32768
2169 #define OVN_MAX_MULTICAST 65535
2171 struct multicast_group
{
2173 uint16_t key
; /* OVN_MIN_MULTICAST...OVN_MAX_MULTICAST. */
2176 #define MC_FLOOD "_MC_flood"
2177 static const struct multicast_group mc_flood
= { MC_FLOOD
, 65535 };
2179 #define MC_UNKNOWN "_MC_unknown"
2180 static const struct multicast_group mc_unknown
= { MC_UNKNOWN
, 65534 };
2183 multicast_group_equal(const struct multicast_group
*a
,
2184 const struct multicast_group
*b
)
2186 return !strcmp(a
->name
, b
->name
) && a
->key
== b
->key
;
2189 /* Multicast group entry. */
2190 struct ovn_multicast
{
2191 struct hmap_node hmap_node
; /* Index on 'datapath' and 'key'. */
2192 struct ovn_datapath
*datapath
;
2193 const struct multicast_group
*group
;
2195 struct ovn_port
**ports
;
2196 size_t n_ports
, allocated_ports
;
2200 ovn_multicast_hash(const struct ovn_datapath
*datapath
,
2201 const struct multicast_group
*group
)
2203 return hash_pointer(datapath
, group
->key
);
2206 static struct ovn_multicast
*
2207 ovn_multicast_find(struct hmap
*mcgroups
, struct ovn_datapath
*datapath
,
2208 const struct multicast_group
*group
)
2210 struct ovn_multicast
*mc
;
2212 HMAP_FOR_EACH_WITH_HASH (mc
, hmap_node
,
2213 ovn_multicast_hash(datapath
, group
), mcgroups
) {
2214 if (mc
->datapath
== datapath
2215 && multicast_group_equal(mc
->group
, group
)) {
2223 ovn_multicast_add(struct hmap
*mcgroups
, const struct multicast_group
*group
,
2224 struct ovn_port
*port
)
2226 struct ovn_datapath
*od
= port
->od
;
2227 struct ovn_multicast
*mc
= ovn_multicast_find(mcgroups
, od
, group
);
2229 mc
= xmalloc(sizeof *mc
);
2230 hmap_insert(mcgroups
, &mc
->hmap_node
, ovn_multicast_hash(od
, group
));
2234 mc
->allocated_ports
= 4;
2235 mc
->ports
= xmalloc(mc
->allocated_ports
* sizeof *mc
->ports
);
2237 if (mc
->n_ports
>= mc
->allocated_ports
) {
2238 mc
->ports
= x2nrealloc(mc
->ports
, &mc
->allocated_ports
,
2241 mc
->ports
[mc
->n_ports
++] = port
;
2245 ovn_multicast_destroy(struct hmap
*mcgroups
, struct ovn_multicast
*mc
)
2248 hmap_remove(mcgroups
, &mc
->hmap_node
);
2255 ovn_multicast_update_sbrec(const struct ovn_multicast
*mc
,
2256 const struct sbrec_multicast_group
*sb
)
2258 struct sbrec_port_binding
**ports
= xmalloc(mc
->n_ports
* sizeof *ports
);
2259 for (size_t i
= 0; i
< mc
->n_ports
; i
++) {
2260 ports
[i
] = CONST_CAST(struct sbrec_port_binding
*, mc
->ports
[i
]->sb
);
2262 sbrec_multicast_group_set_ports(sb
, ports
, mc
->n_ports
);
2266 /* Logical flow generation.
2268 * This code generates the Logical_Flow table in the southbound database, as a
2269 * function of most of the northbound database.
2273 struct hmap_node hmap_node
;
2275 struct ovn_datapath
*od
;
2276 enum ovn_stage stage
;
2285 ovn_lflow_hash(const struct ovn_lflow
*lflow
)
2287 return ovn_logical_flow_hash(&lflow
->od
->sb
->header_
.uuid
,
2288 ovn_stage_get_table(lflow
->stage
),
2289 ovn_stage_get_pipeline_name(lflow
->stage
),
2290 lflow
->priority
, lflow
->match
,
2295 ovn_lflow_equal(const struct ovn_lflow
*a
, const struct ovn_lflow
*b
)
2297 return (a
->od
== b
->od
2298 && a
->stage
== b
->stage
2299 && a
->priority
== b
->priority
2300 && !strcmp(a
->match
, b
->match
)
2301 && !strcmp(a
->actions
, b
->actions
));
2305 ovn_lflow_init(struct ovn_lflow
*lflow
, struct ovn_datapath
*od
,
2306 enum ovn_stage stage
, uint16_t priority
,
2307 char *match
, char *actions
, char *stage_hint
,
2311 lflow
->stage
= stage
;
2312 lflow
->priority
= priority
;
2313 lflow
->match
= match
;
2314 lflow
->actions
= actions
;
2315 lflow
->stage_hint
= stage_hint
;
2316 lflow
->where
= where
;
2319 /* Adds a row with the specified contents to the Logical_Flow table. */
2321 ovn_lflow_add_at(struct hmap
*lflow_map
, struct ovn_datapath
*od
,
2322 enum ovn_stage stage
, uint16_t priority
,
2323 const char *match
, const char *actions
,
2324 const char *stage_hint
, const char *where
)
2326 ovs_assert(ovn_stage_to_datapath_type(stage
) == ovn_datapath_get_type(od
));
2328 struct ovn_lflow
*lflow
= xmalloc(sizeof *lflow
);
2329 ovn_lflow_init(lflow
, od
, stage
, priority
,
2330 xstrdup(match
), xstrdup(actions
),
2331 nullable_xstrdup(stage_hint
), where
);
2332 hmap_insert(lflow_map
, &lflow
->hmap_node
, ovn_lflow_hash(lflow
));
2335 /* Adds a row with the specified contents to the Logical_Flow table. */
2336 #define ovn_lflow_add_with_hint(LFLOW_MAP, OD, STAGE, PRIORITY, MATCH, \
2337 ACTIONS, STAGE_HINT) \
2338 ovn_lflow_add_at(LFLOW_MAP, OD, STAGE, PRIORITY, MATCH, ACTIONS, \
2339 STAGE_HINT, OVS_SOURCE_LOCATOR)
2341 #define ovn_lflow_add(LFLOW_MAP, OD, STAGE, PRIORITY, MATCH, ACTIONS) \
2342 ovn_lflow_add_with_hint(LFLOW_MAP, OD, STAGE, PRIORITY, MATCH, \
2345 static struct ovn_lflow
*
2346 ovn_lflow_find(struct hmap
*lflows
, struct ovn_datapath
*od
,
2347 enum ovn_stage stage
, uint16_t priority
,
2348 const char *match
, const char *actions
, uint32_t hash
)
2350 struct ovn_lflow target
;
2351 ovn_lflow_init(&target
, od
, stage
, priority
,
2352 CONST_CAST(char *, match
), CONST_CAST(char *, actions
),
2355 struct ovn_lflow
*lflow
;
2356 HMAP_FOR_EACH_WITH_HASH (lflow
, hmap_node
, hash
, lflows
) {
2357 if (ovn_lflow_equal(lflow
, &target
)) {
2365 ovn_lflow_destroy(struct hmap
*lflows
, struct ovn_lflow
*lflow
)
2368 hmap_remove(lflows
, &lflow
->hmap_node
);
2370 free(lflow
->actions
);
2371 free(lflow
->stage_hint
);
2376 /* Appends port security constraints on L2 address field 'eth_addr_field'
2377 * (e.g. "eth.src" or "eth.dst") to 'match'. 'ps_addrs', with 'n_ps_addrs'
2378 * elements, is the collection of port_security constraints from an
2379 * OVN_NB Logical_Switch_Port row generated by extract_lsp_addresses(). */
2381 build_port_security_l2(const char *eth_addr_field
,
2382 struct lport_addresses
*ps_addrs
,
2383 unsigned int n_ps_addrs
,
2390 ds_put_format(match
, " && %s == {", eth_addr_field
);
2392 for (size_t i
= 0; i
< n_ps_addrs
; i
++) {
2393 ds_put_format(match
, "%s ", ps_addrs
[i
].ea_s
);
2395 ds_chomp(match
, ' ');
2396 ds_put_cstr(match
, "}");
2400 build_port_security_ipv6_nd_flow(
2401 struct ds
*match
, struct eth_addr ea
, struct ipv6_netaddr
*ipv6_addrs
,
2404 ds_put_format(match
, " && ip6 && nd && ((nd.sll == "ETH_ADDR_FMT
" || "
2405 "nd.sll == "ETH_ADDR_FMT
") || ((nd.tll == "ETH_ADDR_FMT
" || "
2406 "nd.tll == "ETH_ADDR_FMT
")", ETH_ADDR_ARGS(eth_addr_zero
),
2407 ETH_ADDR_ARGS(ea
), ETH_ADDR_ARGS(eth_addr_zero
),
2409 if (!n_ipv6_addrs
) {
2410 ds_put_cstr(match
, "))");
2414 char ip6_str
[INET6_ADDRSTRLEN
+ 1];
2415 struct in6_addr lla
;
2416 in6_generate_lla(ea
, &lla
);
2417 memset(ip6_str
, 0, sizeof(ip6_str
));
2418 ipv6_string_mapped(ip6_str
, &lla
);
2419 ds_put_format(match
, " && (nd.target == %s", ip6_str
);
2421 for(int i
= 0; i
< n_ipv6_addrs
; i
++) {
2422 memset(ip6_str
, 0, sizeof(ip6_str
));
2423 ipv6_string_mapped(ip6_str
, &ipv6_addrs
[i
].addr
);
2424 ds_put_format(match
, " || nd.target == %s", ip6_str
);
2427 ds_put_format(match
, ")))");
2431 build_port_security_ipv6_flow(
2432 enum ovn_pipeline pipeline
, struct ds
*match
, struct eth_addr ea
,
2433 struct ipv6_netaddr
*ipv6_addrs
, int n_ipv6_addrs
)
2435 char ip6_str
[INET6_ADDRSTRLEN
+ 1];
2437 ds_put_format(match
, " && %s == {",
2438 pipeline
== P_IN
? "ip6.src" : "ip6.dst");
2440 /* Allow link-local address. */
2441 struct in6_addr lla
;
2442 in6_generate_lla(ea
, &lla
);
2443 ipv6_string_mapped(ip6_str
, &lla
);
2444 ds_put_format(match
, "%s, ", ip6_str
);
2446 /* Allow ip6.dst=ff00::/8 for multicast packets */
2447 if (pipeline
== P_OUT
) {
2448 ds_put_cstr(match
, "ff00::/8, ");
2450 for(int i
= 0; i
< n_ipv6_addrs
; i
++) {
2451 ipv6_string_mapped(ip6_str
, &ipv6_addrs
[i
].addr
);
2452 ds_put_format(match
, "%s, ", ip6_str
);
2454 /* Replace ", " by "}". */
2455 ds_chomp(match
, ' ');
2456 ds_chomp(match
, ',');
2457 ds_put_cstr(match
, "}");
2461 * Build port security constraints on ARP and IPv6 ND fields
2462 * and add logical flows to S_SWITCH_IN_PORT_SEC_ND stage.
2464 * For each port security of the logical port, following
2465 * logical flows are added
2466 * - If the port security has no IP (both IPv4 and IPv6) or
2467 * if it has IPv4 address(es)
2468 * - Priority 90 flow to allow ARP packets for known MAC addresses
2469 * in the eth.src and arp.spa fields. If the port security
2470 * has IPv4 addresses, allow known IPv4 addresses in the arp.tpa field.
2472 * - If the port security has no IP (both IPv4 and IPv6) or
2473 * if it has IPv6 address(es)
2474 * - Priority 90 flow to allow IPv6 ND packets for known MAC addresses
2475 * in the eth.src and nd.sll/nd.tll fields. If the port security
2476 * has IPv6 addresses, allow known IPv6 addresses in the nd.target field
2477 * for IPv6 Neighbor Advertisement packet.
2479 * - Priority 80 flow to drop ARP and IPv6 ND packets.
2482 build_port_security_nd(struct ovn_port
*op
, struct hmap
*lflows
)
2484 struct ds match
= DS_EMPTY_INITIALIZER
;
2486 for (size_t i
= 0; i
< op
->n_ps_addrs
; i
++) {
2487 struct lport_addresses
*ps
= &op
->ps_addrs
[i
];
2489 bool no_ip
= !(ps
->n_ipv4_addrs
|| ps
->n_ipv6_addrs
);
2492 if (ps
->n_ipv4_addrs
|| no_ip
) {
2493 ds_put_format(&match
,
2494 "inport == %s && eth.src == %s && arp.sha == %s",
2495 op
->json_key
, ps
->ea_s
, ps
->ea_s
);
2497 if (ps
->n_ipv4_addrs
) {
2498 ds_put_cstr(&match
, " && arp.spa == {");
2499 for (size_t j
= 0; j
< ps
->n_ipv4_addrs
; j
++) {
2500 /* When the netmask is applied, if the host portion is
2501 * non-zero, the host can only use the specified
2502 * address in the arp.spa. If zero, the host is allowed
2503 * to use any address in the subnet. */
2504 if (ps
->ipv4_addrs
[j
].plen
== 32
2505 || ps
->ipv4_addrs
[j
].addr
& ~ps
->ipv4_addrs
[j
].mask
) {
2506 ds_put_cstr(&match
, ps
->ipv4_addrs
[j
].addr_s
);
2508 ds_put_format(&match
, "%s/%d",
2509 ps
->ipv4_addrs
[j
].network_s
,
2510 ps
->ipv4_addrs
[j
].plen
);
2512 ds_put_cstr(&match
, ", ");
2514 ds_chomp(&match
, ' ');
2515 ds_chomp(&match
, ',');
2516 ds_put_cstr(&match
, "}");
2518 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_PORT_SEC_ND
, 90,
2519 ds_cstr(&match
), "next;");
2522 if (ps
->n_ipv6_addrs
|| no_ip
) {
2524 ds_put_format(&match
, "inport == %s && eth.src == %s",
2525 op
->json_key
, ps
->ea_s
);
2526 build_port_security_ipv6_nd_flow(&match
, ps
->ea
, ps
->ipv6_addrs
,
2528 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_PORT_SEC_ND
, 90,
2529 ds_cstr(&match
), "next;");
2534 ds_put_format(&match
, "inport == %s && (arp || nd)", op
->json_key
);
2535 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_PORT_SEC_ND
, 80,
2536 ds_cstr(&match
), "drop;");
2541 * Build port security constraints on IPv4 and IPv6 src and dst fields
2542 * and add logical flows to S_SWITCH_(IN/OUT)_PORT_SEC_IP stage.
2544 * For each port security of the logical port, following
2545 * logical flows are added
2546 * - If the port security has IPv4 addresses,
2547 * - Priority 90 flow to allow IPv4 packets for known IPv4 addresses
2549 * - If the port security has IPv6 addresses,
2550 * - Priority 90 flow to allow IPv6 packets for known IPv6 addresses
2552 * - If the port security has IPv4 addresses or IPv6 addresses or both
2553 * - Priority 80 flow to drop all IPv4 and IPv6 traffic
2556 build_port_security_ip(enum ovn_pipeline pipeline
, struct ovn_port
*op
,
2557 struct hmap
*lflows
)
2559 char *port_direction
;
2560 enum ovn_stage stage
;
2561 if (pipeline
== P_IN
) {
2562 port_direction
= "inport";
2563 stage
= S_SWITCH_IN_PORT_SEC_IP
;
2565 port_direction
= "outport";
2566 stage
= S_SWITCH_OUT_PORT_SEC_IP
;
2569 for (size_t i
= 0; i
< op
->n_ps_addrs
; i
++) {
2570 struct lport_addresses
*ps
= &op
->ps_addrs
[i
];
2572 if (!(ps
->n_ipv4_addrs
|| ps
->n_ipv6_addrs
)) {
2576 if (ps
->n_ipv4_addrs
) {
2577 struct ds match
= DS_EMPTY_INITIALIZER
;
2578 if (pipeline
== P_IN
) {
2579 /* Permit use of the unspecified address for DHCP discovery */
2580 struct ds dhcp_match
= DS_EMPTY_INITIALIZER
;
2581 ds_put_format(&dhcp_match
, "inport == %s"
2583 " && ip4.src == 0.0.0.0"
2584 " && ip4.dst == 255.255.255.255"
2585 " && udp.src == 68 && udp.dst == 67",
2586 op
->json_key
, ps
->ea_s
);
2587 ovn_lflow_add(lflows
, op
->od
, stage
, 90,
2588 ds_cstr(&dhcp_match
), "next;");
2589 ds_destroy(&dhcp_match
);
2590 ds_put_format(&match
, "inport == %s && eth.src == %s"
2591 " && ip4.src == {", op
->json_key
,
2594 ds_put_format(&match
, "outport == %s && eth.dst == %s"
2595 " && ip4.dst == {255.255.255.255, 224.0.0.0/4, ",
2596 op
->json_key
, ps
->ea_s
);
2599 for (int j
= 0; j
< ps
->n_ipv4_addrs
; j
++) {
2600 ovs_be32 mask
= ps
->ipv4_addrs
[j
].mask
;
2601 /* When the netmask is applied, if the host portion is
2602 * non-zero, the host can only use the specified
2603 * address. If zero, the host is allowed to use any
2604 * address in the subnet.
2606 if (ps
->ipv4_addrs
[j
].plen
== 32
2607 || ps
->ipv4_addrs
[j
].addr
& ~mask
) {
2608 ds_put_format(&match
, "%s", ps
->ipv4_addrs
[j
].addr_s
);
2609 if (pipeline
== P_OUT
&& ps
->ipv4_addrs
[j
].plen
!= 32) {
2610 /* Host is also allowed to receive packets to the
2611 * broadcast address in the specified subnet. */
2612 ds_put_format(&match
, ", %s",
2613 ps
->ipv4_addrs
[j
].bcast_s
);
2616 /* host portion is zero */
2617 ds_put_format(&match
, "%s/%d", ps
->ipv4_addrs
[j
].network_s
,
2618 ps
->ipv4_addrs
[j
].plen
);
2620 ds_put_cstr(&match
, ", ");
2623 /* Replace ", " by "}". */
2624 ds_chomp(&match
, ' ');
2625 ds_chomp(&match
, ',');
2626 ds_put_cstr(&match
, "}");
2627 ovn_lflow_add(lflows
, op
->od
, stage
, 90, ds_cstr(&match
), "next;");
2631 if (ps
->n_ipv6_addrs
) {
2632 struct ds match
= DS_EMPTY_INITIALIZER
;
2633 if (pipeline
== P_IN
) {
2634 /* Permit use of unspecified address for duplicate address
2636 struct ds dad_match
= DS_EMPTY_INITIALIZER
;
2637 ds_put_format(&dad_match
, "inport == %s"
2640 " && ip6.dst == ff02::/16"
2641 " && icmp6.type == {131, 135, 143}", op
->json_key
,
2643 ovn_lflow_add(lflows
, op
->od
, stage
, 90,
2644 ds_cstr(&dad_match
), "next;");
2645 ds_destroy(&dad_match
);
2647 ds_put_format(&match
, "%s == %s && %s == %s",
2648 port_direction
, op
->json_key
,
2649 pipeline
== P_IN
? "eth.src" : "eth.dst", ps
->ea_s
);
2650 build_port_security_ipv6_flow(pipeline
, &match
, ps
->ea
,
2651 ps
->ipv6_addrs
, ps
->n_ipv6_addrs
);
2652 ovn_lflow_add(lflows
, op
->od
, stage
, 90,
2653 ds_cstr(&match
), "next;");
2657 char *match
= xasprintf("%s == %s && %s == %s && ip",
2658 port_direction
, op
->json_key
,
2659 pipeline
== P_IN
? "eth.src" : "eth.dst",
2661 ovn_lflow_add(lflows
, op
->od
, stage
, 80, match
, "drop;");
2668 lsp_is_enabled(const struct nbrec_logical_switch_port
*lsp
)
2670 return !lsp
->enabled
|| *lsp
->enabled
;
2674 lsp_is_up(const struct nbrec_logical_switch_port
*lsp
)
2676 return !lsp
->up
|| *lsp
->up
;
2680 build_dhcpv4_action(struct ovn_port
*op
, ovs_be32 offer_ip
,
2681 struct ds
*options_action
, struct ds
*response_action
,
2682 struct ds
*ipv4_addr_match
)
2684 if (!op
->nbsp
->dhcpv4_options
) {
2685 /* CMS has disabled native DHCPv4 for this lport. */
2689 ovs_be32 host_ip
, mask
;
2690 char *error
= ip_parse_masked(op
->nbsp
->dhcpv4_options
->cidr
, &host_ip
,
2692 if (error
|| ((offer_ip
^ host_ip
) & mask
)) {
2694 * - cidr defined is invalid or
2695 * - the offer ip of the logical port doesn't belong to the cidr
2696 * defined in the DHCPv4 options.
2702 const char *server_ip
= smap_get(
2703 &op
->nbsp
->dhcpv4_options
->options
, "server_id");
2704 const char *server_mac
= smap_get(
2705 &op
->nbsp
->dhcpv4_options
->options
, "server_mac");
2706 const char *lease_time
= smap_get(
2707 &op
->nbsp
->dhcpv4_options
->options
, "lease_time");
2709 if (!(server_ip
&& server_mac
&& lease_time
)) {
2710 /* "server_id", "server_mac" and "lease_time" should be
2711 * present in the dhcp_options. */
2712 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2713 VLOG_WARN_RL(&rl
, "Required DHCPv4 options not defined for lport - %s",
2718 struct smap dhcpv4_options
= SMAP_INITIALIZER(&dhcpv4_options
);
2719 smap_clone(&dhcpv4_options
, &op
->nbsp
->dhcpv4_options
->options
);
2721 /* server_mac is not DHCPv4 option, delete it from the smap. */
2722 smap_remove(&dhcpv4_options
, "server_mac");
2723 char *netmask
= xasprintf(IP_FMT
, IP_ARGS(mask
));
2724 smap_add(&dhcpv4_options
, "netmask", netmask
);
2727 ds_put_format(options_action
,
2728 REGBIT_DHCP_OPTS_RESULT
" = put_dhcp_opts(offerip = "
2729 IP_FMT
", ", IP_ARGS(offer_ip
));
2731 /* We're not using SMAP_FOR_EACH because we want a consistent order of the
2732 * options on different architectures (big or little endian, SSE4.2) */
2733 const struct smap_node
**sorted_opts
= smap_sort(&dhcpv4_options
);
2734 for (size_t i
= 0; i
< smap_count(&dhcpv4_options
); i
++) {
2735 const struct smap_node
*node
= sorted_opts
[i
];
2736 ds_put_format(options_action
, "%s = %s, ", node
->key
, node
->value
);
2740 ds_chomp(options_action
, ' ');
2741 ds_chomp(options_action
, ',');
2742 ds_put_cstr(options_action
, "); next;");
2744 ds_put_format(response_action
, "eth.dst = eth.src; eth.src = %s; "
2745 "ip4.dst = "IP_FMT
"; ip4.src = %s; udp.src = 67; "
2746 "udp.dst = 68; outport = inport; flags.loopback = 1; "
2748 server_mac
, IP_ARGS(offer_ip
), server_ip
);
2750 ds_put_format(ipv4_addr_match
,
2751 "ip4.src == "IP_FMT
" && ip4.dst == {%s, 255.255.255.255}",
2752 IP_ARGS(offer_ip
), server_ip
);
2753 smap_destroy(&dhcpv4_options
);
2758 build_dhcpv6_action(struct ovn_port
*op
, struct in6_addr
*offer_ip
,
2759 struct ds
*options_action
, struct ds
*response_action
)
2761 if (!op
->nbsp
->dhcpv6_options
) {
2762 /* CMS has disabled native DHCPv6 for this lport. */
2766 struct in6_addr host_ip
, mask
;
2768 char *error
= ipv6_parse_masked(op
->nbsp
->dhcpv6_options
->cidr
, &host_ip
,
2774 struct in6_addr ip6_mask
= ipv6_addr_bitxor(offer_ip
, &host_ip
);
2775 ip6_mask
= ipv6_addr_bitand(&ip6_mask
, &mask
);
2776 if (!ipv6_mask_is_any(&ip6_mask
)) {
2777 /* offer_ip doesn't belongs to the cidr defined in lport's DHCPv6
2782 const struct smap
*options_map
= &op
->nbsp
->dhcpv6_options
->options
;
2783 /* "server_id" should be the MAC address. */
2784 const char *server_mac
= smap_get(options_map
, "server_id");
2786 if (!server_mac
|| !eth_addr_from_string(server_mac
, &ea
)) {
2787 /* "server_id" should be present in the dhcpv6_options. */
2788 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
2789 VLOG_WARN_RL(&rl
, "server_id not present in the DHCPv6 options"
2790 " for lport %s", op
->json_key
);
2794 /* Get the link local IP of the DHCPv6 server from the server MAC. */
2795 struct in6_addr lla
;
2796 in6_generate_lla(ea
, &lla
);
2798 char server_ip
[INET6_ADDRSTRLEN
+ 1];
2799 ipv6_string_mapped(server_ip
, &lla
);
2801 char ia_addr
[INET6_ADDRSTRLEN
+ 1];
2802 ipv6_string_mapped(ia_addr
, offer_ip
);
2804 ds_put_format(options_action
,
2805 REGBIT_DHCP_OPTS_RESULT
" = put_dhcpv6_opts(");
2807 /* Check whether the dhcpv6 options should be configured as stateful.
2808 * Only reply with ia_addr option for dhcpv6 stateful address mode. */
2809 if (!smap_get_bool(options_map
, "dhcpv6_stateless", false)) {
2810 ipv6_string_mapped(ia_addr
, offer_ip
);
2811 ds_put_format(options_action
, "ia_addr = %s, ", ia_addr
);
2814 /* We're not using SMAP_FOR_EACH because we want a consistent order of the
2815 * options on different architectures (big or little endian, SSE4.2) */
2816 const struct smap_node
**sorted_opts
= smap_sort(options_map
);
2817 for (size_t i
= 0; i
< smap_count(options_map
); i
++) {
2818 const struct smap_node
*node
= sorted_opts
[i
];
2819 if (strcmp(node
->key
, "dhcpv6_stateless")) {
2820 ds_put_format(options_action
, "%s = %s, ", node
->key
, node
->value
);
2825 ds_chomp(options_action
, ' ');
2826 ds_chomp(options_action
, ',');
2827 ds_put_cstr(options_action
, "); next;");
2829 ds_put_format(response_action
, "eth.dst = eth.src; eth.src = %s; "
2830 "ip6.dst = ip6.src; ip6.src = %s; udp.src = 547; "
2831 "udp.dst = 546; outport = inport; flags.loopback = 1; "
2833 server_mac
, server_ip
);
2839 has_stateful_acl(struct ovn_datapath
*od
)
2841 for (size_t i
= 0; i
< od
->nbs
->n_acls
; i
++) {
2842 struct nbrec_acl
*acl
= od
->nbs
->acls
[i
];
2843 if (!strcmp(acl
->action
, "allow-related")) {
2852 build_pre_acls(struct ovn_datapath
*od
, struct hmap
*lflows
)
2854 bool has_stateful
= has_stateful_acl(od
);
2856 /* Ingress and Egress Pre-ACL Table (Priority 0): Packets are
2857 * allowed by default. */
2858 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PRE_ACL
, 0, "1", "next;");
2859 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_PRE_ACL
, 0, "1", "next;");
2861 /* If there are any stateful ACL rules in this datapath, we must
2862 * send all IP packets through the conntrack action, which handles
2863 * defragmentation, in order to match L4 headers. */
2865 for (size_t i
= 0; i
< od
->n_router_ports
; i
++) {
2866 struct ovn_port
*op
= od
->router_ports
[i
];
2867 /* Can't use ct() for router ports. Consider the
2868 * following configuration: lp1(10.0.0.2) on
2869 * hostA--ls1--lr0--ls2--lp2(10.0.1.2) on hostB, For a
2870 * ping from lp1 to lp2, First, the response will go
2871 * through ct() with a zone for lp2 in the ls2 ingress
2872 * pipeline on hostB. That ct zone knows about this
2873 * connection. Next, it goes through ct() with the zone
2874 * for the router port in the egress pipeline of ls2 on
2875 * hostB. This zone does not know about the connection,
2876 * as the icmp request went through the logical router
2877 * on hostA, not hostB. This would only work with
2878 * distributed conntrack state across all chassis. */
2879 struct ds match_in
= DS_EMPTY_INITIALIZER
;
2880 struct ds match_out
= DS_EMPTY_INITIALIZER
;
2882 ds_put_format(&match_in
, "ip && inport == %s", op
->json_key
);
2883 ds_put_format(&match_out
, "ip && outport == %s", op
->json_key
);
2884 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PRE_ACL
, 110,
2885 ds_cstr(&match_in
), "next;");
2886 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_PRE_ACL
, 110,
2887 ds_cstr(&match_out
), "next;");
2889 ds_destroy(&match_in
);
2890 ds_destroy(&match_out
);
2892 if (od
->localnet_port
) {
2893 struct ds match_in
= DS_EMPTY_INITIALIZER
;
2894 struct ds match_out
= DS_EMPTY_INITIALIZER
;
2896 ds_put_format(&match_in
, "ip && inport == %s",
2897 od
->localnet_port
->json_key
);
2898 ds_put_format(&match_out
, "ip && outport == %s",
2899 od
->localnet_port
->json_key
);
2900 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PRE_ACL
, 110,
2901 ds_cstr(&match_in
), "next;");
2902 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_PRE_ACL
, 110,
2903 ds_cstr(&match_out
), "next;");
2905 ds_destroy(&match_in
);
2906 ds_destroy(&match_out
);
2909 /* Ingress and Egress Pre-ACL Table (Priority 110).
2911 * Not to do conntrack on ND and ICMP destination
2912 * unreachable packets. */
2913 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PRE_ACL
, 110,
2914 "nd || nd_rs || nd_ra || icmp4.type == 3 || "
2915 "icmp6.type == 1 || (tcp && tcp.flags == 4)",
2917 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_PRE_ACL
, 110,
2918 "nd || nd_rs || nd_ra || icmp4.type == 3 || "
2919 "icmp6.type == 1 || (tcp && tcp.flags == 4)",
2922 /* Ingress and Egress Pre-ACL Table (Priority 100).
2924 * Regardless of whether the ACL is "from-lport" or "to-lport",
2925 * we need rules in both the ingress and egress table, because
2926 * the return traffic needs to be followed.
2928 * 'REGBIT_CONNTRACK_DEFRAG' is set to let the pre-stateful table send
2929 * it to conntrack for tracking and defragmentation. */
2930 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PRE_ACL
, 100, "ip",
2931 REGBIT_CONNTRACK_DEFRAG
" = 1; next;");
2932 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_PRE_ACL
, 100, "ip",
2933 REGBIT_CONNTRACK_DEFRAG
" = 1; next;");
2937 /* For a 'key' of the form "IP:port" or just "IP", sets 'port' and
2938 * 'ip_address'. The caller must free() the memory allocated for
2941 ip_address_and_port_from_lb_key(const char *key
, char **ip_address
,
2942 uint16_t *port
, int *addr_family
)
2944 struct sockaddr_storage ss
;
2945 if (!inet_parse_active(key
, 0, &ss
)) {
2946 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
2947 VLOG_WARN_RL(&rl
, "bad ip address or port for load balancer key %s",
2952 struct ds s
= DS_EMPTY_INITIALIZER
;
2953 ss_format_address_nobracks(&ss
, &s
);
2954 *ip_address
= ds_steal_cstr(&s
);
2956 *port
= ss_get_port(&ss
);
2958 *addr_family
= ss
.ss_family
;
2962 * Returns true if logical switch is configured with DNS records, false
2966 ls_has_dns_records(const struct nbrec_logical_switch
*nbs
)
2968 for (size_t i
= 0; i
< nbs
->n_dns_records
; i
++) {
2969 if (!smap_is_empty(&nbs
->dns_records
[i
]->records
)) {
2978 build_pre_lb(struct ovn_datapath
*od
, struct hmap
*lflows
)
2980 /* Allow all packets to go to next tables by default. */
2981 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PRE_LB
, 0, "1", "next;");
2982 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_PRE_LB
, 0, "1", "next;");
2984 struct sset all_ips
= SSET_INITIALIZER(&all_ips
);
2985 bool vip_configured
= false;
2986 int addr_family
= AF_INET
;
2987 for (int i
= 0; i
< od
->nbs
->n_load_balancer
; i
++) {
2988 struct nbrec_load_balancer
*lb
= od
->nbs
->load_balancer
[i
];
2989 struct smap
*vips
= &lb
->vips
;
2990 struct smap_node
*node
;
2992 SMAP_FOR_EACH (node
, vips
) {
2993 vip_configured
= true;
2995 /* node->key contains IP:port or just IP. */
2996 char *ip_address
= NULL
;
2998 ip_address_and_port_from_lb_key(node
->key
, &ip_address
, &port
,
3004 if (!sset_contains(&all_ips
, ip_address
)) {
3005 sset_add(&all_ips
, ip_address
);
3010 /* Ignore L4 port information in the key because fragmented packets
3011 * may not have L4 information. The pre-stateful table will send
3012 * the packet through ct() action to de-fragment. In stateful
3013 * table, we will eventually look at L4 information. */
3017 /* 'REGBIT_CONNTRACK_DEFRAG' is set to let the pre-stateful table send
3018 * packet to conntrack for defragmentation. */
3019 const char *ip_address
;
3020 SSET_FOR_EACH(ip_address
, &all_ips
) {
3023 if (addr_family
== AF_INET
) {
3024 match
= xasprintf("ip && ip4.dst == %s", ip_address
);
3026 match
= xasprintf("ip && ip6.dst == %s", ip_address
);
3028 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PRE_LB
,
3029 100, match
, REGBIT_CONNTRACK_DEFRAG
" = 1; next;");
3033 sset_destroy(&all_ips
);
3035 if (vip_configured
) {
3036 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_PRE_LB
,
3037 100, "ip", REGBIT_CONNTRACK_DEFRAG
" = 1; next;");
3042 build_pre_stateful(struct ovn_datapath
*od
, struct hmap
*lflows
)
3044 /* Ingress and Egress pre-stateful Table (Priority 0): Packets are
3045 * allowed by default. */
3046 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PRE_STATEFUL
, 0, "1", "next;");
3047 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_PRE_STATEFUL
, 0, "1", "next;");
3049 /* If REGBIT_CONNTRACK_DEFRAG is set as 1, then the packets should be
3050 * sent to conntrack for tracking and defragmentation. */
3051 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PRE_STATEFUL
, 100,
3052 REGBIT_CONNTRACK_DEFRAG
" == 1", "ct_next;");
3053 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_PRE_STATEFUL
, 100,
3054 REGBIT_CONNTRACK_DEFRAG
" == 1", "ct_next;");
3058 build_acl_log(struct ds
*actions
, const struct nbrec_acl
*acl
)
3064 ds_put_cstr(actions
, "log(");
3067 ds_put_format(actions
, "name=\"%s\", ", acl
->name
);
3070 /* If a severity level isn't specified, default to "info". */
3071 if (acl
->severity
) {
3072 ds_put_format(actions
, "severity=%s, ", acl
->severity
);
3074 ds_put_format(actions
, "severity=info, ");
3077 if (!strcmp(acl
->action
, "drop")) {
3078 ds_put_cstr(actions
, "verdict=drop, ");
3079 } else if (!strcmp(acl
->action
, "reject")) {
3080 ds_put_cstr(actions
, "verdict=reject, ");
3081 } else if (!strcmp(acl
->action
, "allow")
3082 || !strcmp(acl
->action
, "allow-related")) {
3083 ds_put_cstr(actions
, "verdict=allow, ");
3086 ds_chomp(actions
, ' ');
3087 ds_chomp(actions
, ',');
3088 ds_put_cstr(actions
, "); ");
3092 build_reject_acl_rules(struct ovn_datapath
*od
, struct hmap
*lflows
,
3093 enum ovn_stage stage
, struct nbrec_acl
*acl
,
3094 struct ds
*extra_match
, struct ds
*extra_actions
)
3096 struct ds match
= DS_EMPTY_INITIALIZER
;
3097 struct ds actions
= DS_EMPTY_INITIALIZER
;
3098 bool ingress
= (stage
== S_SWITCH_IN_ACL
);
3101 build_acl_log(&actions
, acl
);
3102 if (extra_match
->length
> 0) {
3103 ds_put_format(&match
, "(%s) && ", extra_match
->string
);
3105 ds_put_format(&match
, "ip4 && tcp && (%s)", acl
->match
);
3106 ds_put_format(&actions
, "reg0 = 0; "
3107 "eth.dst <-> eth.src; ip4.dst <-> ip4.src; "
3108 "tcp_reset { outport <-> inport; %s };",
3109 ingress
? "output;" : "next(pipeline=ingress,table=0);");
3110 ovn_lflow_add(lflows
, od
, stage
, acl
->priority
+ OVN_ACL_PRI_OFFSET
+ 10,
3111 ds_cstr(&match
), ds_cstr(&actions
));
3114 build_acl_log(&actions
, acl
);
3115 if (extra_match
->length
> 0) {
3116 ds_put_format(&match
, "(%s) && ", extra_match
->string
);
3118 ds_put_format(&match
, "ip6 && tcp && (%s)", acl
->match
);
3119 ds_put_format(&actions
, "reg0 = 0; "
3120 "eth.dst <-> eth.src; ip6.dst <-> ip6.src; "
3121 "tcp_reset { outport <-> inport; %s };",
3122 ingress
? "output;" : "next(pipeline=ingress,table=0);");
3123 ovn_lflow_add(lflows
, od
, stage
, acl
->priority
+ OVN_ACL_PRI_OFFSET
+ 10,
3124 ds_cstr(&match
), ds_cstr(&actions
));
3129 build_acl_log(&actions
, acl
);
3130 if (extra_match
->length
> 0) {
3131 ds_put_format(&match
, "(%s) && ", extra_match
->string
);
3133 ds_put_format(&match
, "ip4 && (%s)", acl
->match
);
3134 if (extra_actions
->length
> 0) {
3135 ds_put_format(&actions
, "%s ", extra_actions
->string
);
3137 ds_put_format(&actions
, "reg0 = 0; "
3138 "eth.dst <-> eth.src; ip4.dst <-> ip4.src; "
3139 "icmp4 { outport <-> inport; %s };",
3140 ingress
? "output;" : "next(pipeline=ingress,table=0);");
3141 ovn_lflow_add(lflows
, od
, stage
, acl
->priority
+ OVN_ACL_PRI_OFFSET
,
3142 ds_cstr(&match
), ds_cstr(&actions
));
3145 build_acl_log(&actions
, acl
);
3146 if (extra_match
->length
> 0) {
3147 ds_put_format(&match
, "(%s) && ", extra_match
->string
);
3149 ds_put_format(&match
, "ip6 && (%s)", acl
->match
);
3150 if (extra_actions
->length
> 0) {
3151 ds_put_format(&actions
, "%s ", extra_actions
->string
);
3153 ds_put_format(&actions
, "reg0 = 0; icmp6 { "
3154 "eth.dst <-> eth.src; ip6.dst <-> ip6.src; "
3155 "outport <-> inport; %s };",
3156 ingress
? "output;" : "next(pipeline=ingress,table=0);");
3157 ovn_lflow_add(lflows
, od
, stage
, acl
->priority
+ OVN_ACL_PRI_OFFSET
,
3158 ds_cstr(&match
), ds_cstr(&actions
));
3161 ds_destroy(&actions
);
3165 consider_acl(struct hmap
*lflows
, struct ovn_datapath
*od
,
3166 struct nbrec_acl
*acl
, bool has_stateful
)
3168 bool ingress
= !strcmp(acl
->direction
, "from-lport") ? true :false;
3169 enum ovn_stage stage
= ingress
? S_SWITCH_IN_ACL
: S_SWITCH_OUT_ACL
;
3171 char *stage_hint
= xasprintf("%08x", acl
->header_
.uuid
.parts
[0]);
3172 if (!strcmp(acl
->action
, "allow")
3173 || !strcmp(acl
->action
, "allow-related")) {
3174 /* If there are any stateful flows, we must even commit "allow"
3175 * actions. This is because, while the initiater's
3176 * direction may not have any stateful rules, the server's
3177 * may and then its return traffic would not have an
3178 * associated conntrack entry and would return "+invalid". */
3179 if (!has_stateful
) {
3180 struct ds actions
= DS_EMPTY_INITIALIZER
;
3181 build_acl_log(&actions
, acl
);
3182 ds_put_cstr(&actions
, "next;");
3183 ovn_lflow_add_with_hint(lflows
, od
, stage
,
3184 acl
->priority
+ OVN_ACL_PRI_OFFSET
,
3185 acl
->match
, ds_cstr(&actions
),
3187 ds_destroy(&actions
);
3189 struct ds match
= DS_EMPTY_INITIALIZER
;
3190 struct ds actions
= DS_EMPTY_INITIALIZER
;
3192 /* Commit the connection tracking entry if it's a new
3193 * connection that matches this ACL. After this commit,
3194 * the reply traffic is allowed by a flow we create at
3195 * priority 65535, defined earlier.
3197 * It's also possible that a known connection was marked for
3198 * deletion after a policy was deleted, but the policy was
3199 * re-added while that connection is still known. We catch
3200 * that case here and un-set ct_label.blocked (which will be done
3201 * by ct_commit in the "stateful" stage) to indicate that the
3202 * connection should be allowed to resume.
3204 ds_put_format(&match
, "((ct.new && !ct.est)"
3205 " || (!ct.new && ct.est && !ct.rpl "
3206 "&& ct_label.blocked == 1)) "
3207 "&& (%s)", acl
->match
);
3208 ds_put_cstr(&actions
, REGBIT_CONNTRACK_COMMIT
" = 1; ");
3209 build_acl_log(&actions
, acl
);
3210 ds_put_cstr(&actions
, "next;");
3211 ovn_lflow_add_with_hint(lflows
, od
, stage
,
3212 acl
->priority
+ OVN_ACL_PRI_OFFSET
,
3217 /* Match on traffic in the request direction for an established
3218 * connection tracking entry that has not been marked for
3219 * deletion. There is no need to commit here, so we can just
3220 * proceed to the next table. We use this to ensure that this
3221 * connection is still allowed by the currently defined
3225 ds_put_format(&match
,
3226 "!ct.new && ct.est && !ct.rpl"
3227 " && ct_label.blocked == 0 && (%s)",
3230 build_acl_log(&actions
, acl
);
3231 ds_put_cstr(&actions
, "next;");
3232 ovn_lflow_add_with_hint(lflows
, od
, stage
,
3233 acl
->priority
+ OVN_ACL_PRI_OFFSET
,
3234 ds_cstr(&match
), ds_cstr(&actions
),
3238 ds_destroy(&actions
);
3240 } else if (!strcmp(acl
->action
, "drop")
3241 || !strcmp(acl
->action
, "reject")) {
3242 struct ds match
= DS_EMPTY_INITIALIZER
;
3243 struct ds actions
= DS_EMPTY_INITIALIZER
;
3245 /* The implementation of "drop" differs if stateful ACLs are in
3246 * use for this datapath. In that case, the actions differ
3247 * depending on whether the connection was previously committed
3248 * to the connection tracker with ct_commit. */
3250 /* If the packet is not part of an established connection, then
3251 * we can simply reject/drop it. */
3253 "(!ct.est || (ct.est && ct_label.blocked == 1))");
3254 if (!strcmp(acl
->action
, "reject")) {
3255 build_reject_acl_rules(od
, lflows
, stage
, acl
, &match
,
3258 ds_put_format(&match
, " && (%s)", acl
->match
);
3259 build_acl_log(&actions
, acl
);
3260 ds_put_cstr(&actions
, "/* drop */");
3261 ovn_lflow_add(lflows
, od
, stage
,
3262 acl
->priority
+ OVN_ACL_PRI_OFFSET
,
3263 ds_cstr(&match
), ds_cstr(&actions
));
3265 /* For an existing connection without ct_label set, we've
3266 * encountered a policy change. ACLs previously allowed
3267 * this connection and we committed the connection tracking
3268 * entry. Current policy says that we should drop this
3269 * connection. First, we set bit 0 of ct_label to indicate
3270 * that this connection is set for deletion. By not
3271 * specifying "next;", we implicitly drop the packet after
3272 * updating conntrack state. We would normally defer
3273 * ct_commit() to the "stateful" stage, but since we're
3274 * rejecting/dropping the packet, we go ahead and do it here.
3278 ds_put_cstr(&match
, "ct.est && ct_label.blocked == 0");
3279 ds_put_cstr(&actions
, "ct_commit(ct_label=1/1); ");
3280 if (!strcmp(acl
->action
, "reject")) {
3281 build_reject_acl_rules(od
, lflows
, stage
, acl
, &match
,
3284 ds_put_format(&match
, " && (%s)", acl
->match
);
3285 build_acl_log(&actions
, acl
);
3286 ds_put_cstr(&actions
, "/* drop */");
3287 ovn_lflow_add(lflows
, od
, stage
,
3288 acl
->priority
+ OVN_ACL_PRI_OFFSET
,
3289 ds_cstr(&match
), ds_cstr(&actions
));
3292 /* There are no stateful ACLs in use on this datapath,
3293 * so a "reject/drop" ACL is simply the "reject/drop"
3294 * logical flow action in all cases. */
3295 if (!strcmp(acl
->action
, "reject")) {
3296 build_reject_acl_rules(od
, lflows
, stage
, acl
, &match
,
3299 build_acl_log(&actions
, acl
);
3300 ds_put_cstr(&actions
, "/* drop */");
3301 ovn_lflow_add(lflows
, od
, stage
,
3302 acl
->priority
+ OVN_ACL_PRI_OFFSET
,
3303 acl
->match
, ds_cstr(&actions
));
3307 ds_destroy(&actions
);
3312 struct ovn_port_group_ls
{
3313 struct hmap_node key_node
; /* Index on 'key'. */
3314 struct uuid key
; /* nb_ls->header_.uuid. */
3315 const struct nbrec_logical_switch
*nb_ls
;
3318 struct ovn_port_group
{
3319 struct hmap_node key_node
; /* Index on 'key'. */
3320 struct uuid key
; /* nb_pg->header_.uuid. */
3321 const struct nbrec_port_group
*nb_pg
;
3322 struct hmap nb_lswitches
; /* NB lswitches related to the port group */
3323 size_t n_acls
; /* Number of ACLs applied to the port group */
3324 struct nbrec_acl
**acls
; /* ACLs applied to the port group */
3327 static struct ovn_port_group
*
3328 ovn_port_group_create(struct hmap
*pgs
,
3329 const struct nbrec_port_group
*nb_pg
)
3331 struct ovn_port_group
*pg
= xzalloc(sizeof *pg
);
3332 pg
->key
= nb_pg
->header_
.uuid
;
3334 pg
->n_acls
= nb_pg
->n_acls
;
3335 pg
->acls
= nb_pg
->acls
;
3336 hmap_init(&pg
->nb_lswitches
);
3337 hmap_insert(pgs
, &pg
->key_node
, uuid_hash(&pg
->key
));
3342 ovn_port_group_ls_add(struct ovn_port_group
*pg
,
3343 const struct nbrec_logical_switch
*nb_ls
)
3345 struct ovn_port_group_ls
*pg_ls
= xzalloc(sizeof *pg_ls
);
3346 pg_ls
->key
= nb_ls
->header_
.uuid
;
3347 pg_ls
->nb_ls
= nb_ls
;
3348 hmap_insert(&pg
->nb_lswitches
, &pg_ls
->key_node
, uuid_hash(&pg_ls
->key
));
3351 static struct ovn_port_group_ls
*
3352 ovn_port_group_ls_find(struct ovn_port_group
*pg
, const struct uuid
*ls_uuid
)
3354 struct ovn_port_group_ls
*pg_ls
;
3356 HMAP_FOR_EACH_WITH_HASH (pg_ls
, key_node
, uuid_hash(ls_uuid
),
3357 &pg
->nb_lswitches
) {
3358 if (uuid_equals(ls_uuid
, &pg_ls
->key
)) {
3366 ovn_port_group_destroy(struct hmap
*pgs
, struct ovn_port_group
*pg
)
3369 hmap_remove(pgs
, &pg
->key_node
);
3370 struct ovn_port_group_ls
*ls
;
3371 HMAP_FOR_EACH_POP (ls
, key_node
, &pg
->nb_lswitches
) {
3374 hmap_destroy(&pg
->nb_lswitches
);
3380 build_port_group_lswitches(struct northd_context
*ctx
, struct hmap
*pgs
,
3385 const struct nbrec_port_group
*nb_pg
;
3386 NBREC_PORT_GROUP_FOR_EACH (nb_pg
, ctx
->ovnnb_idl
) {
3387 struct ovn_port_group
*pg
= ovn_port_group_create(pgs
, nb_pg
);
3388 for (size_t i
= 0; i
< nb_pg
->n_ports
; i
++) {
3389 struct ovn_port
*op
= ovn_port_find(ports
, nb_pg
->ports
[i
]->name
);
3391 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
3392 VLOG_ERR_RL(&rl
, "lport %s in port group %s not found.",
3393 nb_pg
->ports
[i
]->name
,
3399 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
3400 VLOG_WARN_RL(&rl
, "lport %s in port group %s has no lswitch.",
3401 nb_pg
->ports
[i
]->name
,
3406 struct ovn_port_group_ls
*pg_ls
=
3407 ovn_port_group_ls_find(pg
, &op
->od
->nbs
->header_
.uuid
);
3409 ovn_port_group_ls_add(pg
, op
->od
->nbs
);
3416 build_acls(struct ovn_datapath
*od
, struct hmap
*lflows
,
3417 struct hmap
*port_groups
)
3419 bool has_stateful
= has_stateful_acl(od
);
3421 /* Ingress and Egress ACL Table (Priority 0): Packets are allowed by
3422 * default. A related rule at priority 1 is added below if there
3423 * are any stateful ACLs in this datapath. */
3424 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_ACL
, 0, "1", "next;");
3425 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_ACL
, 0, "1", "next;");
3428 /* Ingress and Egress ACL Table (Priority 1).
3430 * By default, traffic is allowed. This is partially handled by
3431 * the Priority 0 ACL flows added earlier, but we also need to
3432 * commit IP flows. This is because, while the initiater's
3433 * direction may not have any stateful rules, the server's may
3434 * and then its return traffic would not have an associated
3435 * conntrack entry and would return "+invalid".
3437 * We use "ct_commit" for a connection that is not already known
3438 * by the connection tracker. Once a connection is committed,
3439 * subsequent packets will hit the flow at priority 0 that just
3442 * We also check for established connections that have ct_label.blocked
3443 * set on them. That's a connection that was disallowed, but is
3444 * now allowed by policy again since it hit this default-allow flow.
3445 * We need to set ct_label.blocked=0 to let the connection continue,
3446 * which will be done by ct_commit() in the "stateful" stage.
3447 * Subsequent packets will hit the flow at priority 0 that just
3449 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_ACL
, 1,
3450 "ip && (!ct.est || (ct.est && ct_label.blocked == 1))",
3451 REGBIT_CONNTRACK_COMMIT
" = 1; next;");
3452 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_ACL
, 1,
3453 "ip && (!ct.est || (ct.est && ct_label.blocked == 1))",
3454 REGBIT_CONNTRACK_COMMIT
" = 1; next;");
3456 /* Ingress and Egress ACL Table (Priority 65535).
3458 * Always drop traffic that's in an invalid state. Also drop
3459 * reply direction packets for connections that have been marked
3460 * for deletion (bit 0 of ct_label is set).
3462 * This is enforced at a higher priority than ACLs can be defined. */
3463 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_ACL
, UINT16_MAX
,
3464 "ct.inv || (ct.est && ct.rpl && ct_label.blocked == 1)",
3466 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_ACL
, UINT16_MAX
,
3467 "ct.inv || (ct.est && ct.rpl && ct_label.blocked == 1)",
3470 /* Ingress and Egress ACL Table (Priority 65535).
3472 * Allow reply traffic that is part of an established
3473 * conntrack entry that has not been marked for deletion
3474 * (bit 0 of ct_label). We only match traffic in the
3475 * reply direction because we want traffic in the request
3476 * direction to hit the currently defined policy from ACLs.
3478 * This is enforced at a higher priority than ACLs can be defined. */
3479 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_ACL
, UINT16_MAX
,
3480 "ct.est && !ct.rel && !ct.new && !ct.inv "
3481 "&& ct.rpl && ct_label.blocked == 0",
3483 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_ACL
, UINT16_MAX
,
3484 "ct.est && !ct.rel && !ct.new && !ct.inv "
3485 "&& ct.rpl && ct_label.blocked == 0",
3488 /* Ingress and Egress ACL Table (Priority 65535).
3490 * Allow traffic that is related to an existing conntrack entry that
3491 * has not been marked for deletion (bit 0 of ct_label).
3493 * This is enforced at a higher priority than ACLs can be defined.
3495 * NOTE: This does not support related data sessions (eg,
3496 * a dynamically negotiated FTP data channel), but will allow
3497 * related traffic such as an ICMP Port Unreachable through
3498 * that's generated from a non-listening UDP port. */
3499 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_ACL
, UINT16_MAX
,
3500 "!ct.est && ct.rel && !ct.new && !ct.inv "
3501 "&& ct_label.blocked == 0",
3503 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_ACL
, UINT16_MAX
,
3504 "!ct.est && ct.rel && !ct.new && !ct.inv "
3505 "&& ct_label.blocked == 0",
3508 /* Ingress and Egress ACL Table (Priority 65535).
3510 * Not to do conntrack on ND packets. */
3511 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_ACL
, UINT16_MAX
, "nd", "next;");
3512 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_ACL
, UINT16_MAX
, "nd", "next;");
3515 /* Ingress or Egress ACL Table (Various priorities). */
3516 for (size_t i
= 0; i
< od
->nbs
->n_acls
; i
++) {
3517 struct nbrec_acl
*acl
= od
->nbs
->acls
[i
];
3518 consider_acl(lflows
, od
, acl
, has_stateful
);
3520 struct ovn_port_group
*pg
;
3521 HMAP_FOR_EACH (pg
, key_node
, port_groups
) {
3522 if (ovn_port_group_ls_find(pg
, &od
->nbs
->header_
.uuid
)) {
3523 for (size_t i
= 0; i
< pg
->n_acls
; i
++) {
3524 consider_acl(lflows
, od
, pg
->acls
[i
], has_stateful
);
3529 /* Add 34000 priority flow to allow DHCP reply from ovn-controller to all
3530 * logical ports of the datapath if the CMS has configured DHCPv4 options.
3532 for (size_t i
= 0; i
< od
->nbs
->n_ports
; i
++) {
3533 if (od
->nbs
->ports
[i
]->dhcpv4_options
) {
3534 const char *server_id
= smap_get(
3535 &od
->nbs
->ports
[i
]->dhcpv4_options
->options
, "server_id");
3536 const char *server_mac
= smap_get(
3537 &od
->nbs
->ports
[i
]->dhcpv4_options
->options
, "server_mac");
3538 const char *lease_time
= smap_get(
3539 &od
->nbs
->ports
[i
]->dhcpv4_options
->options
, "lease_time");
3540 if (server_id
&& server_mac
&& lease_time
) {
3541 struct ds match
= DS_EMPTY_INITIALIZER
;
3542 const char *actions
=
3543 has_stateful
? "ct_commit; next;" : "next;";
3544 ds_put_format(&match
, "outport == \"%s\" && eth.src == %s "
3545 "&& ip4.src == %s && udp && udp.src == 67 "
3546 "&& udp.dst == 68", od
->nbs
->ports
[i
]->name
,
3547 server_mac
, server_id
);
3549 lflows
, od
, S_SWITCH_OUT_ACL
, 34000, ds_cstr(&match
),
3555 if (od
->nbs
->ports
[i
]->dhcpv6_options
) {
3556 const char *server_mac
= smap_get(
3557 &od
->nbs
->ports
[i
]->dhcpv6_options
->options
, "server_id");
3559 if (server_mac
&& eth_addr_from_string(server_mac
, &ea
)) {
3560 /* Get the link local IP of the DHCPv6 server from the
3562 struct in6_addr lla
;
3563 in6_generate_lla(ea
, &lla
);
3565 char server_ip
[INET6_ADDRSTRLEN
+ 1];
3566 ipv6_string_mapped(server_ip
, &lla
);
3568 struct ds match
= DS_EMPTY_INITIALIZER
;
3569 const char *actions
= has_stateful
? "ct_commit; next;" :
3571 ds_put_format(&match
, "outport == \"%s\" && eth.src == %s "
3572 "&& ip6.src == %s && udp && udp.src == 547 "
3573 "&& udp.dst == 546", od
->nbs
->ports
[i
]->name
,
3574 server_mac
, server_ip
);
3576 lflows
, od
, S_SWITCH_OUT_ACL
, 34000, ds_cstr(&match
),
3583 /* Add a 34000 priority flow to advance the DNS reply from ovn-controller,
3584 * if the CMS has configured DNS records for the datapath.
3586 if (ls_has_dns_records(od
->nbs
)) {
3587 const char *actions
= has_stateful
? "ct_commit; next;" : "next;";
3589 lflows
, od
, S_SWITCH_OUT_ACL
, 34000, "udp.src == 53",
3595 build_qos(struct ovn_datapath
*od
, struct hmap
*lflows
) {
3596 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_QOS_MARK
, 0, "1", "next;");
3597 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_QOS_MARK
, 0, "1", "next;");
3598 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_QOS_METER
, 0, "1", "next;");
3599 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_QOS_METER
, 0, "1", "next;");
3601 for (size_t i
= 0; i
< od
->nbs
->n_qos_rules
; i
++) {
3602 struct nbrec_qos
*qos
= od
->nbs
->qos_rules
[i
];
3603 bool ingress
= !strcmp(qos
->direction
, "from-lport") ? true :false;
3604 enum ovn_stage stage
= ingress
? S_SWITCH_IN_QOS_MARK
: S_SWITCH_OUT_QOS_MARK
;
3608 for (size_t j
= 0; j
< qos
->n_action
; j
++) {
3609 if (!strcmp(qos
->key_action
[j
], "dscp")) {
3610 struct ds dscp_action
= DS_EMPTY_INITIALIZER
;
3612 ds_put_format(&dscp_action
, "ip.dscp = %"PRId64
"; next;",
3613 qos
->value_action
[j
]);
3614 ovn_lflow_add(lflows
, od
, stage
,
3616 qos
->match
, ds_cstr(&dscp_action
));
3617 ds_destroy(&dscp_action
);
3621 for (size_t n
= 0; n
< qos
->n_bandwidth
; n
++) {
3622 if (!strcmp(qos
->key_bandwidth
[n
], "rate")) {
3623 rate
= qos
->value_bandwidth
[n
];
3624 } else if (!strcmp(qos
->key_bandwidth
[n
], "burst")) {
3625 burst
= qos
->value_bandwidth
[n
];
3629 struct ds meter_action
= DS_EMPTY_INITIALIZER
;
3630 stage
= ingress
? S_SWITCH_IN_QOS_METER
: S_SWITCH_OUT_QOS_METER
;
3632 ds_put_format(&meter_action
,
3633 "set_meter(%"PRId64
", %"PRId64
"); next;",
3636 ds_put_format(&meter_action
,
3637 "set_meter(%"PRId64
"); next;",
3641 /* Ingress and Egress QoS Meter Table.
3643 * We limit the bandwidth of this flow by adding a meter table.
3645 ovn_lflow_add(lflows
, od
, stage
,
3647 qos
->match
, ds_cstr(&meter_action
));
3648 ds_destroy(&meter_action
);
3654 build_lb(struct ovn_datapath
*od
, struct hmap
*lflows
)
3656 /* Ingress and Egress LB Table (Priority 0): Packets are allowed by
3658 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_LB
, 0, "1", "next;");
3659 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_LB
, 0, "1", "next;");
3661 if (od
->nbs
->load_balancer
) {
3662 /* Ingress and Egress LB Table (Priority 65535).
3664 * Send established traffic through conntrack for just NAT. */
3665 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_LB
, UINT16_MAX
,
3666 "ct.est && !ct.rel && !ct.new && !ct.inv",
3667 REGBIT_CONNTRACK_NAT
" = 1; next;");
3668 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_LB
, UINT16_MAX
,
3669 "ct.est && !ct.rel && !ct.new && !ct.inv",
3670 REGBIT_CONNTRACK_NAT
" = 1; next;");
3675 build_stateful(struct ovn_datapath
*od
, struct hmap
*lflows
)
3677 /* Ingress and Egress stateful Table (Priority 0): Packets are
3678 * allowed by default. */
3679 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_STATEFUL
, 0, "1", "next;");
3680 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_STATEFUL
, 0, "1", "next;");
3682 /* If REGBIT_CONNTRACK_COMMIT is set as 1, then the packets should be
3683 * committed to conntrack. We always set ct_label.blocked to 0 here as
3684 * any packet that makes it this far is part of a connection we
3685 * want to allow to continue. */
3686 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_STATEFUL
, 100,
3687 REGBIT_CONNTRACK_COMMIT
" == 1", "ct_commit(ct_label=0/1); next;");
3688 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_STATEFUL
, 100,
3689 REGBIT_CONNTRACK_COMMIT
" == 1", "ct_commit(ct_label=0/1); next;");
3691 /* If REGBIT_CONNTRACK_NAT is set as 1, then packets should just be sent
3692 * through nat (without committing).
3694 * REGBIT_CONNTRACK_COMMIT is set for new connections and
3695 * REGBIT_CONNTRACK_NAT is set for established connections. So they
3698 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_STATEFUL
, 100,
3699 REGBIT_CONNTRACK_NAT
" == 1", "ct_lb;");
3700 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_STATEFUL
, 100,
3701 REGBIT_CONNTRACK_NAT
" == 1", "ct_lb;");
3703 /* Load balancing rules for new connections get committed to conntrack
3704 * table. So even if REGBIT_CONNTRACK_COMMIT is set in a previous table
3705 * a higher priority rule for load balancing below also commits the
3706 * connection, so it is okay if we do not hit the above match on
3707 * REGBIT_CONNTRACK_COMMIT. */
3708 for (int i
= 0; i
< od
->nbs
->n_load_balancer
; i
++) {
3709 struct nbrec_load_balancer
*lb
= od
->nbs
->load_balancer
[i
];
3710 struct smap
*vips
= &lb
->vips
;
3711 struct smap_node
*node
;
3713 SMAP_FOR_EACH (node
, vips
) {
3717 /* node->key contains IP:port or just IP. */
3718 char *ip_address
= NULL
;
3719 ip_address_and_port_from_lb_key(node
->key
, &ip_address
, &port
,
3725 /* New connections in Ingress table. */
3726 char *action
= xasprintf("ct_lb(%s);", node
->value
);
3727 struct ds match
= DS_EMPTY_INITIALIZER
;
3728 if (addr_family
== AF_INET
) {
3729 ds_put_format(&match
, "ct.new && ip4.dst == %s", ip_address
);
3731 ds_put_format(&match
, "ct.new && ip6.dst == %s", ip_address
);
3734 if (lb
->protocol
&& !strcmp(lb
->protocol
, "udp")) {
3735 ds_put_format(&match
, " && udp.dst == %d", port
);
3737 ds_put_format(&match
, " && tcp.dst == %d", port
);
3739 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_STATEFUL
,
3740 120, ds_cstr(&match
), action
);
3742 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_STATEFUL
,
3743 110, ds_cstr(&match
), action
);
3754 build_lswitch_flows(struct hmap
*datapaths
, struct hmap
*ports
,
3755 struct hmap
*port_groups
, struct hmap
*lflows
,
3756 struct hmap
*mcgroups
)
3758 /* This flow table structure is documented in ovn-northd(8), so please
3759 * update ovn-northd.8.xml if you change anything. */
3761 struct ds match
= DS_EMPTY_INITIALIZER
;
3762 struct ds actions
= DS_EMPTY_INITIALIZER
;
3764 /* Build pre-ACL and ACL tables for both ingress and egress.
3765 * Ingress tables 3 through 10. Egress tables 0 through 7. */
3766 struct ovn_datapath
*od
;
3767 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
3772 build_pre_acls(od
, lflows
);
3773 build_pre_lb(od
, lflows
);
3774 build_pre_stateful(od
, lflows
);
3775 build_acls(od
, lflows
, port_groups
);
3776 build_qos(od
, lflows
);
3777 build_lb(od
, lflows
);
3778 build_stateful(od
, lflows
);
3781 /* Logical switch ingress table 0: Admission control framework (priority
3783 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
3788 /* Logical VLANs not supported. */
3789 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PORT_SEC_L2
, 100, "vlan.present",
3792 /* Broadcast/multicast source address is invalid. */
3793 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PORT_SEC_L2
, 100, "eth.src[40]",
3796 /* Port security flows have priority 50 (see below) and will continue
3797 * to the next table if packet source is acceptable. */
3800 /* Logical switch ingress table 0: Ingress port security - L2
3802 * Ingress table 1: Ingress port security - IP (priority 90 and 80)
3803 * Ingress table 2: Ingress port security - ND (priority 90 and 80)
3805 struct ovn_port
*op
;
3806 HMAP_FOR_EACH (op
, key_node
, ports
) {
3811 if (!lsp_is_enabled(op
->nbsp
)) {
3812 /* Drop packets from disabled logical ports (since logical flow
3813 * tables are default-drop). */
3819 ds_put_format(&match
, "inport == %s", op
->json_key
);
3820 build_port_security_l2("eth.src", op
->ps_addrs
, op
->n_ps_addrs
,
3823 const char *queue_id
= smap_get(&op
->sb
->options
, "qdisc_queue_id");
3825 ds_put_format(&actions
, "set_queue(%s); ", queue_id
);
3827 ds_put_cstr(&actions
, "next;");
3828 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_PORT_SEC_L2
, 50,
3829 ds_cstr(&match
), ds_cstr(&actions
));
3831 if (op
->nbsp
->n_port_security
) {
3832 build_port_security_ip(P_IN
, op
, lflows
);
3833 build_port_security_nd(op
, lflows
);
3837 /* Ingress table 1 and 2: Port security - IP and ND, by default goto next.
3839 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
3844 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PORT_SEC_ND
, 0, "1", "next;");
3845 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_PORT_SEC_IP
, 0, "1", "next;");
3848 /* Ingress table 11: ARP/ND responder, skip requests coming from localnet
3849 * and vtep ports. (priority 100); see ovn-northd.8.xml for the
3851 HMAP_FOR_EACH (op
, key_node
, ports
) {
3856 if ((!strcmp(op
->nbsp
->type
, "localnet")) ||
3857 (!strcmp(op
->nbsp
->type
, "vtep"))) {
3859 ds_put_format(&match
, "inport == %s", op
->json_key
);
3860 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_ARP_ND_RSP
, 100,
3861 ds_cstr(&match
), "next;");
3865 /* Ingress table 11: ARP/ND responder, reply for known IPs.
3867 HMAP_FOR_EACH (op
, key_node
, ports
) {
3873 * Add ARP/ND reply flows if either the
3875 * - port type is router or
3876 * - port type is localport
3878 if (!lsp_is_up(op
->nbsp
) && strcmp(op
->nbsp
->type
, "router") &&
3879 strcmp(op
->nbsp
->type
, "localport")) {
3883 for (size_t i
= 0; i
< op
->n_lsp_addrs
; i
++) {
3884 for (size_t j
= 0; j
< op
->lsp_addrs
[i
].n_ipv4_addrs
; j
++) {
3886 ds_put_format(&match
, "arp.tpa == %s && arp.op == 1",
3887 op
->lsp_addrs
[i
].ipv4_addrs
[j
].addr_s
);
3889 ds_put_format(&actions
,
3890 "eth.dst = eth.src; "
3892 "arp.op = 2; /* ARP reply */ "
3893 "arp.tha = arp.sha; "
3895 "arp.tpa = arp.spa; "
3897 "outport = inport; "
3898 "flags.loopback = 1; "
3900 op
->lsp_addrs
[i
].ea_s
, op
->lsp_addrs
[i
].ea_s
,
3901 op
->lsp_addrs
[i
].ipv4_addrs
[j
].addr_s
);
3902 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_ARP_ND_RSP
, 50,
3903 ds_cstr(&match
), ds_cstr(&actions
));
3905 /* Do not reply to an ARP request from the port that owns the
3906 * address (otherwise a DHCP client that ARPs to check for a
3907 * duplicate address will fail). Instead, forward it the usual
3910 * (Another alternative would be to simply drop the packet. If
3911 * everything is working as it is configured, then this would
3912 * produce equivalent results, since no one should reply to the
3913 * request. But ARPing for one's own IP address is intended to
3914 * detect situations where the network is not working as
3915 * configured, so dropping the request would frustrate that
3917 ds_put_format(&match
, " && inport == %s", op
->json_key
);
3918 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_ARP_ND_RSP
, 100,
3919 ds_cstr(&match
), "next;");
3922 /* For ND solicitations, we need to listen for both the
3923 * unicast IPv6 address and its all-nodes multicast address,
3924 * but always respond with the unicast IPv6 address. */
3925 for (size_t j
= 0; j
< op
->lsp_addrs
[i
].n_ipv6_addrs
; j
++) {
3927 ds_put_format(&match
,
3928 "nd_ns && ip6.dst == {%s, %s} && nd.target == %s",
3929 op
->lsp_addrs
[i
].ipv6_addrs
[j
].addr_s
,
3930 op
->lsp_addrs
[i
].ipv6_addrs
[j
].sn_addr_s
,
3931 op
->lsp_addrs
[i
].ipv6_addrs
[j
].addr_s
);
3934 ds_put_format(&actions
,
3940 "outport = inport; "
3941 "flags.loopback = 1; "
3944 op
->lsp_addrs
[i
].ea_s
,
3945 op
->lsp_addrs
[i
].ipv6_addrs
[j
].addr_s
,
3946 op
->lsp_addrs
[i
].ipv6_addrs
[j
].addr_s
,
3947 op
->lsp_addrs
[i
].ea_s
);
3948 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_ARP_ND_RSP
, 50,
3949 ds_cstr(&match
), ds_cstr(&actions
));
3951 /* Do not reply to a solicitation from the port that owns the
3952 * address (otherwise DAD detection will fail). */
3953 ds_put_format(&match
, " && inport == %s", op
->json_key
);
3954 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_ARP_ND_RSP
, 100,
3955 ds_cstr(&match
), "next;");
3960 /* Ingress table 11: ARP/ND responder, by default goto next.
3962 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
3967 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_ARP_ND_RSP
, 0, "1", "next;");
3970 /* Logical switch ingress table 12 and 13: DHCP options and response
3971 * priority 100 flows. */
3972 HMAP_FOR_EACH (op
, key_node
, ports
) {
3977 if (!lsp_is_enabled(op
->nbsp
) || !strcmp(op
->nbsp
->type
, "router")) {
3978 /* Don't add the DHCP flows if the port is not enabled or if the
3979 * port is a router port. */
3983 if (!op
->nbsp
->dhcpv4_options
&& !op
->nbsp
->dhcpv6_options
) {
3984 /* CMS has disabled both native DHCPv4 and DHCPv6 for this lport.
3989 for (size_t i
= 0; i
< op
->n_lsp_addrs
; i
++) {
3990 for (size_t j
= 0; j
< op
->lsp_addrs
[i
].n_ipv4_addrs
; j
++) {
3991 struct ds options_action
= DS_EMPTY_INITIALIZER
;
3992 struct ds response_action
= DS_EMPTY_INITIALIZER
;
3993 struct ds ipv4_addr_match
= DS_EMPTY_INITIALIZER
;
3994 if (build_dhcpv4_action(
3995 op
, op
->lsp_addrs
[i
].ipv4_addrs
[j
].addr
,
3996 &options_action
, &response_action
, &ipv4_addr_match
)) {
3999 &match
, "inport == %s && eth.src == %s && "
4000 "ip4.src == 0.0.0.0 && ip4.dst == 255.255.255.255 && "
4001 "udp.src == 68 && udp.dst == 67", op
->json_key
,
4002 op
->lsp_addrs
[i
].ea_s
);
4004 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_DHCP_OPTIONS
,
4005 100, ds_cstr(&match
),
4006 ds_cstr(&options_action
));
4008 /* Allow ip4.src = OFFER_IP and
4009 * ip4.dst = {SERVER_IP, 255.255.255.255} for the below
4011 * - When the client wants to renew the IP by sending
4012 * the DHCPREQUEST to the server ip.
4013 * - When the client wants to renew the IP by
4014 * broadcasting the DHCPREQUEST.
4017 &match
, "inport == %s && eth.src == %s && "
4018 "%s && udp.src == 68 && udp.dst == 67", op
->json_key
,
4019 op
->lsp_addrs
[i
].ea_s
, ds_cstr(&ipv4_addr_match
));
4021 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_DHCP_OPTIONS
,
4022 100, ds_cstr(&match
),
4023 ds_cstr(&options_action
));
4026 /* If REGBIT_DHCP_OPTS_RESULT is set, it means the
4027 * put_dhcp_opts action is successful. */
4029 &match
, "inport == %s && eth.src == %s && "
4030 "ip4 && udp.src == 68 && udp.dst == 67"
4031 " && "REGBIT_DHCP_OPTS_RESULT
, op
->json_key
,
4032 op
->lsp_addrs
[i
].ea_s
);
4033 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_DHCP_RESPONSE
,
4034 100, ds_cstr(&match
),
4035 ds_cstr(&response_action
));
4036 ds_destroy(&options_action
);
4037 ds_destroy(&response_action
);
4038 ds_destroy(&ipv4_addr_match
);
4043 for (size_t j
= 0; j
< op
->lsp_addrs
[i
].n_ipv6_addrs
; j
++) {
4044 struct ds options_action
= DS_EMPTY_INITIALIZER
;
4045 struct ds response_action
= DS_EMPTY_INITIALIZER
;
4046 if (build_dhcpv6_action(
4047 op
, &op
->lsp_addrs
[i
].ipv6_addrs
[j
].addr
,
4048 &options_action
, &response_action
)) {
4051 &match
, "inport == %s && eth.src == %s"
4052 " && ip6.dst == ff02::1:2 && udp.src == 546 &&"
4053 " udp.dst == 547", op
->json_key
,
4054 op
->lsp_addrs
[i
].ea_s
);
4056 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_DHCP_OPTIONS
, 100,
4057 ds_cstr(&match
), ds_cstr(&options_action
));
4059 /* If REGBIT_DHCP_OPTS_RESULT is set to 1, it means the
4060 * put_dhcpv6_opts action is successful */
4061 ds_put_cstr(&match
, " && "REGBIT_DHCP_OPTS_RESULT
);
4062 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_DHCP_RESPONSE
, 100,
4063 ds_cstr(&match
), ds_cstr(&response_action
));
4064 ds_destroy(&options_action
);
4065 ds_destroy(&response_action
);
4072 /* Logical switch ingress table 14 and 15: DNS lookup and response
4073 * priority 100 flows.
4075 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
4076 if (!od
->nbs
|| !ls_has_dns_records(od
->nbs
)) {
4080 struct ds action
= DS_EMPTY_INITIALIZER
;
4083 ds_put_cstr(&match
, "udp.dst == 53");
4084 ds_put_format(&action
,
4085 REGBIT_DNS_LOOKUP_RESULT
" = dns_lookup(); next;");
4086 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_DNS_LOOKUP
, 100,
4087 ds_cstr(&match
), ds_cstr(&action
));
4089 ds_put_cstr(&match
, " && "REGBIT_DNS_LOOKUP_RESULT
);
4090 ds_put_format(&action
, "eth.dst <-> eth.src; ip4.src <-> ip4.dst; "
4091 "udp.dst = udp.src; udp.src = 53; outport = inport; "
4092 "flags.loopback = 1; output;");
4093 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_DNS_RESPONSE
, 100,
4094 ds_cstr(&match
), ds_cstr(&action
));
4096 ds_put_format(&action
, "eth.dst <-> eth.src; ip6.src <-> ip6.dst; "
4097 "udp.dst = udp.src; udp.src = 53; outport = inport; "
4098 "flags.loopback = 1; output;");
4099 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_DNS_RESPONSE
, 100,
4100 ds_cstr(&match
), ds_cstr(&action
));
4101 ds_destroy(&action
);
4104 /* Ingress table 12 and 13: DHCP options and response, by default goto
4105 * next. (priority 0).
4106 * Ingress table 14 and 15: DNS lookup and response, by default goto next.
4109 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
4114 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_DHCP_OPTIONS
, 0, "1", "next;");
4115 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_DHCP_RESPONSE
, 0, "1", "next;");
4116 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_DNS_LOOKUP
, 0, "1", "next;");
4117 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_DNS_RESPONSE
, 0, "1", "next;");
4120 /* Ingress table 16: Destination lookup, broadcast and multicast handling
4121 * (priority 100). */
4122 HMAP_FOR_EACH (op
, key_node
, ports
) {
4127 if (lsp_is_enabled(op
->nbsp
)) {
4128 ovn_multicast_add(mcgroups
, &mc_flood
, op
);
4131 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
4136 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_L2_LKUP
, 100, "eth.mcast",
4137 "outport = \""MC_FLOOD
"\"; output;");
4140 /* Ingress table 16: Destination lookup, unicast handling (priority 50), */
4141 HMAP_FOR_EACH (op
, key_node
, ports
) {
4146 for (size_t i
= 0; i
< op
->nbsp
->n_addresses
; i
++) {
4147 /* Addresses are owned by the logical port.
4148 * Ethernet address followed by zero or more IPv4
4149 * or IPv6 addresses (or both). */
4150 struct eth_addr mac
;
4151 if (ovs_scan(op
->nbsp
->addresses
[i
],
4152 ETH_ADDR_SCAN_FMT
, ETH_ADDR_SCAN_ARGS(mac
))) {
4154 ds_put_format(&match
, "eth.dst == "ETH_ADDR_FMT
,
4155 ETH_ADDR_ARGS(mac
));
4158 ds_put_format(&actions
, "outport = %s; output;", op
->json_key
);
4159 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_L2_LKUP
, 50,
4160 ds_cstr(&match
), ds_cstr(&actions
));
4161 } else if (!strcmp(op
->nbsp
->addresses
[i
], "unknown")) {
4162 if (lsp_is_enabled(op
->nbsp
)) {
4163 ovn_multicast_add(mcgroups
, &mc_unknown
, op
);
4164 op
->od
->has_unknown
= true;
4166 } else if (is_dynamic_lsp_address(op
->nbsp
->addresses
[i
])) {
4167 if (!op
->nbsp
->dynamic_addresses
4168 || !ovs_scan(op
->nbsp
->dynamic_addresses
,
4169 ETH_ADDR_SCAN_FMT
, ETH_ADDR_SCAN_ARGS(mac
))) {
4173 ds_put_format(&match
, "eth.dst == "ETH_ADDR_FMT
,
4174 ETH_ADDR_ARGS(mac
));
4177 ds_put_format(&actions
, "outport = %s; output;", op
->json_key
);
4178 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_L2_LKUP
, 50,
4179 ds_cstr(&match
), ds_cstr(&actions
));
4180 } else if (!strcmp(op
->nbsp
->addresses
[i
], "router")) {
4181 if (!op
->peer
|| !op
->peer
->nbrp
4182 || !ovs_scan(op
->peer
->nbrp
->mac
,
4183 ETH_ADDR_SCAN_FMT
, ETH_ADDR_SCAN_ARGS(mac
))) {
4187 ds_put_format(&match
, "eth.dst == "ETH_ADDR_FMT
,
4188 ETH_ADDR_ARGS(mac
));
4189 if (op
->peer
->od
->l3dgw_port
4190 && op
->peer
== op
->peer
->od
->l3dgw_port
4191 && op
->peer
->od
->l3redirect_port
) {
4192 /* The destination lookup flow for the router's
4193 * distributed gateway port MAC address should only be
4194 * programmed on the "redirect-chassis". */
4195 ds_put_format(&match
, " && is_chassis_resident(%s)",
4196 op
->peer
->od
->l3redirect_port
->json_key
);
4200 ds_put_format(&actions
, "outport = %s; output;", op
->json_key
);
4201 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_L2_LKUP
, 50,
4202 ds_cstr(&match
), ds_cstr(&actions
));
4204 /* Add ethernet addresses specified in NAT rules on
4205 * distributed logical routers. */
4206 if (op
->peer
->od
->l3dgw_port
4207 && op
->peer
== op
->peer
->od
->l3dgw_port
) {
4208 for (int j
= 0; j
< op
->peer
->od
->nbr
->n_nat
; j
++) {
4209 const struct nbrec_nat
*nat
4210 = op
->peer
->od
->nbr
->nat
[j
];
4211 if (!strcmp(nat
->type
, "dnat_and_snat")
4212 && nat
->logical_port
&& nat
->external_mac
4213 && eth_addr_from_string(nat
->external_mac
, &mac
)) {
4216 ds_put_format(&match
, "eth.dst == "ETH_ADDR_FMT
4217 " && is_chassis_resident(\"%s\")",
4222 ds_put_format(&actions
, "outport = %s; output;",
4224 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_IN_L2_LKUP
,
4225 50, ds_cstr(&match
),
4231 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
4234 "%s: invalid syntax '%s' in addresses column",
4235 op
->nbsp
->name
, op
->nbsp
->addresses
[i
]);
4240 /* Ingress table 16: Destination lookup for unknown MACs (priority 0). */
4241 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
4246 if (od
->has_unknown
) {
4247 ovn_lflow_add(lflows
, od
, S_SWITCH_IN_L2_LKUP
, 0, "1",
4248 "outport = \""MC_UNKNOWN
"\"; output;");
4252 /* Egress tables 8: Egress port security - IP (priority 0)
4253 * Egress table 9: Egress port security L2 - multicast/broadcast
4254 * (priority 100). */
4255 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
4260 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_PORT_SEC_IP
, 0, "1", "next;");
4261 ovn_lflow_add(lflows
, od
, S_SWITCH_OUT_PORT_SEC_L2
, 100, "eth.mcast",
4265 /* Egress table 8: Egress port security - IP (priorities 90 and 80)
4266 * if port security enabled.
4268 * Egress table 9: Egress port security - L2 (priorities 50 and 150).
4270 * Priority 50 rules implement port security for enabled logical port.
4272 * Priority 150 rules drop packets to disabled logical ports, so that they
4273 * don't even receive multicast or broadcast packets. */
4274 HMAP_FOR_EACH (op
, key_node
, ports
) {
4280 ds_put_format(&match
, "outport == %s", op
->json_key
);
4281 if (lsp_is_enabled(op
->nbsp
)) {
4282 build_port_security_l2("eth.dst", op
->ps_addrs
, op
->n_ps_addrs
,
4284 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_OUT_PORT_SEC_L2
, 50,
4285 ds_cstr(&match
), "output;");
4287 ovn_lflow_add(lflows
, op
->od
, S_SWITCH_OUT_PORT_SEC_L2
, 150,
4288 ds_cstr(&match
), "drop;");
4291 if (op
->nbsp
->n_port_security
) {
4292 build_port_security_ip(P_OUT
, op
, lflows
);
4297 ds_destroy(&actions
);
4301 lrport_is_enabled(const struct nbrec_logical_router_port
*lrport
)
4303 return !lrport
->enabled
|| *lrport
->enabled
;
4306 /* Returns a string of the IP address of the router port 'op' that
4307 * overlaps with 'ip_s". If one is not found, returns NULL.
4309 * The caller must not free the returned string. */
4311 find_lrp_member_ip(const struct ovn_port
*op
, const char *ip_s
)
4313 bool is_ipv4
= strchr(ip_s
, '.') ? true : false;
4318 if (!ip_parse(ip_s
, &ip
)) {
4319 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
4320 VLOG_WARN_RL(&rl
, "bad ip address %s", ip_s
);
4324 for (int i
= 0; i
< op
->lrp_networks
.n_ipv4_addrs
; i
++) {
4325 const struct ipv4_netaddr
*na
= &op
->lrp_networks
.ipv4_addrs
[i
];
4327 if (!((na
->network
^ ip
) & na
->mask
)) {
4328 /* There should be only 1 interface that matches the
4329 * supplied IP. Otherwise, it's a configuration error,
4330 * because subnets of a router's interfaces should NOT
4336 struct in6_addr ip6
;
4338 if (!ipv6_parse(ip_s
, &ip6
)) {
4339 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
4340 VLOG_WARN_RL(&rl
, "bad ipv6 address %s", ip_s
);
4344 for (int i
= 0; i
< op
->lrp_networks
.n_ipv6_addrs
; i
++) {
4345 const struct ipv6_netaddr
*na
= &op
->lrp_networks
.ipv6_addrs
[i
];
4346 struct in6_addr xor_addr
= ipv6_addr_bitxor(&na
->network
, &ip6
);
4347 struct in6_addr and_addr
= ipv6_addr_bitand(&xor_addr
, &na
->mask
);
4349 if (ipv6_is_zero(&and_addr
)) {
4350 /* There should be only 1 interface that matches the
4351 * supplied IP. Otherwise, it's a configuration error,
4352 * because subnets of a router's interfaces should NOT
4363 add_route(struct hmap
*lflows
, const struct ovn_port
*op
,
4364 const char *lrp_addr_s
, const char *network_s
, int plen
,
4365 const char *gateway
, const char *policy
)
4367 bool is_ipv4
= strchr(network_s
, '.') ? true : false;
4368 struct ds match
= DS_EMPTY_INITIALIZER
;
4372 if (policy
&& !strcmp(policy
, "src-ip")) {
4374 priority
= plen
* 2;
4377 priority
= (plen
* 2) + 1;
4380 /* IPv6 link-local addresses must be scoped to the local router port. */
4382 struct in6_addr network
;
4383 ovs_assert(ipv6_parse(network_s
, &network
));
4384 if (in6_is_lla(&network
)) {
4385 ds_put_format(&match
, "inport == %s && ", op
->json_key
);
4388 ds_put_format(&match
, "ip%s.%s == %s/%d", is_ipv4
? "4" : "6", dir
,
4391 struct ds actions
= DS_EMPTY_INITIALIZER
;
4392 ds_put_format(&actions
, "ip.ttl--; %sreg0 = ", is_ipv4
? "" : "xx");
4395 ds_put_cstr(&actions
, gateway
);
4397 ds_put_format(&actions
, "ip%s.dst", is_ipv4
? "4" : "6");
4399 ds_put_format(&actions
, "; "
4403 "flags.loopback = 1; "
4405 is_ipv4
? "" : "xx",
4407 op
->lrp_networks
.ea_s
,
4410 /* The priority here is calculated to implement longest-prefix-match
4412 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_IP_ROUTING
, priority
,
4413 ds_cstr(&match
), ds_cstr(&actions
));
4415 ds_destroy(&actions
);
4419 build_static_route_flow(struct hmap
*lflows
, struct ovn_datapath
*od
,
4421 const struct nbrec_logical_router_static_route
*route
)
4424 const char *lrp_addr_s
= NULL
;
4428 /* Verify that the next hop is an IP address with an all-ones mask. */
4429 char *error
= ip_parse_cidr(route
->nexthop
, &nexthop
, &plen
);
4432 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
4433 VLOG_WARN_RL(&rl
, "bad next hop mask %s", route
->nexthop
);
4440 struct in6_addr ip6
;
4441 error
= ipv6_parse_cidr(route
->nexthop
, &ip6
, &plen
);
4444 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
4445 VLOG_WARN_RL(&rl
, "bad next hop mask %s", route
->nexthop
);
4450 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
4451 VLOG_WARN_RL(&rl
, "bad next hop ip address %s", route
->nexthop
);
4460 /* Verify that ip prefix is a valid IPv4 address. */
4461 error
= ip_parse_cidr(route
->ip_prefix
, &prefix
, &plen
);
4463 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
4464 VLOG_WARN_RL(&rl
, "bad 'ip_prefix' in static routes %s",
4469 prefix_s
= xasprintf(IP_FMT
, IP_ARGS(prefix
& be32_prefix_mask(plen
)));
4471 /* Verify that ip prefix is a valid IPv6 address. */
4472 struct in6_addr prefix
;
4473 error
= ipv6_parse_cidr(route
->ip_prefix
, &prefix
, &plen
);
4475 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
4476 VLOG_WARN_RL(&rl
, "bad 'ip_prefix' in static routes %s",
4481 struct in6_addr mask
= ipv6_create_mask(plen
);
4482 struct in6_addr network
= ipv6_addr_bitand(&prefix
, &mask
);
4483 prefix_s
= xmalloc(INET6_ADDRSTRLEN
);
4484 inet_ntop(AF_INET6
, &network
, prefix_s
, INET6_ADDRSTRLEN
);
4487 /* Find the outgoing port. */
4488 struct ovn_port
*out_port
= NULL
;
4489 if (route
->output_port
) {
4490 out_port
= ovn_port_find(ports
, route
->output_port
);
4492 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
4493 VLOG_WARN_RL(&rl
, "Bad out port %s for static route %s",
4494 route
->output_port
, route
->ip_prefix
);
4497 lrp_addr_s
= find_lrp_member_ip(out_port
, route
->nexthop
);
4499 /* There are no IP networks configured on the router's port via
4500 * which 'route->nexthop' is theoretically reachable. But since
4501 * 'out_port' has been specified, we honor it by trying to reach
4502 * 'route->nexthop' via the first IP address of 'out_port'.
4503 * (There are cases, e.g in GCE, where each VM gets a /32 IP
4504 * address and the default gateway is still reachable from it.) */
4506 if (out_port
->lrp_networks
.n_ipv4_addrs
) {
4507 lrp_addr_s
= out_port
->lrp_networks
.ipv4_addrs
[0].addr_s
;
4510 if (out_port
->lrp_networks
.n_ipv6_addrs
) {
4511 lrp_addr_s
= out_port
->lrp_networks
.ipv6_addrs
[0].addr_s
;
4516 /* output_port is not specified, find the
4517 * router port matching the next hop. */
4519 for (i
= 0; i
< od
->nbr
->n_ports
; i
++) {
4520 struct nbrec_logical_router_port
*lrp
= od
->nbr
->ports
[i
];
4521 out_port
= ovn_port_find(ports
, lrp
->name
);
4523 /* This should not happen. */
4527 lrp_addr_s
= find_lrp_member_ip(out_port
, route
->nexthop
);
4534 if (!out_port
|| !lrp_addr_s
) {
4535 /* There is no matched out port. */
4536 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
4537 VLOG_WARN_RL(&rl
, "No path for static route %s; next hop %s",
4538 route
->ip_prefix
, route
->nexthop
);
4542 char *policy
= route
->policy
? route
->policy
: "dst-ip";
4543 add_route(lflows
, out_port
, lrp_addr_s
, prefix_s
, plen
, route
->nexthop
,
4551 op_put_v4_networks(struct ds
*ds
, const struct ovn_port
*op
, bool add_bcast
)
4553 if (!add_bcast
&& op
->lrp_networks
.n_ipv4_addrs
== 1) {
4554 ds_put_format(ds
, "%s", op
->lrp_networks
.ipv4_addrs
[0].addr_s
);
4558 ds_put_cstr(ds
, "{");
4559 for (int i
= 0; i
< op
->lrp_networks
.n_ipv4_addrs
; i
++) {
4560 ds_put_format(ds
, "%s, ", op
->lrp_networks
.ipv4_addrs
[i
].addr_s
);
4562 ds_put_format(ds
, "%s, ", op
->lrp_networks
.ipv4_addrs
[i
].bcast_s
);
4567 ds_put_cstr(ds
, "}");
4571 op_put_v6_networks(struct ds
*ds
, const struct ovn_port
*op
)
4573 if (op
->lrp_networks
.n_ipv6_addrs
== 1) {
4574 ds_put_format(ds
, "%s", op
->lrp_networks
.ipv6_addrs
[0].addr_s
);
4578 ds_put_cstr(ds
, "{");
4579 for (int i
= 0; i
< op
->lrp_networks
.n_ipv6_addrs
; i
++) {
4580 ds_put_format(ds
, "%s, ", op
->lrp_networks
.ipv6_addrs
[i
].addr_s
);
4584 ds_put_cstr(ds
, "}");
4588 get_force_snat_ip(struct ovn_datapath
*od
, const char *key_type
, ovs_be32
*ip
)
4590 char *key
= xasprintf("%s_force_snat_ip", key_type
);
4591 const char *ip_address
= smap_get(&od
->nbr
->options
, key
);
4596 char *error
= ip_parse_masked(ip_address
, ip
, &mask
);
4597 if (error
|| mask
!= OVS_BE32_MAX
) {
4598 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
4599 VLOG_WARN_RL(&rl
, "bad ip %s in options of router "UUID_FMT
"",
4600 ip_address
, UUID_ARGS(&od
->key
));
4613 add_router_lb_flow(struct hmap
*lflows
, struct ovn_datapath
*od
,
4614 struct ds
*match
, struct ds
*actions
, int priority
,
4615 const char *lb_force_snat_ip
, char *backend_ips
,
4616 bool is_udp
, int addr_family
)
4618 /* A match and actions for new connections. */
4619 char *new_match
= xasprintf("ct.new && %s", ds_cstr(match
));
4620 if (lb_force_snat_ip
) {
4621 char *new_actions
= xasprintf("flags.force_snat_for_lb = 1; %s",
4623 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_DNAT
, priority
, new_match
,
4627 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_DNAT
, priority
, new_match
,
4631 /* A match and actions for established connections. */
4632 char *est_match
= xasprintf("ct.est && %s", ds_cstr(match
));
4633 if (lb_force_snat_ip
) {
4634 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_DNAT
, priority
, est_match
,
4635 "flags.force_snat_for_lb = 1; ct_dnat;");
4637 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_DNAT
, priority
, est_match
,
4644 if (!od
->l3dgw_port
|| !od
->l3redirect_port
|| !backend_ips
4645 || addr_family
!= AF_INET
) {
4649 /* Add logical flows to UNDNAT the load balanced reverse traffic in
4650 * the router egress pipleine stage - S_ROUTER_OUT_UNDNAT if the logical
4651 * router has a gateway router port associated.
4653 struct ds undnat_match
= DS_EMPTY_INITIALIZER
;
4654 ds_put_cstr(&undnat_match
, "ip4 && (");
4655 char *start
, *next
, *ip_str
;
4656 start
= next
= xstrdup(backend_ips
);
4657 ip_str
= strsep(&next
, ",");
4658 bool backend_ips_found
= false;
4659 while (ip_str
&& ip_str
[0]) {
4660 char *ip_address
= NULL
;
4663 ip_address_and_port_from_lb_key(ip_str
, &ip_address
, &port
,
4669 ds_put_format(&undnat_match
, "(ip4.src == %s", ip_address
);
4672 ds_put_format(&undnat_match
, " && %s.src == %d) || ",
4673 is_udp
? "udp" : "tcp", port
);
4675 ds_put_cstr(&undnat_match
, ") || ");
4677 ip_str
= strsep(&next
, ",");
4678 backend_ips_found
= true;
4682 if (!backend_ips_found
) {
4683 ds_destroy(&undnat_match
);
4686 ds_chomp(&undnat_match
, ' ');
4687 ds_chomp(&undnat_match
, '|');
4688 ds_chomp(&undnat_match
, '|');
4689 ds_chomp(&undnat_match
, ' ');
4690 ds_put_format(&undnat_match
, ") && outport == %s && "
4691 "is_chassis_resident(%s)", od
->l3dgw_port
->json_key
,
4692 od
->l3redirect_port
->json_key
);
4693 if (lb_force_snat_ip
) {
4694 ovn_lflow_add(lflows
, od
, S_ROUTER_OUT_UNDNAT
, 120,
4695 ds_cstr(&undnat_match
),
4696 "flags.force_snat_for_lb = 1; ct_dnat;");
4698 ovn_lflow_add(lflows
, od
, S_ROUTER_OUT_UNDNAT
, 120,
4699 ds_cstr(&undnat_match
), "ct_dnat;");
4702 ds_destroy(&undnat_match
);
4705 #define ND_RA_MAX_INTERVAL_MAX 1800
4706 #define ND_RA_MAX_INTERVAL_MIN 4
4708 #define ND_RA_MIN_INTERVAL_MAX(max) ((max) * 3 / 4)
4709 #define ND_RA_MIN_INTERVAL_MIN 3
4712 copy_ra_to_sb(struct ovn_port
*op
, const char *address_mode
)
4714 struct smap options
;
4715 smap_clone(&options
, &op
->sb
->options
);
4717 smap_add(&options
, "ipv6_ra_send_periodic", "true");
4718 smap_add(&options
, "ipv6_ra_address_mode", address_mode
);
4720 int max_interval
= smap_get_int(&op
->nbrp
->ipv6_ra_configs
,
4721 "max_interval", ND_RA_MAX_INTERVAL_DEFAULT
);
4722 if (max_interval
> ND_RA_MAX_INTERVAL_MAX
) {
4723 max_interval
= ND_RA_MAX_INTERVAL_MAX
;
4725 if (max_interval
< ND_RA_MAX_INTERVAL_MIN
) {
4726 max_interval
= ND_RA_MAX_INTERVAL_MIN
;
4728 smap_add_format(&options
, "ipv6_ra_max_interval", "%d", max_interval
);
4730 int min_interval
= smap_get_int(&op
->nbrp
->ipv6_ra_configs
,
4731 "min_interval", nd_ra_min_interval_default(max_interval
));
4732 if (min_interval
> ND_RA_MIN_INTERVAL_MAX(max_interval
)) {
4733 min_interval
= ND_RA_MIN_INTERVAL_MAX(max_interval
);
4735 if (min_interval
< ND_RA_MIN_INTERVAL_MIN
) {
4736 min_interval
= ND_RA_MIN_INTERVAL_MIN
;
4738 smap_add_format(&options
, "ipv6_ra_min_interval", "%d", min_interval
);
4740 int mtu
= smap_get_int(&op
->nbrp
->ipv6_ra_configs
, "mtu", ND_MTU_DEFAULT
);
4741 /* RFC 2460 requires the MTU for IPv6 to be at least 1280 */
4742 if (mtu
&& mtu
>= 1280) {
4743 smap_add_format(&options
, "ipv6_ra_mtu", "%d", mtu
);
4746 struct ds s
= DS_EMPTY_INITIALIZER
;
4747 for (int i
= 0; i
< op
->lrp_networks
.n_ipv6_addrs
; ++i
) {
4748 struct ipv6_netaddr
*addrs
= &op
->lrp_networks
.ipv6_addrs
[i
];
4749 if (in6_is_lla(&addrs
->network
)) {
4750 smap_add(&options
, "ipv6_ra_src_addr", addrs
->addr_s
);
4753 ds_put_format(&s
, "%s/%u ", addrs
->network_s
, addrs
->plen
);
4755 /* Remove trailing space */
4757 smap_add(&options
, "ipv6_ra_prefixes", ds_cstr(&s
));
4760 smap_add(&options
, "ipv6_ra_src_eth", op
->lrp_networks
.ea_s
);
4762 sbrec_port_binding_set_options(op
->sb
, &options
);
4763 smap_destroy(&options
);
4767 build_lrouter_flows(struct hmap
*datapaths
, struct hmap
*ports
,
4768 struct hmap
*lflows
)
4770 /* This flow table structure is documented in ovn-northd(8), so please
4771 * update ovn-northd.8.xml if you change anything. */
4773 struct ds match
= DS_EMPTY_INITIALIZER
;
4774 struct ds actions
= DS_EMPTY_INITIALIZER
;
4776 /* Logical router ingress table 0: Admission control framework. */
4777 struct ovn_datapath
*od
;
4778 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
4783 /* Logical VLANs not supported.
4784 * Broadcast/multicast source address is invalid. */
4785 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_ADMISSION
, 100,
4786 "vlan.present || eth.src[40]", "drop;");
4789 /* Logical router ingress table 0: match (priority 50). */
4790 struct ovn_port
*op
;
4791 HMAP_FOR_EACH (op
, key_node
, ports
) {
4796 if (!lrport_is_enabled(op
->nbrp
)) {
4797 /* Drop packets from disabled logical ports (since logical flow
4798 * tables are default-drop). */
4803 /* No ingress packets should be received on a chassisredirect
4809 ds_put_format(&match
, "eth.mcast && inport == %s", op
->json_key
);
4810 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_ADMISSION
, 50,
4811 ds_cstr(&match
), "next;");
4814 ds_put_format(&match
, "eth.dst == %s && inport == %s",
4815 op
->lrp_networks
.ea_s
, op
->json_key
);
4816 if (op
->od
->l3dgw_port
&& op
== op
->od
->l3dgw_port
4817 && op
->od
->l3redirect_port
) {
4818 /* Traffic with eth.dst = l3dgw_port->lrp_networks.ea_s
4819 * should only be received on the "redirect-chassis". */
4820 ds_put_format(&match
, " && is_chassis_resident(%s)",
4821 op
->od
->l3redirect_port
->json_key
);
4823 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_ADMISSION
, 50,
4824 ds_cstr(&match
), "next;");
4827 /* Logical router ingress table 1: IP Input. */
4828 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
4833 /* L3 admission control: drop multicast and broadcast source, localhost
4834 * source or destination, and zero network source or destination
4835 * (priority 100). */
4836 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_IP_INPUT
, 100,
4838 "ip4.src == 255.255.255.255 || "
4839 "ip4.src == 127.0.0.0/8 || "
4840 "ip4.dst == 127.0.0.0/8 || "
4841 "ip4.src == 0.0.0.0/8 || "
4842 "ip4.dst == 0.0.0.0/8",
4845 /* ARP reply handling. Use ARP replies to populate the logical
4846 * router's ARP table. */
4847 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_IP_INPUT
, 90, "arp.op == 2",
4848 "put_arp(inport, arp.spa, arp.sha);");
4850 /* Drop Ethernet local broadcast. By definition this traffic should
4851 * not be forwarded.*/
4852 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_IP_INPUT
, 50,
4853 "eth.bcast", "drop;");
4857 * XXX Need to send ICMP time exceeded if !ip.later_frag. */
4859 ds_put_cstr(&match
, "ip4 && ip.ttl == {0, 1}");
4860 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_IP_INPUT
, 30,
4861 ds_cstr(&match
), "drop;");
4863 /* ND advertisement handling. Use advertisements to populate
4864 * the logical router's ARP/ND table. */
4865 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_IP_INPUT
, 90, "nd_na",
4866 "put_nd(inport, nd.target, nd.tll);");
4868 /* Lean from neighbor solicitations that were not directed at
4869 * us. (A priority-90 flow will respond to requests to us and
4870 * learn the sender's mac address. */
4871 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_IP_INPUT
, 80, "nd_ns",
4872 "put_nd(inport, ip6.src, nd.sll);");
4874 /* Pass other traffic not already handled to the next table for
4876 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_IP_INPUT
, 0, "1", "next;");
4879 /* Logical router ingress table 1: IP Input for IPv4. */
4880 HMAP_FOR_EACH (op
, key_node
, ports
) {
4886 /* No ingress packets are accepted on a chassisredirect
4887 * port, so no need to program flows for that port. */
4891 if (op
->lrp_networks
.n_ipv4_addrs
) {
4892 /* L3 admission control: drop packets that originate from an
4893 * IPv4 address owned by the router or a broadcast address
4894 * known to the router (priority 100). */
4896 ds_put_cstr(&match
, "ip4.src == ");
4897 op_put_v4_networks(&match
, op
, true);
4898 ds_put_cstr(&match
, " && "REGBIT_EGRESS_LOOPBACK
" == 0");
4899 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_IP_INPUT
, 100,
4900 ds_cstr(&match
), "drop;");
4902 /* ICMP echo reply. These flows reply to ICMP echo requests
4903 * received for the router's IP address. Since packets only
4904 * get here as part of the logical router datapath, the inport
4905 * (i.e. the incoming locally attached net) does not matter.
4906 * The ip.ttl also does not matter (RFC1812 section 4.2.2.9) */
4908 ds_put_cstr(&match
, "ip4.dst == ");
4909 op_put_v4_networks(&match
, op
, false);
4910 ds_put_cstr(&match
, " && icmp4.type == 8 && icmp4.code == 0");
4913 ds_put_format(&actions
,
4914 "ip4.dst <-> ip4.src; "
4917 "flags.loopback = 1; "
4919 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_IP_INPUT
, 90,
4920 ds_cstr(&match
), ds_cstr(&actions
));
4923 /* ARP reply. These flows reply to ARP requests for the router's own
4925 for (int i
= 0; i
< op
->lrp_networks
.n_ipv4_addrs
; i
++) {
4927 ds_put_format(&match
,
4928 "inport == %s && arp.tpa == %s && arp.op == 1",
4929 op
->json_key
, op
->lrp_networks
.ipv4_addrs
[i
].addr_s
);
4930 if (op
->od
->l3dgw_port
&& op
== op
->od
->l3dgw_port
4931 && op
->od
->l3redirect_port
) {
4932 /* Traffic with eth.src = l3dgw_port->lrp_networks.ea_s
4933 * should only be sent from the "redirect-chassis", so that
4934 * upstream MAC learning points to the "redirect-chassis".
4935 * Also need to avoid generation of multiple ARP responses
4936 * from different chassis. */
4937 ds_put_format(&match
, " && is_chassis_resident(%s)",
4938 op
->od
->l3redirect_port
->json_key
);
4942 ds_put_format(&actions
,
4943 "eth.dst = eth.src; "
4945 "arp.op = 2; /* ARP reply */ "
4946 "arp.tha = arp.sha; "
4948 "arp.tpa = arp.spa; "
4951 "flags.loopback = 1; "
4953 op
->lrp_networks
.ea_s
,
4954 op
->lrp_networks
.ea_s
,
4955 op
->lrp_networks
.ipv4_addrs
[i
].addr_s
,
4957 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_IP_INPUT
, 90,
4958 ds_cstr(&match
), ds_cstr(&actions
));
4961 /* A set to hold all load-balancer vips that need ARP responses. */
4962 struct sset all_ips
= SSET_INITIALIZER(&all_ips
);
4964 get_router_load_balancer_ips(op
->od
, &all_ips
, &addr_family
);
4966 const char *ip_address
;
4967 SSET_FOR_EACH(ip_address
, &all_ips
) {
4969 if (addr_family
== AF_INET
) {
4970 ds_put_format(&match
,
4971 "inport == %s && arp.tpa == %s && arp.op == 1",
4972 op
->json_key
, ip_address
);
4974 ds_put_format(&match
,
4975 "inport == %s && nd_ns && nd.target == %s",
4976 op
->json_key
, ip_address
);
4980 if (addr_family
== AF_INET
) {
4981 ds_put_format(&actions
,
4982 "eth.dst = eth.src; "
4984 "arp.op = 2; /* ARP reply */ "
4985 "arp.tha = arp.sha; "
4987 "arp.tpa = arp.spa; "
4990 "flags.loopback = 1; "
4992 op
->lrp_networks
.ea_s
,
4993 op
->lrp_networks
.ea_s
,
4997 ds_put_format(&actions
,
5003 "outport = inport; "
5004 "flags.loopback = 1; "
5007 op
->lrp_networks
.ea_s
,
5010 op
->lrp_networks
.ea_s
);
5012 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_IP_INPUT
, 90,
5013 ds_cstr(&match
), ds_cstr(&actions
));
5016 sset_destroy(&all_ips
);
5018 /* A gateway router can have 2 SNAT IP addresses to force DNATed and
5019 * LBed traffic respectively to be SNATed. In addition, there can be
5020 * a number of SNAT rules in the NAT table. */
5021 ovs_be32
*snat_ips
= xmalloc(sizeof *snat_ips
*
5022 (op
->od
->nbr
->n_nat
+ 2));
5023 size_t n_snat_ips
= 0;
5026 const char *dnat_force_snat_ip
= get_force_snat_ip(op
->od
, "dnat",
5028 if (dnat_force_snat_ip
) {
5029 snat_ips
[n_snat_ips
++] = snat_ip
;
5032 const char *lb_force_snat_ip
= get_force_snat_ip(op
->od
, "lb",
5034 if (lb_force_snat_ip
) {
5035 snat_ips
[n_snat_ips
++] = snat_ip
;
5038 for (int i
= 0; i
< op
->od
->nbr
->n_nat
; i
++) {
5039 const struct nbrec_nat
*nat
;
5041 nat
= op
->od
->nbr
->nat
[i
];
5044 if (!ip_parse(nat
->external_ip
, &ip
) || !ip
) {
5045 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
5046 VLOG_WARN_RL(&rl
, "bad ip address %s in nat configuration "
5047 "for router %s", nat
->external_ip
, op
->key
);
5051 if (!strcmp(nat
->type
, "snat")) {
5052 snat_ips
[n_snat_ips
++] = ip
;
5056 /* ARP handling for external IP addresses.
5058 * DNAT IP addresses are external IP addresses that need ARP
5061 ds_put_format(&match
,
5062 "inport == %s && arp.tpa == "IP_FMT
" && arp.op == 1",
5063 op
->json_key
, IP_ARGS(ip
));
5066 ds_put_format(&actions
,
5067 "eth.dst = eth.src; "
5068 "arp.op = 2; /* ARP reply */ "
5069 "arp.tha = arp.sha; ");
5071 if (op
->od
->l3dgw_port
&& op
== op
->od
->l3dgw_port
) {
5072 struct eth_addr mac
;
5073 if (nat
->external_mac
&&
5074 eth_addr_from_string(nat
->external_mac
, &mac
)
5075 && nat
->logical_port
) {
5076 /* distributed NAT case, use nat->external_mac */
5077 ds_put_format(&actions
,
5078 "eth.src = "ETH_ADDR_FMT
"; "
5079 "arp.sha = "ETH_ADDR_FMT
"; ",
5081 ETH_ADDR_ARGS(mac
));
5082 /* Traffic with eth.src = nat->external_mac should only be
5083 * sent from the chassis where nat->logical_port is
5084 * resident, so that upstream MAC learning points to the
5085 * correct chassis. Also need to avoid generation of
5086 * multiple ARP responses from different chassis. */
5087 ds_put_format(&match
, " && is_chassis_resident(\"%s\")",
5090 ds_put_format(&actions
,
5093 op
->lrp_networks
.ea_s
,
5094 op
->lrp_networks
.ea_s
);
5095 /* Traffic with eth.src = l3dgw_port->lrp_networks.ea_s
5096 * should only be sent from the "redirect-chassis", so that
5097 * upstream MAC learning points to the "redirect-chassis".
5098 * Also need to avoid generation of multiple ARP responses
5099 * from different chassis. */
5100 if (op
->od
->l3redirect_port
) {
5101 ds_put_format(&match
, " && is_chassis_resident(%s)",
5102 op
->od
->l3redirect_port
->json_key
);
5106 ds_put_format(&actions
,
5109 op
->lrp_networks
.ea_s
,
5110 op
->lrp_networks
.ea_s
);
5112 ds_put_format(&actions
,
5113 "arp.tpa = arp.spa; "
5114 "arp.spa = "IP_FMT
"; "
5116 "flags.loopback = 1; "
5120 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_IP_INPUT
, 90,
5121 ds_cstr(&match
), ds_cstr(&actions
));
5125 ds_put_cstr(&match
, "ip4.dst == {");
5126 bool has_drop_ips
= false;
5127 for (int i
= 0; i
< op
->lrp_networks
.n_ipv4_addrs
; i
++) {
5128 bool snat_ip_is_router_ip
= false;
5129 for (int j
= 0; j
< n_snat_ips
; j
++) {
5130 /* Packets to SNAT IPs should not be dropped. */
5131 if (op
->lrp_networks
.ipv4_addrs
[i
].addr
== snat_ips
[j
]) {
5132 snat_ip_is_router_ip
= true;
5136 if (snat_ip_is_router_ip
) {
5139 ds_put_format(&match
, "%s, ",
5140 op
->lrp_networks
.ipv4_addrs
[i
].addr_s
);
5141 has_drop_ips
= true;
5143 ds_chomp(&match
, ' ');
5144 ds_chomp(&match
, ',');
5145 ds_put_cstr(&match
, "}");
5148 /* Drop IP traffic to this router. */
5149 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_IP_INPUT
, 60,
5150 ds_cstr(&match
), "drop;");
5156 /* Logical router ingress table 1: IP Input for IPv6. */
5157 HMAP_FOR_EACH (op
, key_node
, ports
) {
5163 /* No ingress packets are accepted on a chassisredirect
5164 * port, so no need to program flows for that port. */
5168 if (op
->lrp_networks
.n_ipv6_addrs
) {
5169 /* L3 admission control: drop packets that originate from an
5170 * IPv6 address owned by the router (priority 100). */
5172 ds_put_cstr(&match
, "ip6.src == ");
5173 op_put_v6_networks(&match
, op
);
5174 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_IP_INPUT
, 100,
5175 ds_cstr(&match
), "drop;");
5177 /* ICMPv6 echo reply. These flows reply to echo requests
5178 * received for the router's IP address. */
5180 ds_put_cstr(&match
, "ip6.dst == ");
5181 op_put_v6_networks(&match
, op
);
5182 ds_put_cstr(&match
, " && icmp6.type == 128 && icmp6.code == 0");
5185 ds_put_cstr(&actions
,
5186 "ip6.dst <-> ip6.src; "
5188 "icmp6.type = 129; "
5189 "flags.loopback = 1; "
5191 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_IP_INPUT
, 90,
5192 ds_cstr(&match
), ds_cstr(&actions
));
5194 /* Drop IPv6 traffic to this router. */
5196 ds_put_cstr(&match
, "ip6.dst == ");
5197 op_put_v6_networks(&match
, op
);
5198 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_IP_INPUT
, 60,
5199 ds_cstr(&match
), "drop;");
5202 /* ND reply. These flows reply to ND solicitations for the
5203 * router's own IP address. */
5204 for (int i
= 0; i
< op
->lrp_networks
.n_ipv6_addrs
; i
++) {
5206 ds_put_format(&match
,
5207 "inport == %s && nd_ns && ip6.dst == {%s, %s} "
5208 "&& nd.target == %s",
5210 op
->lrp_networks
.ipv6_addrs
[i
].addr_s
,
5211 op
->lrp_networks
.ipv6_addrs
[i
].sn_addr_s
,
5212 op
->lrp_networks
.ipv6_addrs
[i
].addr_s
);
5213 if (op
->od
->l3dgw_port
&& op
== op
->od
->l3dgw_port
5214 && op
->od
->l3redirect_port
) {
5215 /* Traffic with eth.src = l3dgw_port->lrp_networks.ea_s
5216 * should only be sent from the "redirect-chassis", so that
5217 * upstream MAC learning points to the "redirect-chassis".
5218 * Also need to avoid generation of multiple ND replies
5219 * from different chassis. */
5220 ds_put_format(&match
, " && is_chassis_resident(%s)",
5221 op
->od
->l3redirect_port
->json_key
);
5225 ds_put_format(&actions
,
5226 "put_nd(inport, ip6.src, nd.sll); "
5232 "outport = inport; "
5233 "flags.loopback = 1; "
5236 op
->lrp_networks
.ea_s
,
5237 op
->lrp_networks
.ipv6_addrs
[i
].addr_s
,
5238 op
->lrp_networks
.ipv6_addrs
[i
].addr_s
,
5239 op
->lrp_networks
.ea_s
);
5240 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_IP_INPUT
, 90,
5241 ds_cstr(&match
), ds_cstr(&actions
));
5245 /* NAT, Defrag and load balancing. */
5246 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
5251 /* Packets are allowed by default. */
5252 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_DEFRAG
, 0, "1", "next;");
5253 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_UNSNAT
, 0, "1", "next;");
5254 ovn_lflow_add(lflows
, od
, S_ROUTER_OUT_SNAT
, 0, "1", "next;");
5255 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_DNAT
, 0, "1", "next;");
5256 ovn_lflow_add(lflows
, od
, S_ROUTER_OUT_UNDNAT
, 0, "1", "next;");
5257 ovn_lflow_add(lflows
, od
, S_ROUTER_OUT_EGR_LOOP
, 0, "1", "next;");
5259 /* NAT rules are only valid on Gateway routers and routers with
5260 * l3dgw_port (router has a port with "redirect-chassis"
5262 if (!smap_get(&od
->nbr
->options
, "chassis") && !od
->l3dgw_port
) {
5267 const char *dnat_force_snat_ip
= get_force_snat_ip(od
, "dnat",
5269 const char *lb_force_snat_ip
= get_force_snat_ip(od
, "lb",
5272 for (int i
= 0; i
< od
->nbr
->n_nat
; i
++) {
5273 const struct nbrec_nat
*nat
;
5275 nat
= od
->nbr
->nat
[i
];
5279 char *error
= ip_parse_masked(nat
->external_ip
, &ip
, &mask
);
5280 if (error
|| mask
!= OVS_BE32_MAX
) {
5281 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(5, 1);
5282 VLOG_WARN_RL(&rl
, "bad external ip %s for nat",
5288 /* Check the validity of nat->logical_ip. 'logical_ip' can
5289 * be a subnet when the type is "snat". */
5290 error
= ip_parse_masked(nat
->logical_ip
, &ip
, &mask
);
5291 if (!strcmp(nat
->type
, "snat")) {
5293 static struct vlog_rate_limit rl
=
5294 VLOG_RATE_LIMIT_INIT(5, 1);
5295 VLOG_WARN_RL(&rl
, "bad ip network or ip %s for snat "
5296 "in router "UUID_FMT
"",
5297 nat
->logical_ip
, UUID_ARGS(&od
->key
));
5302 if (error
|| mask
!= OVS_BE32_MAX
) {
5303 static struct vlog_rate_limit rl
=
5304 VLOG_RATE_LIMIT_INIT(5, 1);
5305 VLOG_WARN_RL(&rl
, "bad ip %s for dnat in router "
5306 ""UUID_FMT
"", nat
->logical_ip
, UUID_ARGS(&od
->key
));
5312 /* For distributed router NAT, determine whether this NAT rule
5313 * satisfies the conditions for distributed NAT processing. */
5314 bool distributed
= false;
5315 struct eth_addr mac
;
5316 if (od
->l3dgw_port
&& !strcmp(nat
->type
, "dnat_and_snat") &&
5317 nat
->logical_port
&& nat
->external_mac
) {
5318 if (eth_addr_from_string(nat
->external_mac
, &mac
)) {
5321 static struct vlog_rate_limit rl
=
5322 VLOG_RATE_LIMIT_INIT(5, 1);
5323 VLOG_WARN_RL(&rl
, "bad mac %s for dnat in router "
5324 ""UUID_FMT
"", nat
->external_mac
, UUID_ARGS(&od
->key
));
5329 /* Ingress UNSNAT table: It is for already established connections'
5330 * reverse traffic. i.e., SNAT has already been done in egress
5331 * pipeline and now the packet has entered the ingress pipeline as
5332 * part of a reply. We undo the SNAT here.
5334 * Undoing SNAT has to happen before DNAT processing. This is
5335 * because when the packet was DNATed in ingress pipeline, it did
5336 * not know about the possibility of eventual additional SNAT in
5337 * egress pipeline. */
5338 if (!strcmp(nat
->type
, "snat")
5339 || !strcmp(nat
->type
, "dnat_and_snat")) {
5340 if (!od
->l3dgw_port
) {
5341 /* Gateway router. */
5343 ds_put_format(&match
, "ip && ip4.dst == %s",
5345 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_UNSNAT
, 90,
5346 ds_cstr(&match
), "ct_snat;");
5348 /* Distributed router. */
5350 /* Traffic received on l3dgw_port is subject to NAT. */
5352 ds_put_format(&match
, "ip && ip4.dst == %s"
5355 od
->l3dgw_port
->json_key
);
5356 if (!distributed
&& od
->l3redirect_port
) {
5357 /* Flows for NAT rules that are centralized are only
5358 * programmed on the "redirect-chassis". */
5359 ds_put_format(&match
, " && is_chassis_resident(%s)",
5360 od
->l3redirect_port
->json_key
);
5362 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_UNSNAT
, 100,
5363 ds_cstr(&match
), "ct_snat;");
5365 /* Traffic received on other router ports must be
5366 * redirected to the central instance of the l3dgw_port
5367 * for NAT processing. */
5369 ds_put_format(&match
, "ip && ip4.dst == %s",
5371 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_UNSNAT
, 50,
5373 REGBIT_NAT_REDIRECT
" = 1; next;");
5377 /* Ingress DNAT table: Packets enter the pipeline with destination
5378 * IP address that needs to be DNATted from a external IP address
5379 * to a logical IP address. */
5380 if (!strcmp(nat
->type
, "dnat")
5381 || !strcmp(nat
->type
, "dnat_and_snat")) {
5382 if (!od
->l3dgw_port
) {
5383 /* Gateway router. */
5384 /* Packet when it goes from the initiator to destination.
5385 * We need to set flags.loopback because the router can
5386 * send the packet back through the same interface. */
5388 ds_put_format(&match
, "ip && ip4.dst == %s",
5391 if (dnat_force_snat_ip
) {
5392 /* Indicate to the future tables that a DNAT has taken
5393 * place and a force SNAT needs to be done in the
5394 * Egress SNAT table. */
5395 ds_put_format(&actions
,
5396 "flags.force_snat_for_dnat = 1; ");
5398 ds_put_format(&actions
, "flags.loopback = 1; ct_dnat(%s);",
5400 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_DNAT
, 100,
5401 ds_cstr(&match
), ds_cstr(&actions
));
5403 /* Distributed router. */
5405 /* Traffic received on l3dgw_port is subject to NAT. */
5407 ds_put_format(&match
, "ip && ip4.dst == %s"
5410 od
->l3dgw_port
->json_key
);
5411 if (!distributed
&& od
->l3redirect_port
) {
5412 /* Flows for NAT rules that are centralized are only
5413 * programmed on the "redirect-chassis". */
5414 ds_put_format(&match
, " && is_chassis_resident(%s)",
5415 od
->l3redirect_port
->json_key
);
5418 ds_put_format(&actions
, "ct_dnat(%s);",
5420 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_DNAT
, 100,
5421 ds_cstr(&match
), ds_cstr(&actions
));
5423 /* Traffic received on other router ports must be
5424 * redirected to the central instance of the l3dgw_port
5425 * for NAT processing. */
5427 ds_put_format(&match
, "ip && ip4.dst == %s",
5429 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_DNAT
, 50,
5431 REGBIT_NAT_REDIRECT
" = 1; next;");
5435 /* Egress UNDNAT table: It is for already established connections'
5436 * reverse traffic. i.e., DNAT has already been done in ingress
5437 * pipeline and now the packet has entered the egress pipeline as
5438 * part of a reply. We undo the DNAT here.
5440 * Note that this only applies for NAT on a distributed router.
5441 * Undo DNAT on a gateway router is done in the ingress DNAT
5442 * pipeline stage. */
5443 if (od
->l3dgw_port
&& (!strcmp(nat
->type
, "dnat")
5444 || !strcmp(nat
->type
, "dnat_and_snat"))) {
5446 ds_put_format(&match
, "ip && ip4.src == %s"
5447 " && outport == %s",
5449 od
->l3dgw_port
->json_key
);
5450 if (!distributed
&& od
->l3redirect_port
) {
5451 /* Flows for NAT rules that are centralized are only
5452 * programmed on the "redirect-chassis". */
5453 ds_put_format(&match
, " && is_chassis_resident(%s)",
5454 od
->l3redirect_port
->json_key
);
5458 ds_put_format(&actions
, "eth.src = "ETH_ADDR_FMT
"; ",
5459 ETH_ADDR_ARGS(mac
));
5461 ds_put_format(&actions
, "ct_dnat;");
5462 ovn_lflow_add(lflows
, od
, S_ROUTER_OUT_UNDNAT
, 100,
5463 ds_cstr(&match
), ds_cstr(&actions
));
5466 /* Egress SNAT table: Packets enter the egress pipeline with
5467 * source ip address that needs to be SNATted to a external ip
5469 if (!strcmp(nat
->type
, "snat")
5470 || !strcmp(nat
->type
, "dnat_and_snat")) {
5471 if (!od
->l3dgw_port
) {
5472 /* Gateway router. */
5474 ds_put_format(&match
, "ip && ip4.src == %s",
5477 ds_put_format(&actions
, "ct_snat(%s);", nat
->external_ip
);
5479 /* The priority here is calculated such that the
5480 * nat->logical_ip with the longest mask gets a higher
5482 ovn_lflow_add(lflows
, od
, S_ROUTER_OUT_SNAT
,
5483 count_1bits(ntohl(mask
)) + 1,
5484 ds_cstr(&match
), ds_cstr(&actions
));
5486 /* Distributed router. */
5488 ds_put_format(&match
, "ip && ip4.src == %s"
5489 " && outport == %s",
5491 od
->l3dgw_port
->json_key
);
5492 if (!distributed
&& od
->l3redirect_port
) {
5493 /* Flows for NAT rules that are centralized are only
5494 * programmed on the "redirect-chassis". */
5495 ds_put_format(&match
, " && is_chassis_resident(%s)",
5496 od
->l3redirect_port
->json_key
);
5500 ds_put_format(&actions
, "eth.src = "ETH_ADDR_FMT
"; ",
5501 ETH_ADDR_ARGS(mac
));
5503 ds_put_format(&actions
, "ct_snat(%s);", nat
->external_ip
);
5505 /* The priority here is calculated such that the
5506 * nat->logical_ip with the longest mask gets a higher
5508 ovn_lflow_add(lflows
, od
, S_ROUTER_OUT_SNAT
,
5509 count_1bits(ntohl(mask
)) + 1,
5510 ds_cstr(&match
), ds_cstr(&actions
));
5514 /* Logical router ingress table 0:
5515 * For NAT on a distributed router, add rules allowing
5516 * ingress traffic with eth.dst matching nat->external_mac
5517 * on the l3dgw_port instance where nat->logical_port is
5521 ds_put_format(&match
,
5522 "eth.dst == "ETH_ADDR_FMT
" && inport == %s"
5523 " && is_chassis_resident(\"%s\")",
5525 od
->l3dgw_port
->json_key
,
5527 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_ADMISSION
, 50,
5528 ds_cstr(&match
), "next;");
5531 /* Ingress Gateway Redirect Table: For NAT on a distributed
5532 * router, add flows that are specific to a NAT rule. These
5533 * flows indicate the presence of an applicable NAT rule that
5534 * can be applied in a distributed manner. */
5537 ds_put_format(&match
, "ip4.src == %s && outport == %s",
5539 od
->l3dgw_port
->json_key
);
5540 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_GW_REDIRECT
, 100,
5541 ds_cstr(&match
), "next;");
5544 /* Egress Loopback table: For NAT on a distributed router.
5545 * If packets in the egress pipeline on the distributed
5546 * gateway port have ip.dst matching a NAT external IP, then
5547 * loop a clone of the packet back to the beginning of the
5548 * ingress pipeline with inport = outport. */
5549 if (od
->l3dgw_port
) {
5550 /* Distributed router. */
5552 ds_put_format(&match
, "ip4.dst == %s && outport == %s",
5554 od
->l3dgw_port
->json_key
);
5556 ds_put_format(&actions
,
5557 "clone { ct_clear; "
5558 "inport = outport; outport = \"\"; "
5559 "flags = 0; flags.loopback = 1; ");
5560 for (int j
= 0; j
< MFF_N_LOG_REGS
; j
++) {
5561 ds_put_format(&actions
, "reg%d = 0; ", j
);
5563 ds_put_format(&actions
, REGBIT_EGRESS_LOOPBACK
" = 1; "
5564 "next(pipeline=ingress, table=0); };");
5565 ovn_lflow_add(lflows
, od
, S_ROUTER_OUT_EGR_LOOP
, 100,
5566 ds_cstr(&match
), ds_cstr(&actions
));
5570 /* Handle force SNAT options set in the gateway router. */
5571 if (dnat_force_snat_ip
&& !od
->l3dgw_port
) {
5572 /* If a packet with destination IP address as that of the
5573 * gateway router (as set in options:dnat_force_snat_ip) is seen,
5576 ds_put_format(&match
, "ip && ip4.dst == %s", dnat_force_snat_ip
);
5577 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_UNSNAT
, 110,
5578 ds_cstr(&match
), "ct_snat;");
5580 /* Higher priority rules to force SNAT with the IP addresses
5581 * configured in the Gateway router. This only takes effect
5582 * when the packet has already been DNATed once. */
5584 ds_put_format(&match
, "flags.force_snat_for_dnat == 1 && ip");
5586 ds_put_format(&actions
, "ct_snat(%s);", dnat_force_snat_ip
);
5587 ovn_lflow_add(lflows
, od
, S_ROUTER_OUT_SNAT
, 100,
5588 ds_cstr(&match
), ds_cstr(&actions
));
5590 if (lb_force_snat_ip
&& !od
->l3dgw_port
) {
5591 /* If a packet with destination IP address as that of the
5592 * gateway router (as set in options:lb_force_snat_ip) is seen,
5595 ds_put_format(&match
, "ip && ip4.dst == %s", lb_force_snat_ip
);
5596 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_UNSNAT
, 100,
5597 ds_cstr(&match
), "ct_snat;");
5599 /* Load balanced traffic will have flags.force_snat_for_lb set.
5602 ds_put_format(&match
, "flags.force_snat_for_lb == 1 && ip");
5604 ds_put_format(&actions
, "ct_snat(%s);", lb_force_snat_ip
);
5605 ovn_lflow_add(lflows
, od
, S_ROUTER_OUT_SNAT
, 100,
5606 ds_cstr(&match
), ds_cstr(&actions
));
5609 if (!od
->l3dgw_port
) {
5610 /* For gateway router, re-circulate every packet through
5611 * the DNAT zone. This helps with the following.
5613 * Any packet that needs to be unDNATed in the reverse
5614 * direction gets unDNATed. Ideally this could be done in
5615 * the egress pipeline. But since the gateway router
5616 * does not have any feature that depends on the source
5617 * ip address being external IP address for IP routing,
5618 * we can do it here, saving a future re-circulation. */
5619 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_DNAT
, 50,
5620 "ip", "flags.loopback = 1; ct_dnat;");
5622 /* For NAT on a distributed router, add flows to Ingress
5623 * IP Routing table, Ingress ARP Resolution table, and
5624 * Ingress Gateway Redirect Table that are not specific to a
5627 /* The highest priority IN_IP_ROUTING rule matches packets
5628 * with REGBIT_NAT_REDIRECT (set in DNAT or UNSNAT stages),
5629 * with action "ip.ttl--; next;". The IN_GW_REDIRECT table
5630 * will take care of setting the outport. */
5631 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_IP_ROUTING
, 300,
5632 REGBIT_NAT_REDIRECT
" == 1", "ip.ttl--; next;");
5634 /* The highest priority IN_ARP_RESOLVE rule matches packets
5635 * with REGBIT_NAT_REDIRECT (set in DNAT or UNSNAT stages),
5636 * then sets eth.dst to the distributed gateway port's
5637 * ethernet address. */
5639 ds_put_format(&actions
, "eth.dst = %s; next;",
5640 od
->l3dgw_port
->lrp_networks
.ea_s
);
5641 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_ARP_RESOLVE
, 200,
5642 REGBIT_NAT_REDIRECT
" == 1", ds_cstr(&actions
));
5644 /* The highest priority IN_GW_REDIRECT rule redirects packets
5645 * with REGBIT_NAT_REDIRECT (set in DNAT or UNSNAT stages) to
5646 * the central instance of the l3dgw_port for NAT processing. */
5648 ds_put_format(&actions
, "outport = %s; next;",
5649 od
->l3redirect_port
->json_key
);
5650 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_GW_REDIRECT
, 200,
5651 REGBIT_NAT_REDIRECT
" == 1", ds_cstr(&actions
));
5654 /* Load balancing and packet defrag are only valid on
5655 * Gateway routers or router with gateway port. */
5656 if (!smap_get(&od
->nbr
->options
, "chassis") && !od
->l3dgw_port
) {
5660 /* A set to hold all ips that need defragmentation and tracking. */
5661 struct sset all_ips
= SSET_INITIALIZER(&all_ips
);
5663 for (int i
= 0; i
< od
->nbr
->n_load_balancer
; i
++) {
5664 struct nbrec_load_balancer
*lb
= od
->nbr
->load_balancer
[i
];
5665 struct smap
*vips
= &lb
->vips
;
5666 struct smap_node
*node
;
5668 SMAP_FOR_EACH (node
, vips
) {
5672 /* node->key contains IP:port or just IP. */
5673 char *ip_address
= NULL
;
5674 ip_address_and_port_from_lb_key(node
->key
, &ip_address
, &port
,
5680 if (!sset_contains(&all_ips
, ip_address
)) {
5681 sset_add(&all_ips
, ip_address
);
5682 /* If there are any load balancing rules, we should send
5683 * the packet to conntrack for defragmentation and
5684 * tracking. This helps with two things.
5686 * 1. With tracking, we can send only new connections to
5687 * pick a DNAT ip address from a group.
5688 * 2. If there are L4 ports in load balancing rules, we
5689 * need the defragmentation to match on L4 ports. */
5691 if (addr_family
== AF_INET
) {
5692 ds_put_format(&match
, "ip && ip4.dst == %s",
5695 ds_put_format(&match
, "ip && ip6.dst == %s",
5698 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_DEFRAG
,
5699 100, ds_cstr(&match
), "ct_next;");
5702 /* Higher priority rules are added for load-balancing in DNAT
5703 * table. For every match (on a VIP[:port]), we add two flows
5704 * via add_router_lb_flow(). One flow is for specific matching
5705 * on ct.new with an action of "ct_lb($targets);". The other
5706 * flow is for ct.est with an action of "ct_dnat;". */
5708 ds_put_format(&actions
, "ct_lb(%s);", node
->value
);
5711 if (addr_family
== AF_INET
) {
5712 ds_put_format(&match
, "ip && ip4.dst == %s",
5715 ds_put_format(&match
, "ip && ip6.dst == %s",
5721 bool is_udp
= lb
->protocol
&& !strcmp(lb
->protocol
, "udp") ?
5725 ds_put_format(&match
, " && udp && udp.dst == %d",
5728 ds_put_format(&match
, " && tcp && tcp.dst == %d",
5734 if (od
->l3redirect_port
) {
5735 ds_put_format(&match
, " && is_chassis_resident(%s)",
5736 od
->l3redirect_port
->json_key
);
5738 add_router_lb_flow(lflows
, od
, &match
, &actions
, prio
,
5739 lb_force_snat_ip
, node
->value
, is_udp
,
5743 sset_destroy(&all_ips
);
5746 /* Logical router ingress table 5 and 6: IPv6 Router Adv (RA) options and
5748 HMAP_FOR_EACH (op
, key_node
, ports
) {
5749 if (!op
->nbrp
|| op
->nbrp
->peer
|| !op
->peer
) {
5753 if (!op
->lrp_networks
.n_ipv6_addrs
) {
5757 const char *address_mode
= smap_get(
5758 &op
->nbrp
->ipv6_ra_configs
, "address_mode");
5760 if (!address_mode
) {
5763 if (strcmp(address_mode
, "slaac") &&
5764 strcmp(address_mode
, "dhcpv6_stateful") &&
5765 strcmp(address_mode
, "dhcpv6_stateless")) {
5766 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
5767 VLOG_WARN_RL(&rl
, "Invalid address mode [%s] defined",
5772 if (smap_get_bool(&op
->nbrp
->ipv6_ra_configs
, "send_periodic",
5774 copy_ra_to_sb(op
, address_mode
);
5778 ds_put_format(&match
, "inport == %s && ip6.dst == ff02::2 && nd_rs",
5782 const char *mtu_s
= smap_get(
5783 &op
->nbrp
->ipv6_ra_configs
, "mtu");
5785 /* As per RFC 2460, 1280 is minimum IPv6 MTU. */
5786 uint32_t mtu
= (mtu_s
&& atoi(mtu_s
) >= 1280) ? atoi(mtu_s
) : 0;
5788 ds_put_format(&actions
, REGBIT_ND_RA_OPTS_RESULT
" = put_nd_ra_opts("
5789 "addr_mode = \"%s\", slla = %s",
5790 address_mode
, op
->lrp_networks
.ea_s
);
5792 ds_put_format(&actions
, ", mtu = %u", mtu
);
5795 bool add_rs_response_flow
= false;
5797 for (size_t i
= 0; i
< op
->lrp_networks
.n_ipv6_addrs
; i
++) {
5798 if (in6_is_lla(&op
->lrp_networks
.ipv6_addrs
[i
].network
)) {
5802 /* Add the prefix option if the address mode is slaac or
5803 * dhcpv6_stateless. */
5804 if (strcmp(address_mode
, "dhcpv6_stateful")) {
5805 ds_put_format(&actions
, ", prefix = %s/%u",
5806 op
->lrp_networks
.ipv6_addrs
[i
].network_s
,
5807 op
->lrp_networks
.ipv6_addrs
[i
].plen
);
5809 add_rs_response_flow
= true;
5812 if (add_rs_response_flow
) {
5813 ds_put_cstr(&actions
, "); next;");
5814 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_ND_RA_OPTIONS
, 50,
5815 ds_cstr(&match
), ds_cstr(&actions
));
5818 ds_put_format(&match
, "inport == %s && ip6.dst == ff02::2 && "
5819 "nd_ra && "REGBIT_ND_RA_OPTS_RESULT
, op
->json_key
);
5821 char ip6_str
[INET6_ADDRSTRLEN
+ 1];
5822 struct in6_addr lla
;
5823 in6_generate_lla(op
->lrp_networks
.ea
, &lla
);
5824 memset(ip6_str
, 0, sizeof(ip6_str
));
5825 ipv6_string_mapped(ip6_str
, &lla
);
5826 ds_put_format(&actions
, "eth.dst = eth.src; eth.src = %s; "
5827 "ip6.dst = ip6.src; ip6.src = %s; "
5828 "outport = inport; flags.loopback = 1; "
5830 op
->lrp_networks
.ea_s
, ip6_str
);
5831 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_IN_ND_RA_RESPONSE
, 50,
5832 ds_cstr(&match
), ds_cstr(&actions
));
5836 /* Logical router ingress table 5, 6: RS responder, by default goto next.
5838 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
5843 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_ND_RA_OPTIONS
, 0, "1", "next;");
5844 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_ND_RA_RESPONSE
, 0, "1", "next;");
5847 /* Logical router ingress table 7: IP Routing.
5849 * A packet that arrives at this table is an IP packet that should be
5850 * routed to the address in 'ip[46].dst'. This table sets outport to
5851 * the correct output port, eth.src to the output port's MAC
5852 * address, and '[xx]reg0' to the next-hop IP address (leaving
5853 * 'ip[46].dst', the packet’s final destination, unchanged), and
5854 * advances to the next table for ARP/ND resolution. */
5855 HMAP_FOR_EACH (op
, key_node
, ports
) {
5860 for (int i
= 0; i
< op
->lrp_networks
.n_ipv4_addrs
; i
++) {
5861 add_route(lflows
, op
, op
->lrp_networks
.ipv4_addrs
[i
].addr_s
,
5862 op
->lrp_networks
.ipv4_addrs
[i
].network_s
,
5863 op
->lrp_networks
.ipv4_addrs
[i
].plen
, NULL
, NULL
);
5866 for (int i
= 0; i
< op
->lrp_networks
.n_ipv6_addrs
; i
++) {
5867 add_route(lflows
, op
, op
->lrp_networks
.ipv6_addrs
[i
].addr_s
,
5868 op
->lrp_networks
.ipv6_addrs
[i
].network_s
,
5869 op
->lrp_networks
.ipv6_addrs
[i
].plen
, NULL
, NULL
);
5873 /* Convert the static routes to flows. */
5874 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
5879 for (int i
= 0; i
< od
->nbr
->n_static_routes
; i
++) {
5880 const struct nbrec_logical_router_static_route
*route
;
5882 route
= od
->nbr
->static_routes
[i
];
5883 build_static_route_flow(lflows
, od
, ports
, route
);
5887 /* XXX destination unreachable */
5889 /* Local router ingress table 8: ARP Resolution.
5891 * Any packet that reaches this table is an IP packet whose next-hop IP
5892 * address is in reg0. (ip4.dst is the final destination.) This table
5893 * resolves the IP address in reg0 into an output port in outport and an
5894 * Ethernet address in eth.dst. */
5895 HMAP_FOR_EACH (op
, key_node
, ports
) {
5896 if (op
->nbsp
&& !lsp_is_enabled(op
->nbsp
)) {
5901 /* This is a logical router port. If next-hop IP address in
5902 * '[xx]reg0' matches IP address of this router port, then
5903 * the packet is intended to eventually be sent to this
5904 * logical port. Set the destination mac address using this
5905 * port's mac address.
5907 * The packet is still in peer's logical pipeline. So the match
5908 * should be on peer's outport. */
5909 if (op
->peer
&& op
->nbrp
->peer
) {
5910 if (op
->lrp_networks
.n_ipv4_addrs
) {
5912 ds_put_format(&match
, "outport == %s && reg0 == ",
5913 op
->peer
->json_key
);
5914 op_put_v4_networks(&match
, op
, false);
5917 ds_put_format(&actions
, "eth.dst = %s; next;",
5918 op
->lrp_networks
.ea_s
);
5919 ovn_lflow_add(lflows
, op
->peer
->od
, S_ROUTER_IN_ARP_RESOLVE
,
5920 100, ds_cstr(&match
), ds_cstr(&actions
));
5923 if (op
->lrp_networks
.n_ipv6_addrs
) {
5925 ds_put_format(&match
, "outport == %s && xxreg0 == ",
5926 op
->peer
->json_key
);
5927 op_put_v6_networks(&match
, op
);
5930 ds_put_format(&actions
, "eth.dst = %s; next;",
5931 op
->lrp_networks
.ea_s
);
5932 ovn_lflow_add(lflows
, op
->peer
->od
, S_ROUTER_IN_ARP_RESOLVE
,
5933 100, ds_cstr(&match
), ds_cstr(&actions
));
5936 } else if (op
->od
->n_router_ports
&& strcmp(op
->nbsp
->type
, "router")) {
5937 /* This is a logical switch port that backs a VM or a container.
5938 * Extract its addresses. For each of the address, go through all
5939 * the router ports attached to the switch (to which this port
5940 * connects) and if the address in question is reachable from the
5941 * router port, add an ARP/ND entry in that router's pipeline. */
5943 for (size_t i
= 0; i
< op
->n_lsp_addrs
; i
++) {
5944 const char *ea_s
= op
->lsp_addrs
[i
].ea_s
;
5945 for (size_t j
= 0; j
< op
->lsp_addrs
[i
].n_ipv4_addrs
; j
++) {
5946 const char *ip_s
= op
->lsp_addrs
[i
].ipv4_addrs
[j
].addr_s
;
5947 for (size_t k
= 0; k
< op
->od
->n_router_ports
; k
++) {
5948 /* Get the Logical_Router_Port that the
5949 * Logical_Switch_Port is connected to, as
5951 const char *peer_name
= smap_get(
5952 &op
->od
->router_ports
[k
]->nbsp
->options
,
5958 struct ovn_port
*peer
= ovn_port_find(ports
, peer_name
);
5959 if (!peer
|| !peer
->nbrp
) {
5963 if (!find_lrp_member_ip(peer
, ip_s
)) {
5968 ds_put_format(&match
, "outport == %s && reg0 == %s",
5969 peer
->json_key
, ip_s
);
5972 ds_put_format(&actions
, "eth.dst = %s; next;", ea_s
);
5973 ovn_lflow_add(lflows
, peer
->od
,
5974 S_ROUTER_IN_ARP_RESOLVE
, 100,
5975 ds_cstr(&match
), ds_cstr(&actions
));
5979 for (size_t j
= 0; j
< op
->lsp_addrs
[i
].n_ipv6_addrs
; j
++) {
5980 const char *ip_s
= op
->lsp_addrs
[i
].ipv6_addrs
[j
].addr_s
;
5981 for (size_t k
= 0; k
< op
->od
->n_router_ports
; k
++) {
5982 /* Get the Logical_Router_Port that the
5983 * Logical_Switch_Port is connected to, as
5985 const char *peer_name
= smap_get(
5986 &op
->od
->router_ports
[k
]->nbsp
->options
,
5992 struct ovn_port
*peer
= ovn_port_find(ports
, peer_name
);
5993 if (!peer
|| !peer
->nbrp
) {
5997 if (!find_lrp_member_ip(peer
, ip_s
)) {
6002 ds_put_format(&match
, "outport == %s && xxreg0 == %s",
6003 peer
->json_key
, ip_s
);
6006 ds_put_format(&actions
, "eth.dst = %s; next;", ea_s
);
6007 ovn_lflow_add(lflows
, peer
->od
,
6008 S_ROUTER_IN_ARP_RESOLVE
, 100,
6009 ds_cstr(&match
), ds_cstr(&actions
));
6013 } else if (!strcmp(op
->nbsp
->type
, "router")) {
6014 /* This is a logical switch port that connects to a router. */
6016 /* The peer of this switch port is the router port for which
6017 * we need to add logical flows such that it can resolve
6018 * ARP entries for all the other router ports connected to
6019 * the switch in question. */
6021 const char *peer_name
= smap_get(&op
->nbsp
->options
,
6027 struct ovn_port
*peer
= ovn_port_find(ports
, peer_name
);
6028 if (!peer
|| !peer
->nbrp
) {
6032 for (size_t i
= 0; i
< op
->od
->n_router_ports
; i
++) {
6033 const char *router_port_name
= smap_get(
6034 &op
->od
->router_ports
[i
]->nbsp
->options
,
6036 struct ovn_port
*router_port
= ovn_port_find(ports
,
6038 if (!router_port
|| !router_port
->nbrp
) {
6042 /* Skip the router port under consideration. */
6043 if (router_port
== peer
) {
6047 if (router_port
->lrp_networks
.n_ipv4_addrs
) {
6049 ds_put_format(&match
, "outport == %s && reg0 == ",
6051 op_put_v4_networks(&match
, router_port
, false);
6054 ds_put_format(&actions
, "eth.dst = %s; next;",
6055 router_port
->lrp_networks
.ea_s
);
6056 ovn_lflow_add(lflows
, peer
->od
, S_ROUTER_IN_ARP_RESOLVE
,
6057 100, ds_cstr(&match
), ds_cstr(&actions
));
6060 if (router_port
->lrp_networks
.n_ipv6_addrs
) {
6062 ds_put_format(&match
, "outport == %s && xxreg0 == ",
6064 op_put_v6_networks(&match
, router_port
);
6067 ds_put_format(&actions
, "eth.dst = %s; next;",
6068 router_port
->lrp_networks
.ea_s
);
6069 ovn_lflow_add(lflows
, peer
->od
, S_ROUTER_IN_ARP_RESOLVE
,
6070 100, ds_cstr(&match
), ds_cstr(&actions
));
6076 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
6081 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_ARP_RESOLVE
, 0, "ip4",
6082 "get_arp(outport, reg0); next;");
6084 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_ARP_RESOLVE
, 0, "ip6",
6085 "get_nd(outport, xxreg0); next;");
6088 /* Logical router ingress table 9: Gateway redirect.
6090 * For traffic with outport equal to the l3dgw_port
6091 * on a distributed router, this table redirects a subset
6092 * of the traffic to the l3redirect_port which represents
6093 * the central instance of the l3dgw_port.
6095 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
6099 if (od
->l3dgw_port
&& od
->l3redirect_port
) {
6100 /* For traffic with outport == l3dgw_port, if the
6101 * packet did not match any higher priority redirect
6102 * rule, then the traffic is redirected to the central
6103 * instance of the l3dgw_port. */
6105 ds_put_format(&match
, "outport == %s",
6106 od
->l3dgw_port
->json_key
);
6108 ds_put_format(&actions
, "outport = %s; next;",
6109 od
->l3redirect_port
->json_key
);
6110 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_GW_REDIRECT
, 50,
6111 ds_cstr(&match
), ds_cstr(&actions
));
6113 /* If the Ethernet destination has not been resolved,
6114 * redirect to the central instance of the l3dgw_port.
6115 * Such traffic will be replaced by an ARP request or ND
6116 * Neighbor Solicitation in the ARP request ingress
6117 * table, before being redirected to the central instance.
6119 ds_put_format(&match
, " && eth.dst == 00:00:00:00:00:00");
6120 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_GW_REDIRECT
, 150,
6121 ds_cstr(&match
), ds_cstr(&actions
));
6124 /* Packets are allowed by default. */
6125 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_GW_REDIRECT
, 0, "1", "next;");
6128 /* Local router ingress table 10: ARP request.
6130 * In the common case where the Ethernet destination has been resolved,
6131 * this table outputs the packet (priority 0). Otherwise, it composes
6132 * and sends an ARP/IPv6 NA request (priority 100). */
6133 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
6138 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_ARP_REQUEST
, 100,
6139 "eth.dst == 00:00:00:00:00:00",
6141 "eth.dst = ff:ff:ff:ff:ff:ff; "
6144 "arp.op = 1; " /* ARP request */
6147 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_ARP_REQUEST
, 100,
6148 "eth.dst == 00:00:00:00:00:00",
6150 "nd.target = xxreg0; "
6153 ovn_lflow_add(lflows
, od
, S_ROUTER_IN_ARP_REQUEST
, 0, "1", "output;");
6156 /* Logical router egress table 1: Delivery (priority 100).
6158 * Priority 100 rules deliver packets to enabled logical ports. */
6159 HMAP_FOR_EACH (op
, key_node
, ports
) {
6164 if (!lrport_is_enabled(op
->nbrp
)) {
6165 /* Drop packets to disabled logical ports (since logical flow
6166 * tables are default-drop). */
6171 /* No egress packets should be processed in the context of
6172 * a chassisredirect port. The chassisredirect port should
6173 * be replaced by the l3dgw port in the local output
6174 * pipeline stage before egress processing. */
6179 ds_put_format(&match
, "outport == %s", op
->json_key
);
6180 ovn_lflow_add(lflows
, op
->od
, S_ROUTER_OUT_DELIVERY
, 100,
6181 ds_cstr(&match
), "output;");
6185 ds_destroy(&actions
);
6188 /* Updates the Logical_Flow and Multicast_Group tables in the OVN_SB database,
6189 * constructing their contents based on the OVN_NB database. */
6191 build_lflows(struct northd_context
*ctx
, struct hmap
*datapaths
,
6192 struct hmap
*ports
, struct hmap
*port_groups
)
6194 struct hmap lflows
= HMAP_INITIALIZER(&lflows
);
6195 struct hmap mcgroups
= HMAP_INITIALIZER(&mcgroups
);
6197 build_lswitch_flows(datapaths
, ports
, port_groups
, &lflows
, &mcgroups
);
6198 build_lrouter_flows(datapaths
, ports
, &lflows
);
6200 /* Push changes to the Logical_Flow table to database. */
6201 const struct sbrec_logical_flow
*sbflow
, *next_sbflow
;
6202 SBREC_LOGICAL_FLOW_FOR_EACH_SAFE (sbflow
, next_sbflow
, ctx
->ovnsb_idl
) {
6203 struct ovn_datapath
*od
6204 = ovn_datapath_from_sbrec(datapaths
, sbflow
->logical_datapath
);
6206 sbrec_logical_flow_delete(sbflow
);
6210 enum ovn_datapath_type dp_type
= od
->nbs
? DP_SWITCH
: DP_ROUTER
;
6211 enum ovn_pipeline pipeline
6212 = !strcmp(sbflow
->pipeline
, "ingress") ? P_IN
: P_OUT
;
6213 struct ovn_lflow
*lflow
= ovn_lflow_find(
6214 &lflows
, od
, ovn_stage_build(dp_type
, pipeline
, sbflow
->table_id
),
6215 sbflow
->priority
, sbflow
->match
, sbflow
->actions
, sbflow
->hash
);
6217 ovn_lflow_destroy(&lflows
, lflow
);
6219 sbrec_logical_flow_delete(sbflow
);
6222 struct ovn_lflow
*lflow
, *next_lflow
;
6223 HMAP_FOR_EACH_SAFE (lflow
, next_lflow
, hmap_node
, &lflows
) {
6224 const char *pipeline
= ovn_stage_get_pipeline_name(lflow
->stage
);
6225 uint8_t table
= ovn_stage_get_table(lflow
->stage
);
6227 sbflow
= sbrec_logical_flow_insert(ctx
->ovnsb_txn
);
6228 sbrec_logical_flow_set_logical_datapath(sbflow
, lflow
->od
->sb
);
6229 sbrec_logical_flow_set_pipeline(sbflow
, pipeline
);
6230 sbrec_logical_flow_set_table_id(sbflow
, table
);
6231 sbrec_logical_flow_set_priority(sbflow
, lflow
->priority
);
6232 sbrec_logical_flow_set_match(sbflow
, lflow
->match
);
6233 sbrec_logical_flow_set_actions(sbflow
, lflow
->actions
);
6235 /* Trim the source locator lflow->where, which looks something like
6236 * "ovn/northd/ovn-northd.c:1234", down to just the part following the
6237 * last slash, e.g. "ovn-northd.c:1234". */
6238 const char *slash
= strrchr(lflow
->where
, '/');
6240 const char *backslash
= strrchr(lflow
->where
, '\\');
6241 if (!slash
|| backslash
> slash
) {
6245 const char *where
= slash
? slash
+ 1 : lflow
->where
;
6247 struct smap ids
= SMAP_INITIALIZER(&ids
);
6248 smap_add(&ids
, "stage-name", ovn_stage_to_str(lflow
->stage
));
6249 smap_add(&ids
, "source", where
);
6250 if (lflow
->stage_hint
) {
6251 smap_add(&ids
, "stage-hint", lflow
->stage_hint
);
6253 sbrec_logical_flow_set_external_ids(sbflow
, &ids
);
6256 ovn_lflow_destroy(&lflows
, lflow
);
6258 hmap_destroy(&lflows
);
6260 /* Push changes to the Multicast_Group table to database. */
6261 const struct sbrec_multicast_group
*sbmc
, *next_sbmc
;
6262 SBREC_MULTICAST_GROUP_FOR_EACH_SAFE (sbmc
, next_sbmc
, ctx
->ovnsb_idl
) {
6263 struct ovn_datapath
*od
= ovn_datapath_from_sbrec(datapaths
,
6266 sbrec_multicast_group_delete(sbmc
);
6270 struct multicast_group group
= { .name
= sbmc
->name
,
6271 .key
= sbmc
->tunnel_key
};
6272 struct ovn_multicast
*mc
= ovn_multicast_find(&mcgroups
, od
, &group
);
6274 ovn_multicast_update_sbrec(mc
, sbmc
);
6275 ovn_multicast_destroy(&mcgroups
, mc
);
6277 sbrec_multicast_group_delete(sbmc
);
6280 struct ovn_multicast
*mc
, *next_mc
;
6281 HMAP_FOR_EACH_SAFE (mc
, next_mc
, hmap_node
, &mcgroups
) {
6282 sbmc
= sbrec_multicast_group_insert(ctx
->ovnsb_txn
);
6283 sbrec_multicast_group_set_datapath(sbmc
, mc
->datapath
->sb
);
6284 sbrec_multicast_group_set_name(sbmc
, mc
->group
->name
);
6285 sbrec_multicast_group_set_tunnel_key(sbmc
, mc
->group
->key
);
6286 ovn_multicast_update_sbrec(mc
, sbmc
);
6287 ovn_multicast_destroy(&mcgroups
, mc
);
6289 hmap_destroy(&mcgroups
);
6293 sync_address_set(struct northd_context
*ctx
, const char *name
,
6294 const char **addrs
, size_t n_addrs
,
6295 struct shash
*sb_address_sets
)
6297 const struct sbrec_address_set
*sb_address_set
;
6298 sb_address_set
= shash_find_and_delete(sb_address_sets
,
6300 if (!sb_address_set
) {
6301 sb_address_set
= sbrec_address_set_insert(ctx
->ovnsb_txn
);
6302 sbrec_address_set_set_name(sb_address_set
, name
);
6305 sbrec_address_set_set_addresses(sb_address_set
,
6309 /* OVN_Southbound Address_Set table contains same records as in north
6310 * bound, plus the records generated from Port_Group table in north bound.
6312 * There are 2 records generated from each port group, one for IPv4, and
6313 * one for IPv6, named in the format: <port group name>_ip4 and
6314 * <port group name>_ip6 respectively. MAC addresses are ignored.
6316 * We always update OVN_Southbound to match the Address_Set and Port_Group
6317 * in OVN_Northbound, so that the address sets used in Logical_Flows in
6318 * OVN_Southbound is checked against the proper set.*/
6320 sync_address_sets(struct northd_context
*ctx
)
6322 struct shash sb_address_sets
= SHASH_INITIALIZER(&sb_address_sets
);
6324 const struct sbrec_address_set
*sb_address_set
;
6325 SBREC_ADDRESS_SET_FOR_EACH (sb_address_set
, ctx
->ovnsb_idl
) {
6326 shash_add(&sb_address_sets
, sb_address_set
->name
, sb_address_set
);
6329 /* sync port group generated address sets first */
6330 const struct nbrec_port_group
*nb_port_group
;
6331 NBREC_PORT_GROUP_FOR_EACH (nb_port_group
, ctx
->ovnnb_idl
) {
6332 char **ipv4_addrs
= xcalloc(1, sizeof *ipv4_addrs
);
6333 size_t n_ipv4_addrs
= 0;
6334 size_t n_ipv4_addrs_buf
= 1;
6335 char **ipv6_addrs
= xcalloc(1, sizeof *ipv6_addrs
);
6336 size_t n_ipv6_addrs
= 0;
6337 size_t n_ipv6_addrs_buf
= 1;
6338 for (size_t i
= 0; i
< nb_port_group
->n_ports
; i
++) {
6339 for (size_t j
= 0; j
< nb_port_group
->ports
[i
]->n_addresses
; j
++) {
6340 struct lport_addresses laddrs
;
6341 extract_lsp_addresses(nb_port_group
->ports
[i
]->addresses
[j
],
6343 while (n_ipv4_addrs_buf
< n_ipv4_addrs
+ laddrs
.n_ipv4_addrs
) {
6344 n_ipv4_addrs_buf
*= 2;
6345 ipv4_addrs
= xrealloc(ipv4_addrs
,
6346 n_ipv4_addrs_buf
* sizeof *ipv4_addrs
);
6348 for (size_t k
= 0; k
< laddrs
.n_ipv4_addrs
; k
++) {
6349 ipv4_addrs
[n_ipv4_addrs
++] =
6350 xstrdup(laddrs
.ipv4_addrs
[k
].addr_s
);
6352 while (n_ipv6_addrs_buf
< n_ipv6_addrs
+ laddrs
.n_ipv6_addrs
) {
6353 n_ipv6_addrs_buf
*= 2;
6354 ipv6_addrs
= xrealloc(ipv6_addrs
,
6355 n_ipv6_addrs_buf
* sizeof *ipv6_addrs
);
6357 for (size_t k
= 0; k
< laddrs
.n_ipv6_addrs
; k
++) {
6358 ipv6_addrs
[n_ipv6_addrs
++] =
6359 xstrdup(laddrs
.ipv6_addrs
[k
].addr_s
);
6361 destroy_lport_addresses(&laddrs
);
6364 char *ipv4_addrs_name
= xasprintf("%s_ip4", nb_port_group
->name
);
6365 char *ipv6_addrs_name
= xasprintf("%s_ip6", nb_port_group
->name
);
6366 sync_address_set(ctx
, ipv4_addrs_name
, (const char **)ipv4_addrs
,
6367 n_ipv4_addrs
, &sb_address_sets
);
6368 sync_address_set(ctx
, ipv6_addrs_name
, (const char **)ipv6_addrs
,
6369 n_ipv6_addrs
, &sb_address_sets
);
6370 free(ipv4_addrs_name
);
6371 free(ipv6_addrs_name
);
6372 for (size_t i
= 0; i
< n_ipv4_addrs
; i
++) {
6373 free(ipv4_addrs
[i
]);
6376 for (size_t i
= 0; i
< n_ipv6_addrs
; i
++) {
6377 free(ipv6_addrs
[i
]);
6382 /* sync user defined address sets, which may overwrite port group
6383 * generated address sets if same name is used */
6384 const struct nbrec_address_set
*nb_address_set
;
6385 NBREC_ADDRESS_SET_FOR_EACH (nb_address_set
, ctx
->ovnnb_idl
) {
6386 sync_address_set(ctx
, nb_address_set
->name
,
6387 /* "char **" is not compatible with "const char **" */
6388 (const char **)nb_address_set
->addresses
,
6389 nb_address_set
->n_addresses
, &sb_address_sets
);
6392 struct shash_node
*node
, *next
;
6393 SHASH_FOR_EACH_SAFE (node
, next
, &sb_address_sets
) {
6394 sbrec_address_set_delete(node
->data
);
6395 shash_delete(&sb_address_sets
, node
);
6397 shash_destroy(&sb_address_sets
);
6400 /* Each port group in Port_Group table in OVN_Northbound has a corresponding
6401 * entry in Port_Group table in OVN_Southbound. In OVN_Northbound the entries
6402 * contains lport uuids, while in OVN_Southbound we store the lport names.
6405 sync_port_groups(struct northd_context
*ctx
)
6407 struct shash sb_port_groups
= SHASH_INITIALIZER(&sb_port_groups
);
6409 const struct sbrec_port_group
*sb_port_group
;
6410 SBREC_PORT_GROUP_FOR_EACH (sb_port_group
, ctx
->ovnsb_idl
) {
6411 shash_add(&sb_port_groups
, sb_port_group
->name
, sb_port_group
);
6414 const struct nbrec_port_group
*nb_port_group
;
6415 NBREC_PORT_GROUP_FOR_EACH (nb_port_group
, ctx
->ovnnb_idl
) {
6416 sb_port_group
= shash_find_and_delete(&sb_port_groups
,
6417 nb_port_group
->name
);
6418 if (!sb_port_group
) {
6419 sb_port_group
= sbrec_port_group_insert(ctx
->ovnsb_txn
);
6420 sbrec_port_group_set_name(sb_port_group
, nb_port_group
->name
);
6423 const char **nb_port_names
= xcalloc(nb_port_group
->n_ports
,
6424 sizeof *nb_port_names
);
6426 for (i
= 0; i
< nb_port_group
->n_ports
; i
++) {
6427 nb_port_names
[i
] = nb_port_group
->ports
[i
]->name
;
6429 sbrec_port_group_set_ports(sb_port_group
,
6431 nb_port_group
->n_ports
);
6432 free(nb_port_names
);
6435 struct shash_node
*node
, *next
;
6436 SHASH_FOR_EACH_SAFE (node
, next
, &sb_port_groups
) {
6437 sbrec_port_group_delete(node
->data
);
6438 shash_delete(&sb_port_groups
, node
);
6440 shash_destroy(&sb_port_groups
);
6444 * struct 'dns_info' is used to sync the DNS records between OVN Northbound db
6445 * and Southbound db.
6448 struct hmap_node hmap_node
;
6449 const struct nbrec_dns
*nb_dns
; /* DNS record in the Northbound db. */
6450 const struct sbrec_dns
*sb_dns
; /* DNS record in the Soutbound db. */
6452 /* Datapaths to which the DNS entry is associated with it. */
6453 const struct sbrec_datapath_binding
**sbs
;
6457 static inline struct dns_info
*
6458 get_dns_info_from_hmap(struct hmap
*dns_map
, struct uuid
*uuid
)
6460 struct dns_info
*dns_info
;
6461 size_t hash
= uuid_hash(uuid
);
6462 HMAP_FOR_EACH_WITH_HASH (dns_info
, hmap_node
, hash
, dns_map
) {
6463 if (uuid_equals(&dns_info
->nb_dns
->header_
.uuid
, uuid
)) {
6472 sync_dns_entries(struct northd_context
*ctx
, struct hmap
*datapaths
)
6474 struct hmap dns_map
= HMAP_INITIALIZER(&dns_map
);
6475 struct ovn_datapath
*od
;
6476 HMAP_FOR_EACH (od
, key_node
, datapaths
) {
6477 if (!od
->nbs
|| !od
->nbs
->n_dns_records
) {
6481 for (size_t i
= 0; i
< od
->nbs
->n_dns_records
; i
++) {
6482 struct dns_info
*dns_info
= get_dns_info_from_hmap(
6483 &dns_map
, &od
->nbs
->dns_records
[i
]->header_
.uuid
);
6485 size_t hash
= uuid_hash(
6486 &od
->nbs
->dns_records
[i
]->header_
.uuid
);
6487 dns_info
= xzalloc(sizeof *dns_info
);;
6488 dns_info
->nb_dns
= od
->nbs
->dns_records
[i
];
6489 hmap_insert(&dns_map
, &dns_info
->hmap_node
, hash
);
6493 dns_info
->sbs
= xrealloc(dns_info
->sbs
,
6494 dns_info
->n_sbs
* sizeof *dns_info
->sbs
);
6495 dns_info
->sbs
[dns_info
->n_sbs
- 1] = od
->sb
;
6499 const struct sbrec_dns
*sbrec_dns
, *next
;
6500 SBREC_DNS_FOR_EACH_SAFE (sbrec_dns
, next
, ctx
->ovnsb_idl
) {
6501 const char *nb_dns_uuid
= smap_get(&sbrec_dns
->external_ids
, "dns_id");
6502 struct uuid dns_uuid
;
6503 if (!nb_dns_uuid
|| !uuid_from_string(&dns_uuid
, nb_dns_uuid
)) {
6504 sbrec_dns_delete(sbrec_dns
);
6508 struct dns_info
*dns_info
=
6509 get_dns_info_from_hmap(&dns_map
, &dns_uuid
);
6511 dns_info
->sb_dns
= sbrec_dns
;
6513 sbrec_dns_delete(sbrec_dns
);
6517 struct dns_info
*dns_info
;
6518 HMAP_FOR_EACH_POP (dns_info
, hmap_node
, &dns_map
) {
6519 if (!dns_info
->sb_dns
) {
6520 sbrec_dns
= sbrec_dns_insert(ctx
->ovnsb_txn
);
6521 dns_info
->sb_dns
= sbrec_dns
;
6522 char *dns_id
= xasprintf(
6523 UUID_FMT
, UUID_ARGS(&dns_info
->nb_dns
->header_
.uuid
));
6524 const struct smap external_ids
=
6525 SMAP_CONST1(&external_ids
, "dns_id", dns_id
);
6526 sbrec_dns_set_external_ids(sbrec_dns
, &external_ids
);
6530 /* Set the datapaths and records. If nothing has changed, then
6531 * this will be a no-op.
6533 sbrec_dns_set_datapaths(
6535 (struct sbrec_datapath_binding
**)dns_info
->sbs
,
6537 sbrec_dns_set_records(dns_info
->sb_dns
, &dns_info
->nb_dns
->records
);
6538 free(dns_info
->sbs
);
6541 hmap_destroy(&dns_map
);
6547 ovnnb_db_run(struct northd_context
*ctx
, struct chassis_index
*chassis_index
,
6548 struct ovsdb_idl_loop
*sb_loop
)
6550 if (!ctx
->ovnsb_txn
|| !ctx
->ovnnb_txn
) {
6553 struct hmap datapaths
, ports
, port_groups
;
6554 build_datapaths(ctx
, &datapaths
);
6555 build_ports(ctx
, &datapaths
, chassis_index
, &ports
);
6556 build_ipam(&datapaths
, &ports
);
6557 build_port_group_lswitches(ctx
, &port_groups
, &ports
);
6558 build_lflows(ctx
, &datapaths
, &ports
, &port_groups
);
6560 sync_address_sets(ctx
);
6561 sync_port_groups(ctx
);
6562 sync_dns_entries(ctx
, &datapaths
);
6564 struct ovn_port_group
*pg
, *next_pg
;
6565 HMAP_FOR_EACH_SAFE (pg
, next_pg
, key_node
, &port_groups
) {
6566 ovn_port_group_destroy(&port_groups
, pg
);
6568 hmap_destroy(&port_groups
);
6570 struct ovn_datapath
*dp
, *next_dp
;
6571 HMAP_FOR_EACH_SAFE (dp
, next_dp
, key_node
, &datapaths
) {
6572 ovn_datapath_destroy(&datapaths
, dp
);
6574 hmap_destroy(&datapaths
);
6576 struct ovn_port
*port
, *next_port
;
6577 HMAP_FOR_EACH_SAFE (port
, next_port
, key_node
, &ports
) {
6578 ovn_port_destroy(&ports
, port
);
6580 hmap_destroy(&ports
);
6582 /* Copy nb_cfg from northbound to southbound database.
6584 * Also set up to update sb_cfg once our southbound transaction commits. */
6585 const struct nbrec_nb_global
*nb
= nbrec_nb_global_first(ctx
->ovnnb_idl
);
6587 nb
= nbrec_nb_global_insert(ctx
->ovnnb_txn
);
6589 const struct sbrec_sb_global
*sb
= sbrec_sb_global_first(ctx
->ovnsb_idl
);
6591 sb
= sbrec_sb_global_insert(ctx
->ovnsb_txn
);
6593 sbrec_sb_global_set_nb_cfg(sb
, nb
->nb_cfg
);
6594 sb_loop
->next_cfg
= nb
->nb_cfg
;
6596 cleanup_macam(&macam
);
6599 /* Handle changes to the 'chassis' column of the 'Port_Binding' table. When
6600 * this column is not empty, it means we need to set the corresponding logical
6601 * port as 'up' in the northbound DB. */
6603 update_logical_port_status(struct northd_context
*ctx
)
6605 struct hmap lports_hmap
;
6606 const struct sbrec_port_binding
*sb
;
6607 const struct nbrec_logical_switch_port
*nbsp
;
6609 struct lport_hash_node
{
6610 struct hmap_node node
;
6611 const struct nbrec_logical_switch_port
*nbsp
;
6614 hmap_init(&lports_hmap
);
6616 NBREC_LOGICAL_SWITCH_PORT_FOR_EACH(nbsp
, ctx
->ovnnb_idl
) {
6617 hash_node
= xzalloc(sizeof *hash_node
);
6618 hash_node
->nbsp
= nbsp
;
6619 hmap_insert(&lports_hmap
, &hash_node
->node
, hash_string(nbsp
->name
, 0));
6622 SBREC_PORT_BINDING_FOR_EACH(sb
, ctx
->ovnsb_idl
) {
6624 HMAP_FOR_EACH_WITH_HASH(hash_node
, node
,
6625 hash_string(sb
->logical_port
, 0),
6627 if (!strcmp(sb
->logical_port
, hash_node
->nbsp
->name
)) {
6628 nbsp
= hash_node
->nbsp
;
6634 /* The logical port doesn't exist for this port binding. This can
6635 * happen under normal circumstances when ovn-northd hasn't gotten
6636 * around to pruning the Port_Binding yet. */
6640 bool up
= (sb
->chassis
|| !strcmp(nbsp
->type
, "router"));
6641 if (!nbsp
->up
|| *nbsp
->up
!= up
) {
6642 nbrec_logical_switch_port_set_up(nbsp
, &up
, 1);
6646 HMAP_FOR_EACH_POP(hash_node
, node
, &lports_hmap
) {
6649 hmap_destroy(&lports_hmap
);
6652 static struct gen_opts_map supported_dhcp_opts
[] = {
6656 DHCP_OPT_DNS_SERVER
,
6657 DHCP_OPT_LOG_SERVER
,
6658 DHCP_OPT_LPR_SERVER
,
6659 DHCP_OPT_SWAP_SERVER
,
6660 DHCP_OPT_POLICY_FILTER
,
6661 DHCP_OPT_ROUTER_SOLICITATION
,
6662 DHCP_OPT_NIS_SERVER
,
6663 DHCP_OPT_NTP_SERVER
,
6665 DHCP_OPT_TFTP_SERVER
,
6666 DHCP_OPT_CLASSLESS_STATIC_ROUTE
,
6667 DHCP_OPT_MS_CLASSLESS_STATIC_ROUTE
,
6668 DHCP_OPT_IP_FORWARD_ENABLE
,
6669 DHCP_OPT_ROUTER_DISCOVERY
,
6670 DHCP_OPT_ETHERNET_ENCAP
,
6671 DHCP_OPT_DEFAULT_TTL
,
6674 DHCP_OPT_LEASE_TIME
,
6679 static struct gen_opts_map supported_dhcpv6_opts
[] = {
6681 DHCPV6_OPT_SERVER_ID
,
6682 DHCPV6_OPT_DOMAIN_SEARCH
,
6683 DHCPV6_OPT_DNS_SERVER
6687 check_and_add_supported_dhcp_opts_to_sb_db(struct northd_context
*ctx
)
6689 struct hmap dhcp_opts_to_add
= HMAP_INITIALIZER(&dhcp_opts_to_add
);
6690 for (size_t i
= 0; (i
< sizeof(supported_dhcp_opts
) /
6691 sizeof(supported_dhcp_opts
[0])); i
++) {
6692 hmap_insert(&dhcp_opts_to_add
, &supported_dhcp_opts
[i
].hmap_node
,
6693 dhcp_opt_hash(supported_dhcp_opts
[i
].name
));
6696 const struct sbrec_dhcp_options
*opt_row
, *opt_row_next
;
6697 SBREC_DHCP_OPTIONS_FOR_EACH_SAFE(opt_row
, opt_row_next
, ctx
->ovnsb_idl
) {
6698 struct gen_opts_map
*dhcp_opt
=
6699 dhcp_opts_find(&dhcp_opts_to_add
, opt_row
->name
);
6701 hmap_remove(&dhcp_opts_to_add
, &dhcp_opt
->hmap_node
);
6703 sbrec_dhcp_options_delete(opt_row
);
6707 struct gen_opts_map
*opt
;
6708 HMAP_FOR_EACH (opt
, hmap_node
, &dhcp_opts_to_add
) {
6709 struct sbrec_dhcp_options
*sbrec_dhcp_option
=
6710 sbrec_dhcp_options_insert(ctx
->ovnsb_txn
);
6711 sbrec_dhcp_options_set_name(sbrec_dhcp_option
, opt
->name
);
6712 sbrec_dhcp_options_set_code(sbrec_dhcp_option
, opt
->code
);
6713 sbrec_dhcp_options_set_type(sbrec_dhcp_option
, opt
->type
);
6716 hmap_destroy(&dhcp_opts_to_add
);
6720 check_and_add_supported_dhcpv6_opts_to_sb_db(struct northd_context
*ctx
)
6722 struct hmap dhcpv6_opts_to_add
= HMAP_INITIALIZER(&dhcpv6_opts_to_add
);
6723 for (size_t i
= 0; (i
< sizeof(supported_dhcpv6_opts
) /
6724 sizeof(supported_dhcpv6_opts
[0])); i
++) {
6725 hmap_insert(&dhcpv6_opts_to_add
, &supported_dhcpv6_opts
[i
].hmap_node
,
6726 dhcp_opt_hash(supported_dhcpv6_opts
[i
].name
));
6729 const struct sbrec_dhcpv6_options
*opt_row
, *opt_row_next
;
6730 SBREC_DHCPV6_OPTIONS_FOR_EACH_SAFE(opt_row
, opt_row_next
, ctx
->ovnsb_idl
) {
6731 struct gen_opts_map
*dhcp_opt
=
6732 dhcp_opts_find(&dhcpv6_opts_to_add
, opt_row
->name
);
6734 hmap_remove(&dhcpv6_opts_to_add
, &dhcp_opt
->hmap_node
);
6736 sbrec_dhcpv6_options_delete(opt_row
);
6740 struct gen_opts_map
*opt
;
6741 HMAP_FOR_EACH(opt
, hmap_node
, &dhcpv6_opts_to_add
) {
6742 struct sbrec_dhcpv6_options
*sbrec_dhcpv6_option
=
6743 sbrec_dhcpv6_options_insert(ctx
->ovnsb_txn
);
6744 sbrec_dhcpv6_options_set_name(sbrec_dhcpv6_option
, opt
->name
);
6745 sbrec_dhcpv6_options_set_code(sbrec_dhcpv6_option
, opt
->code
);
6746 sbrec_dhcpv6_options_set_type(sbrec_dhcpv6_option
, opt
->type
);
6749 hmap_destroy(&dhcpv6_opts_to_add
);
6752 static const char *rbac_chassis_auth
[] =
6754 static const char *rbac_chassis_update
[] =
6755 {"nb_cfg", "external_ids", "encaps", "vtep_logical_switches"};
6757 static const char *rbac_encap_auth
[] =
6759 static const char *rbac_encap_update
[] =
6760 {"type", "options", "ip"};
6762 static const char *rbac_port_binding_auth
[] =
6764 static const char *rbac_port_binding_update
[] =
6767 static const char *rbac_mac_binding_auth
[] =
6769 static const char *rbac_mac_binding_update
[] =
6770 {"logical_port", "ip", "mac", "datapath"};
6772 static struct rbac_perm_cfg
{
6777 const char **update
;
6779 const struct sbrec_rbac_permission
*row
;
6780 } rbac_perm_cfg
[] = {
6783 .auth
= rbac_chassis_auth
,
6784 .n_auth
= ARRAY_SIZE(rbac_chassis_auth
),
6786 .update
= rbac_chassis_update
,
6787 .n_update
= ARRAY_SIZE(rbac_chassis_update
),
6791 .auth
= rbac_encap_auth
,
6792 .n_auth
= ARRAY_SIZE(rbac_encap_auth
),
6794 .update
= rbac_encap_update
,
6795 .n_update
= ARRAY_SIZE(rbac_encap_update
),
6798 .table
= "Port_Binding",
6799 .auth
= rbac_port_binding_auth
,
6800 .n_auth
= ARRAY_SIZE(rbac_port_binding_auth
),
6802 .update
= rbac_port_binding_update
,
6803 .n_update
= ARRAY_SIZE(rbac_port_binding_update
),
6806 .table
= "MAC_Binding",
6807 .auth
= rbac_mac_binding_auth
,
6808 .n_auth
= ARRAY_SIZE(rbac_mac_binding_auth
),
6810 .update
= rbac_mac_binding_update
,
6811 .n_update
= ARRAY_SIZE(rbac_mac_binding_update
),
6825 ovn_rbac_validate_perm(const struct sbrec_rbac_permission
*perm
)
6827 struct rbac_perm_cfg
*pcfg
;
6830 for (pcfg
= rbac_perm_cfg
; pcfg
->table
; pcfg
++) {
6831 if (!strcmp(perm
->table
, pcfg
->table
)) {
6838 if (perm
->n_authorization
!= pcfg
->n_auth
||
6839 perm
->n_update
!= pcfg
->n_update
) {
6842 if (perm
->insert_delete
!= pcfg
->insdel
) {
6845 /* verify perm->authorization vs. pcfg->auth */
6847 for (i
= 0; i
< pcfg
->n_auth
; i
++) {
6848 for (j
= 0; j
< perm
->n_authorization
; j
++) {
6849 if (!strcmp(pcfg
->auth
[i
], perm
->authorization
[j
])) {
6855 if (n_found
!= pcfg
->n_auth
) {
6859 /* verify perm->update vs. pcfg->update */
6861 for (i
= 0; i
< pcfg
->n_update
; i
++) {
6862 for (j
= 0; j
< perm
->n_update
; j
++) {
6863 if (!strcmp(pcfg
->update
[i
], perm
->update
[j
])) {
6869 if (n_found
!= pcfg
->n_update
) {
6873 /* Success, db state matches expected state */
6879 ovn_rbac_create_perm(struct rbac_perm_cfg
*pcfg
,
6880 struct northd_context
*ctx
,
6881 const struct sbrec_rbac_role
*rbac_role
)
6883 struct sbrec_rbac_permission
*rbac_perm
;
6885 rbac_perm
= sbrec_rbac_permission_insert(ctx
->ovnsb_txn
);
6886 sbrec_rbac_permission_set_table(rbac_perm
, pcfg
->table
);
6887 sbrec_rbac_permission_set_authorization(rbac_perm
,
6890 sbrec_rbac_permission_set_insert_delete(rbac_perm
, pcfg
->insdel
);
6891 sbrec_rbac_permission_set_update(rbac_perm
,
6894 sbrec_rbac_role_update_permissions_setkey(rbac_role
, pcfg
->table
,
6899 check_and_update_rbac(struct northd_context
*ctx
)
6901 const struct sbrec_rbac_role
*rbac_role
= NULL
;
6902 const struct sbrec_rbac_permission
*perm_row
, *perm_next
;
6903 const struct sbrec_rbac_role
*role_row
, *role_row_next
;
6904 struct rbac_perm_cfg
*pcfg
;
6906 for (pcfg
= rbac_perm_cfg
; pcfg
->table
; pcfg
++) {
6910 SBREC_RBAC_PERMISSION_FOR_EACH_SAFE (perm_row
, perm_next
, ctx
->ovnsb_idl
) {
6911 if (!ovn_rbac_validate_perm(perm_row
)) {
6912 sbrec_rbac_permission_delete(perm_row
);
6915 SBREC_RBAC_ROLE_FOR_EACH_SAFE (role_row
, role_row_next
, ctx
->ovnsb_idl
) {
6916 if (strcmp(role_row
->name
, "ovn-controller")) {
6917 sbrec_rbac_role_delete(role_row
);
6919 rbac_role
= role_row
;
6924 rbac_role
= sbrec_rbac_role_insert(ctx
->ovnsb_txn
);
6925 sbrec_rbac_role_set_name(rbac_role
, "ovn-controller");
6928 for (pcfg
= rbac_perm_cfg
; pcfg
->table
; pcfg
++) {
6930 ovn_rbac_create_perm(pcfg
, ctx
, rbac_role
);
6935 /* Updates the sb_cfg and hv_cfg columns in the northbound NB_Global table. */
6937 update_northbound_cfg(struct northd_context
*ctx
,
6938 struct ovsdb_idl_loop
*sb_loop
)
6940 /* Update northbound sb_cfg if appropriate. */
6941 const struct nbrec_nb_global
*nbg
= nbrec_nb_global_first(ctx
->ovnnb_idl
);
6942 int64_t sb_cfg
= sb_loop
->cur_cfg
;
6943 if (nbg
&& sb_cfg
&& nbg
->sb_cfg
!= sb_cfg
) {
6944 nbrec_nb_global_set_sb_cfg(nbg
, sb_cfg
);
6947 /* Update northbound hv_cfg if appropriate. */
6949 /* Find minimum nb_cfg among all chassis. */
6950 const struct sbrec_chassis
*chassis
;
6951 int64_t hv_cfg
= nbg
->nb_cfg
;
6952 SBREC_CHASSIS_FOR_EACH (chassis
, ctx
->ovnsb_idl
) {
6953 if (chassis
->nb_cfg
< hv_cfg
) {
6954 hv_cfg
= chassis
->nb_cfg
;
6958 /* Update hv_cfg. */
6959 if (nbg
->hv_cfg
!= hv_cfg
) {
6960 nbrec_nb_global_set_hv_cfg(nbg
, hv_cfg
);
6965 /* Handle a fairly small set of changes in the southbound database. */
6967 ovnsb_db_run(struct northd_context
*ctx
, struct ovsdb_idl_loop
*sb_loop
)
6969 if (!ctx
->ovnnb_txn
|| !ovsdb_idl_has_ever_connected(ctx
->ovnsb_idl
)) {
6973 update_logical_port_status(ctx
);
6974 update_northbound_cfg(ctx
, sb_loop
);
6978 parse_options(int argc OVS_UNUSED
, char *argv
[] OVS_UNUSED
)
6981 DAEMON_OPTION_ENUMS
,
6985 static const struct option long_options
[] = {
6986 {"ovnsb-db", required_argument
, NULL
, 'd'},
6987 {"ovnnb-db", required_argument
, NULL
, 'D'},
6988 {"unixctl", required_argument
, NULL
, 'u'},
6989 {"help", no_argument
, NULL
, 'h'},
6990 {"options", no_argument
, NULL
, 'o'},
6991 {"version", no_argument
, NULL
, 'V'},
6992 DAEMON_LONG_OPTIONS
,
6994 STREAM_SSL_LONG_OPTIONS
,
6997 char *short_options
= ovs_cmdl_long_options_to_short_options(long_options
);
7002 c
= getopt_long(argc
, argv
, short_options
, long_options
, NULL
);
7008 DAEMON_OPTION_HANDLERS
;
7009 VLOG_OPTION_HANDLERS
;
7010 STREAM_SSL_OPTION_HANDLERS
;
7021 unixctl_path
= optarg
;
7029 ovs_cmdl_print_options(long_options
);
7033 ovs_print_version(0, 0);
7042 ovnsb_db
= default_sb_db();
7046 ovnnb_db
= default_nb_db();
7049 free(short_options
);
7053 add_column_noalert(struct ovsdb_idl
*idl
,
7054 const struct ovsdb_idl_column
*column
)
7056 ovsdb_idl_add_column(idl
, column
);
7057 ovsdb_idl_omit_alert(idl
, column
);
7061 main(int argc
, char *argv
[])
7063 int res
= EXIT_SUCCESS
;
7064 struct unixctl_server
*unixctl
;
7068 fatal_ignore_sigpipe();
7069 ovs_cmdl_proctitle_init(argc
, argv
);
7070 set_program_name(argv
[0]);
7071 service_start(&argc
, &argv
);
7072 parse_options(argc
, argv
);
7074 daemonize_start(false);
7076 retval
= unixctl_server_create(unixctl_path
, &unixctl
);
7080 unixctl_command_register("exit", "", 0, 0, ovn_northd_exit
, &exiting
);
7082 daemonize_complete();
7084 /* We want to detect (almost) all changes to the ovn-nb db. */
7085 struct ovsdb_idl_loop ovnnb_idl_loop
= OVSDB_IDL_LOOP_INITIALIZER(
7086 ovsdb_idl_create(ovnnb_db
, &nbrec_idl_class
, true, true));
7087 ovsdb_idl_omit_alert(ovnnb_idl_loop
.idl
, &nbrec_nb_global_col_sb_cfg
);
7088 ovsdb_idl_omit_alert(ovnnb_idl_loop
.idl
, &nbrec_nb_global_col_hv_cfg
);
7090 /* We want to detect only selected changes to the ovn-sb db. */
7091 struct ovsdb_idl_loop ovnsb_idl_loop
= OVSDB_IDL_LOOP_INITIALIZER(
7092 ovsdb_idl_create(ovnsb_db
, &sbrec_idl_class
, false, true));
7094 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_sb_global
);
7095 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_sb_global_col_nb_cfg
);
7097 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_logical_flow
);
7098 add_column_noalert(ovnsb_idl_loop
.idl
,
7099 &sbrec_logical_flow_col_logical_datapath
);
7100 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_logical_flow_col_pipeline
);
7101 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_logical_flow_col_table_id
);
7102 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_logical_flow_col_priority
);
7103 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_logical_flow_col_match
);
7104 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_logical_flow_col_actions
);
7106 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_multicast_group
);
7107 add_column_noalert(ovnsb_idl_loop
.idl
,
7108 &sbrec_multicast_group_col_datapath
);
7109 add_column_noalert(ovnsb_idl_loop
.idl
,
7110 &sbrec_multicast_group_col_tunnel_key
);
7111 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_multicast_group_col_name
);
7112 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_multicast_group_col_ports
);
7114 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_datapath_binding
);
7115 add_column_noalert(ovnsb_idl_loop
.idl
,
7116 &sbrec_datapath_binding_col_tunnel_key
);
7117 add_column_noalert(ovnsb_idl_loop
.idl
,
7118 &sbrec_datapath_binding_col_external_ids
);
7120 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_port_binding
);
7121 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_port_binding_col_datapath
);
7122 add_column_noalert(ovnsb_idl_loop
.idl
,
7123 &sbrec_port_binding_col_logical_port
);
7124 add_column_noalert(ovnsb_idl_loop
.idl
,
7125 &sbrec_port_binding_col_tunnel_key
);
7126 add_column_noalert(ovnsb_idl_loop
.idl
,
7127 &sbrec_port_binding_col_parent_port
);
7128 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_port_binding_col_tag
);
7129 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_port_binding_col_type
);
7130 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_port_binding_col_options
);
7131 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_port_binding_col_mac
);
7132 add_column_noalert(ovnsb_idl_loop
.idl
,
7133 &sbrec_port_binding_col_nat_addresses
);
7134 ovsdb_idl_add_column(ovnsb_idl_loop
.idl
, &sbrec_port_binding_col_chassis
);
7135 ovsdb_idl_add_column(ovnsb_idl_loop
.idl
,
7136 &sbrec_port_binding_col_gateway_chassis
);
7137 ovsdb_idl_add_column(ovnsb_idl_loop
.idl
,
7138 &sbrec_gateway_chassis_col_chassis
);
7139 ovsdb_idl_add_column(ovnsb_idl_loop
.idl
, &sbrec_gateway_chassis_col_name
);
7140 ovsdb_idl_add_column(ovnsb_idl_loop
.idl
,
7141 &sbrec_gateway_chassis_col_priority
);
7142 ovsdb_idl_add_column(ovnsb_idl_loop
.idl
,
7143 &sbrec_gateway_chassis_col_external_ids
);
7144 ovsdb_idl_add_column(ovnsb_idl_loop
.idl
,
7145 &sbrec_gateway_chassis_col_options
);
7146 add_column_noalert(ovnsb_idl_loop
.idl
,
7147 &sbrec_port_binding_col_external_ids
);
7148 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_mac_binding
);
7149 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_mac_binding_col_datapath
);
7150 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_mac_binding_col_ip
);
7151 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_mac_binding_col_mac
);
7152 add_column_noalert(ovnsb_idl_loop
.idl
,
7153 &sbrec_mac_binding_col_logical_port
);
7154 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_dhcp_options
);
7155 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_dhcp_options_col_code
);
7156 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_dhcp_options_col_type
);
7157 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_dhcp_options_col_name
);
7158 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_dhcpv6_options
);
7159 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_dhcpv6_options_col_code
);
7160 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_dhcpv6_options_col_type
);
7161 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_dhcpv6_options_col_name
);
7162 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_address_set
);
7163 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_address_set_col_name
);
7164 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_address_set_col_addresses
);
7165 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_port_group
);
7166 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_port_group_col_name
);
7167 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_port_group_col_ports
);
7169 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_dns
);
7170 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_dns_col_datapaths
);
7171 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_dns_col_records
);
7172 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_dns_col_external_ids
);
7174 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_rbac_role
);
7175 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_rbac_role_col_name
);
7176 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_rbac_role_col_permissions
);
7178 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_rbac_permission
);
7179 add_column_noalert(ovnsb_idl_loop
.idl
,
7180 &sbrec_rbac_permission_col_table
);
7181 add_column_noalert(ovnsb_idl_loop
.idl
,
7182 &sbrec_rbac_permission_col_authorization
);
7183 add_column_noalert(ovnsb_idl_loop
.idl
,
7184 &sbrec_rbac_permission_col_insert_delete
);
7185 add_column_noalert(ovnsb_idl_loop
.idl
, &sbrec_rbac_permission_col_update
);
7187 ovsdb_idl_add_table(ovnsb_idl_loop
.idl
, &sbrec_table_chassis
);
7188 ovsdb_idl_add_column(ovnsb_idl_loop
.idl
, &sbrec_chassis_col_nb_cfg
);
7189 ovsdb_idl_add_column(ovnsb_idl_loop
.idl
, &sbrec_chassis_col_name
);
7191 /* Ensure that only a single ovn-northd is active in the deployment by
7192 * acquiring a lock called "ovn_northd" on the southbound database
7193 * and then only performing DB transactions if the lock is held. */
7194 ovsdb_idl_set_lock(ovnsb_idl_loop
.idl
, "ovn_northd");
7195 bool had_lock
= false;
7200 struct northd_context ctx
= {
7201 .ovnnb_idl
= ovnnb_idl_loop
.idl
,
7202 .ovnnb_txn
= ovsdb_idl_loop_run(&ovnnb_idl_loop
),
7203 .ovnsb_idl
= ovnsb_idl_loop
.idl
,
7204 .ovnsb_txn
= ovsdb_idl_loop_run(&ovnsb_idl_loop
),
7207 if (!had_lock
&& ovsdb_idl_has_lock(ovnsb_idl_loop
.idl
)) {
7208 VLOG_INFO("ovn-northd lock acquired. "
7209 "This ovn-northd instance is now active.");
7211 } else if (had_lock
&& !ovsdb_idl_has_lock(ovnsb_idl_loop
.idl
)) {
7212 VLOG_INFO("ovn-northd lock lost. "
7213 "This ovn-northd instance is now on standby.");
7217 struct chassis_index chassis_index
;
7218 bool destroy_chassis_index
= false;
7219 if (ovsdb_idl_has_lock(ovnsb_idl_loop
.idl
)) {
7220 chassis_index_init(&chassis_index
, ctx
.ovnsb_idl
);
7221 destroy_chassis_index
= true;
7223 ovnnb_db_run(&ctx
, &chassis_index
, &ovnsb_idl_loop
);
7224 ovnsb_db_run(&ctx
, &ovnsb_idl_loop
);
7225 if (ctx
.ovnsb_txn
) {
7226 check_and_add_supported_dhcp_opts_to_sb_db(&ctx
);
7227 check_and_add_supported_dhcpv6_opts_to_sb_db(&ctx
);
7228 check_and_update_rbac(&ctx
);
7232 unixctl_server_run(unixctl
);
7233 unixctl_server_wait(unixctl
);
7235 poll_immediate_wake();
7237 ovsdb_idl_loop_commit_and_wait(&ovnnb_idl_loop
);
7238 ovsdb_idl_loop_commit_and_wait(&ovnsb_idl_loop
);
7241 if (should_service_stop()) {
7245 if (destroy_chassis_index
) {
7246 chassis_index_destroy(&chassis_index
);
7250 unixctl_server_destroy(unixctl
);
7251 ovsdb_idl_loop_destroy(&ovnnb_idl_loop
);
7252 ovsdb_idl_loop_destroy(&ovnsb_idl_loop
);
7259 ovn_northd_exit(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
7260 const char *argv
[] OVS_UNUSED
, void *exiting_
)
7262 bool *exiting
= exiting_
;
7265 unixctl_command_reply(conn
, NULL
);