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 #define RIB_KERNEL_ROUTE(R) \
882 ((R)->type == ZEBRA_ROUTE_KERNEL)
884 /* This function verifies reachability of one given nexthop, which can be
885 * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored
886 * in nexthop->flags field. If the 4th parameter, 'set', is non-zero,
887 * nexthop->ifindex will be updated appropriately as well.
888 * An existing route map can turn (otherwise active) nexthop into inactive, but
891 * The return value is the final value of 'ACTIVE' flag.
894 static unsigned nexthop_active_check(struct route_node
*rn
,
895 struct route_entry
*re
,
896 struct nexthop
*nexthop
, int set
)
898 struct interface
*ifp
;
899 route_map_result_t ret
= RMAP_MATCH
;
901 char buf
[SRCDEST2STR_BUFFER
];
902 const struct prefix
*p
, *src_p
;
903 struct zebra_vrf
*zvrf
;
905 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
907 if (rn
->p
.family
== AF_INET
)
909 else if (rn
->p
.family
== AF_INET6
)
913 switch (nexthop
->type
) {
914 case NEXTHOP_TYPE_IFINDEX
:
915 ifp
= if_lookup_by_index(nexthop
->ifindex
, nexthop
->vrf_id
);
916 if (ifp
&& if_is_operative(ifp
))
917 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
919 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
921 case NEXTHOP_TYPE_IPV4
:
922 case NEXTHOP_TYPE_IPV4_IFINDEX
:
924 if (nexthop_active(AFI_IP
, re
, nexthop
, set
, rn
))
925 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
927 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
929 case NEXTHOP_TYPE_IPV6
:
931 if (nexthop_active(AFI_IP6
, re
, nexthop
, set
, rn
))
932 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
934 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
936 case NEXTHOP_TYPE_IPV6_IFINDEX
:
937 /* RFC 5549, v4 prefix with v6 NH */
938 if (rn
->p
.family
!= AF_INET
)
940 if (IN6_IS_ADDR_LINKLOCAL(&nexthop
->gate
.ipv6
)) {
941 ifp
= if_lookup_by_index(nexthop
->ifindex
,
943 if (ifp
&& if_is_operative(ifp
))
944 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
946 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
948 if (nexthop_active(AFI_IP6
, re
, nexthop
, set
, rn
))
949 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
951 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
954 case NEXTHOP_TYPE_BLACKHOLE
:
955 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
960 if (!CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
)) {
961 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
962 zlog_debug("\t%s: Unable to find a active nexthop",
963 __PRETTY_FUNCTION__
);
967 /* XXX: What exactly do those checks do? Do we support
968 * e.g. IPv4 routes with IPv6 nexthops or vice versa?
970 if (RIB_SYSTEM_ROUTE(re
) || (family
== AFI_IP
&& p
->family
!= AF_INET
)
971 || (family
== AFI_IP6
&& p
->family
!= AF_INET6
))
972 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
974 /* The original code didn't determine the family correctly
975 * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi
976 * from the rib_table_info in those cases.
977 * Possibly it may be better to use only the rib_table_info
981 rib_table_info_t
*info
;
983 info
= srcdest_rnode_table_info(rn
);
987 memset(&nexthop
->rmap_src
.ipv6
, 0, sizeof(union g_addr
));
989 zvrf
= zebra_vrf_lookup_by_id(nexthop
->vrf_id
);
991 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
992 zlog_debug("\t%s: zvrf is NULL", __PRETTY_FUNCTION__
);
993 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
996 /* It'll get set if required inside */
997 ret
= zebra_route_map_check(family
, re
->type
, re
->instance
, p
,
998 nexthop
, zvrf
, re
->tag
);
999 if (ret
== RMAP_DENYMATCH
) {
1000 if (IS_ZEBRA_DEBUG_RIB
) {
1001 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1003 "%u:%s: Filtering out with NH out %s due to route map",
1005 ifindex2ifname(nexthop
->ifindex
,
1008 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
1010 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
1013 /* Iterate over all nexthops of the given RIB entry and refresh their
1014 * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any
1015 * nexthop is found to toggle the ACTIVE flag, the whole re structure
1016 * is flagged with ROUTE_ENTRY_CHANGED. The 4th 'set' argument is
1017 * transparently passed to nexthop_active_check().
1019 * Return value is the new number of active nexthops.
1022 static int nexthop_active_update(struct route_node
*rn
, struct route_entry
*re
,
1025 struct nexthop
*nexthop
;
1026 union g_addr prev_src
;
1027 unsigned int prev_active
, new_active
, old_num_nh
;
1028 ifindex_t prev_index
;
1030 old_num_nh
= re
->nexthop_active_num
;
1032 re
->nexthop_active_num
= 0;
1033 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1035 for (nexthop
= re
->ng
.nexthop
; nexthop
; nexthop
= nexthop
->next
) {
1036 /* No protocol daemon provides src and so we're skipping
1038 prev_src
= nexthop
->rmap_src
;
1039 prev_active
= CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
1040 prev_index
= nexthop
->ifindex
;
1041 if ((new_active
= nexthop_active_check(rn
, re
, nexthop
, set
)))
1042 re
->nexthop_active_num
++;
1043 /* Don't allow src setting on IPv6 addr for now */
1044 if (prev_active
!= new_active
|| prev_index
!= nexthop
->ifindex
1045 || ((nexthop
->type
>= NEXTHOP_TYPE_IFINDEX
1046 && nexthop
->type
< NEXTHOP_TYPE_IPV6
)
1047 && prev_src
.ipv4
.s_addr
1048 != nexthop
->rmap_src
.ipv4
.s_addr
)
1049 || ((nexthop
->type
>= NEXTHOP_TYPE_IPV6
1050 && nexthop
->type
< NEXTHOP_TYPE_BLACKHOLE
)
1051 && !(IPV6_ADDR_SAME(&prev_src
.ipv6
,
1052 &nexthop
->rmap_src
.ipv6
)))) {
1053 SET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1054 SET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1058 if (old_num_nh
!= re
->nexthop_active_num
)
1059 SET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1061 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
)) {
1062 SET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1065 return re
->nexthop_active_num
;
1069 * Is this RIB labeled-unicast? It must be of type BGP and all paths
1070 * (nexthops) must have a label.
1072 int zebra_rib_labeled_unicast(struct route_entry
*re
)
1074 struct nexthop
*nexthop
= NULL
;
1076 if (re
->type
!= ZEBRA_ROUTE_BGP
)
1079 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1080 if (!nexthop
->nh_label
|| !nexthop
->nh_label
->num_labels
)
1086 void kernel_route_rib_pass_fail(struct route_node
*rn
, const struct prefix
*p
,
1087 struct route_entry
*re
,
1088 enum zebra_dplane_status res
)
1090 struct nexthop
*nexthop
;
1091 char buf
[PREFIX_STRLEN
];
1094 dest
= rib_dest_from_rnode(rn
);
1097 case ZEBRA_DPLANE_INSTALL_SUCCESS
:
1098 dest
->selected_fib
= re
;
1099 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1100 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
))
1103 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
))
1104 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1106 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1108 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_INSTALLED
);
1110 case ZEBRA_DPLANE_INSTALL_FAILURE
:
1112 * I am not sure this is the right thing to do here
1113 * but the code always set selected_fib before
1114 * this assignment was moved here.
1116 dest
->selected_fib
= re
;
1118 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_FAIL_INSTALL
);
1119 flog_err(EC_ZEBRA_DP_INSTALL_FAIL
,
1120 "%u:%s: Route install failed", re
->vrf_id
,
1121 prefix2str(p
, buf
, sizeof(buf
)));
1123 case ZEBRA_DPLANE_DELETE_SUCCESS
:
1125 * The case where selected_fib is not re is
1126 * when we have received a system route
1127 * that is overriding our installed route
1128 * as such we should leave the selected_fib
1131 if (dest
->selected_fib
== re
)
1132 dest
->selected_fib
= NULL
;
1133 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1134 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1136 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_REMOVED
);
1138 case ZEBRA_DPLANE_DELETE_FAILURE
:
1140 * Should we set this to NULL if the
1143 dest
->selected_fib
= NULL
;
1144 flog_err(EC_ZEBRA_DP_DELETE_FAIL
,
1145 "%u:%s: Route Deletion failure", re
->vrf_id
,
1146 prefix2str(p
, buf
, sizeof(buf
)));
1148 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_REMOVE_FAIL
);
1150 case ZEBRA_DPLANE_STATUS_NONE
:
1155 /* Update flag indicates whether this is a "replace" or not. Currently, this
1156 * is only used for IPv4.
1158 void rib_install_kernel(struct route_node
*rn
, struct route_entry
*re
,
1159 struct route_entry
*old
)
1161 struct nexthop
*nexthop
;
1162 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1163 const struct prefix
*p
, *src_p
;
1164 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1166 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1168 if (info
->safi
!= SAFI_UNICAST
) {
1169 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1170 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1173 struct nexthop
*prev
;
1175 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1176 UNSET_FLAG (nexthop
->flags
, NEXTHOP_FLAG_DUPLICATE
);
1177 for (ALL_NEXTHOPS(re
->ng
, prev
)) {
1178 if (prev
== nexthop
)
1180 if (nexthop_same_firsthop(nexthop
, prev
)) {
1181 SET_FLAG(nexthop
->flags
,
1182 NEXTHOP_FLAG_DUPLICATE
);
1190 * If this is a replace to a new RE let the originator of the RE
1191 * know that they've lost
1193 if (old
&& (old
!= re
) && (old
->type
!= re
->type
))
1194 zsend_route_notify_owner(old
, p
, ZAPI_ROUTE_BETTER_ADMIN_WON
);
1197 * Make sure we update the FPM any time we send new information to
1200 hook_call(rib_update
, rn
, "installing in kernel");
1201 switch (kernel_route_rib(rn
, p
, src_p
, old
, re
)) {
1202 case ZEBRA_DPLANE_REQUEST_QUEUED
:
1204 EC_ZEBRA_DP_INVALID_RC
,
1205 "No current known DataPlane interfaces can return this, please fix");
1207 case ZEBRA_DPLANE_REQUEST_FAILURE
:
1209 EC_ZEBRA_DP_INSTALL_FAIL
,
1210 "No current known Rib Install Failure cases, please fix");
1212 case ZEBRA_DPLANE_REQUEST_SUCCESS
:
1220 /* Uninstall the route from kernel. */
1221 void rib_uninstall_kernel(struct route_node
*rn
, struct route_entry
*re
)
1223 struct nexthop
*nexthop
;
1224 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1225 const struct prefix
*p
, *src_p
;
1226 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1228 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1230 if (info
->safi
!= SAFI_UNICAST
) {
1231 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1232 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1237 * Make sure we update the FPM any time we send new information to
1240 hook_call(rib_update
, rn
, "uninstalling from kernel");
1241 switch (kernel_route_rib(rn
, p
, src_p
, re
, NULL
)) {
1242 case ZEBRA_DPLANE_REQUEST_QUEUED
:
1244 EC_ZEBRA_DP_INVALID_RC
,
1245 "No current known DataPlane interfaces can return this, please fix");
1247 case ZEBRA_DPLANE_REQUEST_FAILURE
:
1249 EC_ZEBRA_DP_INSTALL_FAIL
,
1250 "No current known RIB Install Failure cases, please fix");
1252 case ZEBRA_DPLANE_REQUEST_SUCCESS
:
1261 /* Uninstall the route from kernel. */
1262 static void rib_uninstall(struct route_node
*rn
, struct route_entry
*re
)
1264 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1265 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1267 if (dest
&& dest
->selected_fib
== re
) {
1268 if (info
->safi
== SAFI_UNICAST
)
1269 hook_call(rib_update
, rn
, "rib_uninstall");
1271 if (!RIB_SYSTEM_ROUTE(re
))
1272 rib_uninstall_kernel(rn
, re
);
1274 /* If labeled-unicast route, uninstall transit LSP. */
1275 if (zebra_rib_labeled_unicast(re
))
1276 zebra_mpls_lsp_uninstall(info
->zvrf
, rn
, re
);
1279 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)) {
1280 const struct prefix
*p
, *src_p
;
1282 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1284 redistribute_delete(p
, src_p
, re
);
1285 UNSET_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
);
1290 * rib_can_delete_dest
1292 * Returns TRUE if the given dest can be deleted from the table.
1294 static int rib_can_delete_dest(rib_dest_t
*dest
)
1301 * Don't delete the dest if we have to update the FPM about this
1304 if (CHECK_FLAG(dest
->flags
, RIB_DEST_UPDATE_FPM
)
1305 || CHECK_FLAG(dest
->flags
, RIB_DEST_SENT_TO_FPM
))
1314 * Garbage collect the rib dest corresponding to the given route node
1317 * Returns TRUE if the dest was deleted, FALSE otherwise.
1319 int rib_gc_dest(struct route_node
*rn
)
1323 dest
= rib_dest_from_rnode(rn
);
1327 if (!rib_can_delete_dest(dest
))
1330 if (IS_ZEBRA_DEBUG_RIB
) {
1331 struct zebra_vrf
*zvrf
;
1333 zvrf
= rib_dest_vrf(dest
);
1334 rnode_debug(rn
, zvrf_id(zvrf
), "removing dest from table");
1338 XFREE(MTYPE_RIB_DEST
, dest
);
1342 * Release the one reference that we keep on the route node.
1344 route_unlock_node(rn
);
1348 static void rib_process_add_fib(struct zebra_vrf
*zvrf
, struct route_node
*rn
,
1349 struct route_entry
*new)
1351 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1353 hook_call(rib_update
, rn
, "new route selected");
1355 /* Update real nexthop. This may actually determine if nexthop is active
1357 if (!nexthop_active_update(rn
, new, 1)) {
1358 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1362 if (IS_ZEBRA_DEBUG_RIB
) {
1363 char buf
[SRCDEST2STR_BUFFER
];
1364 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1365 zlog_debug("%u:%s: Adding route rn %p, re %p (type %d)",
1366 zvrf_id(zvrf
), buf
, rn
, new, new->type
);
1369 /* If labeled-unicast route, install transit LSP. */
1370 if (zebra_rib_labeled_unicast(new))
1371 zebra_mpls_lsp_install(zvrf
, rn
, new);
1373 if (!RIB_SYSTEM_ROUTE(new))
1374 rib_install_kernel(rn
, new, NULL
);
1376 dest
->selected_fib
= new;
1378 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1381 static void rib_process_del_fib(struct zebra_vrf
*zvrf
, struct route_node
*rn
,
1382 struct route_entry
*old
)
1384 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1385 hook_call(rib_update
, rn
, "removing existing route");
1387 /* Uninstall from kernel. */
1388 if (IS_ZEBRA_DEBUG_RIB
) {
1389 char buf
[SRCDEST2STR_BUFFER
];
1390 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1391 zlog_debug("%u:%s: Deleting route rn %p, re %p (type %d)",
1392 zvrf_id(zvrf
), buf
, rn
, old
, old
->type
);
1395 /* If labeled-unicast route, uninstall transit LSP. */
1396 if (zebra_rib_labeled_unicast(old
))
1397 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1399 if (!RIB_SYSTEM_ROUTE(old
))
1400 rib_uninstall_kernel(rn
, old
);
1403 * We are setting this to NULL here
1404 * because that is what we traditionally
1405 * have been doing. I am not positive
1406 * that this is the right thing to do
1407 * but let's leave the code alone
1408 * for the RIB_SYSTEM_ROUTE case
1410 dest
->selected_fib
= NULL
;
1413 /* Update nexthop for route, reset changed flag. */
1414 /* Note: this code also handles the Linux case when an interface goes
1415 * down, causing the kernel to delete routes without sending DELROUTE
1418 if (!nexthop_active_update(rn
, old
, 1) &&
1419 (RIB_KERNEL_ROUTE(old
)))
1420 SET_FLAG(old
->status
, ROUTE_ENTRY_REMOVED
);
1422 UNSET_FLAG(old
->status
, ROUTE_ENTRY_CHANGED
);
1425 static void rib_process_update_fib(struct zebra_vrf
*zvrf
,
1426 struct route_node
*rn
,
1427 struct route_entry
*old
,
1428 struct route_entry
*new)
1430 struct nexthop
*nexthop
= NULL
;
1432 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1435 * We have to install or update if a new route has been selected or
1436 * something has changed.
1438 if (new != old
|| CHECK_FLAG(new->status
, ROUTE_ENTRY_CHANGED
)) {
1439 hook_call(rib_update
, rn
, "updating existing route");
1441 /* Update the nexthop; we could determine here that nexthop is
1443 if (nexthop_active_update(rn
, new, 1))
1446 /* If nexthop is active, install the selected route, if
1448 * the install succeeds, cleanup flags for prior route, if
1453 if (IS_ZEBRA_DEBUG_RIB
) {
1454 char buf
[SRCDEST2STR_BUFFER
];
1455 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1458 "%u:%s: Updating route rn %p, re %p (type %d) "
1460 zvrf_id(zvrf
), buf
, rn
, new,
1461 new->type
, old
, old
->type
);
1464 "%u:%s: Updating route rn %p, re %p (type %d)",
1465 zvrf_id(zvrf
), buf
, rn
, new,
1469 /* If labeled-unicast route, uninstall transit LSP. */
1470 if (zebra_rib_labeled_unicast(old
))
1471 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1473 /* Non-system route should be installed. */
1474 if (!RIB_SYSTEM_ROUTE(new)) {
1475 /* If labeled-unicast route, install transit
1477 if (zebra_rib_labeled_unicast(new))
1478 zebra_mpls_lsp_install(zvrf
, rn
, new);
1480 rib_install_kernel(rn
, new, old
);
1483 * We do not need to install the
1484 * selected route because it
1485 * is already isntalled by
1486 * the system( ie not us )
1487 * so just mark it as winning
1488 * we do need to ensure that
1489 * if we uninstall a route
1490 * from ourselves we don't
1491 * over write this pointer
1493 dest
->selected_fib
= NULL
;
1495 /* If install succeeded or system route, cleanup flags
1496 * for prior route. */
1498 if (RIB_SYSTEM_ROUTE(new)) {
1499 if (!RIB_SYSTEM_ROUTE(old
))
1500 rib_uninstall_kernel(rn
, old
);
1502 for (nexthop
= old
->ng
.nexthop
; nexthop
;
1503 nexthop
= nexthop
->next
)
1504 UNSET_FLAG(nexthop
->flags
,
1511 * If nexthop for selected route is not active or install
1513 * may need to uninstall and delete for redistribution.
1516 if (IS_ZEBRA_DEBUG_RIB
) {
1517 char buf
[SRCDEST2STR_BUFFER
];
1518 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1521 "%u:%s: Deleting route rn %p, re %p (type %d) "
1522 "old %p (type %d) - nexthop inactive",
1523 zvrf_id(zvrf
), buf
, rn
, new,
1524 new->type
, old
, old
->type
);
1527 "%u:%s: Deleting route rn %p, re %p (type %d) - nexthop inactive",
1528 zvrf_id(zvrf
), buf
, rn
, new,
1532 /* If labeled-unicast route, uninstall transit LSP. */
1533 if (zebra_rib_labeled_unicast(old
))
1534 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1536 if (!RIB_SYSTEM_ROUTE(old
))
1537 rib_uninstall_kernel(rn
, old
);
1539 dest
->selected_fib
= NULL
;
1543 * Same route selected; check if in the FIB and if not,
1545 * is housekeeping code to deal with race conditions in kernel
1547 * netlink reporting interface up before IPv4 or IPv6 protocol
1551 if (!RIB_SYSTEM_ROUTE(new)) {
1552 bool in_fib
= false;
1554 for (ALL_NEXTHOPS(new->ng
, nexthop
))
1555 if (CHECK_FLAG(nexthop
->flags
,
1556 NEXTHOP_FLAG_FIB
)) {
1561 rib_install_kernel(rn
, new, NULL
);
1565 /* Update prior route. */
1567 /* Set real nexthop. */
1568 nexthop_active_update(rn
, old
, 1);
1569 UNSET_FLAG(old
->status
, ROUTE_ENTRY_CHANGED
);
1572 /* Clear changed flag. */
1573 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1576 /* Check if 'alternate' RIB entry is better than 'current'. */
1577 static struct route_entry
*rib_choose_best(struct route_entry
*current
,
1578 struct route_entry
*alternate
)
1580 if (current
== NULL
)
1583 /* filter route selection in following order:
1584 * - connected beats other types
1585 * - if both connected, loopback or vrf wins
1586 * - lower distance beats higher
1587 * - lower metric beats higher for equal distance
1588 * - last, hence oldest, route wins tie break.
1591 /* Connected routes. Check to see if either are a vrf
1592 * or loopback interface. If not, pick the last connected
1593 * route of the set of lowest metric connected routes.
1595 if (alternate
->type
== ZEBRA_ROUTE_CONNECT
) {
1596 if (current
->type
!= ZEBRA_ROUTE_CONNECT
)
1599 /* both are connected. are either loop or vrf? */
1600 struct nexthop
*nexthop
= NULL
;
1602 for (ALL_NEXTHOPS(alternate
->ng
, nexthop
)) {
1603 if (if_is_loopback_or_vrf(if_lookup_by_index(
1604 nexthop
->ifindex
, alternate
->vrf_id
)))
1608 for (ALL_NEXTHOPS(current
->ng
, nexthop
)) {
1609 if (if_is_loopback_or_vrf(if_lookup_by_index(
1610 nexthop
->ifindex
, current
->vrf_id
)))
1614 /* Neither are loop or vrf so pick best metric */
1615 if (alternate
->metric
<= current
->metric
)
1621 if (current
->type
== ZEBRA_ROUTE_CONNECT
)
1624 /* higher distance loses */
1625 if (alternate
->distance
< current
->distance
)
1627 if (current
->distance
< alternate
->distance
)
1630 /* metric tie-breaks equal distance */
1631 if (alternate
->metric
<= current
->metric
)
1637 /* Core function for processing routing information base. */
1638 static void rib_process(struct route_node
*rn
)
1640 struct route_entry
*re
;
1641 struct route_entry
*next
;
1642 struct route_entry
*old_selected
= NULL
;
1643 struct route_entry
*new_selected
= NULL
;
1644 struct route_entry
*old_fib
= NULL
;
1645 struct route_entry
*new_fib
= NULL
;
1646 struct route_entry
*best
= NULL
;
1647 char buf
[SRCDEST2STR_BUFFER
];
1649 struct zebra_vrf
*zvrf
= NULL
;
1650 const struct prefix
*p
, *src_p
;
1652 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1653 vrf_id_t vrf_id
= VRF_UNKNOWN
;
1657 dest
= rib_dest_from_rnode(rn
);
1659 zvrf
= rib_dest_vrf(dest
);
1660 vrf_id
= zvrf_id(zvrf
);
1663 if (IS_ZEBRA_DEBUG_RIB
)
1664 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1666 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1667 zlog_debug("%u:%s: Processing rn %p", vrf_id
, buf
, rn
);
1670 * we can have rn's that have a NULL info pointer
1671 * (dest). As such let's not let the deref happen
1672 * additionally we know RNODE_FOREACH_RE_SAFE
1673 * will not iterate so we are ok.
1676 old_fib
= dest
->selected_fib
;
1678 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
1679 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1681 "%u:%s: Examine re %p (type %d) status %x flags %x "
1682 "dist %d metric %d",
1683 vrf_id
, buf
, re
, re
->type
, re
->status
,
1684 re
->flags
, re
->distance
, re
->metric
);
1686 UNSET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1688 /* Currently selected re. */
1689 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)) {
1690 assert(old_selected
== NULL
);
1694 /* Skip deleted entries from selection */
1695 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
1698 /* Skip unreachable nexthop. */
1699 /* This first call to nexthop_active_update is merely to
1701 * there's any change to nexthops associated with this RIB
1703 * rib_process() can be invoked due to an external event such as
1705 * down or due to next-hop-tracking evaluation. In the latter
1707 * a decision has already been made that the NHs have changed.
1709 * need to invoke a potentially expensive call again. Further,
1711 * the change might be in a recursive NH which is not caught in
1712 * the nexthop_active_update() code. Thus, we might miss changes
1716 if (!CHECK_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
)
1717 && !nexthop_active_update(rn
, re
, 0)) {
1718 if (re
->type
== ZEBRA_ROUTE_TABLE
) {
1719 /* XXX: HERE BE DRAGONS!!!!!
1720 * In all honesty, I have not yet figured out
1722 * does or why the ROUTE_ENTRY_CHANGED test
1724 * or why we need to delete a route here, and
1726 * this concerns both selected and fib route, or
1729 /* This entry was denied by the 'ip protocol
1730 * table' route-map, we
1731 * need to delete it */
1732 if (re
!= old_selected
) {
1733 if (IS_ZEBRA_DEBUG_RIB
)
1735 "%s: %u:%s: imported via import-table but denied "
1736 "by the ip protocol table route-map",
1737 __func__
, vrf_id
, buf
);
1740 SET_FLAG(re
->status
,
1741 ROUTE_ENTRY_REMOVED
);
1747 /* Infinite distance. */
1748 if (re
->distance
== DISTANCE_INFINITY
) {
1749 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1753 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_FIB_OVERRIDE
)) {
1754 best
= rib_choose_best(new_fib
, re
);
1755 if (new_fib
&& best
!= new_fib
)
1756 UNSET_FLAG(new_fib
->status
,
1757 ROUTE_ENTRY_CHANGED
);
1760 best
= rib_choose_best(new_selected
, re
);
1761 if (new_selected
&& best
!= new_selected
)
1762 UNSET_FLAG(new_selected
->status
,
1763 ROUTE_ENTRY_CHANGED
);
1764 new_selected
= best
;
1767 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1768 } /* RNODE_FOREACH_RE */
1770 /* If no FIB override route, use the selected route also for FIB */
1771 if (new_fib
== NULL
)
1772 new_fib
= new_selected
;
1774 /* After the cycle is finished, the following pointers will be set:
1775 * old_selected --- RE entry currently having SELECTED
1776 * new_selected --- RE entry that is newly SELECTED
1777 * old_fib --- RE entry currently in kernel FIB
1778 * new_fib --- RE entry that is newly to be in kernel FIB
1780 * new_selected will get SELECTED flag, and is going to be redistributed
1781 * the zclients. new_fib (which can be new_selected) will be installed
1785 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
1787 "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p",
1788 vrf_id
, buf
, (void *)old_selected
, (void *)new_selected
,
1789 (void *)old_fib
, (void *)new_fib
);
1792 /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if
1793 * fib == selected */
1794 bool selected_changed
= new_selected
&& CHECK_FLAG(new_selected
->status
,
1795 ROUTE_ENTRY_CHANGED
);
1797 /* Update fib according to selection results */
1798 if (new_fib
&& old_fib
)
1799 rib_process_update_fib(zvrf
, rn
, old_fib
, new_fib
);
1801 rib_process_add_fib(zvrf
, rn
, new_fib
);
1803 rib_process_del_fib(zvrf
, rn
, old_fib
);
1805 /* Redistribute SELECTED entry */
1806 if (old_selected
!= new_selected
|| selected_changed
) {
1807 struct nexthop
*nexthop
= NULL
;
1809 /* Check if we have a FIB route for the destination, otherwise,
1810 * don't redistribute it */
1812 for (ALL_NEXTHOPS(new_fib
->ng
, nexthop
)) {
1813 if (CHECK_FLAG(nexthop
->flags
,
1814 NEXTHOP_FLAG_FIB
)) {
1820 new_selected
= NULL
;
1822 if (new_selected
&& new_selected
!= new_fib
) {
1823 nexthop_active_update(rn
, new_selected
, 1);
1824 UNSET_FLAG(new_selected
->status
, ROUTE_ENTRY_CHANGED
);
1829 redistribute_delete(p
, src_p
, old_selected
);
1830 if (old_selected
!= new_selected
)
1831 UNSET_FLAG(old_selected
->flags
,
1832 ZEBRA_FLAG_SELECTED
);
1836 /* Install new or replace existing redistributed entry
1838 SET_FLAG(new_selected
->flags
, ZEBRA_FLAG_SELECTED
);
1839 redistribute_update(p
, src_p
, new_selected
,
1844 /* Remove all RE entries queued for removal */
1845 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
1846 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
1847 if (IS_ZEBRA_DEBUG_RIB
) {
1848 rnode_debug(rn
, vrf_id
, "rn %p, removing re %p",
1849 (void *)rn
, (void *)re
);
1856 * Check if the dest can be deleted now.
1861 /* Take a list of route_node structs and return 1, if there was a record
1862 * picked from it and processed by rib_process(). Don't process more,
1863 * than one RN record; operate only in the specified sub-queue.
1865 static unsigned int process_subq(struct list
*subq
, uint8_t qindex
)
1867 struct listnode
*lnode
= listhead(subq
);
1868 struct route_node
*rnode
;
1870 struct zebra_vrf
*zvrf
= NULL
;
1875 rnode
= listgetdata(lnode
);
1876 dest
= rib_dest_from_rnode(rnode
);
1878 zvrf
= rib_dest_vrf(dest
);
1882 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
1883 char buf
[SRCDEST2STR_BUFFER
];
1884 srcdest_rnode2str(rnode
, buf
, sizeof(buf
));
1885 zlog_debug("%u:%s: rn %p dequeued from sub-queue %u",
1886 zvrf
? zvrf_id(zvrf
) : 0, buf
, rnode
, qindex
);
1890 UNSET_FLAG(rib_dest_from_rnode(rnode
)->flags
,
1891 RIB_ROUTE_QUEUED(qindex
));
1896 zlog_debug ("%s: called for route_node (%p, %d) with no ribs",
1897 __func__
, rnode
, rnode
->lock
);
1898 zlog_backtrace(LOG_DEBUG
);
1901 route_unlock_node(rnode
);
1902 list_delete_node(subq
, lnode
);
1907 * All meta queues have been processed. Trigger next-hop evaluation.
1909 static void meta_queue_process_complete(struct work_queue
*dummy
)
1912 struct zebra_vrf
*zvrf
;
1914 /* Evaluate nexthops for those VRFs which underwent route processing.
1916 * should limit the evaluation to the necessary VRFs in most common
1919 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
1921 if (zvrf
== NULL
|| !(zvrf
->flags
& ZEBRA_VRF_RIB_SCHEDULED
))
1924 zvrf
->flags
&= ~ZEBRA_VRF_RIB_SCHEDULED
;
1925 zebra_evaluate_rnh(zvrf
, AF_INET
, 0, RNH_NEXTHOP_TYPE
, NULL
);
1926 zebra_evaluate_rnh(zvrf
, AF_INET
, 0, RNH_IMPORT_CHECK_TYPE
,
1928 zebra_evaluate_rnh(zvrf
, AF_INET6
, 0, RNH_NEXTHOP_TYPE
, NULL
);
1929 zebra_evaluate_rnh(zvrf
, AF_INET6
, 0, RNH_IMPORT_CHECK_TYPE
,
1933 /* Schedule LSPs for processing, if needed. */
1934 zvrf
= vrf_info_lookup(VRF_DEFAULT
);
1935 if (mpls_should_lsps_be_processed(zvrf
)) {
1936 if (IS_ZEBRA_DEBUG_MPLS
)
1938 "%u: Scheduling all LSPs upon RIB completion",
1940 zebra_mpls_lsp_schedule(zvrf
);
1941 mpls_unmark_lsps_for_processing(zvrf
);
1945 /* Dispatch the meta queue by picking, processing and unlocking the next RN from
1946 * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and
1948 * is pointed to the meta queue structure.
1950 static wq_item_status
meta_queue_process(struct work_queue
*dummy
, void *data
)
1952 struct meta_queue
*mq
= data
;
1955 for (i
= 0; i
< MQ_SIZE
; i
++)
1956 if (process_subq(mq
->subq
[i
], i
)) {
1960 return mq
->size
? WQ_REQUEUE
: WQ_SUCCESS
;
1964 * Map from rib types to queue type (priority) in meta queue
1966 static const uint8_t meta_queue_map
[ZEBRA_ROUTE_MAX
] = {
1967 [ZEBRA_ROUTE_SYSTEM
] = 4,
1968 [ZEBRA_ROUTE_KERNEL
] = 0,
1969 [ZEBRA_ROUTE_CONNECT
] = 0,
1970 [ZEBRA_ROUTE_STATIC
] = 1,
1971 [ZEBRA_ROUTE_RIP
] = 2,
1972 [ZEBRA_ROUTE_RIPNG
] = 2,
1973 [ZEBRA_ROUTE_OSPF
] = 2,
1974 [ZEBRA_ROUTE_OSPF6
] = 2,
1975 [ZEBRA_ROUTE_ISIS
] = 2,
1976 [ZEBRA_ROUTE_BGP
] = 3,
1977 [ZEBRA_ROUTE_PIM
] = 4, // Shouldn't happen but for safety
1978 [ZEBRA_ROUTE_EIGRP
] = 2,
1979 [ZEBRA_ROUTE_NHRP
] = 2,
1980 [ZEBRA_ROUTE_HSLS
] = 4,
1981 [ZEBRA_ROUTE_OLSR
] = 4,
1982 [ZEBRA_ROUTE_TABLE
] = 1,
1983 [ZEBRA_ROUTE_LDP
] = 4,
1984 [ZEBRA_ROUTE_VNC
] = 3,
1985 [ZEBRA_ROUTE_VNC_DIRECT
] = 3,
1986 [ZEBRA_ROUTE_VNC_DIRECT_RH
] = 3,
1987 [ZEBRA_ROUTE_BGP_DIRECT
] = 3,
1988 [ZEBRA_ROUTE_BGP_DIRECT_EXT
] = 3,
1989 [ZEBRA_ROUTE_BABEL
] = 2,
1990 [ZEBRA_ROUTE_ALL
] = 4, // Shouldn't happen but for safety
1993 /* Look into the RN and queue it into one or more priority queues,
1994 * increasing the size for each data push done.
1996 static void rib_meta_queue_add(struct meta_queue
*mq
, struct route_node
*rn
)
1998 struct route_entry
*re
;
2000 RNODE_FOREACH_RE (rn
, re
) {
2001 uint8_t qindex
= meta_queue_map
[re
->type
];
2002 struct zebra_vrf
*zvrf
;
2004 /* Invariant: at this point we always have rn->info set. */
2005 if (CHECK_FLAG(rib_dest_from_rnode(rn
)->flags
,
2006 RIB_ROUTE_QUEUED(qindex
))) {
2007 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2010 "rn %p is already queued in sub-queue %u",
2011 (void *)rn
, qindex
);
2015 SET_FLAG(rib_dest_from_rnode(rn
)->flags
,
2016 RIB_ROUTE_QUEUED(qindex
));
2017 listnode_add(mq
->subq
[qindex
], rn
);
2018 route_lock_node(rn
);
2021 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2022 rnode_debug(rn
, re
->vrf_id
,
2023 "queued rn %p into sub-queue %u",
2024 (void *)rn
, qindex
);
2026 zvrf
= zebra_vrf_lookup_by_id(re
->vrf_id
);
2028 zvrf
->flags
|= ZEBRA_VRF_RIB_SCHEDULED
;
2032 /* Add route_node to work queue and schedule processing */
2033 void rib_queue_add(struct route_node
*rn
)
2037 /* Pointless to queue a route_node with no RIB entries to add or remove
2039 if (!rnode_to_ribs(rn
)) {
2040 zlog_debug("%s: called for route_node (%p, %d) with no ribs",
2041 __func__
, (void *)rn
, rn
->lock
);
2042 zlog_backtrace(LOG_DEBUG
);
2046 if (zebrad
.ribq
== NULL
) {
2047 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2048 "%s: work_queue does not exist!", __func__
);
2053 * The RIB queue should normally be either empty or holding the only
2054 * work_queue_item element. In the latter case this element would
2055 * hold a pointer to the meta queue structure, which must be used to
2056 * actually queue the route nodes to process. So create the MQ
2057 * holder, if necessary, then push the work into it in any case.
2058 * This semantics was introduced after 0.99.9 release.
2060 if (work_queue_empty(zebrad
.ribq
))
2061 work_queue_add(zebrad
.ribq
, zebrad
.mq
);
2063 rib_meta_queue_add(zebrad
.mq
, rn
);
2068 /* Create new meta queue.
2069 A destructor function doesn't seem to be necessary here.
2071 static struct meta_queue
*meta_queue_new(void)
2073 struct meta_queue
*new;
2076 new = XCALLOC(MTYPE_WORK_QUEUE
, sizeof(struct meta_queue
));
2078 for (i
= 0; i
< MQ_SIZE
; i
++) {
2079 new->subq
[i
] = list_new();
2080 assert(new->subq
[i
]);
2086 void meta_queue_free(struct meta_queue
*mq
)
2090 for (i
= 0; i
< MQ_SIZE
; i
++)
2091 list_delete(&mq
->subq
[i
]);
2093 XFREE(MTYPE_WORK_QUEUE
, mq
);
2096 /* initialise zebra rib work queue */
2097 static void rib_queue_init(struct zebra_t
*zebra
)
2102 work_queue_new(zebra
->master
, "route_node processing"))) {
2103 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2104 "%s: could not initialise work queue!", __func__
);
2108 /* fill in the work queue spec */
2109 zebra
->ribq
->spec
.workfunc
= &meta_queue_process
;
2110 zebra
->ribq
->spec
.errorfunc
= NULL
;
2111 zebra
->ribq
->spec
.completion_func
= &meta_queue_process_complete
;
2112 /* XXX: TODO: These should be runtime configurable via vty */
2113 zebra
->ribq
->spec
.max_retries
= 3;
2114 zebra
->ribq
->spec
.hold
= ZEBRA_RIB_PROCESS_HOLD_TIME
;
2116 if (!(zebra
->mq
= meta_queue_new())) {
2117 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2118 "%s: could not initialise meta queue!", __func__
);
2124 /* RIB updates are processed via a queue of pointers to route_nodes.
2126 * The queue length is bounded by the maximal size of the routing table,
2127 * as a route_node will not be requeued, if already queued.
2129 * REs are submitted via rib_addnode or rib_delnode which set minimal
2130 * state, or static_install_route (when an existing RE is updated)
2131 * and then submit route_node to queue for best-path selection later.
2132 * Order of add/delete state changes are preserved for any given RE.
2134 * Deleted REs are reaped during best-path selection.
2137 * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with
2138 * |-------->| | best RE, if required
2140 * static_install->|->rib_addqueue...... -> rib_process
2142 * |-------->| |-> rib_unlink
2143 * |-> set ROUTE_ENTRY_REMOVE |
2144 * rib_delnode (RE freed)
2146 * The 'info' pointer of a route_node points to a rib_dest_t
2147 * ('dest'). Queueing state for a route_node is kept on the dest. The
2148 * dest is created on-demand by rib_link() and is kept around at least
2149 * as long as there are ribs hanging off it (@see rib_gc_dest()).
2151 * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code):
2153 * - route_nodes: refcounted by:
2154 * - dest attached to route_node:
2155 * - managed by: rib_link/rib_gc_dest
2156 * - route_node processing queue
2157 * - managed by: rib_addqueue, rib_process.
2161 /* Add RE to head of the route node. */
2162 static void rib_link(struct route_node
*rn
, struct route_entry
*re
, int process
)
2164 struct route_entry
*head
;
2167 const char *rmap_name
;
2171 dest
= rib_dest_from_rnode(rn
);
2173 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2174 rnode_debug(rn
, re
->vrf_id
, "rn %p adding dest", rn
);
2176 dest
= XCALLOC(MTYPE_RIB_DEST
, sizeof(rib_dest_t
));
2177 route_lock_node(rn
); /* rn route table reference */
2182 head
= dest
->routes
;
2189 afi
= (rn
->p
.family
== AF_INET
)
2191 : (rn
->p
.family
== AF_INET6
) ? AFI_IP6
: AFI_MAX
;
2192 if (is_zebra_import_table_enabled(afi
, re
->table
)) {
2193 rmap_name
= zebra_get_import_table_route_map(afi
, re
->table
);
2194 zebra_add_import_table_entry(rn
, re
, rmap_name
);
2199 static void rib_addnode(struct route_node
*rn
,
2200 struct route_entry
*re
, int process
)
2202 /* RE node has been un-removed before route-node is processed.
2203 * route_node must hence already be on the queue for processing..
2205 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
2206 if (IS_ZEBRA_DEBUG_RIB
)
2207 rnode_debug(rn
, re
->vrf_id
, "rn %p, un-removed re %p",
2208 (void *)rn
, (void *)re
);
2210 UNSET_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
);
2213 rib_link(rn
, re
, process
);
2219 * Detach a rib structure from a route_node.
2221 * Note that a call to rib_unlink() should be followed by a call to
2222 * rib_gc_dest() at some point. This allows a rib_dest_t that is no
2223 * longer required to be deleted.
2225 void rib_unlink(struct route_node
*rn
, struct route_entry
*re
)
2231 if (IS_ZEBRA_DEBUG_RIB
)
2232 rnode_debug(rn
, re
->vrf_id
, "rn %p, re %p", (void *)rn
,
2235 dest
= rib_dest_from_rnode(rn
);
2238 re
->next
->prev
= re
->prev
;
2241 re
->prev
->next
= re
->next
;
2243 dest
->routes
= re
->next
;
2246 if (dest
->selected_fib
== re
)
2247 dest
->selected_fib
= NULL
;
2249 nexthops_free(re
->ng
.nexthop
);
2250 XFREE(MTYPE_RE
, re
);
2253 void rib_delnode(struct route_node
*rn
, struct route_entry
*re
)
2257 if (IS_ZEBRA_DEBUG_RIB
)
2258 rnode_debug(rn
, re
->vrf_id
, "rn %p, re %p, removing",
2259 (void *)rn
, (void *)re
);
2260 SET_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
);
2262 afi
= (rn
->p
.family
== AF_INET
)
2264 : (rn
->p
.family
== AF_INET6
) ? AFI_IP6
: AFI_MAX
;
2265 if (is_zebra_import_table_enabled(afi
, re
->table
)) {
2266 zebra_del_import_table_entry(rn
, re
);
2267 /* Just clean up if non main table */
2268 if (IS_ZEBRA_DEBUG_RIB
) {
2269 char buf
[SRCDEST2STR_BUFFER
];
2270 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
2272 "%u:%s: Freeing route rn %p, re %p (type %d)",
2273 re
->vrf_id
, buf
, rn
, re
, re
->type
);
2282 /* This function dumps the contents of a given RE entry into
2283 * standard debug log. Calling function name and IP prefix in
2284 * question are passed as 1st and 2nd arguments.
2287 void _route_entry_dump(const char *func
, union prefixconstptr pp
,
2288 union prefixconstptr src_pp
,
2289 const struct route_entry
*re
)
2291 const struct prefix
*src_p
= src_pp
.p
;
2292 bool is_srcdst
= src_p
&& src_p
->prefixlen
;
2293 char straddr
[PREFIX_STRLEN
];
2294 char srcaddr
[PREFIX_STRLEN
];
2295 struct nexthop
*nexthop
;
2297 zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func
,
2298 (const void *)re
, prefix2str(pp
, straddr
, sizeof(straddr
)),
2299 is_srcdst
? " from " : "",
2300 is_srcdst
? prefix2str(src_pp
, srcaddr
, sizeof(srcaddr
))
2303 zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d",
2304 func
, (unsigned long)re
->uptime
, re
->type
, re
->instance
,
2307 "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u",
2308 func
, re
->metric
, re
->mtu
, re
->distance
, re
->flags
, re
->status
);
2309 zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func
,
2310 re
->nexthop_num
, re
->nexthop_active_num
);
2312 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
2313 struct interface
*ifp
;
2314 struct vrf
*vrf
= vrf_lookup_by_id(nexthop
->vrf_id
);
2316 switch (nexthop
->type
) {
2317 case NEXTHOP_TYPE_BLACKHOLE
:
2318 sprintf(straddr
, "Blackhole");
2320 case NEXTHOP_TYPE_IFINDEX
:
2321 ifp
= if_lookup_by_index(nexthop
->ifindex
,
2323 sprintf(straddr
, "%s", ifp
? ifp
->name
: "Unknown");
2325 case NEXTHOP_TYPE_IPV4
:
2327 case NEXTHOP_TYPE_IPV4_IFINDEX
:
2328 inet_ntop(AF_INET
, &nexthop
->gate
, straddr
,
2331 case NEXTHOP_TYPE_IPV6
:
2332 case NEXTHOP_TYPE_IPV6_IFINDEX
:
2333 inet_ntop(AF_INET6
, &nexthop
->gate
, straddr
,
2337 zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s", func
,
2338 (nexthop
->rparent
? " NH" : "NH"), straddr
,
2339 nexthop
->ifindex
, vrf
? vrf
->name
: "Unknown",
2341 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
)
2344 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
)
2347 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
)
2351 zlog_debug("%s: dump complete", func
);
2354 /* This is an exported helper to rtm_read() to dump the strange
2355 * RE entry found by rib_lookup_ipv4_route()
2358 void rib_lookup_and_dump(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
2360 struct route_table
*table
;
2361 struct route_node
*rn
;
2362 struct route_entry
*re
;
2363 char prefix_buf
[INET_ADDRSTRLEN
];
2366 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
2368 flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED
,
2369 "%s:%u zebra_vrf_table() returned NULL", __func__
,
2374 /* Scan the RIB table for exactly matching RE entry. */
2375 rn
= route_node_lookup(table
, (struct prefix
*)p
);
2377 /* No route for this prefix. */
2379 zlog_debug("%s:%u lookup failed for %s", __func__
, vrf_id
,
2380 prefix2str((struct prefix
*)p
, prefix_buf
,
2381 sizeof(prefix_buf
)));
2386 route_unlock_node(rn
);
2389 RNODE_FOREACH_RE (rn
, re
) {
2390 zlog_debug("%s:%u rn %p, re %p: %s, %s",
2392 (void *)rn
, (void *)re
,
2393 (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)
2396 (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)
2399 route_entry_dump(p
, NULL
, re
);
2403 /* Check if requested address assignment will fail due to another
2404 * route being installed by zebra in FIB already. Take necessary
2405 * actions, if needed: remove such a route from FIB and deSELECT
2406 * corresponding RE entry. Then put affected RN into RIBQ head.
2408 void rib_lookup_and_pushup(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
2410 struct route_table
*table
;
2411 struct route_node
*rn
;
2412 unsigned changed
= 0;
2415 if (NULL
== (table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
))) {
2416 flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED
,
2417 "%s:%u zebra_vrf_table() returned NULL", __func__
,
2422 /* No matches would be the simplest case. */
2423 if (NULL
== (rn
= route_node_lookup(table
, (struct prefix
*)p
)))
2427 route_unlock_node(rn
);
2429 dest
= rib_dest_from_rnode(rn
);
2430 /* Check all RE entries. In case any changes have to be done, requeue
2431 * the RN into RIBQ head. If the routing message about the new connected
2432 * route (generated by the IP address we are going to assign very soon)
2433 * comes before the RIBQ is processed, the new RE entry will join
2434 * RIBQ record already on head. This is necessary for proper
2436 * of the rest of the RE.
2438 if (dest
->selected_fib
&& !RIB_SYSTEM_ROUTE(dest
->selected_fib
)) {
2440 if (IS_ZEBRA_DEBUG_RIB
) {
2441 char buf
[PREFIX_STRLEN
];
2443 zlog_debug("%u:%s: freeing way for connected prefix",
2444 dest
->selected_fib
->vrf_id
,
2445 prefix2str(&rn
->p
, buf
, sizeof(buf
)));
2446 route_entry_dump(&rn
->p
, NULL
, dest
->selected_fib
);
2448 rib_uninstall(rn
, dest
->selected_fib
);
2454 int rib_add_multipath(afi_t afi
, safi_t safi
, struct prefix
*p
,
2455 struct prefix_ipv6
*src_p
, struct route_entry
*re
)
2457 struct route_table
*table
;
2458 struct route_node
*rn
;
2459 struct route_entry
*same
= NULL
;
2460 struct nexthop
*nexthop
;
2466 assert(!src_p
|| !src_p
->prefixlen
|| afi
== AFI_IP6
);
2469 table
= zebra_vrf_table_with_table_id(afi
, safi
, re
->vrf_id
, re
->table
);
2471 XFREE(MTYPE_RE
, re
);
2475 /* Make it sure prefixlen is applied to the prefix. */
2478 apply_mask_ipv6(src_p
);
2480 /* Set default distance by route type. */
2481 if (re
->distance
== 0) {
2482 re
->distance
= route_distance(re
->type
);
2484 /* iBGP distance is 200. */
2485 if (re
->type
== ZEBRA_ROUTE_BGP
2486 && CHECK_FLAG(re
->flags
, ZEBRA_FLAG_IBGP
))
2490 /* Lookup route node.*/
2491 rn
= srcdest_rnode_get(table
, p
, src_p
);
2494 * If same type of route are installed, treat it as a implicit
2496 * If the user has specified the No route replace semantics
2497 * for the install don't do a route replace.
2499 RNODE_FOREACH_RE (rn
, same
) {
2500 if (CHECK_FLAG(same
->status
, ROUTE_ENTRY_REMOVED
))
2503 if (same
->type
!= re
->type
)
2505 if (same
->instance
!= re
->instance
)
2507 if (same
->type
== ZEBRA_ROUTE_KERNEL
2508 && same
->metric
!= re
->metric
)
2511 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_RR_USE_DISTANCE
) &&
2512 same
->distance
!= re
->distance
)
2516 * We should allow duplicate connected routes
2517 * because of IPv6 link-local routes and unnumbered
2518 * interfaces on Linux.
2520 if (same
->type
!= ZEBRA_ROUTE_CONNECT
)
2524 /* If this route is kernel route, set FIB flag to the route. */
2525 if (RIB_SYSTEM_ROUTE(re
))
2526 for (nexthop
= re
->ng
.nexthop
; nexthop
; nexthop
= nexthop
->next
)
2527 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
2529 /* Link new re to node.*/
2530 if (IS_ZEBRA_DEBUG_RIB
) {
2533 "Inserting route rn %p, re %p (type %d) existing %p",
2534 (void *)rn
, (void *)re
, re
->type
, (void *)same
);
2536 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2537 route_entry_dump(p
, src_p
, re
);
2539 rib_addnode(rn
, re
, 1);
2542 /* Free implicit route.*/
2544 rib_delnode(rn
, same
);
2548 route_unlock_node(rn
);
2552 void rib_delete(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
, int type
,
2553 unsigned short instance
, int flags
, struct prefix
*p
,
2554 struct prefix_ipv6
*src_p
, const struct nexthop
*nh
,
2555 uint32_t table_id
, uint32_t metric
, uint8_t distance
,
2558 struct route_table
*table
;
2559 struct route_node
*rn
;
2560 struct route_entry
*re
;
2561 struct route_entry
*fib
= NULL
;
2562 struct route_entry
*same
= NULL
;
2563 struct nexthop
*rtnh
;
2564 char buf2
[INET6_ADDRSTRLEN
];
2567 assert(!src_p
|| !src_p
->prefixlen
|| afi
== AFI_IP6
);
2570 table
= zebra_vrf_table_with_table_id(afi
, safi
, vrf_id
, table_id
);
2577 apply_mask_ipv6(src_p
);
2579 /* Lookup route node. */
2580 rn
= srcdest_rnode_lookup(table
, p
, src_p
);
2582 char dst_buf
[PREFIX_STRLEN
], src_buf
[PREFIX_STRLEN
];
2584 prefix2str(p
, dst_buf
, sizeof(dst_buf
));
2585 if (src_p
&& src_p
->prefixlen
)
2586 prefix2str(src_p
, src_buf
, sizeof(src_buf
));
2590 if (IS_ZEBRA_DEBUG_RIB
)
2591 zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id
,
2593 (src_buf
[0] != '\0') ? " from " : "",
2598 dest
= rib_dest_from_rnode(rn
);
2599 fib
= dest
->selected_fib
;
2601 /* Lookup same type route. */
2602 RNODE_FOREACH_RE (rn
, re
) {
2603 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
2606 if (re
->type
!= type
)
2608 if (re
->instance
!= instance
)
2610 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_RR_USE_DISTANCE
) &&
2611 distance
!= re
->distance
)
2614 if (re
->type
== ZEBRA_ROUTE_KERNEL
&& re
->metric
!= metric
)
2616 if (re
->type
== ZEBRA_ROUTE_CONNECT
&& (rtnh
= re
->ng
.nexthop
)
2617 && rtnh
->type
== NEXTHOP_TYPE_IFINDEX
&& nh
) {
2618 if (rtnh
->ifindex
!= nh
->ifindex
)
2623 /* Make sure that the route found has the same gateway. */
2629 for (ALL_NEXTHOPS(re
->ng
, rtnh
))
2630 if (nexthop_same_no_recurse(rtnh
, nh
)) {
2638 /* If same type of route can't be found and this message is from
2642 * In the past(HA!) we could get here because
2643 * we were receiving a route delete from the
2644 * kernel and we're not marking the proto
2645 * as coming from it's appropriate originator.
2646 * Now that we are properly noticing the fact
2647 * that the kernel has deleted our route we
2648 * are not going to get called in this path
2649 * I am going to leave this here because
2650 * this might still work this way on non-linux
2651 * platforms as well as some weird state I have
2652 * not properly thought of yet.
2653 * If we can show that this code path is
2654 * dead then we can remove it.
2656 if (fib
&& CHECK_FLAG(flags
, ZEBRA_FLAG_SELFROUTE
)) {
2657 if (IS_ZEBRA_DEBUG_RIB
) {
2660 "rn %p, re %p (type %d) was deleted from kernel, adding",
2661 rn
, fib
, fib
->type
);
2665 for (rtnh
= fib
->ng
.nexthop
; rtnh
;
2667 UNSET_FLAG(rtnh
->flags
,
2671 * This is a non FRR route
2672 * as such we should mark
2675 dest
->selected_fib
= NULL
;
2677 /* This means someone else, other than Zebra,
2679 * a Zebra router from the kernel. We will add
2681 rib_install_kernel(rn
, fib
, NULL
);
2684 if (IS_ZEBRA_DEBUG_RIB
) {
2688 "via %s ifindex %d type %d "
2689 "doesn't exist in rib",
2690 inet_ntop(afi2family(afi
),
2697 "type %d doesn't exist in rib",
2700 route_unlock_node(rn
);
2706 if (fromkernel
&& CHECK_FLAG(flags
, ZEBRA_FLAG_SELFROUTE
)
2708 rib_install_kernel(rn
, same
, NULL
);
2709 route_unlock_node(rn
);
2714 if (CHECK_FLAG(flags
, ZEBRA_FLAG_EVPN_ROUTE
)) {
2715 struct nexthop
*tmp_nh
;
2717 for (ALL_NEXTHOPS(re
->ng
, tmp_nh
)) {
2718 struct ipaddr vtep_ip
;
2720 memset(&vtep_ip
, 0, sizeof(struct ipaddr
));
2721 if (afi
== AFI_IP
) {
2722 vtep_ip
.ipa_type
= IPADDR_V4
;
2723 memcpy(&(vtep_ip
.ipaddr_v4
),
2724 &(tmp_nh
->gate
.ipv4
),
2725 sizeof(struct in_addr
));
2727 vtep_ip
.ipa_type
= IPADDR_V6
;
2728 memcpy(&(vtep_ip
.ipaddr_v6
),
2729 &(tmp_nh
->gate
.ipv6
),
2730 sizeof(struct in6_addr
));
2732 zebra_vxlan_evpn_vrf_route_del(re
->vrf_id
,
2736 rib_delnode(rn
, same
);
2739 route_unlock_node(rn
);
2744 int rib_add(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
, int type
,
2745 unsigned short instance
, int flags
, struct prefix
*p
,
2746 struct prefix_ipv6
*src_p
, const struct nexthop
*nh
,
2747 uint32_t table_id
, uint32_t metric
, uint32_t mtu
, uint8_t distance
,
2750 struct route_entry
*re
;
2751 struct nexthop
*nexthop
;
2753 /* Allocate new route_entry structure. */
2754 re
= XCALLOC(MTYPE_RE
, sizeof(struct route_entry
));
2756 re
->instance
= instance
;
2757 re
->distance
= distance
;
2759 re
->metric
= metric
;
2761 re
->table
= table_id
;
2762 re
->vrf_id
= vrf_id
;
2763 re
->nexthop_num
= 0;
2764 re
->uptime
= time(NULL
);
2768 nexthop
= nexthop_new();
2770 route_entry_nexthop_add(re
, nexthop
);
2772 return rib_add_multipath(afi
, safi
, p
, src_p
, re
);
2775 /* Schedule routes of a particular table (address-family) based on event. */
2776 void rib_update_table(struct route_table
*table
, rib_update_event_t event
)
2778 struct route_node
*rn
;
2779 struct route_entry
*re
, *next
;
2781 /* Walk all routes and queue for processing, if appropriate for
2782 * the trigger event.
2784 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
2786 * If we are looking at a route node and the node
2787 * has already been queued we don't
2788 * need to queue it up again
2790 if (rn
->info
&& CHECK_FLAG(rib_dest_from_rnode(rn
)->flags
,
2791 RIB_ROUTE_ANY_QUEUED
))
2794 case RIB_UPDATE_IF_CHANGE
:
2795 /* Examine all routes that won't get processed by the
2797 * triggered by nexthop evaluation (NHT). This would be
2799 * kernel and certain static routes. Note that NHT will
2801 * triggered upon an interface event as connected routes
2803 * get queued for processing.
2805 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
2808 if (re
->type
!= ZEBRA_ROUTE_SYSTEM
2809 && re
->type
!= ZEBRA_ROUTE_KERNEL
2810 && re
->type
!= ZEBRA_ROUTE_CONNECT
2811 && re
->type
!= ZEBRA_ROUTE_STATIC
)
2814 if (re
->type
!= ZEBRA_ROUTE_STATIC
) {
2819 for (nh
= re
->ng
.nexthop
; nh
; nh
= nh
->next
)
2820 if (!(nh
->type
== NEXTHOP_TYPE_IPV4
2821 || nh
->type
== NEXTHOP_TYPE_IPV6
))
2824 /* If we only have nexthops to a
2833 case RIB_UPDATE_RMAP_CHANGE
:
2834 case RIB_UPDATE_OTHER
:
2835 /* Right now, examine all routes. Can restrict to a
2837 * some cases (TODO).
2839 if (rnode_to_ribs(rn
))
2849 /* RIB update function. */
2850 void rib_update(vrf_id_t vrf_id
, rib_update_event_t event
)
2852 struct route_table
*table
;
2854 /* Process routes of interested address-families. */
2855 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
2857 if (IS_ZEBRA_DEBUG_EVENT
)
2858 zlog_debug("%s : AFI_IP event %d", __func__
, event
);
2859 rib_update_table(table
, event
);
2862 table
= zebra_vrf_table(AFI_IP6
, SAFI_UNICAST
, vrf_id
);
2864 if (IS_ZEBRA_DEBUG_EVENT
)
2865 zlog_debug("%s : AFI_IP6 event %d", __func__
, event
);
2866 rib_update_table(table
, event
);
2870 /* Delete self installed routes after zebra is relaunched. */
2871 void rib_sweep_table(struct route_table
*table
)
2873 struct route_node
*rn
;
2874 struct route_entry
*re
;
2875 struct route_entry
*next
;
2876 struct nexthop
*nexthop
;
2881 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
2882 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
2883 if (IS_ZEBRA_DEBUG_RIB
)
2884 route_entry_dump(&rn
->p
, NULL
, re
);
2886 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
2889 if (!CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELFROUTE
))
2893 * So we are starting up and have received
2894 * routes from the kernel that we have installed
2895 * from a previous run of zebra but not cleaned
2896 * up ( say a kill -9 )
2897 * But since we haven't actually installed
2898 * them yet( we received them from the kernel )
2899 * we don't think they are active.
2900 * So let's pretend they are active to actually
2902 * In all honesty I'm not sure if we should
2903 * mark them as active when we receive them
2904 * This is startup only so probably ok.
2906 * If we ever decide to move rib_sweep_table
2907 * to a different spot (ie startup )
2908 * this decision needs to be revisited
2910 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
2911 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
2913 rib_uninstall_kernel(rn
, re
);
2914 rib_delnode(rn
, re
);
2919 /* Sweep all RIB tables. */
2920 void rib_sweep_route(void)
2923 struct zebra_vrf
*zvrf
;
2925 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
2926 if ((zvrf
= vrf
->info
) == NULL
)
2929 rib_sweep_table(zvrf
->table
[AFI_IP
][SAFI_UNICAST
]);
2930 rib_sweep_table(zvrf
->table
[AFI_IP6
][SAFI_UNICAST
]);
2933 zebra_ns_sweep_route();
2936 /* Remove specific by protocol routes from 'table'. */
2937 unsigned long rib_score_proto_table(uint8_t proto
, unsigned short instance
,
2938 struct route_table
*table
)
2940 struct route_node
*rn
;
2941 struct route_entry
*re
;
2942 struct route_entry
*next
;
2943 unsigned long n
= 0;
2946 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
))
2947 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
2948 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
2950 if (re
->type
== proto
2951 && re
->instance
== instance
) {
2952 rib_delnode(rn
, re
);
2959 /* Remove specific by protocol routes. */
2960 unsigned long rib_score_proto(uint8_t proto
, unsigned short instance
)
2963 struct zebra_vrf
*zvrf
;
2964 unsigned long cnt
= 0;
2966 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
)
2967 if ((zvrf
= vrf
->info
) != NULL
)
2968 cnt
+= rib_score_proto_table(
2970 zvrf
->table
[AFI_IP
][SAFI_UNICAST
])
2971 + rib_score_proto_table(
2973 zvrf
->table
[AFI_IP6
][SAFI_UNICAST
]);
2975 cnt
+= zebra_ns_score_proto(proto
, instance
);
2980 /* Close RIB and clean up kernel routes. */
2981 void rib_close_table(struct route_table
*table
)
2983 struct route_node
*rn
;
2984 rib_table_info_t
*info
;
2990 info
= route_table_get_info(table
);
2992 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
2993 dest
= rib_dest_from_rnode(rn
);
2995 if (dest
&& dest
->selected_fib
) {
2996 if (info
->safi
== SAFI_UNICAST
)
2997 hook_call(rib_update
, rn
, NULL
);
2999 if (!RIB_SYSTEM_ROUTE(dest
->selected_fib
))
3000 rib_uninstall_kernel(rn
, dest
->selected_fib
);
3005 /* Routing information base initialize. */
3008 rib_queue_init(&zebrad
);
3014 * Get the first vrf id that is greater than the given vrf id if any.
3016 * Returns TRUE if a vrf id was found, FALSE otherwise.
3018 static inline int vrf_id_get_next(vrf_id_t vrf_id
, vrf_id_t
*next_id_p
)
3022 vrf
= vrf_lookup_by_id(vrf_id
);
3024 vrf
= RB_NEXT(vrf_id_head
, vrf
);
3026 *next_id_p
= vrf
->vrf_id
;
3035 * rib_tables_iter_next
3037 * Returns the next table in the iteration.
3039 struct route_table
*rib_tables_iter_next(rib_tables_iter_t
*iter
)
3041 struct route_table
*table
;
3044 * Array that helps us go over all AFI/SAFI combinations via one
3051 {AFI_IP
, SAFI_UNICAST
}, {AFI_IP
, SAFI_MULTICAST
},
3052 {AFI_IP
, SAFI_LABELED_UNICAST
}, {AFI_IP6
, SAFI_UNICAST
},
3053 {AFI_IP6
, SAFI_MULTICAST
}, {AFI_IP6
, SAFI_LABELED_UNICAST
},
3058 switch (iter
->state
) {
3060 case RIB_TABLES_ITER_S_INIT
:
3061 iter
->vrf_id
= VRF_DEFAULT
;
3062 iter
->afi_safi_ix
= -1;
3066 case RIB_TABLES_ITER_S_ITERATING
:
3067 iter
->afi_safi_ix
++;
3070 while (iter
->afi_safi_ix
3071 < (int)ZEBRA_NUM_OF(afi_safis
)) {
3072 table
= zebra_vrf_table(
3073 afi_safis
[iter
->afi_safi_ix
].afi
,
3074 afi_safis
[iter
->afi_safi_ix
].safi
,
3079 iter
->afi_safi_ix
++;
3083 * Found another table in this vrf.
3089 * Done with all tables in the current vrf, go to the
3093 if (!vrf_id_get_next(iter
->vrf_id
, &iter
->vrf_id
))
3096 iter
->afi_safi_ix
= 0;
3101 case RIB_TABLES_ITER_S_DONE
:
3106 iter
->state
= RIB_TABLES_ITER_S_ITERATING
;
3108 iter
->state
= RIB_TABLES_ITER_S_DONE
;