1 /* Routing Information Base.
2 * Copyright (C) 1997, 98, 99, 2001 Kunihiro Ishiguro
4 * This file is part of GNU Zebra.
6 * GNU Zebra is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License as published by the
8 * Free Software Foundation; either version 2, or (at your option) any
11 * GNU Zebra is distributed in the hope that it will be useful, but
12 * WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * General Public License for more details.
16 * You should have received a copy of the GNU General Public License along
17 * with this program; see the file COPYING; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
34 #include "sockunion.h"
35 #include "srcdest_table.h"
39 #include "workqueue.h"
41 #include "zebra/connected.h"
42 #include "zebra/debug.h"
43 #include "zebra/interface.h"
44 #include "zebra/redistribute.h"
45 #include "zebra/rib.h"
47 #include "zebra/zapi_msg.h"
48 #include "zebra/zebra_errors.h"
49 #include "zebra/zebra_memory.h"
50 #include "zebra/zebra_ns.h"
51 #include "zebra/zebra_rnh.h"
52 #include "zebra/zebra_routemap.h"
53 #include "zebra/zebra_vrf.h"
54 #include "zebra/zebra_vxlan.h"
55 #include "zebra/zapi_msg.h"
56 #include "zebra/zebra_dplane.h"
59 * Event, list, and mutex for delivery of dataplane results
61 static pthread_mutex_t dplane_mutex
;
62 static struct thread
*t_dplane
;
63 static struct dplane_ctx_q_s rib_dplane_q
;
65 DEFINE_HOOK(rib_update
, (struct route_node
* rn
, const char *reason
),
68 /* Should we allow non Quagga processes to delete our routes */
69 extern int allow_delete
;
71 /* Each route type's string and default distance value. */
75 } route_info
[ZEBRA_ROUTE_MAX
] = {
76 [ZEBRA_ROUTE_SYSTEM
] = {ZEBRA_ROUTE_SYSTEM
, 0},
77 [ZEBRA_ROUTE_KERNEL
] = {ZEBRA_ROUTE_KERNEL
, 0},
78 [ZEBRA_ROUTE_CONNECT
] = {ZEBRA_ROUTE_CONNECT
, 0},
79 [ZEBRA_ROUTE_STATIC
] = {ZEBRA_ROUTE_STATIC
, 1},
80 [ZEBRA_ROUTE_RIP
] = {ZEBRA_ROUTE_RIP
, 120},
81 [ZEBRA_ROUTE_RIPNG
] = {ZEBRA_ROUTE_RIPNG
, 120},
82 [ZEBRA_ROUTE_OSPF
] = {ZEBRA_ROUTE_OSPF
, 110},
83 [ZEBRA_ROUTE_OSPF6
] = {ZEBRA_ROUTE_OSPF6
, 110},
84 [ZEBRA_ROUTE_ISIS
] = {ZEBRA_ROUTE_ISIS
, 115},
85 [ZEBRA_ROUTE_BGP
] = {ZEBRA_ROUTE_BGP
, 20 /* IBGP is 200. */},
86 [ZEBRA_ROUTE_PIM
] = {ZEBRA_ROUTE_PIM
, 255},
87 [ZEBRA_ROUTE_EIGRP
] = {ZEBRA_ROUTE_EIGRP
, 90},
88 [ZEBRA_ROUTE_NHRP
] = {ZEBRA_ROUTE_NHRP
, 10},
89 [ZEBRA_ROUTE_HSLS
] = {ZEBRA_ROUTE_HSLS
, 255},
90 [ZEBRA_ROUTE_OLSR
] = {ZEBRA_ROUTE_OLSR
, 255},
91 [ZEBRA_ROUTE_TABLE
] = {ZEBRA_ROUTE_TABLE
, 150},
92 [ZEBRA_ROUTE_LDP
] = {ZEBRA_ROUTE_LDP
, 150},
93 [ZEBRA_ROUTE_VNC
] = {ZEBRA_ROUTE_VNC
, 20},
94 [ZEBRA_ROUTE_VNC_DIRECT
] = {ZEBRA_ROUTE_VNC_DIRECT
, 20},
95 [ZEBRA_ROUTE_VNC_DIRECT_RH
] = {ZEBRA_ROUTE_VNC_DIRECT_RH
, 20},
96 [ZEBRA_ROUTE_BGP_DIRECT
] = {ZEBRA_ROUTE_BGP_DIRECT
, 20},
97 [ZEBRA_ROUTE_BGP_DIRECT_EXT
] = {ZEBRA_ROUTE_BGP_DIRECT_EXT
, 20},
98 [ZEBRA_ROUTE_BABEL
] = {ZEBRA_ROUTE_BABEL
, 100},
99 [ZEBRA_ROUTE_SHARP
] = {ZEBRA_ROUTE_SHARP
, 150},
101 /* no entry/default: 150 */
104 /* RPF lookup behaviour */
105 static enum multicast_mode ipv4_multicast_mode
= MCAST_NO_CONFIG
;
108 static void __attribute__((format(printf
, 5, 6)))
109 _rnode_zlog(const char *_func
, vrf_id_t vrf_id
, struct route_node
*rn
,
110 int priority
, const char *msgfmt
, ...)
112 char buf
[SRCDEST2STR_BUFFER
+ sizeof(" (MRIB)")];
116 va_start(ap
, msgfmt
);
117 vsnprintf(msgbuf
, sizeof(msgbuf
), msgfmt
, ap
);
121 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
122 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
124 if (info
->safi
== SAFI_MULTICAST
)
125 strcat(buf
, " (MRIB)");
127 snprintf(buf
, sizeof(buf
), "{(route_node *) NULL}");
130 zlog(priority
, "%s: %d:%s: %s", _func
, vrf_id
, buf
, msgbuf
);
133 #define rnode_debug(node, vrf_id, ...) \
134 _rnode_zlog(__func__, vrf_id, node, LOG_DEBUG, __VA_ARGS__)
135 #define rnode_info(node, ...) \
136 _rnode_zlog(__func__, vrf_id, node, LOG_INFO, __VA_ARGS__)
138 uint8_t route_distance(int type
)
142 if ((unsigned)type
>= array_size(route_info
))
145 distance
= route_info
[type
].distance
;
150 int is_zebra_valid_kernel_table(uint32_t table_id
)
153 if ((table_id
== RT_TABLE_UNSPEC
) || (table_id
== RT_TABLE_LOCAL
)
154 || (table_id
== RT_TABLE_COMPAT
))
161 int is_zebra_main_routing_table(uint32_t table_id
)
163 if ((table_id
== RT_TABLE_MAIN
)
164 || (table_id
== zebrad
.rtm_table_default
))
169 int zebra_check_addr(const struct prefix
*p
)
171 if (p
->family
== AF_INET
) {
174 addr
= p
->u
.prefix4
.s_addr
;
177 if (IPV4_NET127(addr
) || IN_CLASSD(addr
)
178 || IPV4_LINKLOCAL(addr
))
181 if (p
->family
== AF_INET6
) {
182 if (IN6_IS_ADDR_LOOPBACK(&p
->u
.prefix6
))
184 if (IN6_IS_ADDR_LINKLOCAL(&p
->u
.prefix6
))
190 /* Add nexthop to the end of a rib node's nexthop list */
191 void route_entry_nexthop_add(struct route_entry
*re
, struct nexthop
*nexthop
)
193 nexthop_add(&re
->ng
.nexthop
, nexthop
);
199 * copy_nexthop - copy a nexthop to the rib structure.
201 void route_entry_copy_nexthops(struct route_entry
*re
, struct nexthop
*nh
)
203 assert(!re
->ng
.nexthop
);
204 copy_nexthops(&re
->ng
.nexthop
, nh
, NULL
);
205 for (struct nexthop
*nexthop
= nh
; nexthop
; nexthop
= nexthop
->next
)
209 /* Delete specified nexthop from the list. */
210 void route_entry_nexthop_delete(struct route_entry
*re
, struct nexthop
*nexthop
)
213 nexthop
->next
->prev
= nexthop
->prev
;
215 nexthop
->prev
->next
= nexthop
->next
;
217 re
->ng
.nexthop
= nexthop
->next
;
222 struct nexthop
*route_entry_nexthop_ifindex_add(struct route_entry
*re
,
226 struct nexthop
*nexthop
;
228 nexthop
= nexthop_new();
229 nexthop
->type
= NEXTHOP_TYPE_IFINDEX
;
230 nexthop
->ifindex
= ifindex
;
231 nexthop
->vrf_id
= nh_vrf_id
;
233 route_entry_nexthop_add(re
, nexthop
);
238 struct nexthop
*route_entry_nexthop_ipv4_add(struct route_entry
*re
,
239 struct in_addr
*ipv4
,
243 struct nexthop
*nexthop
;
245 nexthop
= nexthop_new();
246 nexthop
->type
= NEXTHOP_TYPE_IPV4
;
247 nexthop
->vrf_id
= nh_vrf_id
;
248 nexthop
->gate
.ipv4
= *ipv4
;
250 nexthop
->src
.ipv4
= *src
;
252 route_entry_nexthop_add(re
, nexthop
);
257 struct nexthop
*route_entry_nexthop_ipv4_ifindex_add(struct route_entry
*re
,
258 struct in_addr
*ipv4
,
263 struct nexthop
*nexthop
;
264 struct interface
*ifp
;
266 nexthop
= nexthop_new();
267 nexthop
->vrf_id
= nh_vrf_id
;
268 nexthop
->type
= NEXTHOP_TYPE_IPV4_IFINDEX
;
269 nexthop
->gate
.ipv4
= *ipv4
;
271 nexthop
->src
.ipv4
= *src
;
272 nexthop
->ifindex
= ifindex
;
273 ifp
= if_lookup_by_index(nexthop
->ifindex
, nh_vrf_id
);
274 /*Pending: need to think if null ifp here is ok during bootup?
275 There was a crash because ifp here was coming to be NULL */
277 if (connected_is_unnumbered(ifp
)
278 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_EVPN_ROUTE
)
279 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ONLINK
)) {
280 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
);
283 route_entry_nexthop_add(re
, nexthop
);
288 struct nexthop
*route_entry_nexthop_ipv6_add(struct route_entry
*re
,
289 struct in6_addr
*ipv6
,
292 struct nexthop
*nexthop
;
294 nexthop
= nexthop_new();
295 nexthop
->vrf_id
= nh_vrf_id
;
296 nexthop
->type
= NEXTHOP_TYPE_IPV6
;
297 nexthop
->gate
.ipv6
= *ipv6
;
299 route_entry_nexthop_add(re
, nexthop
);
304 struct nexthop
*route_entry_nexthop_ipv6_ifindex_add(struct route_entry
*re
,
305 struct in6_addr
*ipv6
,
309 struct nexthop
*nexthop
;
311 nexthop
= nexthop_new();
312 nexthop
->vrf_id
= nh_vrf_id
;
313 nexthop
->type
= NEXTHOP_TYPE_IPV6_IFINDEX
;
314 nexthop
->gate
.ipv6
= *ipv6
;
315 nexthop
->ifindex
= ifindex
;
316 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_EVPN_ROUTE
)
317 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ONLINK
)) {
318 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
);
321 route_entry_nexthop_add(re
, nexthop
);
326 struct nexthop
*route_entry_nexthop_blackhole_add(struct route_entry
*re
,
327 enum blackhole_type bh_type
)
329 struct nexthop
*nexthop
;
331 nexthop
= nexthop_new();
332 nexthop
->vrf_id
= VRF_DEFAULT
;
333 nexthop
->type
= NEXTHOP_TYPE_BLACKHOLE
;
334 nexthop
->bh_type
= bh_type
;
336 route_entry_nexthop_add(re
, nexthop
);
341 static void nexthop_set_resolved(afi_t afi
, const struct nexthop
*newhop
,
342 struct nexthop
*nexthop
)
344 struct nexthop
*resolved_hop
;
346 resolved_hop
= nexthop_new();
347 SET_FLAG(resolved_hop
->flags
, NEXTHOP_FLAG_ACTIVE
);
349 resolved_hop
->vrf_id
= nexthop
->vrf_id
;
350 switch (newhop
->type
) {
351 case NEXTHOP_TYPE_IPV4
:
352 case NEXTHOP_TYPE_IPV4_IFINDEX
:
353 /* If the resolving route specifies a gateway, use it */
354 resolved_hop
->type
= newhop
->type
;
355 resolved_hop
->gate
.ipv4
= newhop
->gate
.ipv4
;
357 if (newhop
->ifindex
) {
358 resolved_hop
->type
= NEXTHOP_TYPE_IPV4_IFINDEX
;
359 resolved_hop
->ifindex
= newhop
->ifindex
;
360 if (newhop
->flags
& NEXTHOP_FLAG_ONLINK
)
361 resolved_hop
->flags
|= NEXTHOP_FLAG_ONLINK
;
364 case NEXTHOP_TYPE_IPV6
:
365 case NEXTHOP_TYPE_IPV6_IFINDEX
:
366 resolved_hop
->type
= newhop
->type
;
367 resolved_hop
->gate
.ipv6
= newhop
->gate
.ipv6
;
369 if (newhop
->ifindex
) {
370 resolved_hop
->type
= NEXTHOP_TYPE_IPV6_IFINDEX
;
371 resolved_hop
->ifindex
= newhop
->ifindex
;
374 case NEXTHOP_TYPE_IFINDEX
:
375 /* If the resolving route is an interface route,
376 * it means the gateway we are looking up is connected
377 * to that interface. (The actual network is _not_ onlink).
378 * Therefore, the resolved route should have the original
379 * gateway as nexthop as it is directly connected.
381 * On Linux, we have to set the onlink netlink flag because
382 * otherwise, the kernel won't accept the route.
384 resolved_hop
->flags
|= NEXTHOP_FLAG_ONLINK
;
386 resolved_hop
->type
= NEXTHOP_TYPE_IPV4_IFINDEX
;
387 resolved_hop
->gate
.ipv4
= nexthop
->gate
.ipv4
;
388 } else if (afi
== AFI_IP6
) {
389 resolved_hop
->type
= NEXTHOP_TYPE_IPV6_IFINDEX
;
390 resolved_hop
->gate
.ipv6
= nexthop
->gate
.ipv6
;
392 resolved_hop
->ifindex
= newhop
->ifindex
;
394 case NEXTHOP_TYPE_BLACKHOLE
:
395 resolved_hop
->type
= NEXTHOP_TYPE_BLACKHOLE
;
396 resolved_hop
->bh_type
= nexthop
->bh_type
;
400 /* Copy labels of the resolved route */
401 if (newhop
->nh_label
)
402 nexthop_add_labels(resolved_hop
, newhop
->nh_label_type
,
403 newhop
->nh_label
->num_labels
,
404 &newhop
->nh_label
->label
[0]);
406 resolved_hop
->rparent
= nexthop
;
407 nexthop_add(&nexthop
->resolved
, resolved_hop
);
410 /* If force flag is not set, do not modify falgs at all for uninstall
411 the route from FIB. */
412 static int nexthop_active(afi_t afi
, struct route_entry
*re
,
413 struct nexthop
*nexthop
, int set
,
414 struct route_node
*top
)
417 struct route_table
*table
;
418 struct route_node
*rn
;
419 struct route_entry
*match
= NULL
;
421 struct nexthop
*newhop
;
422 struct interface
*ifp
;
425 if ((nexthop
->type
== NEXTHOP_TYPE_IPV4
)
426 || nexthop
->type
== NEXTHOP_TYPE_IPV6
)
427 nexthop
->ifindex
= 0;
430 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
);
431 nexthops_free(nexthop
->resolved
);
432 nexthop
->resolved
= NULL
;
436 /* Next hops (remote VTEPs) for EVPN routes are fully resolved. */
437 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_EVPN_RVTEP
))
440 /* Skip nexthops that have been filtered out due to route-map */
441 /* The nexthops are specific to this route and so the same */
442 /* nexthop for a different route may not have this flag set */
443 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FILTERED
)) {
444 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
445 zlog_debug("\t%s: Nexthop Filtered",
446 __PRETTY_FUNCTION__
);
451 * Check to see if we should trust the passed in information
452 * for UNNUMBERED interfaces as that we won't find the GW
453 * address in the routing table.
455 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
)) {
456 ifp
= if_lookup_by_index(nexthop
->ifindex
, nexthop
->vrf_id
);
457 if ((ifp
&& connected_is_unnumbered(ifp
))
458 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ONLINK
)) {
459 if (if_is_operative(ifp
))
462 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
464 "\t%s: Onlink and interface %s is not operative",
465 __PRETTY_FUNCTION__
, ifp
->name
);
469 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
471 "\t%s: Interface %s is not unnumbered",
473 ifp
? ifp
->name
: "Unknown");
478 /* Make lookup prefix. */
479 memset(&p
, 0, sizeof(struct prefix
));
483 p
.prefixlen
= IPV4_MAX_PREFIXLEN
;
484 p
.u
.prefix4
= nexthop
->gate
.ipv4
;
488 p
.prefixlen
= IPV6_MAX_PREFIXLEN
;
489 p
.u
.prefix6
= nexthop
->gate
.ipv6
;
492 assert(afi
!= AFI_IP
&& afi
!= AFI_IP6
);
496 table
= zebra_vrf_table(afi
, SAFI_UNICAST
, nexthop
->vrf_id
);
498 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
499 zlog_debug("\t%s: Table not found",
500 __PRETTY_FUNCTION__
);
504 rn
= route_node_match(table
, (struct prefix
*)&p
);
506 route_unlock_node(rn
);
508 /* Lookup should halt if we've matched against ourselves ('top',
509 * if specified) - i.e., we cannot have a nexthop NH1 is
510 * resolved by a route NH1. The exception is if the route is a
513 if (top
&& rn
== top
)
514 if (((afi
== AFI_IP
) && (rn
->p
.prefixlen
!= 32))
515 || ((afi
== AFI_IP6
) && (rn
->p
.prefixlen
!= 128))) {
516 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
518 "\t%s: Matched against ourself and prefix length is not max bit length",
519 __PRETTY_FUNCTION__
);
523 /* Pick up selected route. */
524 /* However, do not resolve over default route unless explicitly
526 if (is_default_prefix(&rn
->p
)
527 && !rnh_resolve_via_default(p
.family
)) {
528 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
530 "\t:%s: Resolved against default route",
531 __PRETTY_FUNCTION__
);
535 dest
= rib_dest_from_rnode(rn
);
536 if (dest
&& dest
->selected_fib
537 && !CHECK_FLAG(dest
->selected_fib
->status
,
539 && dest
->selected_fib
->type
!= ZEBRA_ROUTE_TABLE
)
540 match
= dest
->selected_fib
;
542 /* If there is no selected route or matched route is EGP, go up
547 } while (rn
&& rn
->info
== NULL
);
554 if (match
->type
== ZEBRA_ROUTE_CONNECT
) {
555 /* Directly point connected route. */
556 newhop
= match
->ng
.nexthop
;
558 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
559 || nexthop
->type
== NEXTHOP_TYPE_IPV6
)
560 nexthop
->ifindex
= newhop
->ifindex
;
563 } else if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ALLOW_RECURSION
)) {
565 for (ALL_NEXTHOPS(match
->ng
, newhop
)) {
566 if (!CHECK_FLAG(newhop
->flags
,
569 if (CHECK_FLAG(newhop
->flags
,
570 NEXTHOP_FLAG_RECURSIVE
))
574 SET_FLAG(nexthop
->flags
,
575 NEXTHOP_FLAG_RECURSIVE
);
577 ROUTE_ENTRY_NEXTHOPS_CHANGED
);
578 nexthop_set_resolved(afi
, newhop
,
584 re
->nexthop_mtu
= match
->mtu
;
585 if (!resolved
&& IS_ZEBRA_DEBUG_RIB_DETAILED
)
586 zlog_debug("\t%s: Recursion failed to find",
587 __PRETTY_FUNCTION__
);
589 } else if (re
->type
== ZEBRA_ROUTE_STATIC
) {
591 for (ALL_NEXTHOPS(match
->ng
, newhop
)) {
592 if (!CHECK_FLAG(newhop
->flags
,
597 SET_FLAG(nexthop
->flags
,
598 NEXTHOP_FLAG_RECURSIVE
);
599 nexthop_set_resolved(afi
, newhop
,
605 re
->nexthop_mtu
= match
->mtu
;
607 if (!resolved
&& IS_ZEBRA_DEBUG_RIB_DETAILED
)
609 "\t%s: Static route unable to resolve",
610 __PRETTY_FUNCTION__
);
613 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
614 zlog_debug("\t%s: Route Type %s has not turned on recursion",
616 zebra_route_string(re
->type
));
617 if (re
->type
== ZEBRA_ROUTE_BGP
&&
618 !CHECK_FLAG(re
->flags
, ZEBRA_FLAG_IBGP
))
619 zlog_debug("\tEBGP: see \"disable-ebgp-connected-route-check\" or \"disable-connected-check\"");
624 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
625 zlog_debug("\t%s: Nexthop did not lookup in table",
626 __PRETTY_FUNCTION__
);
630 struct route_entry
*rib_match(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
,
631 union g_addr
*addr
, struct route_node
**rn_out
)
634 struct route_table
*table
;
635 struct route_node
*rn
;
636 struct route_entry
*match
= NULL
;
637 struct nexthop
*newhop
;
640 table
= zebra_vrf_table(afi
, safi
, vrf_id
);
644 memset(&p
, 0, sizeof(struct prefix
));
647 p
.u
.prefix4
= addr
->ipv4
;
648 p
.prefixlen
= IPV4_MAX_PREFIXLEN
;
650 p
.u
.prefix6
= addr
->ipv6
;
651 p
.prefixlen
= IPV6_MAX_PREFIXLEN
;
654 rn
= route_node_match(table
, (struct prefix
*)&p
);
659 route_unlock_node(rn
);
661 dest
= rib_dest_from_rnode(rn
);
662 if (dest
&& dest
->selected_fib
663 && !CHECK_FLAG(dest
->selected_fib
->status
,
664 ROUTE_ENTRY_REMOVED
))
665 match
= dest
->selected_fib
;
667 /* If there is no selected route or matched route is EGP, go up
672 } while (rn
&& rn
->info
== NULL
);
676 if (match
->type
!= ZEBRA_ROUTE_CONNECT
) {
678 for (ALL_NEXTHOPS(match
->ng
, newhop
))
679 if (CHECK_FLAG(newhop
->flags
,
696 struct route_entry
*rib_match_ipv4_multicast(vrf_id_t vrf_id
,
698 struct route_node
**rn_out
)
700 struct route_entry
*re
= NULL
, *mre
= NULL
, *ure
= NULL
;
701 struct route_node
*m_rn
= NULL
, *u_rn
= NULL
;
702 union g_addr gaddr
= {.ipv4
= addr
};
704 switch (ipv4_multicast_mode
) {
705 case MCAST_MRIB_ONLY
:
706 return rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
,
708 case MCAST_URIB_ONLY
:
709 return rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, rn_out
);
710 case MCAST_NO_CONFIG
:
711 case MCAST_MIX_MRIB_FIRST
:
712 re
= mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
,
715 re
= ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
,
718 case MCAST_MIX_DISTANCE
:
719 mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
, &m_rn
);
720 ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, &u_rn
);
722 re
= ure
->distance
< mre
->distance
? ure
: mre
;
728 case MCAST_MIX_PFXLEN
:
729 mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
, &m_rn
);
730 ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, &u_rn
);
732 re
= u_rn
->p
.prefixlen
> m_rn
->p
.prefixlen
? ure
: mre
;
741 *rn_out
= (re
== mre
) ? m_rn
: u_rn
;
743 if (IS_ZEBRA_DEBUG_RIB
) {
745 inet_ntop(AF_INET
, &addr
, buf
, BUFSIZ
);
747 zlog_debug("%s: %s: vrf: %u found %s, using %s",
748 __func__
, buf
, vrf_id
,
749 mre
? (ure
? "MRIB+URIB" : "MRIB")
750 : ure
? "URIB" : "nothing",
751 re
== ure
? "URIB" : re
== mre
? "MRIB" : "none");
756 void multicast_mode_ipv4_set(enum multicast_mode mode
)
758 if (IS_ZEBRA_DEBUG_RIB
)
759 zlog_debug("%s: multicast lookup mode set (%d)", __func__
,
761 ipv4_multicast_mode
= mode
;
764 enum multicast_mode
multicast_mode_ipv4_get(void)
766 return ipv4_multicast_mode
;
769 struct route_entry
*rib_lookup_ipv4(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
771 struct route_table
*table
;
772 struct route_node
*rn
;
773 struct route_entry
*match
= NULL
;
774 struct nexthop
*nexthop
;
778 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
782 rn
= route_node_lookup(table
, (struct prefix
*)p
);
784 /* No route for this prefix. */
789 route_unlock_node(rn
);
790 dest
= rib_dest_from_rnode(rn
);
792 if (dest
&& dest
->selected_fib
793 && !CHECK_FLAG(dest
->selected_fib
->status
, ROUTE_ENTRY_REMOVED
))
794 match
= dest
->selected_fib
;
799 if (match
->type
== ZEBRA_ROUTE_CONNECT
)
802 for (ALL_NEXTHOPS(match
->ng
, nexthop
))
803 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
810 * This clone function, unlike its original rib_lookup_ipv4(), checks
811 * if specified IPv4 route record (prefix/mask -> gate) exists in
812 * the whole RIB and has ROUTE_ENTRY_SELECTED_FIB set.
816 * 0: exact match found
817 * 1: a match was found with a different gate
818 * 2: connected route found
819 * 3: no matches found
821 int rib_lookup_ipv4_route(struct prefix_ipv4
*p
, union sockunion
*qgate
,
824 struct route_table
*table
;
825 struct route_node
*rn
;
826 struct route_entry
*match
= NULL
;
827 struct nexthop
*nexthop
;
832 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
834 return ZEBRA_RIB_LOOKUP_ERROR
;
836 /* Scan the RIB table for exactly matching RIB entry. */
837 rn
= route_node_lookup(table
, (struct prefix
*)p
);
839 /* No route for this prefix. */
841 return ZEBRA_RIB_NOTFOUND
;
844 route_unlock_node(rn
);
845 dest
= rib_dest_from_rnode(rn
);
847 /* Find out if a "selected" RR for the discovered RIB entry exists ever.
849 if (dest
&& dest
->selected_fib
850 && !CHECK_FLAG(dest
->selected_fib
->status
, ROUTE_ENTRY_REMOVED
))
851 match
= dest
->selected_fib
;
853 /* None such found :( */
855 return ZEBRA_RIB_NOTFOUND
;
857 if (match
->type
== ZEBRA_ROUTE_CONNECT
)
858 return ZEBRA_RIB_FOUND_CONNECTED
;
860 /* Ok, we have a cood candidate, let's check it's nexthop list... */
862 for (ALL_NEXTHOPS(match
->ng
, nexthop
))
863 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
)) {
865 if (nexthop
->gate
.ipv4
.s_addr
== sockunion2ip(qgate
))
866 return ZEBRA_RIB_FOUND_EXACT
;
867 if (IS_ZEBRA_DEBUG_RIB
) {
868 char gate_buf
[INET_ADDRSTRLEN
],
869 qgate_buf
[INET_ADDRSTRLEN
];
870 inet_ntop(AF_INET
, &nexthop
->gate
.ipv4
.s_addr
,
871 gate_buf
, INET_ADDRSTRLEN
);
872 inet_ntop(AF_INET
, &sockunion2ip(qgate
),
873 qgate_buf
, INET_ADDRSTRLEN
);
874 zlog_debug("%s: qgate == %s, %s == %s",
876 nexthop
->rparent
? "rgate" : "gate",
882 return ZEBRA_RIB_FOUND_NOGATE
;
884 return ZEBRA_RIB_NOTFOUND
;
887 #define RIB_SYSTEM_ROUTE(R) \
888 ((R)->type == ZEBRA_ROUTE_KERNEL || (R)->type == ZEBRA_ROUTE_CONNECT)
890 #define RIB_KERNEL_ROUTE(R) \
891 ((R)->type == ZEBRA_ROUTE_KERNEL)
893 /* This function verifies reachability of one given nexthop, which can be
894 * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored
895 * in nexthop->flags field. If the 4th parameter, 'set', is non-zero,
896 * nexthop->ifindex will be updated appropriately as well.
897 * An existing route map can turn (otherwise active) nexthop into inactive, but
900 * The return value is the final value of 'ACTIVE' flag.
903 static unsigned nexthop_active_check(struct route_node
*rn
,
904 struct route_entry
*re
,
905 struct nexthop
*nexthop
, int set
)
907 struct interface
*ifp
;
908 route_map_result_t ret
= RMAP_MATCH
;
910 char buf
[SRCDEST2STR_BUFFER
];
911 const struct prefix
*p
, *src_p
;
912 struct zebra_vrf
*zvrf
;
914 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
916 if (rn
->p
.family
== AF_INET
)
918 else if (rn
->p
.family
== AF_INET6
)
922 switch (nexthop
->type
) {
923 case NEXTHOP_TYPE_IFINDEX
:
924 ifp
= if_lookup_by_index(nexthop
->ifindex
, nexthop
->vrf_id
);
925 if (ifp
&& if_is_operative(ifp
))
926 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
928 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
930 case NEXTHOP_TYPE_IPV4
:
931 case NEXTHOP_TYPE_IPV4_IFINDEX
:
933 if (nexthop_active(AFI_IP
, re
, nexthop
, set
, rn
))
934 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
936 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
938 case NEXTHOP_TYPE_IPV6
:
940 if (nexthop_active(AFI_IP6
, re
, nexthop
, set
, rn
))
941 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
943 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
945 case NEXTHOP_TYPE_IPV6_IFINDEX
:
946 /* RFC 5549, v4 prefix with v6 NH */
947 if (rn
->p
.family
!= AF_INET
)
949 if (IN6_IS_ADDR_LINKLOCAL(&nexthop
->gate
.ipv6
)) {
950 ifp
= if_lookup_by_index(nexthop
->ifindex
,
952 if (ifp
&& if_is_operative(ifp
))
953 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
955 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
957 if (nexthop_active(AFI_IP6
, re
, nexthop
, set
, rn
))
958 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
960 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
963 case NEXTHOP_TYPE_BLACKHOLE
:
964 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
969 if (!CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
)) {
970 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
971 zlog_debug("\t%s: Unable to find a active nexthop",
972 __PRETTY_FUNCTION__
);
976 /* XXX: What exactly do those checks do? Do we support
977 * e.g. IPv4 routes with IPv6 nexthops or vice versa?
979 if (RIB_SYSTEM_ROUTE(re
) || (family
== AFI_IP
&& p
->family
!= AF_INET
)
980 || (family
== AFI_IP6
&& p
->family
!= AF_INET6
))
981 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
983 /* The original code didn't determine the family correctly
984 * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi
985 * from the rib_table_info in those cases.
986 * Possibly it may be better to use only the rib_table_info
990 rib_table_info_t
*info
;
992 info
= srcdest_rnode_table_info(rn
);
996 memset(&nexthop
->rmap_src
.ipv6
, 0, sizeof(union g_addr
));
998 zvrf
= zebra_vrf_lookup_by_id(nexthop
->vrf_id
);
1000 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1001 zlog_debug("\t%s: zvrf is NULL", __PRETTY_FUNCTION__
);
1002 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
1005 /* It'll get set if required inside */
1006 ret
= zebra_route_map_check(family
, re
->type
, re
->instance
, p
,
1007 nexthop
, zvrf
, re
->tag
);
1008 if (ret
== RMAP_DENYMATCH
) {
1009 if (IS_ZEBRA_DEBUG_RIB
) {
1010 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1012 "%u:%s: Filtering out with NH out %s due to route map",
1014 ifindex2ifname(nexthop
->ifindex
,
1017 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
1019 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
1022 /* Iterate over all nexthops of the given RIB entry and refresh their
1023 * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any
1024 * nexthop is found to toggle the ACTIVE flag, the whole re structure
1025 * is flagged with ROUTE_ENTRY_CHANGED. The 4th 'set' argument is
1026 * transparently passed to nexthop_active_check().
1028 * Return value is the new number of active nexthops.
1031 static int nexthop_active_update(struct route_node
*rn
, struct route_entry
*re
,
1034 struct nexthop
*nexthop
;
1035 union g_addr prev_src
;
1036 unsigned int prev_active
, new_active
, old_num_nh
;
1037 ifindex_t prev_index
;
1039 old_num_nh
= re
->nexthop_active_num
;
1041 re
->nexthop_active_num
= 0;
1042 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1044 for (nexthop
= re
->ng
.nexthop
; nexthop
; nexthop
= nexthop
->next
) {
1045 /* No protocol daemon provides src and so we're skipping
1047 prev_src
= nexthop
->rmap_src
;
1048 prev_active
= CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
1049 prev_index
= nexthop
->ifindex
;
1050 if ((new_active
= nexthop_active_check(rn
, re
, nexthop
, set
)))
1051 re
->nexthop_active_num
++;
1052 /* Don't allow src setting on IPv6 addr for now */
1053 if (prev_active
!= new_active
|| prev_index
!= nexthop
->ifindex
1054 || ((nexthop
->type
>= NEXTHOP_TYPE_IFINDEX
1055 && nexthop
->type
< NEXTHOP_TYPE_IPV6
)
1056 && prev_src
.ipv4
.s_addr
1057 != nexthop
->rmap_src
.ipv4
.s_addr
)
1058 || ((nexthop
->type
>= NEXTHOP_TYPE_IPV6
1059 && nexthop
->type
< NEXTHOP_TYPE_BLACKHOLE
)
1060 && !(IPV6_ADDR_SAME(&prev_src
.ipv6
,
1061 &nexthop
->rmap_src
.ipv6
)))) {
1062 SET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1063 SET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1067 if (old_num_nh
!= re
->nexthop_active_num
)
1068 SET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1070 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
)) {
1071 SET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1074 return re
->nexthop_active_num
;
1078 * Is this RIB labeled-unicast? It must be of type BGP and all paths
1079 * (nexthops) must have a label.
1081 int zebra_rib_labeled_unicast(struct route_entry
*re
)
1083 struct nexthop
*nexthop
= NULL
;
1085 if (re
->type
!= ZEBRA_ROUTE_BGP
)
1088 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1089 if (!nexthop
->nh_label
|| !nexthop
->nh_label
->num_labels
)
1095 /* Update flag indicates whether this is a "replace" or not. Currently, this
1096 * is only used for IPv4.
1098 void rib_install_kernel(struct route_node
*rn
, struct route_entry
*re
,
1099 struct route_entry
*old
)
1101 struct nexthop
*nexthop
;
1102 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1103 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1104 const struct prefix
*p
, *src_p
;
1105 enum zebra_dplane_result ret
;
1107 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1109 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1111 if (info
->safi
!= SAFI_UNICAST
) {
1112 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1113 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1116 struct nexthop
*prev
;
1118 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1119 UNSET_FLAG (nexthop
->flags
, NEXTHOP_FLAG_DUPLICATE
);
1120 for (ALL_NEXTHOPS(re
->ng
, prev
)) {
1121 if (prev
== nexthop
)
1123 if (nexthop_same_firsthop(nexthop
, prev
)) {
1124 SET_FLAG(nexthop
->flags
,
1125 NEXTHOP_FLAG_DUPLICATE
);
1133 * If this is a replace to a new RE let the originator of the RE
1134 * know that they've lost
1136 if (old
&& (old
!= re
) && (old
->type
!= re
->type
))
1137 zsend_route_notify_owner(old
, p
, ZAPI_ROUTE_BETTER_ADMIN_WON
);
1139 /* Update fib selection */
1140 dest
->selected_fib
= re
;
1143 * Make sure we update the FPM any time we send new information to
1146 hook_call(rib_update
, rn
, "installing in kernel");
1148 /* Send add or update */
1149 if (old
&& (old
!= re
))
1150 ret
= dplane_route_update(rn
, re
, old
);
1152 ret
= dplane_route_add(rn
, re
);
1155 case ZEBRA_DPLANE_REQUEST_QUEUED
:
1157 zvrf
->installs_queued
++;
1159 case ZEBRA_DPLANE_REQUEST_FAILURE
:
1161 char str
[SRCDEST2STR_BUFFER
];
1163 srcdest_rnode2str(rn
, str
, sizeof(str
));
1164 flog_err(EC_ZEBRA_DP_INSTALL_FAIL
,
1165 "%u:%s: Failed to enqueue dataplane install",
1169 case ZEBRA_DPLANE_REQUEST_SUCCESS
:
1178 /* Uninstall the route from kernel. */
1179 void rib_uninstall_kernel(struct route_node
*rn
, struct route_entry
*re
)
1181 struct nexthop
*nexthop
;
1182 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1183 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1185 if (info
->safi
!= SAFI_UNICAST
) {
1186 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1187 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1192 * Make sure we update the FPM any time we send new information to
1195 hook_call(rib_update
, rn
, "uninstalling from kernel");
1197 switch (dplane_route_delete(rn
, re
)) {
1198 case ZEBRA_DPLANE_REQUEST_QUEUED
:
1200 zvrf
->removals_queued
++;
1202 case ZEBRA_DPLANE_REQUEST_FAILURE
:
1204 char str
[SRCDEST2STR_BUFFER
];
1206 srcdest_rnode2str(rn
, str
, sizeof(str
));
1207 flog_err(EC_ZEBRA_DP_INSTALL_FAIL
,
1208 "%u:%s: Failed to enqueue dataplane uninstall",
1212 case ZEBRA_DPLANE_REQUEST_SUCCESS
:
1221 /* Uninstall the route from kernel. */
1222 static void rib_uninstall(struct route_node
*rn
, struct route_entry
*re
)
1224 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1225 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1226 struct nexthop
*nexthop
;
1228 if (dest
&& dest
->selected_fib
== re
) {
1229 if (info
->safi
== SAFI_UNICAST
)
1230 hook_call(rib_update
, rn
, "rib_uninstall");
1232 /* If labeled-unicast route, uninstall transit LSP. */
1233 if (zebra_rib_labeled_unicast(re
))
1234 zebra_mpls_lsp_uninstall(info
->zvrf
, rn
, re
);
1236 if (!RIB_SYSTEM_ROUTE(re
))
1237 rib_uninstall_kernel(rn
, re
);
1239 dest
->selected_fib
= NULL
;
1241 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1242 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1245 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)) {
1246 const struct prefix
*p
, *src_p
;
1248 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1250 redistribute_delete(p
, src_p
, re
);
1251 UNSET_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
);
1256 * rib_can_delete_dest
1258 * Returns TRUE if the given dest can be deleted from the table.
1260 static int rib_can_delete_dest(rib_dest_t
*dest
)
1267 * Don't delete the dest if we have to update the FPM about this
1270 if (CHECK_FLAG(dest
->flags
, RIB_DEST_UPDATE_FPM
)
1271 || CHECK_FLAG(dest
->flags
, RIB_DEST_SENT_TO_FPM
))
1280 * Garbage collect the rib dest corresponding to the given route node
1283 * Returns TRUE if the dest was deleted, FALSE otherwise.
1285 int rib_gc_dest(struct route_node
*rn
)
1289 dest
= rib_dest_from_rnode(rn
);
1293 if (!rib_can_delete_dest(dest
))
1296 if (IS_ZEBRA_DEBUG_RIB
) {
1297 struct zebra_vrf
*zvrf
;
1299 zvrf
= rib_dest_vrf(dest
);
1300 rnode_debug(rn
, zvrf_id(zvrf
), "removing dest from table");
1304 XFREE(MTYPE_RIB_DEST
, dest
);
1308 * Release the one reference that we keep on the route node.
1310 route_unlock_node(rn
);
1314 static void rib_process_add_fib(struct zebra_vrf
*zvrf
, struct route_node
*rn
,
1315 struct route_entry
*new)
1317 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1319 hook_call(rib_update
, rn
, "new route selected");
1321 /* Update real nexthop. This may actually determine if nexthop is active
1323 if (!nexthop_active_update(rn
, new, 1)) {
1324 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1328 if (IS_ZEBRA_DEBUG_RIB
) {
1329 char buf
[SRCDEST2STR_BUFFER
];
1330 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1331 zlog_debug("%u:%s: Adding route rn %p, re %p (type %d)",
1332 zvrf_id(zvrf
), buf
, rn
, new, new->type
);
1335 /* If labeled-unicast route, install transit LSP. */
1336 if (zebra_rib_labeled_unicast(new))
1337 zebra_mpls_lsp_install(zvrf
, rn
, new);
1339 if (!RIB_SYSTEM_ROUTE(new))
1340 rib_install_kernel(rn
, new, NULL
);
1342 dest
->selected_fib
= new;
1344 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1347 static void rib_process_del_fib(struct zebra_vrf
*zvrf
, struct route_node
*rn
,
1348 struct route_entry
*old
)
1350 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1351 hook_call(rib_update
, rn
, "removing existing route");
1353 /* Uninstall from kernel. */
1354 if (IS_ZEBRA_DEBUG_RIB
) {
1355 char buf
[SRCDEST2STR_BUFFER
];
1356 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1357 zlog_debug("%u:%s: Deleting route rn %p, re %p (type %d)",
1358 zvrf_id(zvrf
), buf
, rn
, old
, old
->type
);
1361 /* If labeled-unicast route, uninstall transit LSP. */
1362 if (zebra_rib_labeled_unicast(old
))
1363 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1365 if (!RIB_SYSTEM_ROUTE(old
))
1366 rib_uninstall_kernel(rn
, old
);
1369 * We are setting this to NULL here
1370 * because that is what we traditionally
1371 * have been doing. I am not positive
1372 * that this is the right thing to do
1373 * but let's leave the code alone
1374 * for the RIB_SYSTEM_ROUTE case
1376 dest
->selected_fib
= NULL
;
1379 /* Update nexthop for route, reset changed flag. */
1380 /* Note: this code also handles the Linux case when an interface goes
1381 * down, causing the kernel to delete routes without sending DELROUTE
1384 if (!nexthop_active_update(rn
, old
, 1) &&
1385 (RIB_KERNEL_ROUTE(old
)))
1386 SET_FLAG(old
->status
, ROUTE_ENTRY_REMOVED
);
1388 UNSET_FLAG(old
->status
, ROUTE_ENTRY_CHANGED
);
1391 static void rib_process_update_fib(struct zebra_vrf
*zvrf
,
1392 struct route_node
*rn
,
1393 struct route_entry
*old
,
1394 struct route_entry
*new)
1396 struct nexthop
*nexthop
= NULL
;
1398 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1401 * We have to install or update if a new route has been selected or
1402 * something has changed.
1404 if (new != old
|| CHECK_FLAG(new->status
, ROUTE_ENTRY_CHANGED
)) {
1405 hook_call(rib_update
, rn
, "updating existing route");
1407 /* Update the nexthop; we could determine here that nexthop is
1409 if (nexthop_active_update(rn
, new, 1))
1412 /* If nexthop is active, install the selected route, if
1414 * the install succeeds, cleanup flags for prior route, if
1419 if (IS_ZEBRA_DEBUG_RIB
) {
1420 char buf
[SRCDEST2STR_BUFFER
];
1421 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1424 "%u:%s: Updating route rn %p, re %p (type %d) "
1426 zvrf_id(zvrf
), buf
, rn
, new,
1427 new->type
, old
, old
->type
);
1430 "%u:%s: Updating route rn %p, re %p (type %d)",
1431 zvrf_id(zvrf
), buf
, rn
, new,
1435 /* If labeled-unicast route, uninstall transit LSP. */
1436 if (zebra_rib_labeled_unicast(old
))
1437 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1439 /* Non-system route should be installed. */
1440 if (!RIB_SYSTEM_ROUTE(new)) {
1441 /* If labeled-unicast route, install transit
1443 if (zebra_rib_labeled_unicast(new))
1444 zebra_mpls_lsp_install(zvrf
, rn
, new);
1446 rib_install_kernel(rn
, new, old
);
1449 * We do not need to install the
1450 * selected route because it
1451 * is already isntalled by
1452 * the system( ie not us )
1453 * so just mark it as winning
1454 * we do need to ensure that
1455 * if we uninstall a route
1456 * from ourselves we don't
1457 * over write this pointer
1459 dest
->selected_fib
= NULL
;
1461 /* If install succeeded or system route, cleanup flags
1462 * for prior route. */
1464 if (RIB_SYSTEM_ROUTE(new)) {
1465 if (!RIB_SYSTEM_ROUTE(old
))
1466 rib_uninstall_kernel(rn
, old
);
1468 for (nexthop
= old
->ng
.nexthop
; nexthop
;
1469 nexthop
= nexthop
->next
)
1470 UNSET_FLAG(nexthop
->flags
,
1477 * If nexthop for selected route is not active or install
1479 * may need to uninstall and delete for redistribution.
1482 if (IS_ZEBRA_DEBUG_RIB
) {
1483 char buf
[SRCDEST2STR_BUFFER
];
1484 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1487 "%u:%s: Deleting route rn %p, re %p (type %d) "
1488 "old %p (type %d) - nexthop inactive",
1489 zvrf_id(zvrf
), buf
, rn
, new,
1490 new->type
, old
, old
->type
);
1493 "%u:%s: Deleting route rn %p, re %p (type %d) - nexthop inactive",
1494 zvrf_id(zvrf
), buf
, rn
, new,
1498 /* If labeled-unicast route, uninstall transit LSP. */
1499 if (zebra_rib_labeled_unicast(old
))
1500 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1502 if (!RIB_SYSTEM_ROUTE(old
))
1503 rib_uninstall_kernel(rn
, old
);
1505 dest
->selected_fib
= NULL
;
1509 * Same route selected; check if in the FIB and if not,
1511 * is housekeeping code to deal with race conditions in kernel
1513 * netlink reporting interface up before IPv4 or IPv6 protocol
1517 if (!RIB_SYSTEM_ROUTE(new)) {
1518 bool in_fib
= false;
1520 for (ALL_NEXTHOPS(new->ng
, nexthop
))
1521 if (CHECK_FLAG(nexthop
->flags
,
1522 NEXTHOP_FLAG_FIB
)) {
1527 rib_install_kernel(rn
, new, NULL
);
1531 /* Update prior route. */
1533 /* Set real nexthop. */
1534 nexthop_active_update(rn
, old
, 1);
1535 UNSET_FLAG(old
->status
, ROUTE_ENTRY_CHANGED
);
1538 /* Clear changed flag. */
1539 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1542 /* Check if 'alternate' RIB entry is better than 'current'. */
1543 static struct route_entry
*rib_choose_best(struct route_entry
*current
,
1544 struct route_entry
*alternate
)
1546 if (current
== NULL
)
1549 /* filter route selection in following order:
1550 * - connected beats other types
1551 * - if both connected, loopback or vrf wins
1552 * - lower distance beats higher
1553 * - lower metric beats higher for equal distance
1554 * - last, hence oldest, route wins tie break.
1557 /* Connected routes. Check to see if either are a vrf
1558 * or loopback interface. If not, pick the last connected
1559 * route of the set of lowest metric connected routes.
1561 if (alternate
->type
== ZEBRA_ROUTE_CONNECT
) {
1562 if (current
->type
!= ZEBRA_ROUTE_CONNECT
)
1565 /* both are connected. are either loop or vrf? */
1566 struct nexthop
*nexthop
= NULL
;
1568 for (ALL_NEXTHOPS(alternate
->ng
, nexthop
)) {
1569 if (if_is_loopback_or_vrf(if_lookup_by_index(
1570 nexthop
->ifindex
, alternate
->vrf_id
)))
1574 for (ALL_NEXTHOPS(current
->ng
, nexthop
)) {
1575 if (if_is_loopback_or_vrf(if_lookup_by_index(
1576 nexthop
->ifindex
, current
->vrf_id
)))
1580 /* Neither are loop or vrf so pick best metric */
1581 if (alternate
->metric
<= current
->metric
)
1587 if (current
->type
== ZEBRA_ROUTE_CONNECT
)
1590 /* higher distance loses */
1591 if (alternate
->distance
< current
->distance
)
1593 if (current
->distance
< alternate
->distance
)
1596 /* metric tie-breaks equal distance */
1597 if (alternate
->metric
<= current
->metric
)
1603 /* Core function for processing routing information base. */
1604 static void rib_process(struct route_node
*rn
)
1606 struct route_entry
*re
;
1607 struct route_entry
*next
;
1608 struct route_entry
*old_selected
= NULL
;
1609 struct route_entry
*new_selected
= NULL
;
1610 struct route_entry
*old_fib
= NULL
;
1611 struct route_entry
*new_fib
= NULL
;
1612 struct route_entry
*best
= NULL
;
1613 char buf
[SRCDEST2STR_BUFFER
];
1615 struct zebra_vrf
*zvrf
= NULL
;
1616 const struct prefix
*p
, *src_p
;
1618 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1619 vrf_id_t vrf_id
= VRF_UNKNOWN
;
1623 dest
= rib_dest_from_rnode(rn
);
1625 zvrf
= rib_dest_vrf(dest
);
1626 vrf_id
= zvrf_id(zvrf
);
1629 if (IS_ZEBRA_DEBUG_RIB
)
1630 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1632 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1633 zlog_debug("%u:%s: Processing rn %p", vrf_id
, buf
, rn
);
1636 * we can have rn's that have a NULL info pointer
1637 * (dest). As such let's not let the deref happen
1638 * additionally we know RNODE_FOREACH_RE_SAFE
1639 * will not iterate so we are ok.
1642 old_fib
= dest
->selected_fib
;
1644 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
1645 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1647 "%u:%s: Examine re %p (type %d) status %x flags %x "
1648 "dist %d metric %d",
1649 vrf_id
, buf
, re
, re
->type
, re
->status
,
1650 re
->flags
, re
->distance
, re
->metric
);
1652 UNSET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1654 /* Currently selected re. */
1655 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)) {
1656 assert(old_selected
== NULL
);
1660 /* Skip deleted entries from selection */
1661 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
1664 /* Skip unreachable nexthop. */
1665 /* This first call to nexthop_active_update is merely to
1667 * there's any change to nexthops associated with this RIB
1669 * rib_process() can be invoked due to an external event such as
1671 * down or due to next-hop-tracking evaluation. In the latter
1673 * a decision has already been made that the NHs have changed.
1675 * need to invoke a potentially expensive call again. Further,
1677 * the change might be in a recursive NH which is not caught in
1678 * the nexthop_active_update() code. Thus, we might miss changes
1682 if (!CHECK_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
)
1683 && !nexthop_active_update(rn
, re
, 0)) {
1684 if (re
->type
== ZEBRA_ROUTE_TABLE
) {
1685 /* XXX: HERE BE DRAGONS!!!!!
1686 * In all honesty, I have not yet figured out
1688 * does or why the ROUTE_ENTRY_CHANGED test
1690 * or why we need to delete a route here, and
1692 * this concerns both selected and fib route, or
1695 /* This entry was denied by the 'ip protocol
1696 * table' route-map, we
1697 * need to delete it */
1698 if (re
!= old_selected
) {
1699 if (IS_ZEBRA_DEBUG_RIB
)
1701 "%s: %u:%s: imported via import-table but denied "
1702 "by the ip protocol table route-map",
1703 __func__
, vrf_id
, buf
);
1706 SET_FLAG(re
->status
,
1707 ROUTE_ENTRY_REMOVED
);
1713 /* Infinite distance. */
1714 if (re
->distance
== DISTANCE_INFINITY
) {
1715 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1719 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_FIB_OVERRIDE
)) {
1720 best
= rib_choose_best(new_fib
, re
);
1721 if (new_fib
&& best
!= new_fib
)
1722 UNSET_FLAG(new_fib
->status
,
1723 ROUTE_ENTRY_CHANGED
);
1726 best
= rib_choose_best(new_selected
, re
);
1727 if (new_selected
&& best
!= new_selected
)
1728 UNSET_FLAG(new_selected
->status
,
1729 ROUTE_ENTRY_CHANGED
);
1730 new_selected
= best
;
1733 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1734 } /* RNODE_FOREACH_RE */
1736 /* If no FIB override route, use the selected route also for FIB */
1737 if (new_fib
== NULL
)
1738 new_fib
= new_selected
;
1740 /* After the cycle is finished, the following pointers will be set:
1741 * old_selected --- RE entry currently having SELECTED
1742 * new_selected --- RE entry that is newly SELECTED
1743 * old_fib --- RE entry currently in kernel FIB
1744 * new_fib --- RE entry that is newly to be in kernel FIB
1746 * new_selected will get SELECTED flag, and is going to be redistributed
1747 * the zclients. new_fib (which can be new_selected) will be installed
1751 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
1753 "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p",
1754 vrf_id
, buf
, (void *)old_selected
, (void *)new_selected
,
1755 (void *)old_fib
, (void *)new_fib
);
1758 /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if
1759 * fib == selected */
1760 bool selected_changed
= new_selected
&& CHECK_FLAG(new_selected
->status
,
1761 ROUTE_ENTRY_CHANGED
);
1763 /* Update fib according to selection results */
1764 if (new_fib
&& old_fib
)
1765 rib_process_update_fib(zvrf
, rn
, old_fib
, new_fib
);
1767 rib_process_add_fib(zvrf
, rn
, new_fib
);
1769 rib_process_del_fib(zvrf
, rn
, old_fib
);
1771 /* Update SELECTED entry */
1772 if (old_selected
!= new_selected
|| selected_changed
) {
1774 if (new_selected
&& new_selected
!= new_fib
) {
1775 nexthop_active_update(rn
, new_selected
, 1);
1776 UNSET_FLAG(new_selected
->status
, ROUTE_ENTRY_CHANGED
);
1780 SET_FLAG(new_selected
->flags
, ZEBRA_FLAG_SELECTED
);
1782 /* Special case: new route is system route, so
1783 * dataplane update will not be done - ensure we
1784 * redistribute the route.
1786 if (RIB_SYSTEM_ROUTE(new_selected
))
1787 redistribute_update(p
, src_p
, new_selected
,
1793 redistribute_delete(p
, src_p
, old_selected
);
1794 if (old_selected
!= new_selected
)
1795 UNSET_FLAG(old_selected
->flags
,
1796 ZEBRA_FLAG_SELECTED
);
1800 /* Remove all RE entries queued for removal */
1801 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
1802 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
1803 if (IS_ZEBRA_DEBUG_RIB
) {
1804 rnode_debug(rn
, vrf_id
, "rn %p, removing re %p",
1805 (void *)rn
, (void *)re
);
1812 * Check if the dest can be deleted now.
1818 * Utility to match route with dplane context data
1820 static bool rib_route_match_ctx(const struct route_entry
*re
,
1821 const dplane_ctx_h ctx
, bool is_update
)
1823 bool result
= false;
1827 * In 'update' case, we test info about the 'previous' or
1830 if ((re
->type
== dplane_ctx_get_old_type(ctx
)) &&
1831 (re
->instance
== dplane_ctx_get_old_instance(ctx
))) {
1834 /* TODO -- we're using this extra test, but it's not
1835 * exactly clear why.
1837 if (re
->type
== ZEBRA_ROUTE_STATIC
&&
1838 (re
->distance
!= dplane_ctx_get_old_distance(ctx
) ||
1839 re
->tag
!= dplane_ctx_get_old_tag(ctx
))) {
1846 * Ordinary, single-route case using primary context info
1848 if ((dplane_ctx_get_op(ctx
) != DPLANE_OP_ROUTE_DELETE
) &&
1849 CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
1850 /* Skip route that's been deleted */
1854 if ((re
->type
== dplane_ctx_get_type(ctx
)) &&
1855 (re
->instance
== dplane_ctx_get_instance(ctx
))) {
1858 /* TODO -- we're using this extra test, but it's not
1859 * exactly clear why.
1861 if (re
->type
== ZEBRA_ROUTE_STATIC
&&
1862 (re
->distance
!= dplane_ctx_get_distance(ctx
) ||
1863 re
->tag
!= dplane_ctx_get_tag(ctx
))) {
1875 * TODO - WIP version of route-update processing after async dataplane
1878 static void rib_process_after(dplane_ctx_h ctx
)
1880 struct route_table
*table
= NULL
;
1881 struct zebra_vrf
*zvrf
= NULL
;
1882 struct route_node
*rn
= NULL
;
1883 struct route_entry
*re
= NULL
, *old_re
= NULL
, *rib
;
1884 bool is_update
= false;
1885 struct nexthop
*nexthop
, *ctx_nexthop
;
1886 char dest_str
[PREFIX_STRLEN
] = "";
1887 enum dplane_op_e op
;
1888 enum zebra_dplane_result status
;
1889 const struct prefix
*dest_pfx
, *src_pfx
;
1891 /* Locate rn and re(s) from ctx */
1893 table
= zebra_vrf_table_with_table_id(dplane_ctx_get_afi(ctx
),
1894 dplane_ctx_get_safi(ctx
),
1895 dplane_ctx_get_vrf(ctx
),
1896 dplane_ctx_get_table(ctx
));
1897 if (table
== NULL
) {
1898 if (IS_ZEBRA_DEBUG_DPLANE
) {
1899 zlog_debug("Failed to process dplane results: no table for afi %d, safi %d, vrf %u",
1900 dplane_ctx_get_afi(ctx
),
1901 dplane_ctx_get_safi(ctx
),
1902 dplane_ctx_get_vrf(ctx
));
1907 zvrf
= vrf_info_lookup(dplane_ctx_get_vrf(ctx
));
1909 dest_pfx
= dplane_ctx_get_dest(ctx
);
1911 /* Note well: only capturing the prefix string if debug is enabled here;
1912 * unconditional log messages will have to generate the string.
1914 if (IS_ZEBRA_DEBUG_DPLANE
)
1915 prefix2str(dest_pfx
, dest_str
, sizeof(dest_str
));
1917 src_pfx
= dplane_ctx_get_src(ctx
);
1918 rn
= srcdest_rnode_get(table
, dplane_ctx_get_dest(ctx
),
1919 src_pfx
? (struct prefix_ipv6
*)src_pfx
: NULL
);
1921 if (IS_ZEBRA_DEBUG_DPLANE
) {
1922 zlog_debug("Failed to process dplane results: no route for %u:%s",
1923 dplane_ctx_get_vrf(ctx
), dest_str
);
1928 srcdest_rnode_prefixes(rn
, &dest_pfx
, &src_pfx
);
1930 op
= dplane_ctx_get_op(ctx
);
1931 status
= dplane_ctx_get_status(ctx
);
1933 if (IS_ZEBRA_DEBUG_DPLANE_DETAIL
) {
1934 zlog_debug("%u:%s Processing dplane ctx %p, op %s result %s",
1935 dplane_ctx_get_vrf(ctx
), dest_str
, ctx
,
1936 dplane_op2str(op
), dplane_res2str(status
));
1939 if (op
== DPLANE_OP_ROUTE_DELETE
) {
1941 * In the delete case, the zebra core datastructs were
1942 * updated (or removed) at the time the delete was issued,
1943 * so we're just notifying the route owner.
1945 if (status
== ZEBRA_DPLANE_REQUEST_SUCCESS
) {
1946 zsend_route_notify_owner_ctx(ctx
, ZAPI_ROUTE_REMOVED
);
1951 zsend_route_notify_owner_ctx(ctx
,
1952 ZAPI_ROUTE_FAIL_INSTALL
);
1954 zlog_warn("%u:%s: Route Deletion failure",
1955 dplane_ctx_get_vrf(ctx
),
1956 prefix2str(dest_pfx
,
1957 dest_str
, sizeof(dest_str
)));
1960 /* Nothing more to do in delete case */
1965 * Update is a bit of a special case, where we may have both old and new
1966 * routes to post-process.
1968 is_update
= dplane_ctx_is_update(ctx
);
1971 * Take a pass through the routes, look for matches with the context
1974 RNODE_FOREACH_RE(rn
, rib
) {
1977 if (rib_route_match_ctx(rib
, ctx
, false))
1981 /* Check for old route match */
1982 if (is_update
&& (old_re
== NULL
)) {
1983 if (rib_route_match_ctx(rib
, ctx
, true /*is_update*/))
1987 /* Have we found the routes we need to work on? */
1988 if (re
&& ((!is_update
|| old_re
)))
1993 * Check sequence number(s) to detect stale results before continuing
1995 if (re
&& (re
->dplane_sequence
!= dplane_ctx_get_seq(ctx
))) {
1996 if (IS_ZEBRA_DEBUG_DPLANE_DETAIL
) {
1997 zlog_debug("%u:%s Stale dplane result for re %p",
1998 dplane_ctx_get_vrf(ctx
), dest_str
, re
);
2004 (old_re
->dplane_sequence
!= dplane_ctx_get_old_seq(ctx
))) {
2005 if (IS_ZEBRA_DEBUG_DPLANE_DETAIL
) {
2006 zlog_debug("%u:%s Stale dplane result for old_re %p",
2007 dplane_ctx_get_vrf(ctx
), dest_str
, old_re
);
2013 * Here's sort of a tough one: the route update result is stale.
2014 * Is it better to use the context block info to generate
2015 * redist and owner notification, or is it better to wait
2016 * for the up-to-date result to arrive?
2019 /* TODO -- for now, only expose up-to-date results */
2023 if (status
== ZEBRA_DPLANE_REQUEST_SUCCESS
) {
2024 /* Update zebra nexthop FIB flag for each
2025 * nexthop that was installed.
2027 for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx
), ctx_nexthop
)) {
2029 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
2030 if (nexthop_same(ctx_nexthop
, nexthop
))
2034 if (nexthop
== NULL
)
2037 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
))
2040 if (CHECK_FLAG(ctx_nexthop
->flags
,
2042 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
2044 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
2049 /* Set flag for nexthop tracking processing */
2050 zvrf
->flags
|= ZEBRA_VRF_RIB_SCHEDULED
;
2054 /* TODO -- still calling the redist api using the route_entries,
2055 * and there's a corner-case here: if there's no client
2056 * for the 'new' route, a redist deleting the 'old' route
2057 * will be sent. But if the 'old' context info was stale,
2058 * 'old_re' will be NULL here and that delete will not be sent.
2060 redistribute_update(dest_pfx
, src_pfx
, re
, old_re
);
2062 /* Notify route owner */
2063 zsend_route_notify_owner(re
,
2064 dest_pfx
, ZAPI_ROUTE_INSTALLED
);
2067 zsend_route_notify_owner(re
, dest_pfx
,
2068 ZAPI_ROUTE_FAIL_INSTALL
);
2070 zlog_warn("%u:%s: Route install failed",
2071 dplane_ctx_get_vrf(ctx
),
2072 prefix2str(dest_pfx
,
2073 dest_str
, sizeof(dest_str
)));
2078 /* Return context to dataplane module */
2079 dplane_ctx_fini(&ctx
);
2082 /* Take a list of route_node structs and return 1, if there was a record
2083 * picked from it and processed by rib_process(). Don't process more,
2084 * than one RN record; operate only in the specified sub-queue.
2086 static unsigned int process_subq(struct list
*subq
, uint8_t qindex
)
2088 struct listnode
*lnode
= listhead(subq
);
2089 struct route_node
*rnode
;
2091 struct zebra_vrf
*zvrf
= NULL
;
2096 rnode
= listgetdata(lnode
);
2097 dest
= rib_dest_from_rnode(rnode
);
2099 zvrf
= rib_dest_vrf(dest
);
2103 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
2104 char buf
[SRCDEST2STR_BUFFER
];
2105 srcdest_rnode2str(rnode
, buf
, sizeof(buf
));
2106 zlog_debug("%u:%s: rn %p dequeued from sub-queue %u",
2107 zvrf
? zvrf_id(zvrf
) : 0, buf
, rnode
, qindex
);
2111 UNSET_FLAG(rib_dest_from_rnode(rnode
)->flags
,
2112 RIB_ROUTE_QUEUED(qindex
));
2117 zlog_debug ("%s: called for route_node (%p, %d) with no ribs",
2118 __func__
, rnode
, rnode
->lock
);
2119 zlog_backtrace(LOG_DEBUG
);
2122 route_unlock_node(rnode
);
2123 list_delete_node(subq
, lnode
);
2128 * Perform next-hop tracking processing after RIB updates.
2130 static void do_nht_processing(void)
2133 struct zebra_vrf
*zvrf
;
2135 /* Evaluate nexthops for those VRFs which underwent route processing.
2137 * should limit the evaluation to the necessary VRFs in most common
2140 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
2142 if (zvrf
== NULL
|| !(zvrf
->flags
& ZEBRA_VRF_RIB_SCHEDULED
))
2145 if (IS_ZEBRA_DEBUG_RIB_DETAILED
|| IS_ZEBRA_DEBUG_NHT
)
2146 zlog_debug("NHT processing check for zvrf %s",
2149 zvrf
->flags
&= ~ZEBRA_VRF_RIB_SCHEDULED
;
2150 zebra_evaluate_rnh(zvrf
, AF_INET
, 0, RNH_NEXTHOP_TYPE
, NULL
);
2151 zebra_evaluate_rnh(zvrf
, AF_INET
, 0, RNH_IMPORT_CHECK_TYPE
,
2153 zebra_evaluate_rnh(zvrf
, AF_INET6
, 0, RNH_NEXTHOP_TYPE
, NULL
);
2154 zebra_evaluate_rnh(zvrf
, AF_INET6
, 0, RNH_IMPORT_CHECK_TYPE
,
2158 /* Schedule LSPs for processing, if needed. */
2159 zvrf
= vrf_info_lookup(VRF_DEFAULT
);
2160 if (mpls_should_lsps_be_processed(zvrf
)) {
2161 if (IS_ZEBRA_DEBUG_MPLS
)
2163 "%u: Scheduling all LSPs upon RIB completion",
2165 zebra_mpls_lsp_schedule(zvrf
);
2166 mpls_unmark_lsps_for_processing(zvrf
);
2171 * All meta queues have been processed. Trigger next-hop evaluation.
2173 static void meta_queue_process_complete(struct work_queue
*dummy
)
2175 do_nht_processing();
2178 /* Dispatch the meta queue by picking, processing and unlocking the next RN from
2179 * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and
2181 * is pointed to the meta queue structure.
2183 static wq_item_status
meta_queue_process(struct work_queue
*dummy
, void *data
)
2185 struct meta_queue
*mq
= data
;
2188 for (i
= 0; i
< MQ_SIZE
; i
++)
2189 if (process_subq(mq
->subq
[i
], i
)) {
2193 return mq
->size
? WQ_REQUEUE
: WQ_SUCCESS
;
2197 * Map from rib types to queue type (priority) in meta queue
2199 static const uint8_t meta_queue_map
[ZEBRA_ROUTE_MAX
] = {
2200 [ZEBRA_ROUTE_SYSTEM
] = 4,
2201 [ZEBRA_ROUTE_KERNEL
] = 0,
2202 [ZEBRA_ROUTE_CONNECT
] = 0,
2203 [ZEBRA_ROUTE_STATIC
] = 1,
2204 [ZEBRA_ROUTE_RIP
] = 2,
2205 [ZEBRA_ROUTE_RIPNG
] = 2,
2206 [ZEBRA_ROUTE_OSPF
] = 2,
2207 [ZEBRA_ROUTE_OSPF6
] = 2,
2208 [ZEBRA_ROUTE_ISIS
] = 2,
2209 [ZEBRA_ROUTE_BGP
] = 3,
2210 [ZEBRA_ROUTE_PIM
] = 4, // Shouldn't happen but for safety
2211 [ZEBRA_ROUTE_EIGRP
] = 2,
2212 [ZEBRA_ROUTE_NHRP
] = 2,
2213 [ZEBRA_ROUTE_HSLS
] = 4,
2214 [ZEBRA_ROUTE_OLSR
] = 4,
2215 [ZEBRA_ROUTE_TABLE
] = 1,
2216 [ZEBRA_ROUTE_LDP
] = 4,
2217 [ZEBRA_ROUTE_VNC
] = 3,
2218 [ZEBRA_ROUTE_VNC_DIRECT
] = 3,
2219 [ZEBRA_ROUTE_VNC_DIRECT_RH
] = 3,
2220 [ZEBRA_ROUTE_BGP_DIRECT
] = 3,
2221 [ZEBRA_ROUTE_BGP_DIRECT_EXT
] = 3,
2222 [ZEBRA_ROUTE_BABEL
] = 2,
2223 [ZEBRA_ROUTE_ALL
] = 4, // Shouldn't happen but for safety
2226 /* Look into the RN and queue it into one or more priority queues,
2227 * increasing the size for each data push done.
2229 static void rib_meta_queue_add(struct meta_queue
*mq
, struct route_node
*rn
)
2231 struct route_entry
*re
;
2233 RNODE_FOREACH_RE (rn
, re
) {
2234 uint8_t qindex
= meta_queue_map
[re
->type
];
2235 struct zebra_vrf
*zvrf
;
2237 /* Invariant: at this point we always have rn->info set. */
2238 if (CHECK_FLAG(rib_dest_from_rnode(rn
)->flags
,
2239 RIB_ROUTE_QUEUED(qindex
))) {
2240 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2243 "rn %p is already queued in sub-queue %u",
2244 (void *)rn
, qindex
);
2248 SET_FLAG(rib_dest_from_rnode(rn
)->flags
,
2249 RIB_ROUTE_QUEUED(qindex
));
2250 listnode_add(mq
->subq
[qindex
], rn
);
2251 route_lock_node(rn
);
2254 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2255 rnode_debug(rn
, re
->vrf_id
,
2256 "queued rn %p into sub-queue %u",
2257 (void *)rn
, qindex
);
2259 zvrf
= zebra_vrf_lookup_by_id(re
->vrf_id
);
2261 zvrf
->flags
|= ZEBRA_VRF_RIB_SCHEDULED
;
2265 /* Add route_node to work queue and schedule processing */
2266 void rib_queue_add(struct route_node
*rn
)
2270 /* Pointless to queue a route_node with no RIB entries to add or remove
2272 if (!rnode_to_ribs(rn
)) {
2273 zlog_debug("%s: called for route_node (%p, %d) with no ribs",
2274 __func__
, (void *)rn
, rn
->lock
);
2275 zlog_backtrace(LOG_DEBUG
);
2279 if (zebrad
.ribq
== NULL
) {
2280 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2281 "%s: work_queue does not exist!", __func__
);
2286 * The RIB queue should normally be either empty or holding the only
2287 * work_queue_item element. In the latter case this element would
2288 * hold a pointer to the meta queue structure, which must be used to
2289 * actually queue the route nodes to process. So create the MQ
2290 * holder, if necessary, then push the work into it in any case.
2291 * This semantics was introduced after 0.99.9 release.
2293 if (work_queue_empty(zebrad
.ribq
))
2294 work_queue_add(zebrad
.ribq
, zebrad
.mq
);
2296 rib_meta_queue_add(zebrad
.mq
, rn
);
2301 /* Create new meta queue.
2302 A destructor function doesn't seem to be necessary here.
2304 static struct meta_queue
*meta_queue_new(void)
2306 struct meta_queue
*new;
2309 new = XCALLOC(MTYPE_WORK_QUEUE
, sizeof(struct meta_queue
));
2311 for (i
= 0; i
< MQ_SIZE
; i
++) {
2312 new->subq
[i
] = list_new();
2313 assert(new->subq
[i
]);
2319 void meta_queue_free(struct meta_queue
*mq
)
2323 for (i
= 0; i
< MQ_SIZE
; i
++)
2324 list_delete(&mq
->subq
[i
]);
2326 XFREE(MTYPE_WORK_QUEUE
, mq
);
2329 /* initialise zebra rib work queue */
2330 static void rib_queue_init(struct zebra_t
*zebra
)
2335 work_queue_new(zebra
->master
, "route_node processing"))) {
2336 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2337 "%s: could not initialise work queue!", __func__
);
2341 /* fill in the work queue spec */
2342 zebra
->ribq
->spec
.workfunc
= &meta_queue_process
;
2343 zebra
->ribq
->spec
.errorfunc
= NULL
;
2344 zebra
->ribq
->spec
.completion_func
= &meta_queue_process_complete
;
2345 /* XXX: TODO: These should be runtime configurable via vty */
2346 zebra
->ribq
->spec
.max_retries
= 3;
2347 zebra
->ribq
->spec
.hold
= ZEBRA_RIB_PROCESS_HOLD_TIME
;
2349 if (!(zebra
->mq
= meta_queue_new())) {
2350 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2351 "%s: could not initialise meta queue!", __func__
);
2357 /* RIB updates are processed via a queue of pointers to route_nodes.
2359 * The queue length is bounded by the maximal size of the routing table,
2360 * as a route_node will not be requeued, if already queued.
2362 * REs are submitted via rib_addnode or rib_delnode which set minimal
2363 * state, or static_install_route (when an existing RE is updated)
2364 * and then submit route_node to queue for best-path selection later.
2365 * Order of add/delete state changes are preserved for any given RE.
2367 * Deleted REs are reaped during best-path selection.
2370 * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with
2371 * |-------->| | best RE, if required
2373 * static_install->|->rib_addqueue...... -> rib_process
2375 * |-------->| |-> rib_unlink
2376 * |-> set ROUTE_ENTRY_REMOVE |
2377 * rib_delnode (RE freed)
2379 * The 'info' pointer of a route_node points to a rib_dest_t
2380 * ('dest'). Queueing state for a route_node is kept on the dest. The
2381 * dest is created on-demand by rib_link() and is kept around at least
2382 * as long as there are ribs hanging off it (@see rib_gc_dest()).
2384 * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code):
2386 * - route_nodes: refcounted by:
2387 * - dest attached to route_node:
2388 * - managed by: rib_link/rib_gc_dest
2389 * - route_node processing queue
2390 * - managed by: rib_addqueue, rib_process.
2394 /* Add RE to head of the route node. */
2395 static void rib_link(struct route_node
*rn
, struct route_entry
*re
, int process
)
2397 struct route_entry
*head
;
2400 const char *rmap_name
;
2404 dest
= rib_dest_from_rnode(rn
);
2406 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2407 rnode_debug(rn
, re
->vrf_id
, "rn %p adding dest", rn
);
2409 dest
= XCALLOC(MTYPE_RIB_DEST
, sizeof(rib_dest_t
));
2410 route_lock_node(rn
); /* rn route table reference */
2415 head
= dest
->routes
;
2422 afi
= (rn
->p
.family
== AF_INET
)
2424 : (rn
->p
.family
== AF_INET6
) ? AFI_IP6
: AFI_MAX
;
2425 if (is_zebra_import_table_enabled(afi
, re
->table
)) {
2426 rmap_name
= zebra_get_import_table_route_map(afi
, re
->table
);
2427 zebra_add_import_table_entry(rn
, re
, rmap_name
);
2432 static void rib_addnode(struct route_node
*rn
,
2433 struct route_entry
*re
, int process
)
2435 /* RE node has been un-removed before route-node is processed.
2436 * route_node must hence already be on the queue for processing..
2438 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
2439 if (IS_ZEBRA_DEBUG_RIB
)
2440 rnode_debug(rn
, re
->vrf_id
, "rn %p, un-removed re %p",
2441 (void *)rn
, (void *)re
);
2443 UNSET_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
);
2446 rib_link(rn
, re
, process
);
2452 * Detach a rib structure from a route_node.
2454 * Note that a call to rib_unlink() should be followed by a call to
2455 * rib_gc_dest() at some point. This allows a rib_dest_t that is no
2456 * longer required to be deleted.
2458 void rib_unlink(struct route_node
*rn
, struct route_entry
*re
)
2464 if (IS_ZEBRA_DEBUG_RIB
)
2465 rnode_debug(rn
, re
->vrf_id
, "rn %p, re %p", (void *)rn
,
2468 dest
= rib_dest_from_rnode(rn
);
2471 re
->next
->prev
= re
->prev
;
2474 re
->prev
->next
= re
->next
;
2476 dest
->routes
= re
->next
;
2479 if (dest
->selected_fib
== re
)
2480 dest
->selected_fib
= NULL
;
2482 nexthops_free(re
->ng
.nexthop
);
2483 XFREE(MTYPE_RE
, re
);
2486 void rib_delnode(struct route_node
*rn
, struct route_entry
*re
)
2490 if (IS_ZEBRA_DEBUG_RIB
)
2491 rnode_debug(rn
, re
->vrf_id
, "rn %p, re %p, removing",
2492 (void *)rn
, (void *)re
);
2493 SET_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
);
2495 afi
= (rn
->p
.family
== AF_INET
)
2497 : (rn
->p
.family
== AF_INET6
) ? AFI_IP6
: AFI_MAX
;
2498 if (is_zebra_import_table_enabled(afi
, re
->table
)) {
2499 zebra_del_import_table_entry(rn
, re
);
2500 /* Just clean up if non main table */
2501 if (IS_ZEBRA_DEBUG_RIB
) {
2502 char buf
[SRCDEST2STR_BUFFER
];
2503 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
2505 "%u:%s: Freeing route rn %p, re %p (type %d)",
2506 re
->vrf_id
, buf
, rn
, re
, re
->type
);
2515 /* This function dumps the contents of a given RE entry into
2516 * standard debug log. Calling function name and IP prefix in
2517 * question are passed as 1st and 2nd arguments.
2520 void _route_entry_dump(const char *func
, union prefixconstptr pp
,
2521 union prefixconstptr src_pp
,
2522 const struct route_entry
*re
)
2524 const struct prefix
*src_p
= src_pp
.p
;
2525 bool is_srcdst
= src_p
&& src_p
->prefixlen
;
2526 char straddr
[PREFIX_STRLEN
];
2527 char srcaddr
[PREFIX_STRLEN
];
2528 struct nexthop
*nexthop
;
2530 zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func
,
2531 (const void *)re
, prefix2str(pp
, straddr
, sizeof(straddr
)),
2532 is_srcdst
? " from " : "",
2533 is_srcdst
? prefix2str(src_pp
, srcaddr
, sizeof(srcaddr
))
2536 zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d",
2537 func
, (unsigned long)re
->uptime
, re
->type
, re
->instance
,
2540 "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u",
2541 func
, re
->metric
, re
->mtu
, re
->distance
, re
->flags
, re
->status
);
2542 zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func
,
2543 re
->nexthop_num
, re
->nexthop_active_num
);
2545 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
2546 struct interface
*ifp
;
2547 struct vrf
*vrf
= vrf_lookup_by_id(nexthop
->vrf_id
);
2549 switch (nexthop
->type
) {
2550 case NEXTHOP_TYPE_BLACKHOLE
:
2551 sprintf(straddr
, "Blackhole");
2553 case NEXTHOP_TYPE_IFINDEX
:
2554 ifp
= if_lookup_by_index(nexthop
->ifindex
,
2556 sprintf(straddr
, "%s", ifp
? ifp
->name
: "Unknown");
2558 case NEXTHOP_TYPE_IPV4
:
2560 case NEXTHOP_TYPE_IPV4_IFINDEX
:
2561 inet_ntop(AF_INET
, &nexthop
->gate
, straddr
,
2564 case NEXTHOP_TYPE_IPV6
:
2565 case NEXTHOP_TYPE_IPV6_IFINDEX
:
2566 inet_ntop(AF_INET6
, &nexthop
->gate
, straddr
,
2570 zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s", func
,
2571 (nexthop
->rparent
? " NH" : "NH"), straddr
,
2572 nexthop
->ifindex
, vrf
? vrf
->name
: "Unknown",
2574 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
)
2577 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
)
2580 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
)
2584 zlog_debug("%s: dump complete", func
);
2587 /* This is an exported helper to rtm_read() to dump the strange
2588 * RE entry found by rib_lookup_ipv4_route()
2591 void rib_lookup_and_dump(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
2593 struct route_table
*table
;
2594 struct route_node
*rn
;
2595 struct route_entry
*re
;
2596 char prefix_buf
[INET_ADDRSTRLEN
];
2599 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
2601 flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED
,
2602 "%s:%u zebra_vrf_table() returned NULL", __func__
,
2607 /* Scan the RIB table for exactly matching RE entry. */
2608 rn
= route_node_lookup(table
, (struct prefix
*)p
);
2610 /* No route for this prefix. */
2612 zlog_debug("%s:%u lookup failed for %s", __func__
, vrf_id
,
2613 prefix2str((struct prefix
*)p
, prefix_buf
,
2614 sizeof(prefix_buf
)));
2619 route_unlock_node(rn
);
2622 RNODE_FOREACH_RE (rn
, re
) {
2623 zlog_debug("%s:%u rn %p, re %p: %s, %s",
2625 (void *)rn
, (void *)re
,
2626 (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)
2629 (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)
2632 route_entry_dump(p
, NULL
, re
);
2636 /* Check if requested address assignment will fail due to another
2637 * route being installed by zebra in FIB already. Take necessary
2638 * actions, if needed: remove such a route from FIB and deSELECT
2639 * corresponding RE entry. Then put affected RN into RIBQ head.
2641 void rib_lookup_and_pushup(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
2643 struct route_table
*table
;
2644 struct route_node
*rn
;
2645 unsigned changed
= 0;
2648 if (NULL
== (table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
))) {
2649 flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED
,
2650 "%s:%u zebra_vrf_table() returned NULL", __func__
,
2655 /* No matches would be the simplest case. */
2656 if (NULL
== (rn
= route_node_lookup(table
, (struct prefix
*)p
)))
2660 route_unlock_node(rn
);
2662 dest
= rib_dest_from_rnode(rn
);
2663 /* Check all RE entries. In case any changes have to be done, requeue
2664 * the RN into RIBQ head. If the routing message about the new connected
2665 * route (generated by the IP address we are going to assign very soon)
2666 * comes before the RIBQ is processed, the new RE entry will join
2667 * RIBQ record already on head. This is necessary for proper
2669 * of the rest of the RE.
2671 if (dest
->selected_fib
&& !RIB_SYSTEM_ROUTE(dest
->selected_fib
)) {
2673 if (IS_ZEBRA_DEBUG_RIB
) {
2674 char buf
[PREFIX_STRLEN
];
2676 zlog_debug("%u:%s: freeing way for connected prefix",
2677 dest
->selected_fib
->vrf_id
,
2678 prefix2str(&rn
->p
, buf
, sizeof(buf
)));
2679 route_entry_dump(&rn
->p
, NULL
, dest
->selected_fib
);
2681 rib_uninstall(rn
, dest
->selected_fib
);
2687 int rib_add_multipath(afi_t afi
, safi_t safi
, struct prefix
*p
,
2688 struct prefix_ipv6
*src_p
, struct route_entry
*re
)
2690 struct route_table
*table
;
2691 struct route_node
*rn
;
2692 struct route_entry
*same
= NULL
;
2693 struct nexthop
*nexthop
;
2699 assert(!src_p
|| !src_p
->prefixlen
|| afi
== AFI_IP6
);
2702 table
= zebra_vrf_table_with_table_id(afi
, safi
, re
->vrf_id
, re
->table
);
2704 XFREE(MTYPE_RE
, re
);
2708 /* Make it sure prefixlen is applied to the prefix. */
2711 apply_mask_ipv6(src_p
);
2713 /* Set default distance by route type. */
2714 if (re
->distance
== 0) {
2715 re
->distance
= route_distance(re
->type
);
2717 /* iBGP distance is 200. */
2718 if (re
->type
== ZEBRA_ROUTE_BGP
2719 && CHECK_FLAG(re
->flags
, ZEBRA_FLAG_IBGP
))
2723 /* Lookup route node.*/
2724 rn
= srcdest_rnode_get(table
, p
, src_p
);
2727 * If same type of route are installed, treat it as a implicit
2729 * If the user has specified the No route replace semantics
2730 * for the install don't do a route replace.
2732 RNODE_FOREACH_RE (rn
, same
) {
2733 if (CHECK_FLAG(same
->status
, ROUTE_ENTRY_REMOVED
))
2736 if (same
->type
!= re
->type
)
2738 if (same
->instance
!= re
->instance
)
2740 if (same
->type
== ZEBRA_ROUTE_KERNEL
2741 && same
->metric
!= re
->metric
)
2744 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_RR_USE_DISTANCE
) &&
2745 same
->distance
!= re
->distance
)
2749 * We should allow duplicate connected routes
2750 * because of IPv6 link-local routes and unnumbered
2751 * interfaces on Linux.
2753 if (same
->type
!= ZEBRA_ROUTE_CONNECT
)
2757 /* If this route is kernel route, set FIB flag to the route. */
2758 if (RIB_SYSTEM_ROUTE(re
))
2759 for (nexthop
= re
->ng
.nexthop
; nexthop
; nexthop
= nexthop
->next
)
2760 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
2762 /* Link new re to node.*/
2763 if (IS_ZEBRA_DEBUG_RIB
) {
2766 "Inserting route rn %p, re %p (type %d) existing %p",
2767 (void *)rn
, (void *)re
, re
->type
, (void *)same
);
2769 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2770 route_entry_dump(p
, src_p
, re
);
2772 rib_addnode(rn
, re
, 1);
2775 /* Free implicit route.*/
2777 rib_delnode(rn
, same
);
2781 route_unlock_node(rn
);
2785 void rib_delete(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
, int type
,
2786 unsigned short instance
, int flags
, struct prefix
*p
,
2787 struct prefix_ipv6
*src_p
, const struct nexthop
*nh
,
2788 uint32_t table_id
, uint32_t metric
, uint8_t distance
,
2791 struct route_table
*table
;
2792 struct route_node
*rn
;
2793 struct route_entry
*re
;
2794 struct route_entry
*fib
= NULL
;
2795 struct route_entry
*same
= NULL
;
2796 struct nexthop
*rtnh
;
2797 char buf2
[INET6_ADDRSTRLEN
];
2800 assert(!src_p
|| !src_p
->prefixlen
|| afi
== AFI_IP6
);
2803 table
= zebra_vrf_table_with_table_id(afi
, safi
, vrf_id
, table_id
);
2810 apply_mask_ipv6(src_p
);
2812 /* Lookup route node. */
2813 rn
= srcdest_rnode_lookup(table
, p
, src_p
);
2815 char dst_buf
[PREFIX_STRLEN
], src_buf
[PREFIX_STRLEN
];
2817 prefix2str(p
, dst_buf
, sizeof(dst_buf
));
2818 if (src_p
&& src_p
->prefixlen
)
2819 prefix2str(src_p
, src_buf
, sizeof(src_buf
));
2823 if (IS_ZEBRA_DEBUG_RIB
)
2824 zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id
,
2826 (src_buf
[0] != '\0') ? " from " : "",
2831 dest
= rib_dest_from_rnode(rn
);
2832 fib
= dest
->selected_fib
;
2834 /* Lookup same type route. */
2835 RNODE_FOREACH_RE (rn
, re
) {
2836 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
2839 if (re
->type
!= type
)
2841 if (re
->instance
!= instance
)
2843 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_RR_USE_DISTANCE
) &&
2844 distance
!= re
->distance
)
2847 if (re
->type
== ZEBRA_ROUTE_KERNEL
&& re
->metric
!= metric
)
2849 if (re
->type
== ZEBRA_ROUTE_CONNECT
&& (rtnh
= re
->ng
.nexthop
)
2850 && rtnh
->type
== NEXTHOP_TYPE_IFINDEX
&& nh
) {
2851 if (rtnh
->ifindex
!= nh
->ifindex
)
2856 /* Make sure that the route found has the same gateway. */
2862 for (ALL_NEXTHOPS(re
->ng
, rtnh
))
2863 if (nexthop_same_no_recurse(rtnh
, nh
)) {
2871 /* If same type of route can't be found and this message is from
2875 * In the past(HA!) we could get here because
2876 * we were receiving a route delete from the
2877 * kernel and we're not marking the proto
2878 * as coming from it's appropriate originator.
2879 * Now that we are properly noticing the fact
2880 * that the kernel has deleted our route we
2881 * are not going to get called in this path
2882 * I am going to leave this here because
2883 * this might still work this way on non-linux
2884 * platforms as well as some weird state I have
2885 * not properly thought of yet.
2886 * If we can show that this code path is
2887 * dead then we can remove it.
2889 if (fib
&& CHECK_FLAG(flags
, ZEBRA_FLAG_SELFROUTE
)) {
2890 if (IS_ZEBRA_DEBUG_RIB
) {
2893 "rn %p, re %p (type %d) was deleted from kernel, adding",
2894 rn
, fib
, fib
->type
);
2898 for (rtnh
= fib
->ng
.nexthop
; rtnh
;
2900 UNSET_FLAG(rtnh
->flags
,
2904 * This is a non FRR route
2905 * as such we should mark
2908 dest
->selected_fib
= NULL
;
2910 /* This means someone else, other than Zebra,
2912 * a Zebra router from the kernel. We will add
2914 rib_install_kernel(rn
, fib
, NULL
);
2917 if (IS_ZEBRA_DEBUG_RIB
) {
2921 "via %s ifindex %d type %d "
2922 "doesn't exist in rib",
2923 inet_ntop(afi2family(afi
),
2930 "type %d doesn't exist in rib",
2933 route_unlock_node(rn
);
2939 if (fromkernel
&& CHECK_FLAG(flags
, ZEBRA_FLAG_SELFROUTE
)
2941 rib_install_kernel(rn
, same
, NULL
);
2942 route_unlock_node(rn
);
2947 if (CHECK_FLAG(flags
, ZEBRA_FLAG_EVPN_ROUTE
)) {
2948 struct nexthop
*tmp_nh
;
2950 for (ALL_NEXTHOPS(re
->ng
, tmp_nh
)) {
2951 struct ipaddr vtep_ip
;
2953 memset(&vtep_ip
, 0, sizeof(struct ipaddr
));
2954 if (afi
== AFI_IP
) {
2955 vtep_ip
.ipa_type
= IPADDR_V4
;
2956 memcpy(&(vtep_ip
.ipaddr_v4
),
2957 &(tmp_nh
->gate
.ipv4
),
2958 sizeof(struct in_addr
));
2960 vtep_ip
.ipa_type
= IPADDR_V6
;
2961 memcpy(&(vtep_ip
.ipaddr_v6
),
2962 &(tmp_nh
->gate
.ipv6
),
2963 sizeof(struct in6_addr
));
2965 zebra_vxlan_evpn_vrf_route_del(re
->vrf_id
,
2969 rib_delnode(rn
, same
);
2972 route_unlock_node(rn
);
2977 int rib_add(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
, int type
,
2978 unsigned short instance
, int flags
, struct prefix
*p
,
2979 struct prefix_ipv6
*src_p
, const struct nexthop
*nh
,
2980 uint32_t table_id
, uint32_t metric
, uint32_t mtu
, uint8_t distance
,
2983 struct route_entry
*re
;
2984 struct nexthop
*nexthop
;
2986 /* Allocate new route_entry structure. */
2987 re
= XCALLOC(MTYPE_RE
, sizeof(struct route_entry
));
2989 re
->instance
= instance
;
2990 re
->distance
= distance
;
2992 re
->metric
= metric
;
2994 re
->table
= table_id
;
2995 re
->vrf_id
= vrf_id
;
2996 re
->nexthop_num
= 0;
2997 re
->uptime
= time(NULL
);
3001 nexthop
= nexthop_new();
3003 route_entry_nexthop_add(re
, nexthop
);
3005 return rib_add_multipath(afi
, safi
, p
, src_p
, re
);
3008 /* Schedule routes of a particular table (address-family) based on event. */
3009 void rib_update_table(struct route_table
*table
, rib_update_event_t event
)
3011 struct route_node
*rn
;
3012 struct route_entry
*re
, *next
;
3014 /* Walk all routes and queue for processing, if appropriate for
3015 * the trigger event.
3017 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
3019 * If we are looking at a route node and the node
3020 * has already been queued we don't
3021 * need to queue it up again
3023 if (rn
->info
&& CHECK_FLAG(rib_dest_from_rnode(rn
)->flags
,
3024 RIB_ROUTE_ANY_QUEUED
))
3027 case RIB_UPDATE_IF_CHANGE
:
3028 /* Examine all routes that won't get processed by the
3030 * triggered by nexthop evaluation (NHT). This would be
3032 * kernel and certain static routes. Note that NHT will
3034 * triggered upon an interface event as connected routes
3036 * get queued for processing.
3038 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
3041 if (re
->type
!= ZEBRA_ROUTE_SYSTEM
3042 && re
->type
!= ZEBRA_ROUTE_KERNEL
3043 && re
->type
!= ZEBRA_ROUTE_CONNECT
3044 && re
->type
!= ZEBRA_ROUTE_STATIC
)
3047 if (re
->type
!= ZEBRA_ROUTE_STATIC
) {
3052 for (nh
= re
->ng
.nexthop
; nh
; nh
= nh
->next
)
3053 if (!(nh
->type
== NEXTHOP_TYPE_IPV4
3054 || nh
->type
== NEXTHOP_TYPE_IPV6
))
3057 /* If we only have nexthops to a
3066 case RIB_UPDATE_RMAP_CHANGE
:
3067 case RIB_UPDATE_OTHER
:
3068 /* Right now, examine all routes. Can restrict to a
3070 * some cases (TODO).
3072 if (rnode_to_ribs(rn
))
3082 /* RIB update function. */
3083 void rib_update(vrf_id_t vrf_id
, rib_update_event_t event
)
3085 struct route_table
*table
;
3087 /* Process routes of interested address-families. */
3088 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
3090 if (IS_ZEBRA_DEBUG_EVENT
)
3091 zlog_debug("%s : AFI_IP event %d", __func__
, event
);
3092 rib_update_table(table
, event
);
3095 table
= zebra_vrf_table(AFI_IP6
, SAFI_UNICAST
, vrf_id
);
3097 if (IS_ZEBRA_DEBUG_EVENT
)
3098 zlog_debug("%s : AFI_IP6 event %d", __func__
, event
);
3099 rib_update_table(table
, event
);
3103 /* Delete self installed routes after zebra is relaunched. */
3104 void rib_sweep_table(struct route_table
*table
)
3106 struct route_node
*rn
;
3107 struct route_entry
*re
;
3108 struct route_entry
*next
;
3109 struct nexthop
*nexthop
;
3114 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
3115 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
3116 if (IS_ZEBRA_DEBUG_RIB
)
3117 route_entry_dump(&rn
->p
, NULL
, re
);
3119 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
3122 if (!CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELFROUTE
))
3126 * So we are starting up and have received
3127 * routes from the kernel that we have installed
3128 * from a previous run of zebra but not cleaned
3129 * up ( say a kill -9 )
3130 * But since we haven't actually installed
3131 * them yet( we received them from the kernel )
3132 * we don't think they are active.
3133 * So let's pretend they are active to actually
3135 * In all honesty I'm not sure if we should
3136 * mark them as active when we receive them
3137 * This is startup only so probably ok.
3139 * If we ever decide to move rib_sweep_table
3140 * to a different spot (ie startup )
3141 * this decision needs to be revisited
3143 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
3144 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
3146 rib_uninstall_kernel(rn
, re
);
3147 rib_delnode(rn
, re
);
3152 /* Sweep all RIB tables. */
3153 void rib_sweep_route(void)
3156 struct zebra_vrf
*zvrf
;
3158 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
3159 if ((zvrf
= vrf
->info
) == NULL
)
3162 rib_sweep_table(zvrf
->table
[AFI_IP
][SAFI_UNICAST
]);
3163 rib_sweep_table(zvrf
->table
[AFI_IP6
][SAFI_UNICAST
]);
3166 zebra_ns_sweep_route();
3169 /* Remove specific by protocol routes from 'table'. */
3170 unsigned long rib_score_proto_table(uint8_t proto
, unsigned short instance
,
3171 struct route_table
*table
)
3173 struct route_node
*rn
;
3174 struct route_entry
*re
;
3175 struct route_entry
*next
;
3176 unsigned long n
= 0;
3179 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
))
3180 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
3181 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
3183 if (re
->type
== proto
3184 && re
->instance
== instance
) {
3185 rib_delnode(rn
, re
);
3192 /* Remove specific by protocol routes. */
3193 unsigned long rib_score_proto(uint8_t proto
, unsigned short instance
)
3196 struct zebra_vrf
*zvrf
;
3197 unsigned long cnt
= 0;
3199 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
)
3200 if ((zvrf
= vrf
->info
) != NULL
)
3201 cnt
+= rib_score_proto_table(
3203 zvrf
->table
[AFI_IP
][SAFI_UNICAST
])
3204 + rib_score_proto_table(
3206 zvrf
->table
[AFI_IP6
][SAFI_UNICAST
]);
3208 cnt
+= zebra_ns_score_proto(proto
, instance
);
3213 /* Close RIB and clean up kernel routes. */
3214 void rib_close_table(struct route_table
*table
)
3216 struct route_node
*rn
;
3217 rib_table_info_t
*info
;
3223 info
= route_table_get_info(table
);
3225 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
3226 dest
= rib_dest_from_rnode(rn
);
3228 if (dest
&& dest
->selected_fib
) {
3229 if (info
->safi
== SAFI_UNICAST
)
3230 hook_call(rib_update
, rn
, NULL
);
3232 if (!RIB_SYSTEM_ROUTE(dest
->selected_fib
))
3233 rib_uninstall_kernel(rn
, dest
->selected_fib
);
3241 static int rib_process_dplane_results(struct thread
*thread
)
3246 /* Take lock controlling queue of results */
3247 pthread_mutex_lock(&dplane_mutex
);
3249 /* Dequeue context block */
3250 dplane_ctx_dequeue(&rib_dplane_q
, &ctx
);
3252 pthread_mutex_unlock(&dplane_mutex
);
3255 rib_process_after(ctx
);
3261 /* Check for nexthop tracking processing after finishing with results */
3262 do_nht_processing();
3268 * Results are returned from the dataplane subsystem, in the context of
3269 * the dataplane pthread. We enqueue the results here for processing by
3270 * the main thread later.
3272 static int rib_dplane_results(dplane_ctx_h ctx
)
3274 /* Take lock controlling queue of results */
3275 pthread_mutex_lock(&dplane_mutex
);
3277 /* Enqueue context block */
3278 dplane_ctx_enqueue_tail(&rib_dplane_q
, ctx
);
3280 pthread_mutex_unlock(&dplane_mutex
);
3282 /* Ensure event is signalled to zebra main thread */
3283 thread_add_event(zebrad
.master
, rib_process_dplane_results
, NULL
, 0,
3289 /* Routing information base initialize. */
3292 rib_queue_init(&zebrad
);
3294 /* Init dataplane, and register for results */
3295 pthread_mutex_init(&dplane_mutex
, NULL
);
3296 TAILQ_INIT(&rib_dplane_q
);
3297 zebra_dplane_init();
3298 dplane_results_register(rib_dplane_results
);
3304 * Get the first vrf id that is greater than the given vrf id if any.
3306 * Returns TRUE if a vrf id was found, FALSE otherwise.
3308 static inline int vrf_id_get_next(vrf_id_t vrf_id
, vrf_id_t
*next_id_p
)
3312 vrf
= vrf_lookup_by_id(vrf_id
);
3314 vrf
= RB_NEXT(vrf_id_head
, vrf
);
3316 *next_id_p
= vrf
->vrf_id
;
3325 * rib_tables_iter_next
3327 * Returns the next table in the iteration.
3329 struct route_table
*rib_tables_iter_next(rib_tables_iter_t
*iter
)
3331 struct route_table
*table
;
3334 * Array that helps us go over all AFI/SAFI combinations via one
3341 {AFI_IP
, SAFI_UNICAST
}, {AFI_IP
, SAFI_MULTICAST
},
3342 {AFI_IP
, SAFI_LABELED_UNICAST
}, {AFI_IP6
, SAFI_UNICAST
},
3343 {AFI_IP6
, SAFI_MULTICAST
}, {AFI_IP6
, SAFI_LABELED_UNICAST
},
3348 switch (iter
->state
) {
3350 case RIB_TABLES_ITER_S_INIT
:
3351 iter
->vrf_id
= VRF_DEFAULT
;
3352 iter
->afi_safi_ix
= -1;
3356 case RIB_TABLES_ITER_S_ITERATING
:
3357 iter
->afi_safi_ix
++;
3360 while (iter
->afi_safi_ix
3361 < (int)ZEBRA_NUM_OF(afi_safis
)) {
3362 table
= zebra_vrf_table(
3363 afi_safis
[iter
->afi_safi_ix
].afi
,
3364 afi_safis
[iter
->afi_safi_ix
].safi
,
3369 iter
->afi_safi_ix
++;
3373 * Found another table in this vrf.
3379 * Done with all tables in the current vrf, go to the
3383 if (!vrf_id_get_next(iter
->vrf_id
, &iter
->vrf_id
))
3386 iter
->afi_safi_ix
= 0;
3391 case RIB_TABLES_ITER_S_DONE
:
3396 iter
->state
= RIB_TABLES_ITER_S_ITERATING
;
3398 iter
->state
= RIB_TABLES_ITER_S_DONE
;