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__
);
610 struct route_entry
*rib_match(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
,
611 union g_addr
*addr
, struct route_node
**rn_out
)
614 struct route_table
*table
;
615 struct route_node
*rn
;
616 struct route_entry
*match
= NULL
;
617 struct nexthop
*newhop
;
620 table
= zebra_vrf_table(afi
, safi
, vrf_id
);
624 memset(&p
, 0, sizeof(struct prefix
));
627 p
.u
.prefix4
= addr
->ipv4
;
628 p
.prefixlen
= IPV4_MAX_PREFIXLEN
;
630 p
.u
.prefix6
= addr
->ipv6
;
631 p
.prefixlen
= IPV6_MAX_PREFIXLEN
;
634 rn
= route_node_match(table
, (struct prefix
*)&p
);
639 route_unlock_node(rn
);
641 dest
= rib_dest_from_rnode(rn
);
642 if (dest
&& dest
->selected_fib
643 && !CHECK_FLAG(dest
->selected_fib
->status
,
644 ROUTE_ENTRY_REMOVED
))
645 match
= dest
->selected_fib
;
647 /* If there is no selected route or matched route is EGP, go up
652 } while (rn
&& rn
->info
== NULL
);
656 if (match
->type
!= ZEBRA_ROUTE_CONNECT
) {
658 for (ALL_NEXTHOPS(match
->ng
, newhop
))
659 if (CHECK_FLAG(newhop
->flags
,
676 struct route_entry
*rib_match_ipv4_multicast(vrf_id_t vrf_id
,
678 struct route_node
**rn_out
)
680 struct route_entry
*re
= NULL
, *mre
= NULL
, *ure
= NULL
;
681 struct route_node
*m_rn
= NULL
, *u_rn
= NULL
;
682 union g_addr gaddr
= {.ipv4
= addr
};
684 switch (ipv4_multicast_mode
) {
685 case MCAST_MRIB_ONLY
:
686 return rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
,
688 case MCAST_URIB_ONLY
:
689 return rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, rn_out
);
690 case MCAST_NO_CONFIG
:
691 case MCAST_MIX_MRIB_FIRST
:
692 re
= mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
,
695 re
= ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
,
698 case MCAST_MIX_DISTANCE
:
699 mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
, &m_rn
);
700 ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, &u_rn
);
702 re
= ure
->distance
< mre
->distance
? ure
: mre
;
708 case MCAST_MIX_PFXLEN
:
709 mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
, &m_rn
);
710 ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, &u_rn
);
712 re
= u_rn
->p
.prefixlen
> m_rn
->p
.prefixlen
? ure
: mre
;
721 *rn_out
= (re
== mre
) ? m_rn
: u_rn
;
723 if (IS_ZEBRA_DEBUG_RIB
) {
725 inet_ntop(AF_INET
, &addr
, buf
, BUFSIZ
);
727 zlog_debug("%s: %s: vrf: %u found %s, using %s",
728 __func__
, buf
, vrf_id
,
729 mre
? (ure
? "MRIB+URIB" : "MRIB")
730 : ure
? "URIB" : "nothing",
731 re
== ure
? "URIB" : re
== mre
? "MRIB" : "none");
736 void multicast_mode_ipv4_set(enum multicast_mode mode
)
738 if (IS_ZEBRA_DEBUG_RIB
)
739 zlog_debug("%s: multicast lookup mode set (%d)", __func__
,
741 ipv4_multicast_mode
= mode
;
744 enum multicast_mode
multicast_mode_ipv4_get(void)
746 return ipv4_multicast_mode
;
749 struct route_entry
*rib_lookup_ipv4(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
751 struct route_table
*table
;
752 struct route_node
*rn
;
753 struct route_entry
*match
= NULL
;
754 struct nexthop
*nexthop
;
758 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
762 rn
= route_node_lookup(table
, (struct prefix
*)p
);
764 /* No route for this prefix. */
769 route_unlock_node(rn
);
770 dest
= rib_dest_from_rnode(rn
);
772 if (dest
&& dest
->selected_fib
773 && !CHECK_FLAG(dest
->selected_fib
->status
, ROUTE_ENTRY_REMOVED
))
774 match
= dest
->selected_fib
;
779 if (match
->type
== ZEBRA_ROUTE_CONNECT
)
782 for (ALL_NEXTHOPS(match
->ng
, nexthop
))
783 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
790 * This clone function, unlike its original rib_lookup_ipv4(), checks
791 * if specified IPv4 route record (prefix/mask -> gate) exists in
792 * the whole RIB and has ROUTE_ENTRY_SELECTED_FIB set.
796 * 0: exact match found
797 * 1: a match was found with a different gate
798 * 2: connected route found
799 * 3: no matches found
801 int rib_lookup_ipv4_route(struct prefix_ipv4
*p
, union sockunion
*qgate
,
804 struct route_table
*table
;
805 struct route_node
*rn
;
806 struct route_entry
*match
= NULL
;
807 struct nexthop
*nexthop
;
812 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
814 return ZEBRA_RIB_LOOKUP_ERROR
;
816 /* Scan the RIB table for exactly matching RIB entry. */
817 rn
= route_node_lookup(table
, (struct prefix
*)p
);
819 /* No route for this prefix. */
821 return ZEBRA_RIB_NOTFOUND
;
824 route_unlock_node(rn
);
825 dest
= rib_dest_from_rnode(rn
);
827 /* Find out if a "selected" RR for the discovered RIB entry exists ever.
829 if (dest
&& dest
->selected_fib
830 && !CHECK_FLAG(dest
->selected_fib
->status
, ROUTE_ENTRY_REMOVED
))
831 match
= dest
->selected_fib
;
833 /* None such found :( */
835 return ZEBRA_RIB_NOTFOUND
;
837 if (match
->type
== ZEBRA_ROUTE_CONNECT
)
838 return ZEBRA_RIB_FOUND_CONNECTED
;
840 /* Ok, we have a cood candidate, let's check it's nexthop list... */
842 for (ALL_NEXTHOPS(match
->ng
, nexthop
))
843 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
)) {
845 if (nexthop
->gate
.ipv4
.s_addr
== sockunion2ip(qgate
))
846 return ZEBRA_RIB_FOUND_EXACT
;
847 if (IS_ZEBRA_DEBUG_RIB
) {
848 char gate_buf
[INET_ADDRSTRLEN
],
849 qgate_buf
[INET_ADDRSTRLEN
];
850 inet_ntop(AF_INET
, &nexthop
->gate
.ipv4
.s_addr
,
851 gate_buf
, INET_ADDRSTRLEN
);
852 inet_ntop(AF_INET
, &sockunion2ip(qgate
),
853 qgate_buf
, INET_ADDRSTRLEN
);
854 zlog_debug("%s: qgate == %s, %s == %s",
856 nexthop
->rparent
? "rgate" : "gate",
862 return ZEBRA_RIB_FOUND_NOGATE
;
864 return ZEBRA_RIB_NOTFOUND
;
867 #define RIB_SYSTEM_ROUTE(R) \
868 ((R)->type == ZEBRA_ROUTE_KERNEL || (R)->type == ZEBRA_ROUTE_CONNECT)
870 /* This function verifies reachability of one given nexthop, which can be
871 * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored
872 * in nexthop->flags field. If the 4th parameter, 'set', is non-zero,
873 * nexthop->ifindex will be updated appropriately as well.
874 * An existing route map can turn (otherwise active) nexthop into inactive, but
877 * The return value is the final value of 'ACTIVE' flag.
880 static unsigned nexthop_active_check(struct route_node
*rn
,
881 struct route_entry
*re
,
882 struct nexthop
*nexthop
, int set
)
884 struct interface
*ifp
;
885 route_map_result_t ret
= RMAP_MATCH
;
887 char buf
[SRCDEST2STR_BUFFER
];
888 const struct prefix
*p
, *src_p
;
889 struct zebra_vrf
*zvrf
;
891 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
893 if (rn
->p
.family
== AF_INET
)
895 else if (rn
->p
.family
== AF_INET6
)
899 switch (nexthop
->type
) {
900 case NEXTHOP_TYPE_IFINDEX
:
901 ifp
= if_lookup_by_index(nexthop
->ifindex
, nexthop
->vrf_id
);
902 if (ifp
&& if_is_operative(ifp
))
903 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
905 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
907 case NEXTHOP_TYPE_IPV4
:
908 case NEXTHOP_TYPE_IPV4_IFINDEX
:
910 if (nexthop_active(AFI_IP
, re
, nexthop
, set
, rn
))
911 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
913 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
915 case NEXTHOP_TYPE_IPV6
:
917 if (nexthop_active(AFI_IP6
, re
, nexthop
, set
, rn
))
918 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
920 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
922 case NEXTHOP_TYPE_IPV6_IFINDEX
:
923 /* RFC 5549, v4 prefix with v6 NH */
924 if (rn
->p
.family
!= AF_INET
)
926 if (IN6_IS_ADDR_LINKLOCAL(&nexthop
->gate
.ipv6
)) {
927 ifp
= if_lookup_by_index(nexthop
->ifindex
,
929 if (ifp
&& if_is_operative(ifp
))
930 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
932 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
934 if (nexthop_active(AFI_IP6
, re
, nexthop
, set
, rn
))
935 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
937 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
940 case NEXTHOP_TYPE_BLACKHOLE
:
941 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
946 if (!CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
)) {
947 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
948 zlog_debug("\t%s: Unable to find a active nexthop",
949 __PRETTY_FUNCTION__
);
953 /* XXX: What exactly do those checks do? Do we support
954 * e.g. IPv4 routes with IPv6 nexthops or vice versa?
956 if (RIB_SYSTEM_ROUTE(re
) || (family
== AFI_IP
&& p
->family
!= AF_INET
)
957 || (family
== AFI_IP6
&& p
->family
!= AF_INET6
))
958 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
960 /* The original code didn't determine the family correctly
961 * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi
962 * from the rib_table_info in those cases.
963 * Possibly it may be better to use only the rib_table_info
967 rib_table_info_t
*info
;
969 info
= srcdest_rnode_table_info(rn
);
973 memset(&nexthop
->rmap_src
.ipv6
, 0, sizeof(union g_addr
));
975 zvrf
= zebra_vrf_lookup_by_id(nexthop
->vrf_id
);
977 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
978 zlog_debug("\t%s: zvrf is NULL", __PRETTY_FUNCTION__
);
979 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
982 /* It'll get set if required inside */
983 ret
= zebra_route_map_check(family
, re
->type
, re
->instance
, p
,
984 nexthop
, zvrf
, re
->tag
);
985 if (ret
== RMAP_DENYMATCH
) {
986 if (IS_ZEBRA_DEBUG_RIB
) {
987 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
989 "%u:%s: Filtering out with NH out %s due to route map",
991 ifindex2ifname(nexthop
->ifindex
,
994 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
996 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
999 /* Iterate over all nexthops of the given RIB entry and refresh their
1000 * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any
1001 * nexthop is found to toggle the ACTIVE flag, the whole re structure
1002 * is flagged with ROUTE_ENTRY_CHANGED. The 4th 'set' argument is
1003 * transparently passed to nexthop_active_check().
1005 * Return value is the new number of active nexthops.
1008 static int nexthop_active_update(struct route_node
*rn
, struct route_entry
*re
,
1011 struct nexthop
*nexthop
;
1012 union g_addr prev_src
;
1013 unsigned int prev_active
, new_active
, old_num_nh
;
1014 ifindex_t prev_index
;
1016 old_num_nh
= re
->nexthop_active_num
;
1018 re
->nexthop_active_num
= 0;
1019 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1021 for (nexthop
= re
->ng
.nexthop
; nexthop
; nexthop
= nexthop
->next
) {
1022 /* No protocol daemon provides src and so we're skipping
1024 prev_src
= nexthop
->rmap_src
;
1025 prev_active
= CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
1026 prev_index
= nexthop
->ifindex
;
1027 if ((new_active
= nexthop_active_check(rn
, re
, nexthop
, set
)))
1028 re
->nexthop_active_num
++;
1029 /* Don't allow src setting on IPv6 addr for now */
1030 if (prev_active
!= new_active
|| prev_index
!= nexthop
->ifindex
1031 || ((nexthop
->type
>= NEXTHOP_TYPE_IFINDEX
1032 && nexthop
->type
< NEXTHOP_TYPE_IPV6
)
1033 && prev_src
.ipv4
.s_addr
1034 != nexthop
->rmap_src
.ipv4
.s_addr
)
1035 || ((nexthop
->type
>= NEXTHOP_TYPE_IPV6
1036 && nexthop
->type
< NEXTHOP_TYPE_BLACKHOLE
)
1037 && !(IPV6_ADDR_SAME(&prev_src
.ipv6
,
1038 &nexthop
->rmap_src
.ipv6
)))) {
1039 SET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1040 SET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1044 if (old_num_nh
!= re
->nexthop_active_num
)
1045 SET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1047 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
)) {
1048 SET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1051 return re
->nexthop_active_num
;
1055 * Is this RIB labeled-unicast? It must be of type BGP and all paths
1056 * (nexthops) must have a label.
1058 int zebra_rib_labeled_unicast(struct route_entry
*re
)
1060 struct nexthop
*nexthop
= NULL
;
1062 if (re
->type
!= ZEBRA_ROUTE_BGP
)
1065 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1066 if (!nexthop
->nh_label
|| !nexthop
->nh_label
->num_labels
)
1072 void kernel_route_rib_pass_fail(struct route_node
*rn
, const struct prefix
*p
,
1073 struct route_entry
*re
,
1074 enum zebra_dplane_status res
)
1076 struct nexthop
*nexthop
;
1077 char buf
[PREFIX_STRLEN
];
1080 dest
= rib_dest_from_rnode(rn
);
1083 case ZEBRA_DPLANE_INSTALL_SUCCESS
:
1084 dest
->selected_fib
= re
;
1085 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1086 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
))
1089 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
))
1090 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1092 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1094 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_INSTALLED
);
1096 case ZEBRA_DPLANE_INSTALL_FAILURE
:
1098 * I am not sure this is the right thing to do here
1099 * but the code always set selected_fib before
1100 * this assignment was moved here.
1102 dest
->selected_fib
= re
;
1104 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_FAIL_INSTALL
);
1105 flog_err(EC_ZEBRA_DP_INSTALL_FAIL
,
1106 "%u:%s: Route install failed", re
->vrf_id
,
1107 prefix2str(p
, buf
, sizeof(buf
)));
1109 case ZEBRA_DPLANE_DELETE_SUCCESS
:
1111 * The case where selected_fib is not re is
1112 * when we have received a system route
1113 * that is overriding our installed route
1114 * as such we should leave the selected_fib
1117 if (dest
->selected_fib
== re
)
1118 dest
->selected_fib
= NULL
;
1119 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1120 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1122 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_REMOVED
);
1124 case ZEBRA_DPLANE_DELETE_FAILURE
:
1126 * Should we set this to NULL if the
1129 dest
->selected_fib
= NULL
;
1130 flog_err(EC_ZEBRA_DP_DELETE_FAIL
,
1131 "%u:%s: Route Deletion failure", re
->vrf_id
,
1132 prefix2str(p
, buf
, sizeof(buf
)));
1134 zsend_route_notify_owner(re
, p
, ZAPI_ROUTE_REMOVE_FAIL
);
1136 case ZEBRA_DPLANE_STATUS_NONE
:
1141 /* Update flag indicates whether this is a "replace" or not. Currently, this
1142 * is only used for IPv4.
1144 void rib_install_kernel(struct route_node
*rn
, struct route_entry
*re
,
1145 struct route_entry
*old
)
1147 struct nexthop
*nexthop
;
1148 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1149 const struct prefix
*p
, *src_p
;
1150 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1152 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1154 if (info
->safi
!= SAFI_UNICAST
) {
1155 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1156 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1159 struct nexthop
*prev
;
1161 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1162 UNSET_FLAG (nexthop
->flags
, NEXTHOP_FLAG_DUPLICATE
);
1163 for (ALL_NEXTHOPS(re
->ng
, prev
)) {
1164 if (prev
== nexthop
)
1166 if (nexthop_same_firsthop(nexthop
, prev
)) {
1167 SET_FLAG(nexthop
->flags
,
1168 NEXTHOP_FLAG_DUPLICATE
);
1176 * If this is a replace to a new RE let the originator of the RE
1177 * know that they've lost
1179 if (old
&& (old
!= re
) && (old
->type
!= re
->type
))
1180 zsend_route_notify_owner(old
, p
, ZAPI_ROUTE_BETTER_ADMIN_WON
);
1183 * Make sure we update the FPM any time we send new information to
1186 hook_call(rib_update
, rn
, "installing in kernel");
1187 switch (kernel_route_rib(rn
, p
, src_p
, old
, re
)) {
1188 case ZEBRA_DPLANE_REQUEST_QUEUED
:
1190 EC_ZEBRA_DP_INVALID_RC
,
1191 "No current known DataPlane interfaces can return this, please fix");
1193 case ZEBRA_DPLANE_REQUEST_FAILURE
:
1195 EC_ZEBRA_DP_INSTALL_FAIL
,
1196 "No current known Rib Install Failure cases, please fix");
1198 case ZEBRA_DPLANE_REQUEST_SUCCESS
:
1206 /* Uninstall the route from kernel. */
1207 void rib_uninstall_kernel(struct route_node
*rn
, struct route_entry
*re
)
1209 struct nexthop
*nexthop
;
1210 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1211 const struct prefix
*p
, *src_p
;
1212 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1214 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1216 if (info
->safi
!= SAFI_UNICAST
) {
1217 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1218 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1223 * Make sure we update the FPM any time we send new information to
1226 hook_call(rib_update
, rn
, "uninstalling from kernel");
1227 switch (kernel_route_rib(rn
, p
, src_p
, re
, NULL
)) {
1228 case ZEBRA_DPLANE_REQUEST_QUEUED
:
1230 EC_ZEBRA_DP_INVALID_RC
,
1231 "No current known DataPlane interfaces can return this, please fix");
1233 case ZEBRA_DPLANE_REQUEST_FAILURE
:
1235 EC_ZEBRA_DP_INSTALL_FAIL
,
1236 "No current known RIB Install Failure cases, please fix");
1238 case ZEBRA_DPLANE_REQUEST_SUCCESS
:
1247 /* Uninstall the route from kernel. */
1248 static void rib_uninstall(struct route_node
*rn
, struct route_entry
*re
)
1250 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1251 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1253 if (dest
&& dest
->selected_fib
== re
) {
1254 if (info
->safi
== SAFI_UNICAST
)
1255 hook_call(rib_update
, rn
, "rib_uninstall");
1257 if (!RIB_SYSTEM_ROUTE(re
))
1258 rib_uninstall_kernel(rn
, re
);
1260 /* If labeled-unicast route, uninstall transit LSP. */
1261 if (zebra_rib_labeled_unicast(re
))
1262 zebra_mpls_lsp_uninstall(info
->zvrf
, rn
, re
);
1265 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)) {
1266 const struct prefix
*p
, *src_p
;
1268 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1270 redistribute_delete(p
, src_p
, re
);
1271 UNSET_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
);
1276 * rib_can_delete_dest
1278 * Returns TRUE if the given dest can be deleted from the table.
1280 static int rib_can_delete_dest(rib_dest_t
*dest
)
1287 * Don't delete the dest if we have to update the FPM about this
1290 if (CHECK_FLAG(dest
->flags
, RIB_DEST_UPDATE_FPM
)
1291 || CHECK_FLAG(dest
->flags
, RIB_DEST_SENT_TO_FPM
))
1300 * Garbage collect the rib dest corresponding to the given route node
1303 * Returns TRUE if the dest was deleted, FALSE otherwise.
1305 int rib_gc_dest(struct route_node
*rn
)
1309 dest
= rib_dest_from_rnode(rn
);
1313 if (!rib_can_delete_dest(dest
))
1316 if (IS_ZEBRA_DEBUG_RIB
) {
1317 struct zebra_vrf
*zvrf
;
1319 zvrf
= rib_dest_vrf(dest
);
1320 rnode_debug(rn
, zvrf_id(zvrf
), "removing dest from table");
1324 XFREE(MTYPE_RIB_DEST
, dest
);
1328 * Release the one reference that we keep on the route node.
1330 route_unlock_node(rn
);
1334 static void rib_process_add_fib(struct zebra_vrf
*zvrf
, struct route_node
*rn
,
1335 struct route_entry
*new)
1337 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1339 hook_call(rib_update
, rn
, "new route selected");
1341 /* Update real nexthop. This may actually determine if nexthop is active
1343 if (!nexthop_active_update(rn
, new, 1)) {
1344 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1348 if (IS_ZEBRA_DEBUG_RIB
) {
1349 char buf
[SRCDEST2STR_BUFFER
];
1350 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1351 zlog_debug("%u:%s: Adding route rn %p, re %p (type %d)",
1352 zvrf_id(zvrf
), buf
, rn
, new, new->type
);
1355 /* If labeled-unicast route, install transit LSP. */
1356 if (zebra_rib_labeled_unicast(new))
1357 zebra_mpls_lsp_install(zvrf
, rn
, new);
1359 if (!RIB_SYSTEM_ROUTE(new))
1360 rib_install_kernel(rn
, new, NULL
);
1362 dest
->selected_fib
= new;
1364 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1367 static void rib_process_del_fib(struct zebra_vrf
*zvrf
, struct route_node
*rn
,
1368 struct route_entry
*old
)
1370 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1371 hook_call(rib_update
, rn
, "removing existing route");
1373 /* Uninstall from kernel. */
1374 if (IS_ZEBRA_DEBUG_RIB
) {
1375 char buf
[SRCDEST2STR_BUFFER
];
1376 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1377 zlog_debug("%u:%s: Deleting route rn %p, re %p (type %d)",
1378 zvrf_id(zvrf
), buf
, rn
, old
, old
->type
);
1381 /* If labeled-unicast route, uninstall transit LSP. */
1382 if (zebra_rib_labeled_unicast(old
))
1383 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1385 if (!RIB_SYSTEM_ROUTE(old
))
1386 rib_uninstall_kernel(rn
, old
);
1389 * We are setting this to NULL here
1390 * because that is what we traditionally
1391 * have been doing. I am not positive
1392 * that this is the right thing to do
1393 * but let's leave the code alone
1394 * for the RIB_SYSTEM_ROUTE case
1396 dest
->selected_fib
= NULL
;
1399 /* Update nexthop for route, reset changed flag. */
1400 nexthop_active_update(rn
, old
, 1);
1401 UNSET_FLAG(old
->status
, ROUTE_ENTRY_CHANGED
);
1404 static void rib_process_update_fib(struct zebra_vrf
*zvrf
,
1405 struct route_node
*rn
,
1406 struct route_entry
*old
,
1407 struct route_entry
*new)
1409 struct nexthop
*nexthop
= NULL
;
1411 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1414 * We have to install or update if a new route has been selected or
1415 * something has changed.
1417 if (new != old
|| CHECK_FLAG(new->status
, ROUTE_ENTRY_CHANGED
)) {
1418 hook_call(rib_update
, rn
, "updating existing route");
1420 /* Update the nexthop; we could determine here that nexthop is
1422 if (nexthop_active_update(rn
, new, 1))
1425 /* If nexthop is active, install the selected route, if
1427 * the install succeeds, cleanup flags for prior route, if
1432 if (IS_ZEBRA_DEBUG_RIB
) {
1433 char buf
[SRCDEST2STR_BUFFER
];
1434 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1437 "%u:%s: Updating route rn %p, re %p (type %d) "
1439 zvrf_id(zvrf
), buf
, rn
, new,
1440 new->type
, old
, old
->type
);
1443 "%u:%s: Updating route rn %p, re %p (type %d)",
1444 zvrf_id(zvrf
), buf
, rn
, new,
1448 /* If labeled-unicast route, uninstall transit LSP. */
1449 if (zebra_rib_labeled_unicast(old
))
1450 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1452 /* Non-system route should be installed. */
1453 if (!RIB_SYSTEM_ROUTE(new)) {
1454 /* If labeled-unicast route, install transit
1456 if (zebra_rib_labeled_unicast(new))
1457 zebra_mpls_lsp_install(zvrf
, rn
, new);
1459 rib_install_kernel(rn
, new, old
);
1462 * We do not need to install the
1463 * selected route because it
1464 * is already isntalled by
1465 * the system( ie not us )
1466 * so just mark it as winning
1467 * we do need to ensure that
1468 * if we uninstall a route
1469 * from ourselves we don't
1470 * over write this pointer
1472 dest
->selected_fib
= NULL
;
1474 /* If install succeeded or system route, cleanup flags
1475 * for prior route. */
1477 if (RIB_SYSTEM_ROUTE(new)) {
1478 if (!RIB_SYSTEM_ROUTE(old
))
1479 rib_uninstall_kernel(rn
, old
);
1481 for (nexthop
= old
->ng
.nexthop
; nexthop
;
1482 nexthop
= nexthop
->next
)
1483 UNSET_FLAG(nexthop
->flags
,
1490 * If nexthop for selected route is not active or install
1492 * may need to uninstall and delete for redistribution.
1495 if (IS_ZEBRA_DEBUG_RIB
) {
1496 char buf
[SRCDEST2STR_BUFFER
];
1497 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1500 "%u:%s: Deleting route rn %p, re %p (type %d) "
1501 "old %p (type %d) - nexthop inactive",
1502 zvrf_id(zvrf
), buf
, rn
, new,
1503 new->type
, old
, old
->type
);
1506 "%u:%s: Deleting route rn %p, re %p (type %d) - nexthop inactive",
1507 zvrf_id(zvrf
), buf
, rn
, new,
1511 /* If labeled-unicast route, uninstall transit LSP. */
1512 if (zebra_rib_labeled_unicast(old
))
1513 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1515 if (!RIB_SYSTEM_ROUTE(old
))
1516 rib_uninstall_kernel(rn
, old
);
1518 dest
->selected_fib
= NULL
;
1522 * Same route selected; check if in the FIB and if not,
1524 * is housekeeping code to deal with race conditions in kernel
1526 * netlink reporting interface up before IPv4 or IPv6 protocol
1530 if (!RIB_SYSTEM_ROUTE(new)) {
1531 bool in_fib
= false;
1533 for (ALL_NEXTHOPS(new->ng
, nexthop
))
1534 if (CHECK_FLAG(nexthop
->flags
,
1535 NEXTHOP_FLAG_FIB
)) {
1540 rib_install_kernel(rn
, new, NULL
);
1544 /* Update prior route. */
1546 /* Set real nexthop. */
1547 nexthop_active_update(rn
, old
, 1);
1548 UNSET_FLAG(old
->status
, ROUTE_ENTRY_CHANGED
);
1551 /* Clear changed flag. */
1552 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1555 /* Check if 'alternate' RIB entry is better than 'current'. */
1556 static struct route_entry
*rib_choose_best(struct route_entry
*current
,
1557 struct route_entry
*alternate
)
1559 if (current
== NULL
)
1562 /* filter route selection in following order:
1563 * - connected beats other types
1564 * - if both connected, loopback or vrf wins
1565 * - lower distance beats higher
1566 * - lower metric beats higher for equal distance
1567 * - last, hence oldest, route wins tie break.
1570 /* Connected routes. Check to see if either are a vrf
1571 * or loopback interface. If not, pick the last connected
1572 * route of the set of lowest metric connected routes.
1574 if (alternate
->type
== ZEBRA_ROUTE_CONNECT
) {
1575 if (current
->type
!= ZEBRA_ROUTE_CONNECT
)
1578 /* both are connected. are either loop or vrf? */
1579 struct nexthop
*nexthop
= NULL
;
1581 for (ALL_NEXTHOPS(alternate
->ng
, nexthop
)) {
1582 if (if_is_loopback_or_vrf(if_lookup_by_index(
1583 nexthop
->ifindex
, alternate
->vrf_id
)))
1587 for (ALL_NEXTHOPS(current
->ng
, nexthop
)) {
1588 if (if_is_loopback_or_vrf(if_lookup_by_index(
1589 nexthop
->ifindex
, current
->vrf_id
)))
1593 /* Neither are loop or vrf so pick best metric */
1594 if (alternate
->metric
<= current
->metric
)
1600 if (current
->type
== ZEBRA_ROUTE_CONNECT
)
1603 /* higher distance loses */
1604 if (alternate
->distance
< current
->distance
)
1606 if (current
->distance
< alternate
->distance
)
1609 /* metric tie-breaks equal distance */
1610 if (alternate
->metric
<= current
->metric
)
1616 /* Core function for processing routing information base. */
1617 static void rib_process(struct route_node
*rn
)
1619 struct route_entry
*re
;
1620 struct route_entry
*next
;
1621 struct route_entry
*old_selected
= NULL
;
1622 struct route_entry
*new_selected
= NULL
;
1623 struct route_entry
*old_fib
= NULL
;
1624 struct route_entry
*new_fib
= NULL
;
1625 struct route_entry
*best
= NULL
;
1626 char buf
[SRCDEST2STR_BUFFER
];
1628 struct zebra_vrf
*zvrf
= NULL
;
1629 const struct prefix
*p
, *src_p
;
1631 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1632 vrf_id_t vrf_id
= VRF_UNKNOWN
;
1636 dest
= rib_dest_from_rnode(rn
);
1638 zvrf
= rib_dest_vrf(dest
);
1639 vrf_id
= zvrf_id(zvrf
);
1642 if (IS_ZEBRA_DEBUG_RIB
)
1643 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1645 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1646 zlog_debug("%u:%s: Processing rn %p", vrf_id
, buf
, rn
);
1649 * we can have rn's that have a NULL info pointer
1650 * (dest). As such let's not let the deref happen
1651 * additionally we know RNODE_FOREACH_RE_SAFE
1652 * will not iterate so we are ok.
1655 old_fib
= dest
->selected_fib
;
1657 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
1658 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1660 "%u:%s: Examine re %p (type %d) status %x flags %x "
1661 "dist %d metric %d",
1662 vrf_id
, buf
, re
, re
->type
, re
->status
,
1663 re
->flags
, re
->distance
, re
->metric
);
1665 UNSET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1667 /* Currently selected re. */
1668 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)) {
1669 assert(old_selected
== NULL
);
1673 /* Skip deleted entries from selection */
1674 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
1677 /* Skip unreachable nexthop. */
1678 /* This first call to nexthop_active_update is merely to
1680 * there's any change to nexthops associated with this RIB
1682 * rib_process() can be invoked due to an external event such as
1684 * down or due to next-hop-tracking evaluation. In the latter
1686 * a decision has already been made that the NHs have changed.
1688 * need to invoke a potentially expensive call again. Further,
1690 * the change might be in a recursive NH which is not caught in
1691 * the nexthop_active_update() code. Thus, we might miss changes
1695 if (!CHECK_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
)
1696 && !nexthop_active_update(rn
, re
, 0)) {
1697 if (re
->type
== ZEBRA_ROUTE_TABLE
) {
1698 /* XXX: HERE BE DRAGONS!!!!!
1699 * In all honesty, I have not yet figured out
1701 * does or why the ROUTE_ENTRY_CHANGED test
1703 * or why we need to delete a route here, and
1705 * this concerns both selected and fib route, or
1708 /* This entry was denied by the 'ip protocol
1709 * table' route-map, we
1710 * need to delete it */
1711 if (re
!= old_selected
) {
1712 if (IS_ZEBRA_DEBUG_RIB
)
1714 "%s: %u:%s: imported via import-table but denied "
1715 "by the ip protocol table route-map",
1716 __func__
, vrf_id
, buf
);
1719 SET_FLAG(re
->status
,
1720 ROUTE_ENTRY_REMOVED
);
1726 /* Infinite distance. */
1727 if (re
->distance
== DISTANCE_INFINITY
) {
1728 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1732 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_FIB_OVERRIDE
)) {
1733 best
= rib_choose_best(new_fib
, re
);
1734 if (new_fib
&& best
!= new_fib
)
1735 UNSET_FLAG(new_fib
->status
,
1736 ROUTE_ENTRY_CHANGED
);
1739 best
= rib_choose_best(new_selected
, re
);
1740 if (new_selected
&& best
!= new_selected
)
1741 UNSET_FLAG(new_selected
->status
,
1742 ROUTE_ENTRY_CHANGED
);
1743 new_selected
= best
;
1746 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1747 } /* RNODE_FOREACH_RE */
1749 /* If no FIB override route, use the selected route also for FIB */
1750 if (new_fib
== NULL
)
1751 new_fib
= new_selected
;
1753 /* After the cycle is finished, the following pointers will be set:
1754 * old_selected --- RE entry currently having SELECTED
1755 * new_selected --- RE entry that is newly SELECTED
1756 * old_fib --- RE entry currently in kernel FIB
1757 * new_fib --- RE entry that is newly to be in kernel FIB
1759 * new_selected will get SELECTED flag, and is going to be redistributed
1760 * the zclients. new_fib (which can be new_selected) will be installed
1764 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
1766 "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p",
1767 vrf_id
, buf
, (void *)old_selected
, (void *)new_selected
,
1768 (void *)old_fib
, (void *)new_fib
);
1771 /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if
1772 * fib == selected */
1773 bool selected_changed
= new_selected
&& CHECK_FLAG(new_selected
->status
,
1774 ROUTE_ENTRY_CHANGED
);
1776 /* Update fib according to selection results */
1777 if (new_fib
&& old_fib
)
1778 rib_process_update_fib(zvrf
, rn
, old_fib
, new_fib
);
1780 rib_process_add_fib(zvrf
, rn
, new_fib
);
1782 rib_process_del_fib(zvrf
, rn
, old_fib
);
1784 /* Redistribute SELECTED entry */
1785 if (old_selected
!= new_selected
|| selected_changed
) {
1786 struct nexthop
*nexthop
= NULL
;
1788 /* Check if we have a FIB route for the destination, otherwise,
1789 * don't redistribute it */
1791 for (ALL_NEXTHOPS(new_fib
->ng
, nexthop
)) {
1792 if (CHECK_FLAG(nexthop
->flags
,
1793 NEXTHOP_FLAG_FIB
)) {
1799 new_selected
= NULL
;
1801 if (new_selected
&& new_selected
!= new_fib
) {
1802 nexthop_active_update(rn
, new_selected
, 1);
1803 UNSET_FLAG(new_selected
->status
, ROUTE_ENTRY_CHANGED
);
1808 redistribute_delete(p
, src_p
, old_selected
);
1809 if (old_selected
!= new_selected
)
1810 UNSET_FLAG(old_selected
->flags
,
1811 ZEBRA_FLAG_SELECTED
);
1815 /* Install new or replace existing redistributed entry
1817 SET_FLAG(new_selected
->flags
, ZEBRA_FLAG_SELECTED
);
1818 redistribute_update(p
, src_p
, new_selected
,
1823 /* Remove all RE entries queued for removal */
1824 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
1825 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
1826 if (IS_ZEBRA_DEBUG_RIB
) {
1827 rnode_debug(rn
, vrf_id
, "rn %p, removing re %p",
1828 (void *)rn
, (void *)re
);
1835 * Check if the dest can be deleted now.
1840 /* Take a list of route_node structs and return 1, if there was a record
1841 * picked from it and processed by rib_process(). Don't process more,
1842 * than one RN record; operate only in the specified sub-queue.
1844 static unsigned int process_subq(struct list
*subq
, uint8_t qindex
)
1846 struct listnode
*lnode
= listhead(subq
);
1847 struct route_node
*rnode
;
1849 struct zebra_vrf
*zvrf
= NULL
;
1854 rnode
= listgetdata(lnode
);
1855 dest
= rib_dest_from_rnode(rnode
);
1857 zvrf
= rib_dest_vrf(dest
);
1861 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
1862 char buf
[SRCDEST2STR_BUFFER
];
1863 srcdest_rnode2str(rnode
, buf
, sizeof(buf
));
1864 zlog_debug("%u:%s: rn %p dequeued from sub-queue %u",
1865 zvrf
? zvrf_id(zvrf
) : 0, buf
, rnode
, qindex
);
1869 UNSET_FLAG(rib_dest_from_rnode(rnode
)->flags
,
1870 RIB_ROUTE_QUEUED(qindex
));
1875 zlog_debug ("%s: called for route_node (%p, %d) with no ribs",
1876 __func__
, rnode
, rnode
->lock
);
1877 zlog_backtrace(LOG_DEBUG
);
1880 route_unlock_node(rnode
);
1881 list_delete_node(subq
, lnode
);
1886 * All meta queues have been processed. Trigger next-hop evaluation.
1888 static void meta_queue_process_complete(struct work_queue
*dummy
)
1891 struct zebra_vrf
*zvrf
;
1893 /* Evaluate nexthops for those VRFs which underwent route processing.
1895 * should limit the evaluation to the necessary VRFs in most common
1898 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
1900 if (zvrf
== NULL
|| !(zvrf
->flags
& ZEBRA_VRF_RIB_SCHEDULED
))
1903 zvrf
->flags
&= ~ZEBRA_VRF_RIB_SCHEDULED
;
1904 zebra_evaluate_rnh(zvrf_id(zvrf
), AF_INET
, 0, RNH_NEXTHOP_TYPE
,
1906 zebra_evaluate_rnh(zvrf_id(zvrf
), AF_INET
, 0,
1907 RNH_IMPORT_CHECK_TYPE
, NULL
);
1908 zebra_evaluate_rnh(zvrf_id(zvrf
), AF_INET6
, 0, RNH_NEXTHOP_TYPE
,
1910 zebra_evaluate_rnh(zvrf_id(zvrf
), AF_INET6
, 0,
1911 RNH_IMPORT_CHECK_TYPE
, NULL
);
1914 /* Schedule LSPs for processing, if needed. */
1915 zvrf
= vrf_info_lookup(VRF_DEFAULT
);
1916 if (mpls_should_lsps_be_processed(zvrf
)) {
1917 if (IS_ZEBRA_DEBUG_MPLS
)
1919 "%u: Scheduling all LSPs upon RIB completion",
1921 zebra_mpls_lsp_schedule(zvrf
);
1922 mpls_unmark_lsps_for_processing(zvrf
);
1926 /* Dispatch the meta queue by picking, processing and unlocking the next RN from
1927 * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and
1929 * is pointed to the meta queue structure.
1931 static wq_item_status
meta_queue_process(struct work_queue
*dummy
, void *data
)
1933 struct meta_queue
*mq
= data
;
1936 for (i
= 0; i
< MQ_SIZE
; i
++)
1937 if (process_subq(mq
->subq
[i
], i
)) {
1941 return mq
->size
? WQ_REQUEUE
: WQ_SUCCESS
;
1945 * Map from rib types to queue type (priority) in meta queue
1947 static const uint8_t meta_queue_map
[ZEBRA_ROUTE_MAX
] = {
1948 [ZEBRA_ROUTE_SYSTEM
] = 4,
1949 [ZEBRA_ROUTE_KERNEL
] = 0,
1950 [ZEBRA_ROUTE_CONNECT
] = 0,
1951 [ZEBRA_ROUTE_STATIC
] = 1,
1952 [ZEBRA_ROUTE_RIP
] = 2,
1953 [ZEBRA_ROUTE_RIPNG
] = 2,
1954 [ZEBRA_ROUTE_OSPF
] = 2,
1955 [ZEBRA_ROUTE_OSPF6
] = 2,
1956 [ZEBRA_ROUTE_ISIS
] = 2,
1957 [ZEBRA_ROUTE_BGP
] = 3,
1958 [ZEBRA_ROUTE_PIM
] = 4, // Shouldn't happen but for safety
1959 [ZEBRA_ROUTE_EIGRP
] = 2,
1960 [ZEBRA_ROUTE_NHRP
] = 2,
1961 [ZEBRA_ROUTE_HSLS
] = 4,
1962 [ZEBRA_ROUTE_OLSR
] = 4,
1963 [ZEBRA_ROUTE_TABLE
] = 1,
1964 [ZEBRA_ROUTE_LDP
] = 4,
1965 [ZEBRA_ROUTE_VNC
] = 3,
1966 [ZEBRA_ROUTE_VNC_DIRECT
] = 3,
1967 [ZEBRA_ROUTE_VNC_DIRECT_RH
] = 3,
1968 [ZEBRA_ROUTE_BGP_DIRECT
] = 3,
1969 [ZEBRA_ROUTE_BGP_DIRECT_EXT
] = 3,
1970 [ZEBRA_ROUTE_BABEL
] = 2,
1971 [ZEBRA_ROUTE_ALL
] = 4, // Shouldn't happen but for safety
1974 /* Look into the RN and queue it into one or more priority queues,
1975 * increasing the size for each data push done.
1977 static void rib_meta_queue_add(struct meta_queue
*mq
, struct route_node
*rn
)
1979 struct route_entry
*re
;
1981 RNODE_FOREACH_RE (rn
, re
) {
1982 uint8_t qindex
= meta_queue_map
[re
->type
];
1983 struct zebra_vrf
*zvrf
;
1985 /* Invariant: at this point we always have rn->info set. */
1986 if (CHECK_FLAG(rib_dest_from_rnode(rn
)->flags
,
1987 RIB_ROUTE_QUEUED(qindex
))) {
1988 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1991 "rn %p is already queued in sub-queue %u",
1992 (void *)rn
, qindex
);
1996 SET_FLAG(rib_dest_from_rnode(rn
)->flags
,
1997 RIB_ROUTE_QUEUED(qindex
));
1998 listnode_add(mq
->subq
[qindex
], rn
);
1999 route_lock_node(rn
);
2002 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2003 rnode_debug(rn
, re
->vrf_id
,
2004 "queued rn %p into sub-queue %u",
2005 (void *)rn
, qindex
);
2007 zvrf
= zebra_vrf_lookup_by_id(re
->vrf_id
);
2009 zvrf
->flags
|= ZEBRA_VRF_RIB_SCHEDULED
;
2013 /* Add route_node to work queue and schedule processing */
2014 void rib_queue_add(struct route_node
*rn
)
2018 /* Pointless to queue a route_node with no RIB entries to add or remove
2020 if (!rnode_to_ribs(rn
)) {
2021 zlog_debug("%s: called for route_node (%p, %d) with no ribs",
2022 __func__
, (void *)rn
, rn
->lock
);
2023 zlog_backtrace(LOG_DEBUG
);
2027 if (zebrad
.ribq
== NULL
) {
2028 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2029 "%s: work_queue does not exist!", __func__
);
2034 * The RIB queue should normally be either empty or holding the only
2035 * work_queue_item element. In the latter case this element would
2036 * hold a pointer to the meta queue structure, which must be used to
2037 * actually queue the route nodes to process. So create the MQ
2038 * holder, if necessary, then push the work into it in any case.
2039 * This semantics was introduced after 0.99.9 release.
2041 if (work_queue_empty(zebrad
.ribq
))
2042 work_queue_add(zebrad
.ribq
, zebrad
.mq
);
2044 rib_meta_queue_add(zebrad
.mq
, rn
);
2049 /* Create new meta queue.
2050 A destructor function doesn't seem to be necessary here.
2052 static struct meta_queue
*meta_queue_new(void)
2054 struct meta_queue
*new;
2057 new = XCALLOC(MTYPE_WORK_QUEUE
, sizeof(struct meta_queue
));
2059 for (i
= 0; i
< MQ_SIZE
; i
++) {
2060 new->subq
[i
] = list_new();
2061 assert(new->subq
[i
]);
2067 void meta_queue_free(struct meta_queue
*mq
)
2071 for (i
= 0; i
< MQ_SIZE
; i
++)
2072 list_delete(&mq
->subq
[i
]);
2074 XFREE(MTYPE_WORK_QUEUE
, mq
);
2077 /* initialise zebra rib work queue */
2078 static void rib_queue_init(struct zebra_t
*zebra
)
2083 work_queue_new(zebra
->master
, "route_node processing"))) {
2084 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2085 "%s: could not initialise work queue!", __func__
);
2089 /* fill in the work queue spec */
2090 zebra
->ribq
->spec
.workfunc
= &meta_queue_process
;
2091 zebra
->ribq
->spec
.errorfunc
= NULL
;
2092 zebra
->ribq
->spec
.completion_func
= &meta_queue_process_complete
;
2093 /* XXX: TODO: These should be runtime configurable via vty */
2094 zebra
->ribq
->spec
.max_retries
= 3;
2095 zebra
->ribq
->spec
.hold
= ZEBRA_RIB_PROCESS_HOLD_TIME
;
2097 if (!(zebra
->mq
= meta_queue_new())) {
2098 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2099 "%s: could not initialise meta queue!", __func__
);
2105 /* RIB updates are processed via a queue of pointers to route_nodes.
2107 * The queue length is bounded by the maximal size of the routing table,
2108 * as a route_node will not be requeued, if already queued.
2110 * REs are submitted via rib_addnode or rib_delnode which set minimal
2111 * state, or static_install_route (when an existing RE is updated)
2112 * and then submit route_node to queue for best-path selection later.
2113 * Order of add/delete state changes are preserved for any given RE.
2115 * Deleted REs are reaped during best-path selection.
2118 * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with
2119 * |-------->| | best RE, if required
2121 * static_install->|->rib_addqueue...... -> rib_process
2123 * |-------->| |-> rib_unlink
2124 * |-> set ROUTE_ENTRY_REMOVE |
2125 * rib_delnode (RE freed)
2127 * The 'info' pointer of a route_node points to a rib_dest_t
2128 * ('dest'). Queueing state for a route_node is kept on the dest. The
2129 * dest is created on-demand by rib_link() and is kept around at least
2130 * as long as there are ribs hanging off it (@see rib_gc_dest()).
2132 * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code):
2134 * - route_nodes: refcounted by:
2135 * - dest attached to route_node:
2136 * - managed by: rib_link/rib_gc_dest
2137 * - route_node processing queue
2138 * - managed by: rib_addqueue, rib_process.
2142 /* Add RE to head of the route node. */
2143 static void rib_link(struct route_node
*rn
, struct route_entry
*re
, int process
)
2145 struct route_entry
*head
;
2148 const char *rmap_name
;
2152 dest
= rib_dest_from_rnode(rn
);
2154 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2155 rnode_debug(rn
, re
->vrf_id
, "rn %p adding dest", rn
);
2157 dest
= XCALLOC(MTYPE_RIB_DEST
, sizeof(rib_dest_t
));
2158 route_lock_node(rn
); /* rn route table reference */
2163 head
= dest
->routes
;
2170 afi
= (rn
->p
.family
== AF_INET
)
2172 : (rn
->p
.family
== AF_INET6
) ? AFI_IP6
: AFI_MAX
;
2173 if (is_zebra_import_table_enabled(afi
, re
->table
)) {
2174 rmap_name
= zebra_get_import_table_route_map(afi
, re
->table
);
2175 zebra_add_import_table_entry(rn
, re
, rmap_name
);
2180 static void rib_addnode(struct route_node
*rn
,
2181 struct route_entry
*re
, int process
)
2183 /* RE node has been un-removed before route-node is processed.
2184 * route_node must hence already be on the queue for processing..
2186 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
2187 if (IS_ZEBRA_DEBUG_RIB
)
2188 rnode_debug(rn
, re
->vrf_id
, "rn %p, un-removed re %p",
2189 (void *)rn
, (void *)re
);
2191 UNSET_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
);
2194 rib_link(rn
, re
, process
);
2200 * Detach a rib structure from a route_node.
2202 * Note that a call to rib_unlink() should be followed by a call to
2203 * rib_gc_dest() at some point. This allows a rib_dest_t that is no
2204 * longer required to be deleted.
2206 void rib_unlink(struct route_node
*rn
, struct route_entry
*re
)
2212 if (IS_ZEBRA_DEBUG_RIB
)
2213 rnode_debug(rn
, re
->vrf_id
, "rn %p, re %p", (void *)rn
,
2216 dest
= rib_dest_from_rnode(rn
);
2219 re
->next
->prev
= re
->prev
;
2222 re
->prev
->next
= re
->next
;
2224 dest
->routes
= re
->next
;
2227 if (dest
->selected_fib
== re
)
2228 dest
->selected_fib
= NULL
;
2230 nexthops_free(re
->ng
.nexthop
);
2231 XFREE(MTYPE_RE
, re
);
2234 void rib_delnode(struct route_node
*rn
, struct route_entry
*re
)
2238 if (IS_ZEBRA_DEBUG_RIB
)
2239 rnode_debug(rn
, re
->vrf_id
, "rn %p, re %p, removing",
2240 (void *)rn
, (void *)re
);
2241 SET_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
);
2243 afi
= (rn
->p
.family
== AF_INET
)
2245 : (rn
->p
.family
== AF_INET6
) ? AFI_IP6
: AFI_MAX
;
2246 if (is_zebra_import_table_enabled(afi
, re
->table
)) {
2247 zebra_del_import_table_entry(rn
, re
);
2248 /* Just clean up if non main table */
2249 if (IS_ZEBRA_DEBUG_RIB
) {
2250 char buf
[SRCDEST2STR_BUFFER
];
2251 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
2253 "%u:%s: Freeing route rn %p, re %p (type %d)",
2254 re
->vrf_id
, buf
, rn
, re
, re
->type
);
2263 /* This function dumps the contents of a given RE entry into
2264 * standard debug log. Calling function name and IP prefix in
2265 * question are passed as 1st and 2nd arguments.
2268 void _route_entry_dump(const char *func
, union prefixconstptr pp
,
2269 union prefixconstptr src_pp
,
2270 const struct route_entry
*re
)
2272 const struct prefix
*src_p
= src_pp
.p
;
2273 bool is_srcdst
= src_p
&& src_p
->prefixlen
;
2274 char straddr
[PREFIX_STRLEN
];
2275 char srcaddr
[PREFIX_STRLEN
];
2276 struct nexthop
*nexthop
;
2278 zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func
,
2279 (const void *)re
, prefix2str(pp
, straddr
, sizeof(straddr
)),
2280 is_srcdst
? " from " : "",
2281 is_srcdst
? prefix2str(src_pp
, srcaddr
, sizeof(srcaddr
))
2284 zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d",
2285 func
, (unsigned long)re
->uptime
, re
->type
, re
->instance
,
2288 "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u",
2289 func
, re
->metric
, re
->mtu
, re
->distance
, re
->flags
, re
->status
);
2290 zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func
,
2291 re
->nexthop_num
, re
->nexthop_active_num
);
2293 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
2294 struct interface
*ifp
;
2295 struct vrf
*vrf
= vrf_lookup_by_id(nexthop
->vrf_id
);
2297 switch (nexthop
->type
) {
2298 case NEXTHOP_TYPE_BLACKHOLE
:
2299 sprintf(straddr
, "Blackhole");
2301 case NEXTHOP_TYPE_IFINDEX
:
2302 ifp
= if_lookup_by_index(nexthop
->ifindex
,
2304 sprintf(straddr
, "%s", ifp
? ifp
->name
: "Unknown");
2306 case NEXTHOP_TYPE_IPV4
:
2308 case NEXTHOP_TYPE_IPV4_IFINDEX
:
2309 inet_ntop(AF_INET
, &nexthop
->gate
, straddr
,
2312 case NEXTHOP_TYPE_IPV6
:
2313 case NEXTHOP_TYPE_IPV6_IFINDEX
:
2314 inet_ntop(AF_INET6
, &nexthop
->gate
, straddr
,
2318 zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s", func
,
2319 (nexthop
->rparent
? " NH" : "NH"), straddr
,
2320 nexthop
->ifindex
, vrf
? vrf
->name
: "Unknown",
2322 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
)
2325 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
)
2328 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
)
2332 zlog_debug("%s: dump complete", func
);
2335 /* This is an exported helper to rtm_read() to dump the strange
2336 * RE entry found by rib_lookup_ipv4_route()
2339 void rib_lookup_and_dump(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
2341 struct route_table
*table
;
2342 struct route_node
*rn
;
2343 struct route_entry
*re
;
2344 char prefix_buf
[INET_ADDRSTRLEN
];
2347 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
2349 flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED
,
2350 "%s:%u zebra_vrf_table() returned NULL", __func__
,
2355 /* Scan the RIB table for exactly matching RE entry. */
2356 rn
= route_node_lookup(table
, (struct prefix
*)p
);
2358 /* No route for this prefix. */
2360 zlog_debug("%s:%u lookup failed for %s", __func__
, vrf_id
,
2361 prefix2str((struct prefix
*)p
, prefix_buf
,
2362 sizeof(prefix_buf
)));
2367 route_unlock_node(rn
);
2370 RNODE_FOREACH_RE (rn
, re
) {
2371 zlog_debug("%s:%u rn %p, re %p: %s, %s",
2373 (void *)rn
, (void *)re
,
2374 (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)
2377 (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)
2380 route_entry_dump(p
, NULL
, re
);
2384 /* Check if requested address assignment will fail due to another
2385 * route being installed by zebra in FIB already. Take necessary
2386 * actions, if needed: remove such a route from FIB and deSELECT
2387 * corresponding RE entry. Then put affected RN into RIBQ head.
2389 void rib_lookup_and_pushup(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
2391 struct route_table
*table
;
2392 struct route_node
*rn
;
2393 unsigned changed
= 0;
2396 if (NULL
== (table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
))) {
2397 flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED
,
2398 "%s:%u zebra_vrf_table() returned NULL", __func__
,
2403 /* No matches would be the simplest case. */
2404 if (NULL
== (rn
= route_node_lookup(table
, (struct prefix
*)p
)))
2408 route_unlock_node(rn
);
2410 dest
= rib_dest_from_rnode(rn
);
2411 /* Check all RE entries. In case any changes have to be done, requeue
2412 * the RN into RIBQ head. If the routing message about the new connected
2413 * route (generated by the IP address we are going to assign very soon)
2414 * comes before the RIBQ is processed, the new RE entry will join
2415 * RIBQ record already on head. This is necessary for proper
2417 * of the rest of the RE.
2419 if (dest
->selected_fib
&& !RIB_SYSTEM_ROUTE(dest
->selected_fib
)) {
2421 if (IS_ZEBRA_DEBUG_RIB
) {
2422 char buf
[PREFIX_STRLEN
];
2424 zlog_debug("%u:%s: freeing way for connected prefix",
2425 dest
->selected_fib
->vrf_id
,
2426 prefix2str(&rn
->p
, buf
, sizeof(buf
)));
2427 route_entry_dump(&rn
->p
, NULL
, dest
->selected_fib
);
2429 rib_uninstall(rn
, dest
->selected_fib
);
2435 int rib_add_multipath(afi_t afi
, safi_t safi
, struct prefix
*p
,
2436 struct prefix_ipv6
*src_p
, struct route_entry
*re
)
2438 struct route_table
*table
;
2439 struct route_node
*rn
;
2440 struct route_entry
*same
= NULL
;
2441 struct nexthop
*nexthop
;
2447 assert(!src_p
|| !src_p
->prefixlen
|| afi
== AFI_IP6
);
2450 table
= zebra_vrf_table_with_table_id(afi
, safi
, re
->vrf_id
, re
->table
);
2452 XFREE(MTYPE_RE
, re
);
2456 /* Make it sure prefixlen is applied to the prefix. */
2459 apply_mask_ipv6(src_p
);
2461 /* Set default distance by route type. */
2462 if (re
->distance
== 0) {
2463 re
->distance
= route_distance(re
->type
);
2465 /* iBGP distance is 200. */
2466 if (re
->type
== ZEBRA_ROUTE_BGP
2467 && CHECK_FLAG(re
->flags
, ZEBRA_FLAG_IBGP
))
2471 /* Lookup route node.*/
2472 rn
= srcdest_rnode_get(table
, p
, src_p
);
2475 * If same type of route are installed, treat it as a implicit
2477 * If the user has specified the No route replace semantics
2478 * for the install don't do a route replace.
2480 RNODE_FOREACH_RE (rn
, same
) {
2481 if (CHECK_FLAG(same
->status
, ROUTE_ENTRY_REMOVED
))
2484 if (same
->type
!= re
->type
)
2486 if (same
->instance
!= re
->instance
)
2488 if (same
->type
== ZEBRA_ROUTE_KERNEL
2489 && same
->metric
!= re
->metric
)
2492 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_RR_USE_DISTANCE
) &&
2493 same
->distance
!= re
->distance
)
2497 * We should allow duplicate connected routes
2498 * because of IPv6 link-local routes and unnumbered
2499 * interfaces on Linux.
2501 if (same
->type
!= ZEBRA_ROUTE_CONNECT
)
2505 /* If this route is kernel route, set FIB flag to the route. */
2506 if (RIB_SYSTEM_ROUTE(re
))
2507 for (nexthop
= re
->ng
.nexthop
; nexthop
; nexthop
= nexthop
->next
)
2508 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
2510 /* Link new re to node.*/
2511 if (IS_ZEBRA_DEBUG_RIB
) {
2514 "Inserting route rn %p, re %p (type %d) existing %p",
2515 (void *)rn
, (void *)re
, re
->type
, (void *)same
);
2517 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2518 route_entry_dump(p
, src_p
, re
);
2520 rib_addnode(rn
, re
, 1);
2523 /* Free implicit route.*/
2525 rib_delnode(rn
, same
);
2529 route_unlock_node(rn
);
2533 void rib_delete(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
, int type
,
2534 unsigned short instance
, int flags
, struct prefix
*p
,
2535 struct prefix_ipv6
*src_p
, const struct nexthop
*nh
,
2536 uint32_t table_id
, uint32_t metric
, uint8_t distance
,
2539 struct route_table
*table
;
2540 struct route_node
*rn
;
2541 struct route_entry
*re
;
2542 struct route_entry
*fib
= NULL
;
2543 struct route_entry
*same
= NULL
;
2544 struct nexthop
*rtnh
;
2545 char buf2
[INET6_ADDRSTRLEN
];
2548 assert(!src_p
|| !src_p
->prefixlen
|| afi
== AFI_IP6
);
2551 table
= zebra_vrf_table_with_table_id(afi
, safi
, vrf_id
, table_id
);
2558 apply_mask_ipv6(src_p
);
2560 /* Lookup route node. */
2561 rn
= srcdest_rnode_lookup(table
, p
, src_p
);
2563 char dst_buf
[PREFIX_STRLEN
], src_buf
[PREFIX_STRLEN
];
2565 prefix2str(p
, dst_buf
, sizeof(dst_buf
));
2566 if (src_p
&& src_p
->prefixlen
)
2567 prefix2str(src_p
, src_buf
, sizeof(src_buf
));
2571 if (IS_ZEBRA_DEBUG_RIB
)
2572 zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id
,
2574 (src_buf
[0] != '\0') ? " from " : "",
2579 dest
= rib_dest_from_rnode(rn
);
2580 fib
= dest
->selected_fib
;
2582 /* Lookup same type route. */
2583 RNODE_FOREACH_RE (rn
, re
) {
2584 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
2587 if (re
->type
!= type
)
2589 if (re
->instance
!= instance
)
2591 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_RR_USE_DISTANCE
) &&
2592 distance
!= re
->distance
)
2595 if (re
->type
== ZEBRA_ROUTE_KERNEL
&& re
->metric
!= metric
)
2597 if (re
->type
== ZEBRA_ROUTE_CONNECT
&& (rtnh
= re
->ng
.nexthop
)
2598 && rtnh
->type
== NEXTHOP_TYPE_IFINDEX
&& nh
) {
2599 if (rtnh
->ifindex
!= nh
->ifindex
)
2604 /* Make sure that the route found has the same gateway. */
2610 for (ALL_NEXTHOPS(re
->ng
, rtnh
))
2611 if (nexthop_same_no_recurse(rtnh
, nh
)) {
2619 /* If same type of route can't be found and this message is from
2623 * In the past(HA!) we could get here because
2624 * we were receiving a route delete from the
2625 * kernel and we're not marking the proto
2626 * as coming from it's appropriate originator.
2627 * Now that we are properly noticing the fact
2628 * that the kernel has deleted our route we
2629 * are not going to get called in this path
2630 * I am going to leave this here because
2631 * this might still work this way on non-linux
2632 * platforms as well as some weird state I have
2633 * not properly thought of yet.
2634 * If we can show that this code path is
2635 * dead then we can remove it.
2637 if (fib
&& CHECK_FLAG(flags
, ZEBRA_FLAG_SELFROUTE
)) {
2638 if (IS_ZEBRA_DEBUG_RIB
) {
2641 "rn %p, re %p (type %d) was deleted from kernel, adding",
2642 rn
, fib
, fib
->type
);
2646 for (rtnh
= fib
->ng
.nexthop
; rtnh
;
2648 UNSET_FLAG(rtnh
->flags
,
2652 * This is a non FRR route
2653 * as such we should mark
2656 dest
->selected_fib
= NULL
;
2658 /* This means someone else, other than Zebra,
2660 * a Zebra router from the kernel. We will add
2662 rib_install_kernel(rn
, fib
, NULL
);
2665 if (IS_ZEBRA_DEBUG_RIB
) {
2669 "via %s ifindex %d type %d "
2670 "doesn't exist in rib",
2671 inet_ntop(afi2family(afi
),
2678 "type %d doesn't exist in rib",
2681 route_unlock_node(rn
);
2687 if (fromkernel
&& CHECK_FLAG(flags
, ZEBRA_FLAG_SELFROUTE
)
2689 rib_install_kernel(rn
, same
, NULL
);
2690 route_unlock_node(rn
);
2695 if (CHECK_FLAG(flags
, ZEBRA_FLAG_EVPN_ROUTE
)) {
2696 struct nexthop
*tmp_nh
;
2698 for (ALL_NEXTHOPS(re
->ng
, tmp_nh
)) {
2699 struct ipaddr vtep_ip
;
2701 memset(&vtep_ip
, 0, sizeof(struct ipaddr
));
2702 if (afi
== AFI_IP
) {
2703 vtep_ip
.ipa_type
= IPADDR_V4
;
2704 memcpy(&(vtep_ip
.ipaddr_v4
),
2705 &(tmp_nh
->gate
.ipv4
),
2706 sizeof(struct in_addr
));
2708 vtep_ip
.ipa_type
= IPADDR_V6
;
2709 memcpy(&(vtep_ip
.ipaddr_v6
),
2710 &(tmp_nh
->gate
.ipv6
),
2711 sizeof(struct in6_addr
));
2713 zebra_vxlan_evpn_vrf_route_del(re
->vrf_id
,
2717 rib_delnode(rn
, same
);
2720 route_unlock_node(rn
);
2725 int rib_add(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
, int type
,
2726 unsigned short instance
, int flags
, struct prefix
*p
,
2727 struct prefix_ipv6
*src_p
, const struct nexthop
*nh
,
2728 uint32_t table_id
, uint32_t metric
, uint32_t mtu
, uint8_t distance
,
2731 struct route_entry
*re
;
2732 struct nexthop
*nexthop
;
2734 /* Allocate new route_entry structure. */
2735 re
= XCALLOC(MTYPE_RE
, sizeof(struct route_entry
));
2737 re
->instance
= instance
;
2738 re
->distance
= distance
;
2740 re
->metric
= metric
;
2742 re
->table
= table_id
;
2743 re
->vrf_id
= vrf_id
;
2744 re
->nexthop_num
= 0;
2745 re
->uptime
= time(NULL
);
2749 nexthop
= nexthop_new();
2751 route_entry_nexthop_add(re
, nexthop
);
2753 return rib_add_multipath(afi
, safi
, p
, src_p
, re
);
2756 /* Schedule routes of a particular table (address-family) based on event. */
2757 void rib_update_table(struct route_table
*table
, rib_update_event_t event
)
2759 struct route_node
*rn
;
2760 struct route_entry
*re
, *next
;
2762 /* Walk all routes and queue for processing, if appropriate for
2763 * the trigger event.
2765 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
2767 * If we are looking at a route node and the node
2768 * has already been queued we don't
2769 * need to queue it up again
2771 if (rn
->info
&& CHECK_FLAG(rib_dest_from_rnode(rn
)->flags
,
2772 RIB_ROUTE_ANY_QUEUED
))
2775 case RIB_UPDATE_IF_CHANGE
:
2776 /* Examine all routes that won't get processed by the
2778 * triggered by nexthop evaluation (NHT). This would be
2780 * kernel and certain static routes. Note that NHT will
2782 * triggered upon an interface event as connected routes
2784 * get queued for processing.
2786 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
2789 if (re
->type
!= ZEBRA_ROUTE_SYSTEM
2790 && re
->type
!= ZEBRA_ROUTE_KERNEL
2791 && re
->type
!= ZEBRA_ROUTE_CONNECT
2792 && re
->type
!= ZEBRA_ROUTE_STATIC
)
2795 if (re
->type
!= ZEBRA_ROUTE_STATIC
) {
2800 for (nh
= re
->ng
.nexthop
; nh
; nh
= nh
->next
)
2801 if (!(nh
->type
== NEXTHOP_TYPE_IPV4
2802 || nh
->type
== NEXTHOP_TYPE_IPV6
))
2805 /* If we only have nexthops to a
2814 case RIB_UPDATE_RMAP_CHANGE
:
2815 case RIB_UPDATE_OTHER
:
2816 /* Right now, examine all routes. Can restrict to a
2818 * some cases (TODO).
2820 if (rnode_to_ribs(rn
))
2830 /* RIB update function. */
2831 void rib_update(vrf_id_t vrf_id
, rib_update_event_t event
)
2833 struct route_table
*table
;
2835 /* Process routes of interested address-families. */
2836 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
2838 if (IS_ZEBRA_DEBUG_EVENT
)
2839 zlog_debug("%s : AFI_IP event %d", __func__
, event
);
2840 rib_update_table(table
, event
);
2843 table
= zebra_vrf_table(AFI_IP6
, SAFI_UNICAST
, vrf_id
);
2845 if (IS_ZEBRA_DEBUG_EVENT
)
2846 zlog_debug("%s : AFI_IP6 event %d", __func__
, event
);
2847 rib_update_table(table
, event
);
2851 /* Delete self installed routes after zebra is relaunched. */
2852 void rib_sweep_table(struct route_table
*table
)
2854 struct route_node
*rn
;
2855 struct route_entry
*re
;
2856 struct route_entry
*next
;
2857 struct nexthop
*nexthop
;
2862 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
2863 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
2864 if (IS_ZEBRA_DEBUG_RIB
)
2865 route_entry_dump(&rn
->p
, NULL
, re
);
2867 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
2870 if (!CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELFROUTE
))
2874 * So we are starting up and have received
2875 * routes from the kernel that we have installed
2876 * from a previous run of zebra but not cleaned
2877 * up ( say a kill -9 )
2878 * But since we haven't actually installed
2879 * them yet( we received them from the kernel )
2880 * we don't think they are active.
2881 * So let's pretend they are active to actually
2883 * In all honesty I'm not sure if we should
2884 * mark them as active when we receive them
2885 * This is startup only so probably ok.
2887 * If we ever decide to move rib_sweep_table
2888 * to a different spot (ie startup )
2889 * this decision needs to be revisited
2891 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
2892 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
2894 rib_uninstall_kernel(rn
, re
);
2895 rib_delnode(rn
, re
);
2900 /* Sweep all RIB tables. */
2901 void rib_sweep_route(void)
2904 struct zebra_vrf
*zvrf
;
2906 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
2907 if ((zvrf
= vrf
->info
) == NULL
)
2910 rib_sweep_table(zvrf
->table
[AFI_IP
][SAFI_UNICAST
]);
2911 rib_sweep_table(zvrf
->table
[AFI_IP6
][SAFI_UNICAST
]);
2914 zebra_ns_sweep_route();
2917 /* Remove specific by protocol routes from 'table'. */
2918 unsigned long rib_score_proto_table(uint8_t proto
, unsigned short instance
,
2919 struct route_table
*table
)
2921 struct route_node
*rn
;
2922 struct route_entry
*re
;
2923 struct route_entry
*next
;
2924 unsigned long n
= 0;
2927 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
))
2928 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
2929 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
2931 if (re
->type
== proto
2932 && re
->instance
== instance
) {
2933 rib_delnode(rn
, re
);
2940 /* Remove specific by protocol routes. */
2941 unsigned long rib_score_proto(uint8_t proto
, unsigned short instance
)
2944 struct zebra_vrf
*zvrf
;
2945 unsigned long cnt
= 0;
2947 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
)
2948 if ((zvrf
= vrf
->info
) != NULL
)
2949 cnt
+= rib_score_proto_table(
2951 zvrf
->table
[AFI_IP
][SAFI_UNICAST
])
2952 + rib_score_proto_table(
2954 zvrf
->table
[AFI_IP6
][SAFI_UNICAST
]);
2956 cnt
+= zebra_ns_score_proto(proto
, instance
);
2961 /* Close RIB and clean up kernel routes. */
2962 void rib_close_table(struct route_table
*table
)
2964 struct route_node
*rn
;
2965 rib_table_info_t
*info
;
2971 info
= route_table_get_info(table
);
2973 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
2974 dest
= rib_dest_from_rnode(rn
);
2976 if (dest
&& dest
->selected_fib
) {
2977 if (info
->safi
== SAFI_UNICAST
)
2978 hook_call(rib_update
, rn
, NULL
);
2980 if (!RIB_SYSTEM_ROUTE(dest
->selected_fib
))
2981 rib_uninstall_kernel(rn
, dest
->selected_fib
);
2986 /* Routing information base initialize. */
2989 rib_queue_init(&zebrad
);
2995 * Get the first vrf id that is greater than the given vrf id if any.
2997 * Returns TRUE if a vrf id was found, FALSE otherwise.
2999 static inline int vrf_id_get_next(vrf_id_t vrf_id
, vrf_id_t
*next_id_p
)
3003 vrf
= vrf_lookup_by_id(vrf_id
);
3005 vrf
= RB_NEXT(vrf_id_head
, vrf
);
3007 *next_id_p
= vrf
->vrf_id
;
3016 * rib_tables_iter_next
3018 * Returns the next table in the iteration.
3020 struct route_table
*rib_tables_iter_next(rib_tables_iter_t
*iter
)
3022 struct route_table
*table
;
3025 * Array that helps us go over all AFI/SAFI combinations via one
3032 {AFI_IP
, SAFI_UNICAST
}, {AFI_IP
, SAFI_MULTICAST
},
3033 {AFI_IP
, SAFI_LABELED_UNICAST
}, {AFI_IP6
, SAFI_UNICAST
},
3034 {AFI_IP6
, SAFI_MULTICAST
}, {AFI_IP6
, SAFI_LABELED_UNICAST
},
3039 switch (iter
->state
) {
3041 case RIB_TABLES_ITER_S_INIT
:
3042 iter
->vrf_id
= VRF_DEFAULT
;
3043 iter
->afi_safi_ix
= -1;
3047 case RIB_TABLES_ITER_S_ITERATING
:
3048 iter
->afi_safi_ix
++;
3051 while (iter
->afi_safi_ix
3052 < (int)ZEBRA_NUM_OF(afi_safis
)) {
3053 table
= zebra_vrf_table(
3054 afi_safis
[iter
->afi_safi_ix
].afi
,
3055 afi_safis
[iter
->afi_safi_ix
].safi
,
3060 iter
->afi_safi_ix
++;
3064 * Found another table in this vrf.
3070 * Done with all tables in the current vrf, go to the
3074 if (!vrf_id_get_next(iter
->vrf_id
, &iter
->vrf_id
))
3077 iter
->afi_safi_ix
= 0;
3082 case RIB_TABLES_ITER_S_DONE
:
3087 iter
->state
= RIB_TABLES_ITER_S_ITERATING
;
3089 iter
->state
= RIB_TABLES_ITER_S_DONE
;