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"
56 DEFINE_HOOK(rib_update
, (struct route_node
* rn
, const char *reason
),
59 /* Should we allow non Quagga processes to delete our routes */
60 extern int allow_delete
;
62 /* Each route type's string and default distance value. */
66 } route_info
[ZEBRA_ROUTE_MAX
] = {
67 [ZEBRA_ROUTE_SYSTEM
] = {ZEBRA_ROUTE_SYSTEM
, 0},
68 [ZEBRA_ROUTE_KERNEL
] = {ZEBRA_ROUTE_KERNEL
, 0},
69 [ZEBRA_ROUTE_CONNECT
] = {ZEBRA_ROUTE_CONNECT
, 0},
70 [ZEBRA_ROUTE_STATIC
] = {ZEBRA_ROUTE_STATIC
, 1},
71 [ZEBRA_ROUTE_RIP
] = {ZEBRA_ROUTE_RIP
, 120},
72 [ZEBRA_ROUTE_RIPNG
] = {ZEBRA_ROUTE_RIPNG
, 120},
73 [ZEBRA_ROUTE_OSPF
] = {ZEBRA_ROUTE_OSPF
, 110},
74 [ZEBRA_ROUTE_OSPF6
] = {ZEBRA_ROUTE_OSPF6
, 110},
75 [ZEBRA_ROUTE_ISIS
] = {ZEBRA_ROUTE_ISIS
, 115},
76 [ZEBRA_ROUTE_BGP
] = {ZEBRA_ROUTE_BGP
, 20 /* IBGP is 200. */},
77 [ZEBRA_ROUTE_PIM
] = {ZEBRA_ROUTE_PIM
, 255},
78 [ZEBRA_ROUTE_EIGRP
] = {ZEBRA_ROUTE_EIGRP
, 90},
79 [ZEBRA_ROUTE_NHRP
] = {ZEBRA_ROUTE_NHRP
, 10},
80 [ZEBRA_ROUTE_HSLS
] = {ZEBRA_ROUTE_HSLS
, 255},
81 [ZEBRA_ROUTE_OLSR
] = {ZEBRA_ROUTE_OLSR
, 255},
82 [ZEBRA_ROUTE_TABLE
] = {ZEBRA_ROUTE_TABLE
, 150},
83 [ZEBRA_ROUTE_LDP
] = {ZEBRA_ROUTE_LDP
, 150},
84 [ZEBRA_ROUTE_VNC
] = {ZEBRA_ROUTE_VNC
, 20},
85 [ZEBRA_ROUTE_VNC_DIRECT
] = {ZEBRA_ROUTE_VNC_DIRECT
, 20},
86 [ZEBRA_ROUTE_VNC_DIRECT_RH
] = {ZEBRA_ROUTE_VNC_DIRECT_RH
, 20},
87 [ZEBRA_ROUTE_BGP_DIRECT
] = {ZEBRA_ROUTE_BGP_DIRECT
, 20},
88 [ZEBRA_ROUTE_BGP_DIRECT_EXT
] = {ZEBRA_ROUTE_BGP_DIRECT_EXT
, 20},
89 [ZEBRA_ROUTE_BABEL
] = {ZEBRA_ROUTE_BABEL
, 100},
90 [ZEBRA_ROUTE_SHARP
] = {ZEBRA_ROUTE_SHARP
, 150},
92 /* no entry/default: 150 */
95 /* RPF lookup behaviour */
96 static enum multicast_mode ipv4_multicast_mode
= MCAST_NO_CONFIG
;
99 static void __attribute__((format(printf
, 5, 6)))
100 _rnode_zlog(const char *_func
, vrf_id_t vrf_id
, struct route_node
*rn
,
101 int priority
, const char *msgfmt
, ...)
103 char buf
[SRCDEST2STR_BUFFER
+ sizeof(" (MRIB)")];
107 va_start(ap
, msgfmt
);
108 vsnprintf(msgbuf
, sizeof(msgbuf
), msgfmt
, ap
);
112 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
113 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
115 if (info
->safi
== SAFI_MULTICAST
)
116 strcat(buf
, " (MRIB)");
118 snprintf(buf
, sizeof(buf
), "{(route_node *) NULL}");
121 zlog(priority
, "%s: %d:%s: %s", _func
, vrf_id
, buf
, msgbuf
);
124 #define rnode_debug(node, vrf_id, ...) \
125 _rnode_zlog(__func__, vrf_id, node, LOG_DEBUG, __VA_ARGS__)
126 #define rnode_info(node, ...) \
127 _rnode_zlog(__func__, vrf_id, node, LOG_INFO, __VA_ARGS__)
129 uint8_t route_distance(int type
)
133 if ((unsigned)type
>= array_size(route_info
))
136 distance
= route_info
[type
].distance
;
141 int is_zebra_valid_kernel_table(uint32_t table_id
)
144 if ((table_id
== RT_TABLE_UNSPEC
) || (table_id
== RT_TABLE_LOCAL
)
145 || (table_id
== RT_TABLE_COMPAT
))
152 int is_zebra_main_routing_table(uint32_t table_id
)
154 if ((table_id
== RT_TABLE_MAIN
)
155 || (table_id
== zebrad
.rtm_table_default
))
160 int zebra_check_addr(const struct prefix
*p
)
162 if (p
->family
== AF_INET
) {
165 addr
= p
->u
.prefix4
.s_addr
;
168 if (IPV4_NET127(addr
) || IN_CLASSD(addr
)
169 || IPV4_LINKLOCAL(addr
))
172 if (p
->family
== AF_INET6
) {
173 if (IN6_IS_ADDR_LOOPBACK(&p
->u
.prefix6
))
175 if (IN6_IS_ADDR_LINKLOCAL(&p
->u
.prefix6
))
181 /* Add nexthop to the end of a rib node's nexthop list */
182 void route_entry_nexthop_add(struct route_entry
*re
, struct nexthop
*nexthop
)
184 nexthop_add(&re
->ng
.nexthop
, nexthop
);
190 * copy_nexthop - copy a nexthop to the rib structure.
192 void route_entry_copy_nexthops(struct route_entry
*re
, struct nexthop
*nh
)
194 assert(!re
->ng
.nexthop
);
195 copy_nexthops(&re
->ng
.nexthop
, nh
, NULL
);
196 for (struct nexthop
*nexthop
= nh
; nexthop
; nexthop
= nexthop
->next
)
200 /* Delete specified nexthop from the list. */
201 void route_entry_nexthop_delete(struct route_entry
*re
, struct nexthop
*nexthop
)
204 nexthop
->next
->prev
= nexthop
->prev
;
206 nexthop
->prev
->next
= nexthop
->next
;
208 re
->ng
.nexthop
= nexthop
->next
;
213 struct nexthop
*route_entry_nexthop_ifindex_add(struct route_entry
*re
,
217 struct nexthop
*nexthop
;
219 nexthop
= nexthop_new();
220 nexthop
->type
= NEXTHOP_TYPE_IFINDEX
;
221 nexthop
->ifindex
= ifindex
;
222 nexthop
->vrf_id
= nh_vrf_id
;
224 route_entry_nexthop_add(re
, nexthop
);
229 struct nexthop
*route_entry_nexthop_ipv4_add(struct route_entry
*re
,
230 struct in_addr
*ipv4
,
234 struct nexthop
*nexthop
;
236 nexthop
= nexthop_new();
237 nexthop
->type
= NEXTHOP_TYPE_IPV4
;
238 nexthop
->vrf_id
= nh_vrf_id
;
239 nexthop
->gate
.ipv4
= *ipv4
;
241 nexthop
->src
.ipv4
= *src
;
243 route_entry_nexthop_add(re
, nexthop
);
248 struct nexthop
*route_entry_nexthop_ipv4_ifindex_add(struct route_entry
*re
,
249 struct in_addr
*ipv4
,
254 struct nexthop
*nexthop
;
255 struct interface
*ifp
;
257 nexthop
= nexthop_new();
258 nexthop
->vrf_id
= nh_vrf_id
;
259 nexthop
->type
= NEXTHOP_TYPE_IPV4_IFINDEX
;
260 nexthop
->gate
.ipv4
= *ipv4
;
262 nexthop
->src
.ipv4
= *src
;
263 nexthop
->ifindex
= ifindex
;
264 ifp
= if_lookup_by_index(nexthop
->ifindex
, nh_vrf_id
);
265 /*Pending: need to think if null ifp here is ok during bootup?
266 There was a crash because ifp here was coming to be NULL */
268 if (connected_is_unnumbered(ifp
)
269 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_EVPN_ROUTE
)
270 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ONLINK
)) {
271 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
);
274 route_entry_nexthop_add(re
, nexthop
);
279 struct nexthop
*route_entry_nexthop_ipv6_add(struct route_entry
*re
,
280 struct in6_addr
*ipv6
,
283 struct nexthop
*nexthop
;
285 nexthop
= nexthop_new();
286 nexthop
->vrf_id
= nh_vrf_id
;
287 nexthop
->type
= NEXTHOP_TYPE_IPV6
;
288 nexthop
->gate
.ipv6
= *ipv6
;
290 route_entry_nexthop_add(re
, nexthop
);
295 struct nexthop
*route_entry_nexthop_ipv6_ifindex_add(struct route_entry
*re
,
296 struct in6_addr
*ipv6
,
300 struct nexthop
*nexthop
;
302 nexthop
= nexthop_new();
303 nexthop
->vrf_id
= nh_vrf_id
;
304 nexthop
->type
= NEXTHOP_TYPE_IPV6_IFINDEX
;
305 nexthop
->gate
.ipv6
= *ipv6
;
306 nexthop
->ifindex
= ifindex
;
307 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_EVPN_ROUTE
)
308 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ONLINK
)) {
309 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
);
312 route_entry_nexthop_add(re
, nexthop
);
317 struct nexthop
*route_entry_nexthop_blackhole_add(struct route_entry
*re
,
318 enum blackhole_type bh_type
)
320 struct nexthop
*nexthop
;
322 nexthop
= nexthop_new();
323 nexthop
->vrf_id
= VRF_DEFAULT
;
324 nexthop
->type
= NEXTHOP_TYPE_BLACKHOLE
;
325 nexthop
->bh_type
= bh_type
;
327 route_entry_nexthop_add(re
, nexthop
);
332 static void nexthop_set_resolved(afi_t afi
, const struct nexthop
*newhop
,
333 struct nexthop
*nexthop
)
335 struct nexthop
*resolved_hop
;
337 resolved_hop
= nexthop_new();
338 SET_FLAG(resolved_hop
->flags
, NEXTHOP_FLAG_ACTIVE
);
340 resolved_hop
->vrf_id
= nexthop
->vrf_id
;
341 switch (newhop
->type
) {
342 case NEXTHOP_TYPE_IPV4
:
343 case NEXTHOP_TYPE_IPV4_IFINDEX
:
344 /* If the resolving route specifies a gateway, use it */
345 resolved_hop
->type
= newhop
->type
;
346 resolved_hop
->gate
.ipv4
= newhop
->gate
.ipv4
;
348 if (newhop
->ifindex
) {
349 resolved_hop
->type
= NEXTHOP_TYPE_IPV4_IFINDEX
;
350 resolved_hop
->ifindex
= newhop
->ifindex
;
351 if (newhop
->flags
& NEXTHOP_FLAG_ONLINK
)
352 resolved_hop
->flags
|= NEXTHOP_FLAG_ONLINK
;
355 case NEXTHOP_TYPE_IPV6
:
356 case NEXTHOP_TYPE_IPV6_IFINDEX
:
357 resolved_hop
->type
= newhop
->type
;
358 resolved_hop
->gate
.ipv6
= newhop
->gate
.ipv6
;
360 if (newhop
->ifindex
) {
361 resolved_hop
->type
= NEXTHOP_TYPE_IPV6_IFINDEX
;
362 resolved_hop
->ifindex
= newhop
->ifindex
;
365 case NEXTHOP_TYPE_IFINDEX
:
366 /* If the resolving route is an interface route,
367 * it means the gateway we are looking up is connected
368 * to that interface. (The actual network is _not_ onlink).
369 * Therefore, the resolved route should have the original
370 * gateway as nexthop as it is directly connected.
372 * On Linux, we have to set the onlink netlink flag because
373 * otherwise, the kernel won't accept the route.
375 resolved_hop
->flags
|= NEXTHOP_FLAG_ONLINK
;
377 resolved_hop
->type
= NEXTHOP_TYPE_IPV4_IFINDEX
;
378 resolved_hop
->gate
.ipv4
= nexthop
->gate
.ipv4
;
379 } else if (afi
== AFI_IP6
) {
380 resolved_hop
->type
= NEXTHOP_TYPE_IPV6_IFINDEX
;
381 resolved_hop
->gate
.ipv6
= nexthop
->gate
.ipv6
;
383 resolved_hop
->ifindex
= newhop
->ifindex
;
385 case NEXTHOP_TYPE_BLACKHOLE
:
386 resolved_hop
->type
= NEXTHOP_TYPE_BLACKHOLE
;
387 resolved_hop
->bh_type
= nexthop
->bh_type
;
391 /* Copy labels of the resolved route */
392 if (newhop
->nh_label
)
393 nexthop_add_labels(resolved_hop
, newhop
->nh_label_type
,
394 newhop
->nh_label
->num_labels
,
395 &newhop
->nh_label
->label
[0]);
397 resolved_hop
->rparent
= nexthop
;
398 nexthop_add(&nexthop
->resolved
, resolved_hop
);
401 /* If force flag is not set, do not modify falgs at all for uninstall
402 the route from FIB. */
403 static int nexthop_active(afi_t afi
, struct route_entry
*re
,
404 struct nexthop
*nexthop
, int set
,
405 struct route_node
*top
)
408 struct route_table
*table
;
409 struct route_node
*rn
;
410 struct route_entry
*match
= NULL
;
412 struct nexthop
*newhop
;
413 struct interface
*ifp
;
416 if ((nexthop
->type
== NEXTHOP_TYPE_IPV4
)
417 || nexthop
->type
== NEXTHOP_TYPE_IPV6
)
418 nexthop
->ifindex
= 0;
421 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
);
422 nexthops_free(nexthop
->resolved
);
423 nexthop
->resolved
= NULL
;
427 /* Next hops (remote VTEPs) for EVPN routes are fully resolved. */
428 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_EVPN_RVTEP
))
431 /* Skip nexthops that have been filtered out due to route-map */
432 /* The nexthops are specific to this route and so the same */
433 /* nexthop for a different route may not have this flag set */
434 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FILTERED
)) {
435 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
436 zlog_debug("\t%s: Nexthop Filtered",
437 __PRETTY_FUNCTION__
);
442 * Check to see if we should trust the passed in information
443 * for UNNUMBERED interfaces as that we won't find the GW
444 * address in the routing table.
446 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
)) {
447 ifp
= if_lookup_by_index(nexthop
->ifindex
, nexthop
->vrf_id
);
448 if ((ifp
&& connected_is_unnumbered(ifp
))
449 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ONLINK
)) {
450 if (if_is_operative(ifp
))
453 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
455 "\t%s: Onlink and interface %s is not operative",
456 __PRETTY_FUNCTION__
, ifp
->name
);
460 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
462 "\t%s: Interface %s is not unnumbered",
464 ifp
? ifp
->name
: "Unknown");
469 /* Make lookup prefix. */
470 memset(&p
, 0, sizeof(struct prefix
));
474 p
.prefixlen
= IPV4_MAX_PREFIXLEN
;
475 p
.u
.prefix4
= nexthop
->gate
.ipv4
;
479 p
.prefixlen
= IPV6_MAX_PREFIXLEN
;
480 p
.u
.prefix6
= nexthop
->gate
.ipv6
;
483 assert(afi
!= AFI_IP
&& afi
!= AFI_IP6
);
487 table
= zebra_vrf_table(afi
, SAFI_UNICAST
, nexthop
->vrf_id
);
489 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
490 zlog_debug("\t%s: Table not found",
491 __PRETTY_FUNCTION__
);
495 rn
= route_node_match(table
, (struct prefix
*)&p
);
497 route_unlock_node(rn
);
499 /* Lookup should halt if we've matched against ourselves ('top',
500 * if specified) - i.e., we cannot have a nexthop NH1 is
501 * resolved by a route NH1. The exception is if the route is a
504 if (top
&& rn
== top
)
505 if (((afi
== AFI_IP
) && (rn
->p
.prefixlen
!= 32))
506 || ((afi
== AFI_IP6
) && (rn
->p
.prefixlen
!= 128))) {
507 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
509 "\t%s: Matched against ourself and prefix length is not max bit length",
510 __PRETTY_FUNCTION__
);
514 /* Pick up selected route. */
515 /* However, do not resolve over default route unless explicitly
517 if (is_default_prefix(&rn
->p
)
518 && !rnh_resolve_via_default(p
.family
)) {
519 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
521 "\t:%s: Resolved against default route",
522 __PRETTY_FUNCTION__
);
526 dest
= rib_dest_from_rnode(rn
);
527 if (dest
&& dest
->selected_fib
528 && !CHECK_FLAG(dest
->selected_fib
->status
,
530 && dest
->selected_fib
->type
!= ZEBRA_ROUTE_TABLE
)
531 match
= dest
->selected_fib
;
533 /* If there is no selected route or matched route is EGP, go up
538 } while (rn
&& rn
->info
== NULL
);
545 if (match
->type
== ZEBRA_ROUTE_CONNECT
) {
546 /* Directly point connected route. */
547 newhop
= match
->ng
.nexthop
;
549 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
550 || nexthop
->type
== NEXTHOP_TYPE_IPV6
)
551 nexthop
->ifindex
= newhop
->ifindex
;
554 } else if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ALLOW_RECURSION
)) {
556 for (ALL_NEXTHOPS(match
->ng
, newhop
)) {
557 if (!CHECK_FLAG(newhop
->flags
,
560 if (CHECK_FLAG(newhop
->flags
,
561 NEXTHOP_FLAG_RECURSIVE
))
565 SET_FLAG(nexthop
->flags
,
566 NEXTHOP_FLAG_RECURSIVE
);
568 ROUTE_ENTRY_NEXTHOPS_CHANGED
);
569 nexthop_set_resolved(afi
, newhop
,
575 re
->nexthop_mtu
= match
->mtu
;
576 if (!resolved
&& IS_ZEBRA_DEBUG_RIB_DETAILED
)
577 zlog_debug("\t%s: Recursion failed to find",
578 __PRETTY_FUNCTION__
);
580 } else if (re
->type
== ZEBRA_ROUTE_STATIC
) {
582 for (ALL_NEXTHOPS(match
->ng
, newhop
)) {
583 if (!CHECK_FLAG(newhop
->flags
,
588 SET_FLAG(nexthop
->flags
,
589 NEXTHOP_FLAG_RECURSIVE
);
590 nexthop_set_resolved(afi
, newhop
,
596 re
->nexthop_mtu
= match
->mtu
;
598 if (!resolved
&& IS_ZEBRA_DEBUG_RIB_DETAILED
)
600 "\t%s: Static route unable to resolve",
601 __PRETTY_FUNCTION__
);
604 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
605 zlog_debug("\t%s: Route Type %s has not turned on recursion",
607 zebra_route_string(re
->type
));
608 if (re
->type
== ZEBRA_ROUTE_BGP
&&
609 !CHECK_FLAG(re
->flags
, ZEBRA_FLAG_IBGP
))
610 zlog_debug("\tEBGP: see \"disable-ebgp-connected-route-check\" or \"disable-connected-check\"");
615 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
616 zlog_debug("\t%s: Nexthop did not lookup in table",
617 __PRETTY_FUNCTION__
);
621 struct route_entry
*rib_match(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
,
622 union g_addr
*addr
, struct route_node
**rn_out
)
625 struct route_table
*table
;
626 struct route_node
*rn
;
627 struct route_entry
*match
= NULL
;
628 struct nexthop
*newhop
;
631 table
= zebra_vrf_table(afi
, safi
, vrf_id
);
635 memset(&p
, 0, sizeof(struct prefix
));
638 p
.u
.prefix4
= addr
->ipv4
;
639 p
.prefixlen
= IPV4_MAX_PREFIXLEN
;
641 p
.u
.prefix6
= addr
->ipv6
;
642 p
.prefixlen
= IPV6_MAX_PREFIXLEN
;
645 rn
= route_node_match(table
, (struct prefix
*)&p
);
650 route_unlock_node(rn
);
652 dest
= rib_dest_from_rnode(rn
);
653 if (dest
&& dest
->selected_fib
654 && !CHECK_FLAG(dest
->selected_fib
->status
,
655 ROUTE_ENTRY_REMOVED
))
656 match
= dest
->selected_fib
;
658 /* If there is no selected route or matched route is EGP, go up
663 } while (rn
&& rn
->info
== NULL
);
667 if (match
->type
!= ZEBRA_ROUTE_CONNECT
) {
669 for (ALL_NEXTHOPS(match
->ng
, newhop
))
670 if (CHECK_FLAG(newhop
->flags
,
687 struct route_entry
*rib_match_ipv4_multicast(vrf_id_t vrf_id
,
689 struct route_node
**rn_out
)
691 struct route_entry
*re
= NULL
, *mre
= NULL
, *ure
= NULL
;
692 struct route_node
*m_rn
= NULL
, *u_rn
= NULL
;
693 union g_addr gaddr
= {.ipv4
= addr
};
695 switch (ipv4_multicast_mode
) {
696 case MCAST_MRIB_ONLY
:
697 return rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
,
699 case MCAST_URIB_ONLY
:
700 return rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, rn_out
);
701 case MCAST_NO_CONFIG
:
702 case MCAST_MIX_MRIB_FIRST
:
703 re
= mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
,
706 re
= ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
,
709 case MCAST_MIX_DISTANCE
:
710 mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
, &m_rn
);
711 ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, &u_rn
);
713 re
= ure
->distance
< mre
->distance
? ure
: mre
;
719 case MCAST_MIX_PFXLEN
:
720 mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
, &m_rn
);
721 ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, &u_rn
);
723 re
= u_rn
->p
.prefixlen
> m_rn
->p
.prefixlen
? ure
: mre
;
732 *rn_out
= (re
== mre
) ? m_rn
: u_rn
;
734 if (IS_ZEBRA_DEBUG_RIB
) {
736 inet_ntop(AF_INET
, &addr
, buf
, BUFSIZ
);
738 zlog_debug("%s: %s: vrf: %u found %s, using %s",
739 __func__
, buf
, vrf_id
,
740 mre
? (ure
? "MRIB+URIB" : "MRIB")
741 : ure
? "URIB" : "nothing",
742 re
== ure
? "URIB" : re
== mre
? "MRIB" : "none");
747 void multicast_mode_ipv4_set(enum multicast_mode mode
)
749 if (IS_ZEBRA_DEBUG_RIB
)
750 zlog_debug("%s: multicast lookup mode set (%d)", __func__
,
752 ipv4_multicast_mode
= mode
;
755 enum multicast_mode
multicast_mode_ipv4_get(void)
757 return ipv4_multicast_mode
;
760 struct route_entry
*rib_lookup_ipv4(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
762 struct route_table
*table
;
763 struct route_node
*rn
;
764 struct route_entry
*match
= NULL
;
765 struct nexthop
*nexthop
;
769 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
773 rn
= route_node_lookup(table
, (struct prefix
*)p
);
775 /* No route for this prefix. */
780 route_unlock_node(rn
);
781 dest
= rib_dest_from_rnode(rn
);
783 if (dest
&& dest
->selected_fib
784 && !CHECK_FLAG(dest
->selected_fib
->status
, ROUTE_ENTRY_REMOVED
))
785 match
= dest
->selected_fib
;
790 if (match
->type
== ZEBRA_ROUTE_CONNECT
)
793 for (ALL_NEXTHOPS(match
->ng
, nexthop
))
794 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
801 * This clone function, unlike its original rib_lookup_ipv4(), checks
802 * if specified IPv4 route record (prefix/mask -> gate) exists in
803 * the whole RIB and has ROUTE_ENTRY_SELECTED_FIB set.
807 * 0: exact match found
808 * 1: a match was found with a different gate
809 * 2: connected route found
810 * 3: no matches found
812 int rib_lookup_ipv4_route(struct prefix_ipv4
*p
, union sockunion
*qgate
,
815 struct route_table
*table
;
816 struct route_node
*rn
;
817 struct route_entry
*match
= NULL
;
818 struct nexthop
*nexthop
;
823 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
825 return ZEBRA_RIB_LOOKUP_ERROR
;
827 /* Scan the RIB table for exactly matching RIB entry. */
828 rn
= route_node_lookup(table
, (struct prefix
*)p
);
830 /* No route for this prefix. */
832 return ZEBRA_RIB_NOTFOUND
;
835 route_unlock_node(rn
);
836 dest
= rib_dest_from_rnode(rn
);
838 /* Find out if a "selected" RR for the discovered RIB entry exists ever.
840 if (dest
&& dest
->selected_fib
841 && !CHECK_FLAG(dest
->selected_fib
->status
, ROUTE_ENTRY_REMOVED
))
842 match
= dest
->selected_fib
;
844 /* None such found :( */
846 return ZEBRA_RIB_NOTFOUND
;
848 if (match
->type
== ZEBRA_ROUTE_CONNECT
)
849 return ZEBRA_RIB_FOUND_CONNECTED
;
851 /* Ok, we have a cood candidate, let's check it's nexthop list... */
853 for (ALL_NEXTHOPS(match
->ng
, nexthop
))
854 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
)) {
856 if (nexthop
->gate
.ipv4
.s_addr
== sockunion2ip(qgate
))
857 return ZEBRA_RIB_FOUND_EXACT
;
858 if (IS_ZEBRA_DEBUG_RIB
) {
859 char gate_buf
[INET_ADDRSTRLEN
],
860 qgate_buf
[INET_ADDRSTRLEN
];
861 inet_ntop(AF_INET
, &nexthop
->gate
.ipv4
.s_addr
,
862 gate_buf
, INET_ADDRSTRLEN
);
863 inet_ntop(AF_INET
, &sockunion2ip(qgate
),
864 qgate_buf
, INET_ADDRSTRLEN
);
865 zlog_debug("%s: qgate == %s, %s == %s",
867 nexthop
->rparent
? "rgate" : "gate",
873 return ZEBRA_RIB_FOUND_NOGATE
;
875 return ZEBRA_RIB_NOTFOUND
;
878 #define RIB_SYSTEM_ROUTE(R) \
879 ((R)->type == ZEBRA_ROUTE_KERNEL || (R)->type == ZEBRA_ROUTE_CONNECT)
881 /* This function verifies reachability of one given nexthop, which can be
882 * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored
883 * in nexthop->flags field. If the 4th parameter, 'set', is non-zero,
884 * nexthop->ifindex will be updated appropriately as well.
885 * An existing route map can turn (otherwise active) nexthop into inactive, but
888 * The return value is the final value of 'ACTIVE' flag.
891 static unsigned nexthop_active_check(struct route_node
*rn
,
892 struct route_entry
*re
,
893 struct nexthop
*nexthop
, int set
)
895 struct interface
*ifp
;
896 route_map_result_t ret
= RMAP_MATCH
;
898 char buf
[SRCDEST2STR_BUFFER
];
899 const struct prefix
*p
, *src_p
;
900 struct zebra_vrf
*zvrf
;
902 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
904 if (rn
->p
.family
== AF_INET
)
906 else if (rn
->p
.family
== AF_INET6
)
910 switch (nexthop
->type
) {
911 case NEXTHOP_TYPE_IFINDEX
:
912 ifp
= if_lookup_by_index(nexthop
->ifindex
, nexthop
->vrf_id
);
913 if (ifp
&& if_is_operative(ifp
))
914 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
916 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
918 case NEXTHOP_TYPE_IPV4
:
919 case NEXTHOP_TYPE_IPV4_IFINDEX
:
921 if (nexthop_active(AFI_IP
, re
, nexthop
, set
, rn
))
922 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
924 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
926 case NEXTHOP_TYPE_IPV6
:
928 if (nexthop_active(AFI_IP6
, re
, nexthop
, set
, rn
))
929 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
931 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
933 case NEXTHOP_TYPE_IPV6_IFINDEX
:
934 /* RFC 5549, v4 prefix with v6 NH */
935 if (rn
->p
.family
!= AF_INET
)
937 if (IN6_IS_ADDR_LINKLOCAL(&nexthop
->gate
.ipv6
)) {
938 ifp
= if_lookup_by_index(nexthop
->ifindex
,
940 if (ifp
&& if_is_operative(ifp
))
941 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
943 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
945 if (nexthop_active(AFI_IP6
, re
, nexthop
, set
, rn
))
946 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
948 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
951 case NEXTHOP_TYPE_BLACKHOLE
:
952 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
957 if (!CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
)) {
958 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
959 zlog_debug("\t%s: Unable to find a active nexthop",
960 __PRETTY_FUNCTION__
);
964 /* XXX: What exactly do those checks do? Do we support
965 * e.g. IPv4 routes with IPv6 nexthops or vice versa?
967 if (RIB_SYSTEM_ROUTE(re
) || (family
== AFI_IP
&& p
->family
!= AF_INET
)
968 || (family
== AFI_IP6
&& p
->family
!= AF_INET6
))
969 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
971 /* The original code didn't determine the family correctly
972 * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi
973 * from the rib_table_info in those cases.
974 * Possibly it may be better to use only the rib_table_info
978 rib_table_info_t
*info
;
980 info
= srcdest_rnode_table_info(rn
);
984 memset(&nexthop
->rmap_src
.ipv6
, 0, sizeof(union g_addr
));
986 zvrf
= zebra_vrf_lookup_by_id(nexthop
->vrf_id
);
988 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
989 zlog_debug("\t%s: zvrf is NULL", __PRETTY_FUNCTION__
);
990 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
993 /* It'll get set if required inside */
994 ret
= zebra_route_map_check(family
, re
->type
, re
->instance
, p
,
995 nexthop
, zvrf
, re
->tag
);
996 if (ret
== RMAP_DENYMATCH
) {
997 if (IS_ZEBRA_DEBUG_RIB
) {
998 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1000 "%u:%s: Filtering out with NH out %s due to route map",
1002 ifindex2ifname(nexthop
->ifindex
,
1005 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
1007 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
1010 /* Iterate over all nexthops of the given RIB entry and refresh their
1011 * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any
1012 * nexthop is found to toggle the ACTIVE flag, the whole re structure
1013 * is flagged with ROUTE_ENTRY_CHANGED. The 4th 'set' argument is
1014 * transparently passed to nexthop_active_check().
1016 * Return value is the new number of active nexthops.
1019 static int nexthop_active_update(struct route_node
*rn
, struct route_entry
*re
,
1022 struct nexthop
*nexthop
;
1023 union g_addr prev_src
;
1024 unsigned int prev_active
, new_active
, old_num_nh
;
1025 ifindex_t prev_index
;
1027 old_num_nh
= re
->nexthop_active_num
;
1029 re
->nexthop_active_num
= 0;
1030 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1032 for (nexthop
= re
->ng
.nexthop
; nexthop
; nexthop
= nexthop
->next
) {
1033 /* No protocol daemon provides src and so we're skipping
1035 prev_src
= nexthop
->rmap_src
;
1036 prev_active
= CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
1037 prev_index
= nexthop
->ifindex
;
1038 if ((new_active
= nexthop_active_check(rn
, re
, nexthop
, set
)))
1039 re
->nexthop_active_num
++;
1040 /* Don't allow src setting on IPv6 addr for now */
1041 if (prev_active
!= new_active
|| prev_index
!= nexthop
->ifindex
1042 || ((nexthop
->type
>= NEXTHOP_TYPE_IFINDEX
1043 && nexthop
->type
< NEXTHOP_TYPE_IPV6
)
1044 && prev_src
.ipv4
.s_addr
1045 != nexthop
->rmap_src
.ipv4
.s_addr
)
1046 || ((nexthop
->type
>= NEXTHOP_TYPE_IPV6
1047 && nexthop
->type
< NEXTHOP_TYPE_BLACKHOLE
)
1048 && !(IPV6_ADDR_SAME(&prev_src
.ipv6
,
1049 &nexthop
->rmap_src
.ipv6
)))) {
1050 SET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1051 SET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1055 if (old_num_nh
!= re
->nexthop_active_num
)
1056 SET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1058 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
)) {
1059 SET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1062 return re
->nexthop_active_num
;
1066 * Is this RIB labeled-unicast? It must be of type BGP and all paths
1067 * (nexthops) must have a label.
1069 int zebra_rib_labeled_unicast(struct route_entry
*re
)
1071 struct nexthop
*nexthop
= NULL
;
1073 if (re
->type
!= ZEBRA_ROUTE_BGP
)
1076 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1077 if (!nexthop
->nh_label
|| !nexthop
->nh_label
->num_labels
)
1083 void kernel_route_rib_pass_fail(struct route_node
*rn
, const struct prefix
*p
,
1084 struct route_entry
*re
,
1085 enum zebra_dplane_status res
)
1087 struct nexthop
*nexthop
;
1088 char buf
[PREFIX_STRLEN
];
1091 dest
= rib_dest_from_rnode(rn
);
1094 case ZEBRA_DPLANE_INSTALL_SUCCESS
:
1095 dest
->selected_fib
= re
;
1096 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1097 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
))
1100 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
))
1101 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1103 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1105 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_INSTALLED
);
1107 case ZEBRA_DPLANE_INSTALL_FAILURE
:
1109 * I am not sure this is the right thing to do here
1110 * but the code always set selected_fib before
1111 * this assignment was moved here.
1113 dest
->selected_fib
= re
;
1115 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_FAIL_INSTALL
);
1116 flog_err(EC_ZEBRA_DP_INSTALL_FAIL
,
1117 "%u:%s: Route install failed", re
->vrf_id
,
1118 prefix2str(p
, buf
, sizeof(buf
)));
1120 case ZEBRA_DPLANE_DELETE_SUCCESS
:
1122 * The case where selected_fib is not re is
1123 * when we have received a system route
1124 * that is overriding our installed route
1125 * as such we should leave the selected_fib
1128 if (dest
->selected_fib
== re
)
1129 dest
->selected_fib
= NULL
;
1130 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1131 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1133 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_REMOVED
);
1135 case ZEBRA_DPLANE_DELETE_FAILURE
:
1137 * Should we set this to NULL if the
1140 dest
->selected_fib
= NULL
;
1141 flog_err(EC_ZEBRA_DP_DELETE_FAIL
,
1142 "%u:%s: Route Deletion failure", re
->vrf_id
,
1143 prefix2str(p
, buf
, sizeof(buf
)));
1145 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_REMOVE_FAIL
);
1147 case ZEBRA_DPLANE_STATUS_NONE
:
1152 /* Update flag indicates whether this is a "replace" or not. Currently, this
1153 * is only used for IPv4.
1155 void rib_install_kernel(struct route_node
*rn
, struct route_entry
*re
,
1156 struct route_entry
*old
)
1158 struct nexthop
*nexthop
;
1159 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1160 const struct prefix
*p
, *src_p
;
1161 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1163 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1165 if (info
->safi
!= SAFI_UNICAST
) {
1166 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1167 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1170 struct nexthop
*prev
;
1172 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1173 UNSET_FLAG (nexthop
->flags
, NEXTHOP_FLAG_DUPLICATE
);
1174 for (ALL_NEXTHOPS(re
->ng
, prev
)) {
1175 if (prev
== nexthop
)
1177 if (nexthop_same_firsthop(nexthop
, prev
)) {
1178 SET_FLAG(nexthop
->flags
,
1179 NEXTHOP_FLAG_DUPLICATE
);
1187 * If this is a replace to a new RE let the originator of the RE
1188 * know that they've lost
1190 if (old
&& (old
!= re
) && (old
->type
!= re
->type
))
1191 zsend_route_notify_owner(old
, p
, ZAPI_ROUTE_BETTER_ADMIN_WON
);
1194 * Make sure we update the FPM any time we send new information to
1197 hook_call(rib_update
, rn
, "installing in kernel");
1198 switch (kernel_route_rib(rn
, p
, src_p
, old
, re
)) {
1199 case ZEBRA_DPLANE_REQUEST_QUEUED
:
1201 EC_ZEBRA_DP_INVALID_RC
,
1202 "No current known DataPlane interfaces can return this, please fix");
1204 case ZEBRA_DPLANE_REQUEST_FAILURE
:
1206 EC_ZEBRA_DP_INSTALL_FAIL
,
1207 "No current known Rib Install Failure cases, please fix");
1209 case ZEBRA_DPLANE_REQUEST_SUCCESS
:
1217 /* Uninstall the route from kernel. */
1218 void rib_uninstall_kernel(struct route_node
*rn
, struct route_entry
*re
)
1220 struct nexthop
*nexthop
;
1221 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1222 const struct prefix
*p
, *src_p
;
1223 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1225 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1227 if (info
->safi
!= SAFI_UNICAST
) {
1228 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1229 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1234 * Make sure we update the FPM any time we send new information to
1237 hook_call(rib_update
, rn
, "uninstalling from kernel");
1238 switch (kernel_route_rib(rn
, p
, src_p
, re
, NULL
)) {
1239 case ZEBRA_DPLANE_REQUEST_QUEUED
:
1241 EC_ZEBRA_DP_INVALID_RC
,
1242 "No current known DataPlane interfaces can return this, please fix");
1244 case ZEBRA_DPLANE_REQUEST_FAILURE
:
1246 EC_ZEBRA_DP_INSTALL_FAIL
,
1247 "No current known RIB Install Failure cases, please fix");
1249 case ZEBRA_DPLANE_REQUEST_SUCCESS
:
1258 /* Uninstall the route from kernel. */
1259 static void rib_uninstall(struct route_node
*rn
, struct route_entry
*re
)
1261 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1262 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1264 if (dest
&& dest
->selected_fib
== re
) {
1265 if (info
->safi
== SAFI_UNICAST
)
1266 hook_call(rib_update
, rn
, "rib_uninstall");
1268 if (!RIB_SYSTEM_ROUTE(re
))
1269 rib_uninstall_kernel(rn
, re
);
1271 /* If labeled-unicast route, uninstall transit LSP. */
1272 if (zebra_rib_labeled_unicast(re
))
1273 zebra_mpls_lsp_uninstall(info
->zvrf
, rn
, re
);
1276 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)) {
1277 const struct prefix
*p
, *src_p
;
1279 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1281 redistribute_delete(p
, src_p
, re
);
1282 UNSET_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
);
1287 * rib_can_delete_dest
1289 * Returns TRUE if the given dest can be deleted from the table.
1291 static int rib_can_delete_dest(rib_dest_t
*dest
)
1298 * Don't delete the dest if we have to update the FPM about this
1301 if (CHECK_FLAG(dest
->flags
, RIB_DEST_UPDATE_FPM
)
1302 || CHECK_FLAG(dest
->flags
, RIB_DEST_SENT_TO_FPM
))
1311 * Garbage collect the rib dest corresponding to the given route node
1314 * Returns TRUE if the dest was deleted, FALSE otherwise.
1316 int rib_gc_dest(struct route_node
*rn
)
1320 dest
= rib_dest_from_rnode(rn
);
1324 if (!rib_can_delete_dest(dest
))
1327 if (IS_ZEBRA_DEBUG_RIB
) {
1328 struct zebra_vrf
*zvrf
;
1330 zvrf
= rib_dest_vrf(dest
);
1331 rnode_debug(rn
, zvrf_id(zvrf
), "removing dest from table");
1335 XFREE(MTYPE_RIB_DEST
, dest
);
1339 * Release the one reference that we keep on the route node.
1341 route_unlock_node(rn
);
1345 static void rib_process_add_fib(struct zebra_vrf
*zvrf
, struct route_node
*rn
,
1346 struct route_entry
*new)
1348 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1350 hook_call(rib_update
, rn
, "new route selected");
1352 /* Update real nexthop. This may actually determine if nexthop is active
1354 if (!nexthop_active_update(rn
, new, 1)) {
1355 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1359 if (IS_ZEBRA_DEBUG_RIB
) {
1360 char buf
[SRCDEST2STR_BUFFER
];
1361 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1362 zlog_debug("%u:%s: Adding route rn %p, re %p (type %d)",
1363 zvrf_id(zvrf
), buf
, rn
, new, new->type
);
1366 /* If labeled-unicast route, install transit LSP. */
1367 if (zebra_rib_labeled_unicast(new))
1368 zebra_mpls_lsp_install(zvrf
, rn
, new);
1370 if (!RIB_SYSTEM_ROUTE(new))
1371 rib_install_kernel(rn
, new, NULL
);
1373 dest
->selected_fib
= new;
1375 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1378 static void rib_process_del_fib(struct zebra_vrf
*zvrf
, struct route_node
*rn
,
1379 struct route_entry
*old
)
1381 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1382 hook_call(rib_update
, rn
, "removing existing route");
1384 /* Uninstall from kernel. */
1385 if (IS_ZEBRA_DEBUG_RIB
) {
1386 char buf
[SRCDEST2STR_BUFFER
];
1387 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1388 zlog_debug("%u:%s: Deleting route rn %p, re %p (type %d)",
1389 zvrf_id(zvrf
), buf
, rn
, old
, old
->type
);
1392 /* If labeled-unicast route, uninstall transit LSP. */
1393 if (zebra_rib_labeled_unicast(old
))
1394 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1396 if (!RIB_SYSTEM_ROUTE(old
))
1397 rib_uninstall_kernel(rn
, old
);
1400 * We are setting this to NULL here
1401 * because that is what we traditionally
1402 * have been doing. I am not positive
1403 * that this is the right thing to do
1404 * but let's leave the code alone
1405 * for the RIB_SYSTEM_ROUTE case
1407 dest
->selected_fib
= NULL
;
1410 /* Update nexthop for route, reset changed flag. */
1411 nexthop_active_update(rn
, old
, 1);
1412 UNSET_FLAG(old
->status
, ROUTE_ENTRY_CHANGED
);
1415 static void rib_process_update_fib(struct zebra_vrf
*zvrf
,
1416 struct route_node
*rn
,
1417 struct route_entry
*old
,
1418 struct route_entry
*new)
1420 struct nexthop
*nexthop
= NULL
;
1422 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1425 * We have to install or update if a new route has been selected or
1426 * something has changed.
1428 if (new != old
|| CHECK_FLAG(new->status
, ROUTE_ENTRY_CHANGED
)) {
1429 hook_call(rib_update
, rn
, "updating existing route");
1431 /* Update the nexthop; we could determine here that nexthop is
1433 if (nexthop_active_update(rn
, new, 1))
1436 /* If nexthop is active, install the selected route, if
1438 * the install succeeds, cleanup flags for prior route, if
1443 if (IS_ZEBRA_DEBUG_RIB
) {
1444 char buf
[SRCDEST2STR_BUFFER
];
1445 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1448 "%u:%s: Updating route rn %p, re %p (type %d) "
1450 zvrf_id(zvrf
), buf
, rn
, new,
1451 new->type
, old
, old
->type
);
1454 "%u:%s: Updating route rn %p, re %p (type %d)",
1455 zvrf_id(zvrf
), buf
, rn
, new,
1459 /* If labeled-unicast route, uninstall transit LSP. */
1460 if (zebra_rib_labeled_unicast(old
))
1461 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1463 /* Non-system route should be installed. */
1464 if (!RIB_SYSTEM_ROUTE(new)) {
1465 /* If labeled-unicast route, install transit
1467 if (zebra_rib_labeled_unicast(new))
1468 zebra_mpls_lsp_install(zvrf
, rn
, new);
1470 rib_install_kernel(rn
, new, old
);
1473 * We do not need to install the
1474 * selected route because it
1475 * is already isntalled by
1476 * the system( ie not us )
1477 * so just mark it as winning
1478 * we do need to ensure that
1479 * if we uninstall a route
1480 * from ourselves we don't
1481 * over write this pointer
1483 dest
->selected_fib
= NULL
;
1485 /* If install succeeded or system route, cleanup flags
1486 * for prior route. */
1488 if (RIB_SYSTEM_ROUTE(new)) {
1489 if (!RIB_SYSTEM_ROUTE(old
))
1490 rib_uninstall_kernel(rn
, old
);
1492 for (nexthop
= old
->ng
.nexthop
; nexthop
;
1493 nexthop
= nexthop
->next
)
1494 UNSET_FLAG(nexthop
->flags
,
1501 * If nexthop for selected route is not active or install
1503 * may need to uninstall and delete for redistribution.
1506 if (IS_ZEBRA_DEBUG_RIB
) {
1507 char buf
[SRCDEST2STR_BUFFER
];
1508 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1511 "%u:%s: Deleting route rn %p, re %p (type %d) "
1512 "old %p (type %d) - nexthop inactive",
1513 zvrf_id(zvrf
), buf
, rn
, new,
1514 new->type
, old
, old
->type
);
1517 "%u:%s: Deleting route rn %p, re %p (type %d) - nexthop inactive",
1518 zvrf_id(zvrf
), buf
, rn
, new,
1522 /* If labeled-unicast route, uninstall transit LSP. */
1523 if (zebra_rib_labeled_unicast(old
))
1524 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1526 if (!RIB_SYSTEM_ROUTE(old
))
1527 rib_uninstall_kernel(rn
, old
);
1529 dest
->selected_fib
= NULL
;
1533 * Same route selected; check if in the FIB and if not,
1535 * is housekeeping code to deal with race conditions in kernel
1537 * netlink reporting interface up before IPv4 or IPv6 protocol
1541 if (!RIB_SYSTEM_ROUTE(new)) {
1542 bool in_fib
= false;
1544 for (ALL_NEXTHOPS(new->ng
, nexthop
))
1545 if (CHECK_FLAG(nexthop
->flags
,
1546 NEXTHOP_FLAG_FIB
)) {
1551 rib_install_kernel(rn
, new, NULL
);
1555 /* Update prior route. */
1557 /* Set real nexthop. */
1558 nexthop_active_update(rn
, old
, 1);
1559 UNSET_FLAG(old
->status
, ROUTE_ENTRY_CHANGED
);
1562 /* Clear changed flag. */
1563 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1566 /* Check if 'alternate' RIB entry is better than 'current'. */
1567 static struct route_entry
*rib_choose_best(struct route_entry
*current
,
1568 struct route_entry
*alternate
)
1570 if (current
== NULL
)
1573 /* filter route selection in following order:
1574 * - connected beats other types
1575 * - if both connected, loopback or vrf wins
1576 * - lower distance beats higher
1577 * - lower metric beats higher for equal distance
1578 * - last, hence oldest, route wins tie break.
1581 /* Connected routes. Check to see if either are a vrf
1582 * or loopback interface. If not, pick the last connected
1583 * route of the set of lowest metric connected routes.
1585 if (alternate
->type
== ZEBRA_ROUTE_CONNECT
) {
1586 if (current
->type
!= ZEBRA_ROUTE_CONNECT
)
1589 /* both are connected. are either loop or vrf? */
1590 struct nexthop
*nexthop
= NULL
;
1592 for (ALL_NEXTHOPS(alternate
->ng
, nexthop
)) {
1593 if (if_is_loopback_or_vrf(if_lookup_by_index(
1594 nexthop
->ifindex
, alternate
->vrf_id
)))
1598 for (ALL_NEXTHOPS(current
->ng
, nexthop
)) {
1599 if (if_is_loopback_or_vrf(if_lookup_by_index(
1600 nexthop
->ifindex
, current
->vrf_id
)))
1604 /* Neither are loop or vrf so pick best metric */
1605 if (alternate
->metric
<= current
->metric
)
1611 if (current
->type
== ZEBRA_ROUTE_CONNECT
)
1614 /* higher distance loses */
1615 if (alternate
->distance
< current
->distance
)
1617 if (current
->distance
< alternate
->distance
)
1620 /* metric tie-breaks equal distance */
1621 if (alternate
->metric
<= current
->metric
)
1627 /* Core function for processing routing information base. */
1628 static void rib_process(struct route_node
*rn
)
1630 struct route_entry
*re
;
1631 struct route_entry
*next
;
1632 struct route_entry
*old_selected
= NULL
;
1633 struct route_entry
*new_selected
= NULL
;
1634 struct route_entry
*old_fib
= NULL
;
1635 struct route_entry
*new_fib
= NULL
;
1636 struct route_entry
*best
= NULL
;
1637 char buf
[SRCDEST2STR_BUFFER
];
1639 struct zebra_vrf
*zvrf
= NULL
;
1640 const struct prefix
*p
, *src_p
;
1642 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1643 vrf_id_t vrf_id
= VRF_UNKNOWN
;
1647 dest
= rib_dest_from_rnode(rn
);
1649 zvrf
= rib_dest_vrf(dest
);
1650 vrf_id
= zvrf_id(zvrf
);
1653 if (IS_ZEBRA_DEBUG_RIB
)
1654 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1656 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1657 zlog_debug("%u:%s: Processing rn %p", vrf_id
, buf
, rn
);
1660 * we can have rn's that have a NULL info pointer
1661 * (dest). As such let's not let the deref happen
1662 * additionally we know RNODE_FOREACH_RE_SAFE
1663 * will not iterate so we are ok.
1666 old_fib
= dest
->selected_fib
;
1668 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
1669 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1671 "%u:%s: Examine re %p (type %d) status %x flags %x "
1672 "dist %d metric %d",
1673 vrf_id
, buf
, re
, re
->type
, re
->status
,
1674 re
->flags
, re
->distance
, re
->metric
);
1676 UNSET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1678 /* Currently selected re. */
1679 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)) {
1680 assert(old_selected
== NULL
);
1684 /* Skip deleted entries from selection */
1685 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
1688 /* Skip unreachable nexthop. */
1689 /* This first call to nexthop_active_update is merely to
1691 * there's any change to nexthops associated with this RIB
1693 * rib_process() can be invoked due to an external event such as
1695 * down or due to next-hop-tracking evaluation. In the latter
1697 * a decision has already been made that the NHs have changed.
1699 * need to invoke a potentially expensive call again. Further,
1701 * the change might be in a recursive NH which is not caught in
1702 * the nexthop_active_update() code. Thus, we might miss changes
1706 if (!CHECK_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
)
1707 && !nexthop_active_update(rn
, re
, 0)) {
1708 if (re
->type
== ZEBRA_ROUTE_TABLE
) {
1709 /* XXX: HERE BE DRAGONS!!!!!
1710 * In all honesty, I have not yet figured out
1712 * does or why the ROUTE_ENTRY_CHANGED test
1714 * or why we need to delete a route here, and
1716 * this concerns both selected and fib route, or
1719 /* This entry was denied by the 'ip protocol
1720 * table' route-map, we
1721 * need to delete it */
1722 if (re
!= old_selected
) {
1723 if (IS_ZEBRA_DEBUG_RIB
)
1725 "%s: %u:%s: imported via import-table but denied "
1726 "by the ip protocol table route-map",
1727 __func__
, vrf_id
, buf
);
1730 SET_FLAG(re
->status
,
1731 ROUTE_ENTRY_REMOVED
);
1737 /* Infinite distance. */
1738 if (re
->distance
== DISTANCE_INFINITY
) {
1739 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1743 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_FIB_OVERRIDE
)) {
1744 best
= rib_choose_best(new_fib
, re
);
1745 if (new_fib
&& best
!= new_fib
)
1746 UNSET_FLAG(new_fib
->status
,
1747 ROUTE_ENTRY_CHANGED
);
1750 best
= rib_choose_best(new_selected
, re
);
1751 if (new_selected
&& best
!= new_selected
)
1752 UNSET_FLAG(new_selected
->status
,
1753 ROUTE_ENTRY_CHANGED
);
1754 new_selected
= best
;
1757 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1758 } /* RNODE_FOREACH_RE */
1760 /* If no FIB override route, use the selected route also for FIB */
1761 if (new_fib
== NULL
)
1762 new_fib
= new_selected
;
1764 /* After the cycle is finished, the following pointers will be set:
1765 * old_selected --- RE entry currently having SELECTED
1766 * new_selected --- RE entry that is newly SELECTED
1767 * old_fib --- RE entry currently in kernel FIB
1768 * new_fib --- RE entry that is newly to be in kernel FIB
1770 * new_selected will get SELECTED flag, and is going to be redistributed
1771 * the zclients. new_fib (which can be new_selected) will be installed
1775 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
1777 "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p",
1778 vrf_id
, buf
, (void *)old_selected
, (void *)new_selected
,
1779 (void *)old_fib
, (void *)new_fib
);
1782 /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if
1783 * fib == selected */
1784 bool selected_changed
= new_selected
&& CHECK_FLAG(new_selected
->status
,
1785 ROUTE_ENTRY_CHANGED
);
1787 /* Update fib according to selection results */
1788 if (new_fib
&& old_fib
)
1789 rib_process_update_fib(zvrf
, rn
, old_fib
, new_fib
);
1791 rib_process_add_fib(zvrf
, rn
, new_fib
);
1793 rib_process_del_fib(zvrf
, rn
, old_fib
);
1795 /* Redistribute SELECTED entry */
1796 if (old_selected
!= new_selected
|| selected_changed
) {
1797 struct nexthop
*nexthop
= NULL
;
1799 /* Check if we have a FIB route for the destination, otherwise,
1800 * don't redistribute it */
1802 for (ALL_NEXTHOPS(new_fib
->ng
, nexthop
)) {
1803 if (CHECK_FLAG(nexthop
->flags
,
1804 NEXTHOP_FLAG_FIB
)) {
1810 new_selected
= NULL
;
1812 if (new_selected
&& new_selected
!= new_fib
) {
1813 nexthop_active_update(rn
, new_selected
, 1);
1814 UNSET_FLAG(new_selected
->status
, ROUTE_ENTRY_CHANGED
);
1819 redistribute_delete(p
, src_p
, old_selected
);
1820 if (old_selected
!= new_selected
)
1821 UNSET_FLAG(old_selected
->flags
,
1822 ZEBRA_FLAG_SELECTED
);
1826 /* Install new or replace existing redistributed entry
1828 SET_FLAG(new_selected
->flags
, ZEBRA_FLAG_SELECTED
);
1829 redistribute_update(p
, src_p
, new_selected
,
1834 /* Remove all RE entries queued for removal */
1835 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
1836 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
1837 if (IS_ZEBRA_DEBUG_RIB
) {
1838 rnode_debug(rn
, vrf_id
, "rn %p, removing re %p",
1839 (void *)rn
, (void *)re
);
1846 * Check if the dest can be deleted now.
1851 /* Take a list of route_node structs and return 1, if there was a record
1852 * picked from it and processed by rib_process(). Don't process more,
1853 * than one RN record; operate only in the specified sub-queue.
1855 static unsigned int process_subq(struct list
*subq
, uint8_t qindex
)
1857 struct listnode
*lnode
= listhead(subq
);
1858 struct route_node
*rnode
;
1860 struct zebra_vrf
*zvrf
= NULL
;
1865 rnode
= listgetdata(lnode
);
1866 dest
= rib_dest_from_rnode(rnode
);
1868 zvrf
= rib_dest_vrf(dest
);
1872 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
1873 char buf
[SRCDEST2STR_BUFFER
];
1874 srcdest_rnode2str(rnode
, buf
, sizeof(buf
));
1875 zlog_debug("%u:%s: rn %p dequeued from sub-queue %u",
1876 zvrf
? zvrf_id(zvrf
) : 0, buf
, rnode
, qindex
);
1880 UNSET_FLAG(rib_dest_from_rnode(rnode
)->flags
,
1881 RIB_ROUTE_QUEUED(qindex
));
1886 zlog_debug ("%s: called for route_node (%p, %d) with no ribs",
1887 __func__
, rnode
, rnode
->lock
);
1888 zlog_backtrace(LOG_DEBUG
);
1891 route_unlock_node(rnode
);
1892 list_delete_node(subq
, lnode
);
1897 * All meta queues have been processed. Trigger next-hop evaluation.
1899 static void meta_queue_process_complete(struct work_queue
*dummy
)
1902 struct zebra_vrf
*zvrf
;
1904 /* Evaluate nexthops for those VRFs which underwent route processing.
1906 * should limit the evaluation to the necessary VRFs in most common
1909 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
1911 if (zvrf
== NULL
|| !(zvrf
->flags
& ZEBRA_VRF_RIB_SCHEDULED
))
1914 zvrf
->flags
&= ~ZEBRA_VRF_RIB_SCHEDULED
;
1915 zebra_evaluate_rnh(zvrf
, AF_INET
, 0, RNH_NEXTHOP_TYPE
, NULL
);
1916 zebra_evaluate_rnh(zvrf
, AF_INET
, 0, RNH_IMPORT_CHECK_TYPE
,
1918 zebra_evaluate_rnh(zvrf
, AF_INET6
, 0, RNH_NEXTHOP_TYPE
, NULL
);
1919 zebra_evaluate_rnh(zvrf
, AF_INET6
, 0, RNH_IMPORT_CHECK_TYPE
,
1923 /* Schedule LSPs for processing, if needed. */
1924 zvrf
= vrf_info_lookup(VRF_DEFAULT
);
1925 if (mpls_should_lsps_be_processed(zvrf
)) {
1926 if (IS_ZEBRA_DEBUG_MPLS
)
1928 "%u: Scheduling all LSPs upon RIB completion",
1930 zebra_mpls_lsp_schedule(zvrf
);
1931 mpls_unmark_lsps_for_processing(zvrf
);
1935 /* Dispatch the meta queue by picking, processing and unlocking the next RN from
1936 * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and
1938 * is pointed to the meta queue structure.
1940 static wq_item_status
meta_queue_process(struct work_queue
*dummy
, void *data
)
1942 struct meta_queue
*mq
= data
;
1945 for (i
= 0; i
< MQ_SIZE
; i
++)
1946 if (process_subq(mq
->subq
[i
], i
)) {
1950 return mq
->size
? WQ_REQUEUE
: WQ_SUCCESS
;
1954 * Map from rib types to queue type (priority) in meta queue
1956 static const uint8_t meta_queue_map
[ZEBRA_ROUTE_MAX
] = {
1957 [ZEBRA_ROUTE_SYSTEM
] = 4,
1958 [ZEBRA_ROUTE_KERNEL
] = 0,
1959 [ZEBRA_ROUTE_CONNECT
] = 0,
1960 [ZEBRA_ROUTE_STATIC
] = 1,
1961 [ZEBRA_ROUTE_RIP
] = 2,
1962 [ZEBRA_ROUTE_RIPNG
] = 2,
1963 [ZEBRA_ROUTE_OSPF
] = 2,
1964 [ZEBRA_ROUTE_OSPF6
] = 2,
1965 [ZEBRA_ROUTE_ISIS
] = 2,
1966 [ZEBRA_ROUTE_BGP
] = 3,
1967 [ZEBRA_ROUTE_PIM
] = 4, // Shouldn't happen but for safety
1968 [ZEBRA_ROUTE_EIGRP
] = 2,
1969 [ZEBRA_ROUTE_NHRP
] = 2,
1970 [ZEBRA_ROUTE_HSLS
] = 4,
1971 [ZEBRA_ROUTE_OLSR
] = 4,
1972 [ZEBRA_ROUTE_TABLE
] = 1,
1973 [ZEBRA_ROUTE_LDP
] = 4,
1974 [ZEBRA_ROUTE_VNC
] = 3,
1975 [ZEBRA_ROUTE_VNC_DIRECT
] = 3,
1976 [ZEBRA_ROUTE_VNC_DIRECT_RH
] = 3,
1977 [ZEBRA_ROUTE_BGP_DIRECT
] = 3,
1978 [ZEBRA_ROUTE_BGP_DIRECT_EXT
] = 3,
1979 [ZEBRA_ROUTE_BABEL
] = 2,
1980 [ZEBRA_ROUTE_ALL
] = 4, // Shouldn't happen but for safety
1983 /* Look into the RN and queue it into one or more priority queues,
1984 * increasing the size for each data push done.
1986 static void rib_meta_queue_add(struct meta_queue
*mq
, struct route_node
*rn
)
1988 struct route_entry
*re
;
1990 RNODE_FOREACH_RE (rn
, re
) {
1991 uint8_t qindex
= meta_queue_map
[re
->type
];
1992 struct zebra_vrf
*zvrf
;
1994 /* Invariant: at this point we always have rn->info set. */
1995 if (CHECK_FLAG(rib_dest_from_rnode(rn
)->flags
,
1996 RIB_ROUTE_QUEUED(qindex
))) {
1997 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2000 "rn %p is already queued in sub-queue %u",
2001 (void *)rn
, qindex
);
2005 SET_FLAG(rib_dest_from_rnode(rn
)->flags
,
2006 RIB_ROUTE_QUEUED(qindex
));
2007 listnode_add(mq
->subq
[qindex
], rn
);
2008 route_lock_node(rn
);
2011 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2012 rnode_debug(rn
, re
->vrf_id
,
2013 "queued rn %p into sub-queue %u",
2014 (void *)rn
, qindex
);
2016 zvrf
= zebra_vrf_lookup_by_id(re
->vrf_id
);
2018 zvrf
->flags
|= ZEBRA_VRF_RIB_SCHEDULED
;
2022 /* Add route_node to work queue and schedule processing */
2023 void rib_queue_add(struct route_node
*rn
)
2027 /* Pointless to queue a route_node with no RIB entries to add or remove
2029 if (!rnode_to_ribs(rn
)) {
2030 zlog_debug("%s: called for route_node (%p, %d) with no ribs",
2031 __func__
, (void *)rn
, rn
->lock
);
2032 zlog_backtrace(LOG_DEBUG
);
2036 if (zebrad
.ribq
== NULL
) {
2037 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2038 "%s: work_queue does not exist!", __func__
);
2043 * The RIB queue should normally be either empty or holding the only
2044 * work_queue_item element. In the latter case this element would
2045 * hold a pointer to the meta queue structure, which must be used to
2046 * actually queue the route nodes to process. So create the MQ
2047 * holder, if necessary, then push the work into it in any case.
2048 * This semantics was introduced after 0.99.9 release.
2050 if (work_queue_empty(zebrad
.ribq
))
2051 work_queue_add(zebrad
.ribq
, zebrad
.mq
);
2053 rib_meta_queue_add(zebrad
.mq
, rn
);
2058 /* Create new meta queue.
2059 A destructor function doesn't seem to be necessary here.
2061 static struct meta_queue
*meta_queue_new(void)
2063 struct meta_queue
*new;
2066 new = XCALLOC(MTYPE_WORK_QUEUE
, sizeof(struct meta_queue
));
2068 for (i
= 0; i
< MQ_SIZE
; i
++) {
2069 new->subq
[i
] = list_new();
2070 assert(new->subq
[i
]);
2076 void meta_queue_free(struct meta_queue
*mq
)
2080 for (i
= 0; i
< MQ_SIZE
; i
++)
2081 list_delete(&mq
->subq
[i
]);
2083 XFREE(MTYPE_WORK_QUEUE
, mq
);
2086 /* initialise zebra rib work queue */
2087 static void rib_queue_init(struct zebra_t
*zebra
)
2092 work_queue_new(zebra
->master
, "route_node processing"))) {
2093 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2094 "%s: could not initialise work queue!", __func__
);
2098 /* fill in the work queue spec */
2099 zebra
->ribq
->spec
.workfunc
= &meta_queue_process
;
2100 zebra
->ribq
->spec
.errorfunc
= NULL
;
2101 zebra
->ribq
->spec
.completion_func
= &meta_queue_process_complete
;
2102 /* XXX: TODO: These should be runtime configurable via vty */
2103 zebra
->ribq
->spec
.max_retries
= 3;
2104 zebra
->ribq
->spec
.hold
= ZEBRA_RIB_PROCESS_HOLD_TIME
;
2106 if (!(zebra
->mq
= meta_queue_new())) {
2107 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2108 "%s: could not initialise meta queue!", __func__
);
2114 /* RIB updates are processed via a queue of pointers to route_nodes.
2116 * The queue length is bounded by the maximal size of the routing table,
2117 * as a route_node will not be requeued, if already queued.
2119 * REs are submitted via rib_addnode or rib_delnode which set minimal
2120 * state, or static_install_route (when an existing RE is updated)
2121 * and then submit route_node to queue for best-path selection later.
2122 * Order of add/delete state changes are preserved for any given RE.
2124 * Deleted REs are reaped during best-path selection.
2127 * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with
2128 * |-------->| | best RE, if required
2130 * static_install->|->rib_addqueue...... -> rib_process
2132 * |-------->| |-> rib_unlink
2133 * |-> set ROUTE_ENTRY_REMOVE |
2134 * rib_delnode (RE freed)
2136 * The 'info' pointer of a route_node points to a rib_dest_t
2137 * ('dest'). Queueing state for a route_node is kept on the dest. The
2138 * dest is created on-demand by rib_link() and is kept around at least
2139 * as long as there are ribs hanging off it (@see rib_gc_dest()).
2141 * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code):
2143 * - route_nodes: refcounted by:
2144 * - dest attached to route_node:
2145 * - managed by: rib_link/rib_gc_dest
2146 * - route_node processing queue
2147 * - managed by: rib_addqueue, rib_process.
2151 /* Add RE to head of the route node. */
2152 static void rib_link(struct route_node
*rn
, struct route_entry
*re
, int process
)
2154 struct route_entry
*head
;
2157 const char *rmap_name
;
2161 dest
= rib_dest_from_rnode(rn
);
2163 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2164 rnode_debug(rn
, re
->vrf_id
, "rn %p adding dest", rn
);
2166 dest
= XCALLOC(MTYPE_RIB_DEST
, sizeof(rib_dest_t
));
2167 route_lock_node(rn
); /* rn route table reference */
2172 head
= dest
->routes
;
2179 afi
= (rn
->p
.family
== AF_INET
)
2181 : (rn
->p
.family
== AF_INET6
) ? AFI_IP6
: AFI_MAX
;
2182 if (is_zebra_import_table_enabled(afi
, re
->table
)) {
2183 rmap_name
= zebra_get_import_table_route_map(afi
, re
->table
);
2184 zebra_add_import_table_entry(rn
, re
, rmap_name
);
2189 static void rib_addnode(struct route_node
*rn
,
2190 struct route_entry
*re
, int process
)
2192 /* RE node has been un-removed before route-node is processed.
2193 * route_node must hence already be on the queue for processing..
2195 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
2196 if (IS_ZEBRA_DEBUG_RIB
)
2197 rnode_debug(rn
, re
->vrf_id
, "rn %p, un-removed re %p",
2198 (void *)rn
, (void *)re
);
2200 UNSET_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
);
2203 rib_link(rn
, re
, process
);
2209 * Detach a rib structure from a route_node.
2211 * Note that a call to rib_unlink() should be followed by a call to
2212 * rib_gc_dest() at some point. This allows a rib_dest_t that is no
2213 * longer required to be deleted.
2215 void rib_unlink(struct route_node
*rn
, struct route_entry
*re
)
2221 if (IS_ZEBRA_DEBUG_RIB
)
2222 rnode_debug(rn
, re
->vrf_id
, "rn %p, re %p", (void *)rn
,
2225 dest
= rib_dest_from_rnode(rn
);
2228 re
->next
->prev
= re
->prev
;
2231 re
->prev
->next
= re
->next
;
2233 dest
->routes
= re
->next
;
2236 if (dest
->selected_fib
== re
)
2237 dest
->selected_fib
= NULL
;
2239 nexthops_free(re
->ng
.nexthop
);
2240 XFREE(MTYPE_RE
, re
);
2243 void rib_delnode(struct route_node
*rn
, struct route_entry
*re
)
2247 if (IS_ZEBRA_DEBUG_RIB
)
2248 rnode_debug(rn
, re
->vrf_id
, "rn %p, re %p, removing",
2249 (void *)rn
, (void *)re
);
2250 SET_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
);
2252 afi
= (rn
->p
.family
== AF_INET
)
2254 : (rn
->p
.family
== AF_INET6
) ? AFI_IP6
: AFI_MAX
;
2255 if (is_zebra_import_table_enabled(afi
, re
->table
)) {
2256 zebra_del_import_table_entry(rn
, re
);
2257 /* Just clean up if non main table */
2258 if (IS_ZEBRA_DEBUG_RIB
) {
2259 char buf
[SRCDEST2STR_BUFFER
];
2260 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
2262 "%u:%s: Freeing route rn %p, re %p (type %d)",
2263 re
->vrf_id
, buf
, rn
, re
, re
->type
);
2272 /* This function dumps the contents of a given RE entry into
2273 * standard debug log. Calling function name and IP prefix in
2274 * question are passed as 1st and 2nd arguments.
2277 void _route_entry_dump(const char *func
, union prefixconstptr pp
,
2278 union prefixconstptr src_pp
,
2279 const struct route_entry
*re
)
2281 const struct prefix
*src_p
= src_pp
.p
;
2282 bool is_srcdst
= src_p
&& src_p
->prefixlen
;
2283 char straddr
[PREFIX_STRLEN
];
2284 char srcaddr
[PREFIX_STRLEN
];
2285 struct nexthop
*nexthop
;
2287 zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func
,
2288 (const void *)re
, prefix2str(pp
, straddr
, sizeof(straddr
)),
2289 is_srcdst
? " from " : "",
2290 is_srcdst
? prefix2str(src_pp
, srcaddr
, sizeof(srcaddr
))
2293 zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d",
2294 func
, (unsigned long)re
->uptime
, re
->type
, re
->instance
,
2297 "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u",
2298 func
, re
->metric
, re
->mtu
, re
->distance
, re
->flags
, re
->status
);
2299 zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func
,
2300 re
->nexthop_num
, re
->nexthop_active_num
);
2302 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
2303 struct interface
*ifp
;
2304 struct vrf
*vrf
= vrf_lookup_by_id(nexthop
->vrf_id
);
2306 switch (nexthop
->type
) {
2307 case NEXTHOP_TYPE_BLACKHOLE
:
2308 sprintf(straddr
, "Blackhole");
2310 case NEXTHOP_TYPE_IFINDEX
:
2311 ifp
= if_lookup_by_index(nexthop
->ifindex
,
2313 sprintf(straddr
, "%s", ifp
? ifp
->name
: "Unknown");
2315 case NEXTHOP_TYPE_IPV4
:
2317 case NEXTHOP_TYPE_IPV4_IFINDEX
:
2318 inet_ntop(AF_INET
, &nexthop
->gate
, straddr
,
2321 case NEXTHOP_TYPE_IPV6
:
2322 case NEXTHOP_TYPE_IPV6_IFINDEX
:
2323 inet_ntop(AF_INET6
, &nexthop
->gate
, straddr
,
2327 zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s", func
,
2328 (nexthop
->rparent
? " NH" : "NH"), straddr
,
2329 nexthop
->ifindex
, vrf
? vrf
->name
: "Unknown",
2331 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
)
2334 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
)
2337 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
)
2341 zlog_debug("%s: dump complete", func
);
2344 /* This is an exported helper to rtm_read() to dump the strange
2345 * RE entry found by rib_lookup_ipv4_route()
2348 void rib_lookup_and_dump(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
2350 struct route_table
*table
;
2351 struct route_node
*rn
;
2352 struct route_entry
*re
;
2353 char prefix_buf
[INET_ADDRSTRLEN
];
2356 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
2358 flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED
,
2359 "%s:%u zebra_vrf_table() returned NULL", __func__
,
2364 /* Scan the RIB table for exactly matching RE entry. */
2365 rn
= route_node_lookup(table
, (struct prefix
*)p
);
2367 /* No route for this prefix. */
2369 zlog_debug("%s:%u lookup failed for %s", __func__
, vrf_id
,
2370 prefix2str((struct prefix
*)p
, prefix_buf
,
2371 sizeof(prefix_buf
)));
2376 route_unlock_node(rn
);
2379 RNODE_FOREACH_RE (rn
, re
) {
2380 zlog_debug("%s:%u rn %p, re %p: %s, %s",
2382 (void *)rn
, (void *)re
,
2383 (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)
2386 (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)
2389 route_entry_dump(p
, NULL
, re
);
2393 /* Check if requested address assignment will fail due to another
2394 * route being installed by zebra in FIB already. Take necessary
2395 * actions, if needed: remove such a route from FIB and deSELECT
2396 * corresponding RE entry. Then put affected RN into RIBQ head.
2398 void rib_lookup_and_pushup(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
2400 struct route_table
*table
;
2401 struct route_node
*rn
;
2402 unsigned changed
= 0;
2405 if (NULL
== (table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
))) {
2406 flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED
,
2407 "%s:%u zebra_vrf_table() returned NULL", __func__
,
2412 /* No matches would be the simplest case. */
2413 if (NULL
== (rn
= route_node_lookup(table
, (struct prefix
*)p
)))
2417 route_unlock_node(rn
);
2419 dest
= rib_dest_from_rnode(rn
);
2420 /* Check all RE entries. In case any changes have to be done, requeue
2421 * the RN into RIBQ head. If the routing message about the new connected
2422 * route (generated by the IP address we are going to assign very soon)
2423 * comes before the RIBQ is processed, the new RE entry will join
2424 * RIBQ record already on head. This is necessary for proper
2426 * of the rest of the RE.
2428 if (dest
->selected_fib
&& !RIB_SYSTEM_ROUTE(dest
->selected_fib
)) {
2430 if (IS_ZEBRA_DEBUG_RIB
) {
2431 char buf
[PREFIX_STRLEN
];
2433 zlog_debug("%u:%s: freeing way for connected prefix",
2434 dest
->selected_fib
->vrf_id
,
2435 prefix2str(&rn
->p
, buf
, sizeof(buf
)));
2436 route_entry_dump(&rn
->p
, NULL
, dest
->selected_fib
);
2438 rib_uninstall(rn
, dest
->selected_fib
);
2444 int rib_add_multipath(afi_t afi
, safi_t safi
, struct prefix
*p
,
2445 struct prefix_ipv6
*src_p
, struct route_entry
*re
)
2447 struct route_table
*table
;
2448 struct route_node
*rn
;
2449 struct route_entry
*same
= NULL
;
2450 struct nexthop
*nexthop
;
2456 assert(!src_p
|| !src_p
->prefixlen
|| afi
== AFI_IP6
);
2459 table
= zebra_vrf_table_with_table_id(afi
, safi
, re
->vrf_id
, re
->table
);
2461 XFREE(MTYPE_RE
, re
);
2465 /* Make it sure prefixlen is applied to the prefix. */
2468 apply_mask_ipv6(src_p
);
2470 /* Set default distance by route type. */
2471 if (re
->distance
== 0) {
2472 re
->distance
= route_distance(re
->type
);
2474 /* iBGP distance is 200. */
2475 if (re
->type
== ZEBRA_ROUTE_BGP
2476 && CHECK_FLAG(re
->flags
, ZEBRA_FLAG_IBGP
))
2480 /* Lookup route node.*/
2481 rn
= srcdest_rnode_get(table
, p
, src_p
);
2484 * If same type of route are installed, treat it as a implicit
2486 * If the user has specified the No route replace semantics
2487 * for the install don't do a route replace.
2489 RNODE_FOREACH_RE (rn
, same
) {
2490 if (CHECK_FLAG(same
->status
, ROUTE_ENTRY_REMOVED
))
2493 if (same
->type
!= re
->type
)
2495 if (same
->instance
!= re
->instance
)
2497 if (same
->type
== ZEBRA_ROUTE_KERNEL
2498 && same
->metric
!= re
->metric
)
2501 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_RR_USE_DISTANCE
) &&
2502 same
->distance
!= re
->distance
)
2506 * We should allow duplicate connected routes
2507 * because of IPv6 link-local routes and unnumbered
2508 * interfaces on Linux.
2510 if (same
->type
!= ZEBRA_ROUTE_CONNECT
)
2514 /* If this route is kernel route, set FIB flag to the route. */
2515 if (RIB_SYSTEM_ROUTE(re
))
2516 for (nexthop
= re
->ng
.nexthop
; nexthop
; nexthop
= nexthop
->next
)
2517 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
2519 /* Link new re to node.*/
2520 if (IS_ZEBRA_DEBUG_RIB
) {
2523 "Inserting route rn %p, re %p (type %d) existing %p",
2524 (void *)rn
, (void *)re
, re
->type
, (void *)same
);
2526 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2527 route_entry_dump(p
, src_p
, re
);
2529 rib_addnode(rn
, re
, 1);
2532 /* Free implicit route.*/
2534 rib_delnode(rn
, same
);
2538 route_unlock_node(rn
);
2542 void rib_delete(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
, int type
,
2543 unsigned short instance
, int flags
, struct prefix
*p
,
2544 struct prefix_ipv6
*src_p
, const struct nexthop
*nh
,
2545 uint32_t table_id
, uint32_t metric
, uint8_t distance
,
2548 struct route_table
*table
;
2549 struct route_node
*rn
;
2550 struct route_entry
*re
;
2551 struct route_entry
*fib
= NULL
;
2552 struct route_entry
*same
= NULL
;
2553 struct nexthop
*rtnh
;
2554 char buf2
[INET6_ADDRSTRLEN
];
2557 assert(!src_p
|| !src_p
->prefixlen
|| afi
== AFI_IP6
);
2560 table
= zebra_vrf_table_with_table_id(afi
, safi
, vrf_id
, table_id
);
2567 apply_mask_ipv6(src_p
);
2569 /* Lookup route node. */
2570 rn
= srcdest_rnode_lookup(table
, p
, src_p
);
2572 char dst_buf
[PREFIX_STRLEN
], src_buf
[PREFIX_STRLEN
];
2574 prefix2str(p
, dst_buf
, sizeof(dst_buf
));
2575 if (src_p
&& src_p
->prefixlen
)
2576 prefix2str(src_p
, src_buf
, sizeof(src_buf
));
2580 if (IS_ZEBRA_DEBUG_RIB
)
2581 zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id
,
2583 (src_buf
[0] != '\0') ? " from " : "",
2588 dest
= rib_dest_from_rnode(rn
);
2589 fib
= dest
->selected_fib
;
2591 /* Lookup same type route. */
2592 RNODE_FOREACH_RE (rn
, re
) {
2593 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
2596 if (re
->type
!= type
)
2598 if (re
->instance
!= instance
)
2600 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_RR_USE_DISTANCE
) &&
2601 distance
!= re
->distance
)
2604 if (re
->type
== ZEBRA_ROUTE_KERNEL
&& re
->metric
!= metric
)
2606 if (re
->type
== ZEBRA_ROUTE_CONNECT
&& (rtnh
= re
->ng
.nexthop
)
2607 && rtnh
->type
== NEXTHOP_TYPE_IFINDEX
&& nh
) {
2608 if (rtnh
->ifindex
!= nh
->ifindex
)
2613 /* Make sure that the route found has the same gateway. */
2619 for (ALL_NEXTHOPS(re
->ng
, rtnh
))
2620 if (nexthop_same_no_recurse(rtnh
, nh
)) {
2628 /* If same type of route can't be found and this message is from
2632 * In the past(HA!) we could get here because
2633 * we were receiving a route delete from the
2634 * kernel and we're not marking the proto
2635 * as coming from it's appropriate originator.
2636 * Now that we are properly noticing the fact
2637 * that the kernel has deleted our route we
2638 * are not going to get called in this path
2639 * I am going to leave this here because
2640 * this might still work this way on non-linux
2641 * platforms as well as some weird state I have
2642 * not properly thought of yet.
2643 * If we can show that this code path is
2644 * dead then we can remove it.
2646 if (fib
&& CHECK_FLAG(flags
, ZEBRA_FLAG_SELFROUTE
)) {
2647 if (IS_ZEBRA_DEBUG_RIB
) {
2650 "rn %p, re %p (type %d) was deleted from kernel, adding",
2651 rn
, fib
, fib
->type
);
2655 for (rtnh
= fib
->ng
.nexthop
; rtnh
;
2657 UNSET_FLAG(rtnh
->flags
,
2661 * This is a non FRR route
2662 * as such we should mark
2665 dest
->selected_fib
= NULL
;
2667 /* This means someone else, other than Zebra,
2669 * a Zebra router from the kernel. We will add
2671 rib_install_kernel(rn
, fib
, NULL
);
2674 if (IS_ZEBRA_DEBUG_RIB
) {
2678 "via %s ifindex %d type %d "
2679 "doesn't exist in rib",
2680 inet_ntop(afi2family(afi
),
2687 "type %d doesn't exist in rib",
2690 route_unlock_node(rn
);
2696 if (fromkernel
&& CHECK_FLAG(flags
, ZEBRA_FLAG_SELFROUTE
)
2698 rib_install_kernel(rn
, same
, NULL
);
2699 route_unlock_node(rn
);
2704 if (CHECK_FLAG(flags
, ZEBRA_FLAG_EVPN_ROUTE
)) {
2705 struct nexthop
*tmp_nh
;
2707 for (ALL_NEXTHOPS(re
->ng
, tmp_nh
)) {
2708 struct ipaddr vtep_ip
;
2710 memset(&vtep_ip
, 0, sizeof(struct ipaddr
));
2711 if (afi
== AFI_IP
) {
2712 vtep_ip
.ipa_type
= IPADDR_V4
;
2713 memcpy(&(vtep_ip
.ipaddr_v4
),
2714 &(tmp_nh
->gate
.ipv4
),
2715 sizeof(struct in_addr
));
2717 vtep_ip
.ipa_type
= IPADDR_V6
;
2718 memcpy(&(vtep_ip
.ipaddr_v6
),
2719 &(tmp_nh
->gate
.ipv6
),
2720 sizeof(struct in6_addr
));
2722 zebra_vxlan_evpn_vrf_route_del(re
->vrf_id
,
2726 rib_delnode(rn
, same
);
2729 route_unlock_node(rn
);
2734 int rib_add(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
, int type
,
2735 unsigned short instance
, int flags
, struct prefix
*p
,
2736 struct prefix_ipv6
*src_p
, const struct nexthop
*nh
,
2737 uint32_t table_id
, uint32_t metric
, uint32_t mtu
, uint8_t distance
,
2740 struct route_entry
*re
;
2741 struct nexthop
*nexthop
;
2743 /* Allocate new route_entry structure. */
2744 re
= XCALLOC(MTYPE_RE
, sizeof(struct route_entry
));
2746 re
->instance
= instance
;
2747 re
->distance
= distance
;
2749 re
->metric
= metric
;
2751 re
->table
= table_id
;
2752 re
->vrf_id
= vrf_id
;
2753 re
->nexthop_num
= 0;
2754 re
->uptime
= time(NULL
);
2758 nexthop
= nexthop_new();
2760 route_entry_nexthop_add(re
, nexthop
);
2762 return rib_add_multipath(afi
, safi
, p
, src_p
, re
);
2765 /* Schedule routes of a particular table (address-family) based on event. */
2766 void rib_update_table(struct route_table
*table
, rib_update_event_t event
)
2768 struct route_node
*rn
;
2769 struct route_entry
*re
, *next
;
2771 /* Walk all routes and queue for processing, if appropriate for
2772 * the trigger event.
2774 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
2776 * If we are looking at a route node and the node
2777 * has already been queued we don't
2778 * need to queue it up again
2780 if (rn
->info
&& CHECK_FLAG(rib_dest_from_rnode(rn
)->flags
,
2781 RIB_ROUTE_ANY_QUEUED
))
2784 case RIB_UPDATE_IF_CHANGE
:
2785 /* Examine all routes that won't get processed by the
2787 * triggered by nexthop evaluation (NHT). This would be
2789 * kernel and certain static routes. Note that NHT will
2791 * triggered upon an interface event as connected routes
2793 * get queued for processing.
2795 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
2798 if (re
->type
!= ZEBRA_ROUTE_SYSTEM
2799 && re
->type
!= ZEBRA_ROUTE_KERNEL
2800 && re
->type
!= ZEBRA_ROUTE_CONNECT
2801 && re
->type
!= ZEBRA_ROUTE_STATIC
)
2804 if (re
->type
!= ZEBRA_ROUTE_STATIC
) {
2809 for (nh
= re
->ng
.nexthop
; nh
; nh
= nh
->next
)
2810 if (!(nh
->type
== NEXTHOP_TYPE_IPV4
2811 || nh
->type
== NEXTHOP_TYPE_IPV6
))
2814 /* If we only have nexthops to a
2823 case RIB_UPDATE_RMAP_CHANGE
:
2824 case RIB_UPDATE_OTHER
:
2825 /* Right now, examine all routes. Can restrict to a
2827 * some cases (TODO).
2829 if (rnode_to_ribs(rn
))
2839 /* RIB update function. */
2840 void rib_update(vrf_id_t vrf_id
, rib_update_event_t event
)
2842 struct route_table
*table
;
2844 /* Process routes of interested address-families. */
2845 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
2847 if (IS_ZEBRA_DEBUG_EVENT
)
2848 zlog_debug("%s : AFI_IP event %d", __func__
, event
);
2849 rib_update_table(table
, event
);
2852 table
= zebra_vrf_table(AFI_IP6
, SAFI_UNICAST
, vrf_id
);
2854 if (IS_ZEBRA_DEBUG_EVENT
)
2855 zlog_debug("%s : AFI_IP6 event %d", __func__
, event
);
2856 rib_update_table(table
, event
);
2860 /* Delete self installed routes after zebra is relaunched. */
2861 void rib_sweep_table(struct route_table
*table
)
2863 struct route_node
*rn
;
2864 struct route_entry
*re
;
2865 struct route_entry
*next
;
2866 struct nexthop
*nexthop
;
2871 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
2872 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
2873 if (IS_ZEBRA_DEBUG_RIB
)
2874 route_entry_dump(&rn
->p
, NULL
, re
);
2876 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
2879 if (!CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELFROUTE
))
2883 * So we are starting up and have received
2884 * routes from the kernel that we have installed
2885 * from a previous run of zebra but not cleaned
2886 * up ( say a kill -9 )
2887 * But since we haven't actually installed
2888 * them yet( we received them from the kernel )
2889 * we don't think they are active.
2890 * So let's pretend they are active to actually
2892 * In all honesty I'm not sure if we should
2893 * mark them as active when we receive them
2894 * This is startup only so probably ok.
2896 * If we ever decide to move rib_sweep_table
2897 * to a different spot (ie startup )
2898 * this decision needs to be revisited
2900 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
2901 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
2903 rib_uninstall_kernel(rn
, re
);
2904 rib_delnode(rn
, re
);
2909 /* Sweep all RIB tables. */
2910 void rib_sweep_route(void)
2913 struct zebra_vrf
*zvrf
;
2915 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
2916 if ((zvrf
= vrf
->info
) == NULL
)
2919 rib_sweep_table(zvrf
->table
[AFI_IP
][SAFI_UNICAST
]);
2920 rib_sweep_table(zvrf
->table
[AFI_IP6
][SAFI_UNICAST
]);
2923 zebra_ns_sweep_route();
2926 /* Remove specific by protocol routes from 'table'. */
2927 unsigned long rib_score_proto_table(uint8_t proto
, unsigned short instance
,
2928 struct route_table
*table
)
2930 struct route_node
*rn
;
2931 struct route_entry
*re
;
2932 struct route_entry
*next
;
2933 unsigned long n
= 0;
2936 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
))
2937 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
2938 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
2940 if (re
->type
== proto
2941 && re
->instance
== instance
) {
2942 rib_delnode(rn
, re
);
2949 /* Remove specific by protocol routes. */
2950 unsigned long rib_score_proto(uint8_t proto
, unsigned short instance
)
2953 struct zebra_vrf
*zvrf
;
2954 unsigned long cnt
= 0;
2956 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
)
2957 if ((zvrf
= vrf
->info
) != NULL
)
2958 cnt
+= rib_score_proto_table(
2960 zvrf
->table
[AFI_IP
][SAFI_UNICAST
])
2961 + rib_score_proto_table(
2963 zvrf
->table
[AFI_IP6
][SAFI_UNICAST
]);
2965 cnt
+= zebra_ns_score_proto(proto
, instance
);
2970 /* Close RIB and clean up kernel routes. */
2971 void rib_close_table(struct route_table
*table
)
2973 struct route_node
*rn
;
2974 rib_table_info_t
*info
;
2980 info
= route_table_get_info(table
);
2982 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
2983 dest
= rib_dest_from_rnode(rn
);
2985 if (dest
&& dest
->selected_fib
) {
2986 if (info
->safi
== SAFI_UNICAST
)
2987 hook_call(rib_update
, rn
, NULL
);
2989 if (!RIB_SYSTEM_ROUTE(dest
->selected_fib
))
2990 rib_uninstall_kernel(rn
, dest
->selected_fib
);
2995 /* Routing information base initialize. */
2998 rib_queue_init(&zebrad
);
3004 * Get the first vrf id that is greater than the given vrf id if any.
3006 * Returns TRUE if a vrf id was found, FALSE otherwise.
3008 static inline int vrf_id_get_next(vrf_id_t vrf_id
, vrf_id_t
*next_id_p
)
3012 vrf
= vrf_lookup_by_id(vrf_id
);
3014 vrf
= RB_NEXT(vrf_id_head
, vrf
);
3016 *next_id_p
= vrf
->vrf_id
;
3025 * rib_tables_iter_next
3027 * Returns the next table in the iteration.
3029 struct route_table
*rib_tables_iter_next(rib_tables_iter_t
*iter
)
3031 struct route_table
*table
;
3034 * Array that helps us go over all AFI/SAFI combinations via one
3041 {AFI_IP
, SAFI_UNICAST
}, {AFI_IP
, SAFI_MULTICAST
},
3042 {AFI_IP
, SAFI_LABELED_UNICAST
}, {AFI_IP6
, SAFI_UNICAST
},
3043 {AFI_IP6
, SAFI_MULTICAST
}, {AFI_IP6
, SAFI_LABELED_UNICAST
},
3048 switch (iter
->state
) {
3050 case RIB_TABLES_ITER_S_INIT
:
3051 iter
->vrf_id
= VRF_DEFAULT
;
3052 iter
->afi_safi_ix
= -1;
3056 case RIB_TABLES_ITER_S_ITERATING
:
3057 iter
->afi_safi_ix
++;
3060 while (iter
->afi_safi_ix
3061 < (int)ZEBRA_NUM_OF(afi_safis
)) {
3062 table
= zebra_vrf_table(
3063 afi_safis
[iter
->afi_safi_ix
].afi
,
3064 afi_safis
[iter
->afi_safi_ix
].safi
,
3069 iter
->afi_safi_ix
++;
3073 * Found another table in this vrf.
3079 * Done with all tables in the current vrf, go to the
3083 if (!vrf_id_get_next(iter
->vrf_id
, &iter
->vrf_id
))
3086 iter
->afi_safi_ix
= 0;
3091 case RIB_TABLES_ITER_S_DONE
:
3096 iter
->state
= RIB_TABLES_ITER_S_ITERATING
;
3098 iter
->state
= RIB_TABLES_ITER_S_DONE
;