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/zebra_router.h"
42 #include "zebra/connected.h"
43 #include "zebra/debug.h"
44 #include "zebra/interface.h"
45 #include "zebra/redistribute.h"
46 #include "zebra/rib.h"
48 #include "zebra/zapi_msg.h"
49 #include "zebra/zebra_errors.h"
50 #include "zebra/zebra_memory.h"
51 #include "zebra/zebra_ns.h"
52 #include "zebra/zebra_rnh.h"
53 #include "zebra/zebra_routemap.h"
54 #include "zebra/zebra_vrf.h"
55 #include "zebra/zebra_vxlan.h"
56 #include "zebra/zapi_msg.h"
57 #include "zebra/zebra_dplane.h"
60 * Event, list, and mutex for delivery of dataplane results
62 static pthread_mutex_t dplane_mutex
;
63 static struct thread
*t_dplane
;
64 static struct dplane_ctx_q rib_dplane_q
;
66 DEFINE_HOOK(rib_update
, (struct route_node
* rn
, const char *reason
),
69 /* Should we allow non Quagga processes to delete our routes */
70 extern int allow_delete
;
72 /* Each route type's string and default distance value. */
76 } route_info
[ZEBRA_ROUTE_MAX
] = {
77 [ZEBRA_ROUTE_SYSTEM
] = {ZEBRA_ROUTE_SYSTEM
, 0},
78 [ZEBRA_ROUTE_KERNEL
] = {ZEBRA_ROUTE_KERNEL
, 0},
79 [ZEBRA_ROUTE_CONNECT
] = {ZEBRA_ROUTE_CONNECT
, 0},
80 [ZEBRA_ROUTE_STATIC
] = {ZEBRA_ROUTE_STATIC
, 1},
81 [ZEBRA_ROUTE_RIP
] = {ZEBRA_ROUTE_RIP
, 120},
82 [ZEBRA_ROUTE_RIPNG
] = {ZEBRA_ROUTE_RIPNG
, 120},
83 [ZEBRA_ROUTE_OSPF
] = {ZEBRA_ROUTE_OSPF
, 110},
84 [ZEBRA_ROUTE_OSPF6
] = {ZEBRA_ROUTE_OSPF6
, 110},
85 [ZEBRA_ROUTE_ISIS
] = {ZEBRA_ROUTE_ISIS
, 115},
86 [ZEBRA_ROUTE_BGP
] = {ZEBRA_ROUTE_BGP
, 20 /* IBGP is 200. */},
87 [ZEBRA_ROUTE_PIM
] = {ZEBRA_ROUTE_PIM
, 255},
88 [ZEBRA_ROUTE_EIGRP
] = {ZEBRA_ROUTE_EIGRP
, 90},
89 [ZEBRA_ROUTE_NHRP
] = {ZEBRA_ROUTE_NHRP
, 10},
90 [ZEBRA_ROUTE_HSLS
] = {ZEBRA_ROUTE_HSLS
, 255},
91 [ZEBRA_ROUTE_OLSR
] = {ZEBRA_ROUTE_OLSR
, 255},
92 [ZEBRA_ROUTE_TABLE
] = {ZEBRA_ROUTE_TABLE
, 150},
93 [ZEBRA_ROUTE_LDP
] = {ZEBRA_ROUTE_LDP
, 150},
94 [ZEBRA_ROUTE_VNC
] = {ZEBRA_ROUTE_VNC
, 20},
95 [ZEBRA_ROUTE_VNC_DIRECT
] = {ZEBRA_ROUTE_VNC_DIRECT
, 20},
96 [ZEBRA_ROUTE_VNC_DIRECT_RH
] = {ZEBRA_ROUTE_VNC_DIRECT_RH
, 20},
97 [ZEBRA_ROUTE_BGP_DIRECT
] = {ZEBRA_ROUTE_BGP_DIRECT
, 20},
98 [ZEBRA_ROUTE_BGP_DIRECT_EXT
] = {ZEBRA_ROUTE_BGP_DIRECT_EXT
, 20},
99 [ZEBRA_ROUTE_BABEL
] = {ZEBRA_ROUTE_BABEL
, 100},
100 [ZEBRA_ROUTE_SHARP
] = {ZEBRA_ROUTE_SHARP
, 150},
102 /* no entry/default: 150 */
105 /* RPF lookup behaviour */
106 static enum multicast_mode ipv4_multicast_mode
= MCAST_NO_CONFIG
;
109 static void __attribute__((format(printf
, 5, 6)))
110 _rnode_zlog(const char *_func
, vrf_id_t vrf_id
, struct route_node
*rn
,
111 int priority
, const char *msgfmt
, ...)
113 char buf
[SRCDEST2STR_BUFFER
+ sizeof(" (MRIB)")];
117 va_start(ap
, msgfmt
);
118 vsnprintf(msgbuf
, sizeof(msgbuf
), msgfmt
, ap
);
122 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
123 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
125 if (info
->safi
== SAFI_MULTICAST
)
126 strcat(buf
, " (MRIB)");
128 snprintf(buf
, sizeof(buf
), "{(route_node *) NULL}");
131 zlog(priority
, "%s: %d:%s: %s", _func
, vrf_id
, buf
, msgbuf
);
134 #define rnode_debug(node, vrf_id, ...) \
135 _rnode_zlog(__func__, vrf_id, node, LOG_DEBUG, __VA_ARGS__)
136 #define rnode_info(node, ...) \
137 _rnode_zlog(__func__, vrf_id, node, LOG_INFO, __VA_ARGS__)
139 uint8_t route_distance(int type
)
143 if ((unsigned)type
>= array_size(route_info
))
146 distance
= route_info
[type
].distance
;
151 int is_zebra_valid_kernel_table(uint32_t table_id
)
154 if ((table_id
== RT_TABLE_UNSPEC
) || (table_id
== RT_TABLE_LOCAL
)
155 || (table_id
== RT_TABLE_COMPAT
))
162 int is_zebra_main_routing_table(uint32_t table_id
)
164 if ((table_id
== RT_TABLE_MAIN
)
165 || (table_id
== zebrad
.rtm_table_default
))
170 int zebra_check_addr(const struct prefix
*p
)
172 if (p
->family
== AF_INET
) {
175 addr
= p
->u
.prefix4
.s_addr
;
178 if (IPV4_NET127(addr
) || IN_CLASSD(addr
)
179 || IPV4_LINKLOCAL(addr
))
182 if (p
->family
== AF_INET6
) {
183 if (IN6_IS_ADDR_LOOPBACK(&p
->u
.prefix6
))
185 if (IN6_IS_ADDR_LINKLOCAL(&p
->u
.prefix6
))
191 /* Add nexthop to the end of a rib node's nexthop list */
192 void route_entry_nexthop_add(struct route_entry
*re
, struct nexthop
*nexthop
)
194 nexthop_add(&re
->ng
.nexthop
, nexthop
);
200 * copy_nexthop - copy a nexthop to the rib structure.
202 void route_entry_copy_nexthops(struct route_entry
*re
, struct nexthop
*nh
)
204 assert(!re
->ng
.nexthop
);
205 copy_nexthops(&re
->ng
.nexthop
, nh
, NULL
);
206 for (struct nexthop
*nexthop
= nh
; nexthop
; nexthop
= nexthop
->next
)
210 /* Delete specified nexthop from the list. */
211 void route_entry_nexthop_delete(struct route_entry
*re
, struct nexthop
*nexthop
)
214 nexthop
->next
->prev
= nexthop
->prev
;
216 nexthop
->prev
->next
= nexthop
->next
;
218 re
->ng
.nexthop
= nexthop
->next
;
223 struct nexthop
*route_entry_nexthop_ifindex_add(struct route_entry
*re
,
227 struct nexthop
*nexthop
;
229 nexthop
= nexthop_new();
230 nexthop
->type
= NEXTHOP_TYPE_IFINDEX
;
231 nexthop
->ifindex
= ifindex
;
232 nexthop
->vrf_id
= nh_vrf_id
;
234 route_entry_nexthop_add(re
, nexthop
);
239 struct nexthop
*route_entry_nexthop_ipv4_add(struct route_entry
*re
,
240 struct in_addr
*ipv4
,
244 struct nexthop
*nexthop
;
246 nexthop
= nexthop_new();
247 nexthop
->type
= NEXTHOP_TYPE_IPV4
;
248 nexthop
->vrf_id
= nh_vrf_id
;
249 nexthop
->gate
.ipv4
= *ipv4
;
251 nexthop
->src
.ipv4
= *src
;
253 route_entry_nexthop_add(re
, nexthop
);
258 struct nexthop
*route_entry_nexthop_ipv4_ifindex_add(struct route_entry
*re
,
259 struct in_addr
*ipv4
,
264 struct nexthop
*nexthop
;
265 struct interface
*ifp
;
267 nexthop
= nexthop_new();
268 nexthop
->vrf_id
= nh_vrf_id
;
269 nexthop
->type
= NEXTHOP_TYPE_IPV4_IFINDEX
;
270 nexthop
->gate
.ipv4
= *ipv4
;
272 nexthop
->src
.ipv4
= *src
;
273 nexthop
->ifindex
= ifindex
;
274 ifp
= if_lookup_by_index(nexthop
->ifindex
, nh_vrf_id
);
275 /*Pending: need to think if null ifp here is ok during bootup?
276 There was a crash because ifp here was coming to be NULL */
278 if (connected_is_unnumbered(ifp
)
279 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_EVPN_ROUTE
)
280 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ONLINK
)) {
281 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
);
284 route_entry_nexthop_add(re
, nexthop
);
289 struct nexthop
*route_entry_nexthop_ipv6_add(struct route_entry
*re
,
290 struct in6_addr
*ipv6
,
293 struct nexthop
*nexthop
;
295 nexthop
= nexthop_new();
296 nexthop
->vrf_id
= nh_vrf_id
;
297 nexthop
->type
= NEXTHOP_TYPE_IPV6
;
298 nexthop
->gate
.ipv6
= *ipv6
;
300 route_entry_nexthop_add(re
, nexthop
);
305 struct nexthop
*route_entry_nexthop_ipv6_ifindex_add(struct route_entry
*re
,
306 struct in6_addr
*ipv6
,
310 struct nexthop
*nexthop
;
312 nexthop
= nexthop_new();
313 nexthop
->vrf_id
= nh_vrf_id
;
314 nexthop
->type
= NEXTHOP_TYPE_IPV6_IFINDEX
;
315 nexthop
->gate
.ipv6
= *ipv6
;
316 nexthop
->ifindex
= ifindex
;
317 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_EVPN_ROUTE
)
318 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ONLINK
)) {
319 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
);
322 route_entry_nexthop_add(re
, nexthop
);
327 struct nexthop
*route_entry_nexthop_blackhole_add(struct route_entry
*re
,
328 enum blackhole_type bh_type
)
330 struct nexthop
*nexthop
;
332 nexthop
= nexthop_new();
333 nexthop
->vrf_id
= VRF_DEFAULT
;
334 nexthop
->type
= NEXTHOP_TYPE_BLACKHOLE
;
335 nexthop
->bh_type
= bh_type
;
337 route_entry_nexthop_add(re
, nexthop
);
342 static void nexthop_set_resolved(afi_t afi
, const struct nexthop
*newhop
,
343 struct nexthop
*nexthop
)
345 struct nexthop
*resolved_hop
;
347 resolved_hop
= nexthop_new();
348 SET_FLAG(resolved_hop
->flags
, NEXTHOP_FLAG_ACTIVE
);
350 resolved_hop
->vrf_id
= nexthop
->vrf_id
;
351 switch (newhop
->type
) {
352 case NEXTHOP_TYPE_IPV4
:
353 case NEXTHOP_TYPE_IPV4_IFINDEX
:
354 /* If the resolving route specifies a gateway, use it */
355 resolved_hop
->type
= newhop
->type
;
356 resolved_hop
->gate
.ipv4
= newhop
->gate
.ipv4
;
358 if (newhop
->ifindex
) {
359 resolved_hop
->type
= NEXTHOP_TYPE_IPV4_IFINDEX
;
360 resolved_hop
->ifindex
= newhop
->ifindex
;
363 case NEXTHOP_TYPE_IPV6
:
364 case NEXTHOP_TYPE_IPV6_IFINDEX
:
365 resolved_hop
->type
= newhop
->type
;
366 resolved_hop
->gate
.ipv6
= newhop
->gate
.ipv6
;
368 if (newhop
->ifindex
) {
369 resolved_hop
->type
= NEXTHOP_TYPE_IPV6_IFINDEX
;
370 resolved_hop
->ifindex
= newhop
->ifindex
;
373 case NEXTHOP_TYPE_IFINDEX
:
374 /* If the resolving route is an interface route,
375 * it means the gateway we are looking up is connected
376 * to that interface. (The actual network is _not_ onlink).
377 * Therefore, the resolved route should have the original
378 * gateway as nexthop as it is directly connected.
380 * On Linux, we have to set the onlink netlink flag because
381 * otherwise, the kernel won't accept the route.
383 resolved_hop
->flags
|= NEXTHOP_FLAG_ONLINK
;
385 resolved_hop
->type
= NEXTHOP_TYPE_IPV4_IFINDEX
;
386 resolved_hop
->gate
.ipv4
= nexthop
->gate
.ipv4
;
387 } else if (afi
== AFI_IP6
) {
388 resolved_hop
->type
= NEXTHOP_TYPE_IPV6_IFINDEX
;
389 resolved_hop
->gate
.ipv6
= nexthop
->gate
.ipv6
;
391 resolved_hop
->ifindex
= newhop
->ifindex
;
393 case NEXTHOP_TYPE_BLACKHOLE
:
394 resolved_hop
->type
= NEXTHOP_TYPE_BLACKHOLE
;
395 resolved_hop
->bh_type
= nexthop
->bh_type
;
399 if (newhop
->flags
& NEXTHOP_FLAG_ONLINK
)
400 resolved_hop
->flags
|= NEXTHOP_FLAG_ONLINK
;
402 /* Copy labels of the resolved route */
403 if (newhop
->nh_label
)
404 nexthop_add_labels(resolved_hop
, newhop
->nh_label_type
,
405 newhop
->nh_label
->num_labels
,
406 &newhop
->nh_label
->label
[0]);
408 resolved_hop
->rparent
= nexthop
;
409 nexthop_add(&nexthop
->resolved
, resolved_hop
);
412 /* If force flag is not set, do not modify falgs at all for uninstall
413 the route from FIB. */
414 static int nexthop_active(afi_t afi
, struct route_entry
*re
,
415 struct nexthop
*nexthop
, bool set
,
416 struct route_node
*top
)
419 struct route_table
*table
;
420 struct route_node
*rn
;
421 struct route_entry
*match
= NULL
;
423 struct nexthop
*newhop
;
424 struct interface
*ifp
;
427 if ((nexthop
->type
== NEXTHOP_TYPE_IPV4
)
428 || nexthop
->type
== NEXTHOP_TYPE_IPV6
)
429 nexthop
->ifindex
= 0;
432 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
);
433 nexthops_free(nexthop
->resolved
);
434 nexthop
->resolved
= NULL
;
438 /* Next hops (remote VTEPs) for EVPN routes are fully resolved. */
439 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_EVPN_RVTEP
))
442 /* Skip nexthops that have been filtered out due to route-map */
443 /* The nexthops are specific to this route and so the same */
444 /* nexthop for a different route may not have this flag set */
445 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FILTERED
)) {
446 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
447 zlog_debug("\t%s: Nexthop Filtered",
448 __PRETTY_FUNCTION__
);
453 * Check to see if we should trust the passed in information
454 * for UNNUMBERED interfaces as that we won't find the GW
455 * address in the routing table.
457 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
)) {
458 ifp
= if_lookup_by_index(nexthop
->ifindex
, nexthop
->vrf_id
);
459 if ((ifp
&& connected_is_unnumbered(ifp
))
460 || CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ONLINK
)) {
461 if (if_is_operative(ifp
))
464 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
466 "\t%s: Onlink and interface %s is not operative",
467 __PRETTY_FUNCTION__
, ifp
->name
);
471 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
473 "\t%s: Interface %s is not unnumbered",
475 ifp
? ifp
->name
: "Unknown");
480 /* Make lookup prefix. */
481 memset(&p
, 0, sizeof(struct prefix
));
485 p
.prefixlen
= IPV4_MAX_PREFIXLEN
;
486 p
.u
.prefix4
= nexthop
->gate
.ipv4
;
490 p
.prefixlen
= IPV6_MAX_PREFIXLEN
;
491 p
.u
.prefix6
= nexthop
->gate
.ipv6
;
494 assert(afi
!= AFI_IP
&& afi
!= AFI_IP6
);
498 table
= zebra_vrf_table(afi
, SAFI_UNICAST
, nexthop
->vrf_id
);
500 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
501 zlog_debug("\t%s: Table not found",
502 __PRETTY_FUNCTION__
);
506 rn
= route_node_match(table
, (struct prefix
*)&p
);
508 route_unlock_node(rn
);
510 /* Lookup should halt if we've matched against ourselves ('top',
511 * if specified) - i.e., we cannot have a nexthop NH1 is
512 * resolved by a route NH1. The exception is if the route is a
515 if (top
&& rn
== top
)
516 if (((afi
== AFI_IP
) && (rn
->p
.prefixlen
!= 32))
517 || ((afi
== AFI_IP6
) && (rn
->p
.prefixlen
!= 128))) {
518 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
520 "\t%s: Matched against ourself and prefix length is not max bit length",
521 __PRETTY_FUNCTION__
);
525 /* Pick up selected route. */
526 /* However, do not resolve over default route unless explicitly
528 if (is_default_prefix(&rn
->p
)
529 && !rnh_resolve_via_default(p
.family
)) {
530 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
532 "\t:%s: Resolved against default route",
533 __PRETTY_FUNCTION__
);
537 dest
= rib_dest_from_rnode(rn
);
538 if (dest
&& dest
->selected_fib
539 && !CHECK_FLAG(dest
->selected_fib
->status
,
541 && dest
->selected_fib
->type
!= ZEBRA_ROUTE_TABLE
)
542 match
= dest
->selected_fib
;
544 /* If there is no selected route or matched route is EGP, go up
549 } while (rn
&& rn
->info
== NULL
);
556 if (match
->type
== ZEBRA_ROUTE_CONNECT
) {
557 /* Directly point connected route. */
558 newhop
= match
->ng
.nexthop
;
560 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
561 || nexthop
->type
== NEXTHOP_TYPE_IPV6
)
562 nexthop
->ifindex
= newhop
->ifindex
;
565 } else if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_ALLOW_RECURSION
)) {
567 for (ALL_NEXTHOPS(match
->ng
, newhop
)) {
568 if (!CHECK_FLAG(newhop
->flags
,
571 if (CHECK_FLAG(newhop
->flags
,
572 NEXTHOP_FLAG_RECURSIVE
))
576 SET_FLAG(nexthop
->flags
,
577 NEXTHOP_FLAG_RECURSIVE
);
579 ROUTE_ENTRY_NEXTHOPS_CHANGED
);
580 nexthop_set_resolved(afi
, newhop
,
586 re
->nexthop_mtu
= match
->mtu
;
587 if (!resolved
&& IS_ZEBRA_DEBUG_RIB_DETAILED
)
588 zlog_debug("\t%s: Recursion failed to find",
589 __PRETTY_FUNCTION__
);
591 } else if (re
->type
== ZEBRA_ROUTE_STATIC
) {
593 for (ALL_NEXTHOPS(match
->ng
, newhop
)) {
594 if (!CHECK_FLAG(newhop
->flags
,
599 SET_FLAG(nexthop
->flags
,
600 NEXTHOP_FLAG_RECURSIVE
);
601 nexthop_set_resolved(afi
, newhop
,
607 re
->nexthop_mtu
= match
->mtu
;
609 if (!resolved
&& IS_ZEBRA_DEBUG_RIB_DETAILED
)
611 "\t%s: Static route unable to resolve",
612 __PRETTY_FUNCTION__
);
615 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
616 zlog_debug("\t%s: Route Type %s has not turned on recursion",
618 zebra_route_string(re
->type
));
619 if (re
->type
== ZEBRA_ROUTE_BGP
&&
620 !CHECK_FLAG(re
->flags
, ZEBRA_FLAG_IBGP
))
621 zlog_debug("\tEBGP: see \"disable-ebgp-connected-route-check\" or \"disable-connected-check\"");
626 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
627 zlog_debug("\t%s: Nexthop did not lookup in table",
628 __PRETTY_FUNCTION__
);
632 struct route_entry
*rib_match(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
,
633 union g_addr
*addr
, struct route_node
**rn_out
)
636 struct route_table
*table
;
637 struct route_node
*rn
;
638 struct route_entry
*match
= NULL
;
639 struct nexthop
*newhop
;
642 table
= zebra_vrf_table(afi
, safi
, vrf_id
);
646 memset(&p
, 0, sizeof(struct prefix
));
649 p
.u
.prefix4
= addr
->ipv4
;
650 p
.prefixlen
= IPV4_MAX_PREFIXLEN
;
652 p
.u
.prefix6
= addr
->ipv6
;
653 p
.prefixlen
= IPV6_MAX_PREFIXLEN
;
656 rn
= route_node_match(table
, (struct prefix
*)&p
);
661 route_unlock_node(rn
);
663 dest
= rib_dest_from_rnode(rn
);
664 if (dest
&& dest
->selected_fib
665 && !CHECK_FLAG(dest
->selected_fib
->status
,
666 ROUTE_ENTRY_REMOVED
))
667 match
= dest
->selected_fib
;
669 /* If there is no selected route or matched route is EGP, go up
674 } while (rn
&& rn
->info
== NULL
);
678 if (match
->type
!= ZEBRA_ROUTE_CONNECT
) {
680 for (ALL_NEXTHOPS(match
->ng
, newhop
))
681 if (CHECK_FLAG(newhop
->flags
,
698 struct route_entry
*rib_match_ipv4_multicast(vrf_id_t vrf_id
,
700 struct route_node
**rn_out
)
702 struct route_entry
*re
= NULL
, *mre
= NULL
, *ure
= NULL
;
703 struct route_node
*m_rn
= NULL
, *u_rn
= NULL
;
704 union g_addr gaddr
= {.ipv4
= addr
};
706 switch (ipv4_multicast_mode
) {
707 case MCAST_MRIB_ONLY
:
708 return rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
,
710 case MCAST_URIB_ONLY
:
711 return rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, rn_out
);
712 case MCAST_NO_CONFIG
:
713 case MCAST_MIX_MRIB_FIRST
:
714 re
= mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
,
717 re
= ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
,
720 case MCAST_MIX_DISTANCE
:
721 mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
, &m_rn
);
722 ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, &u_rn
);
724 re
= ure
->distance
< mre
->distance
? ure
: mre
;
730 case MCAST_MIX_PFXLEN
:
731 mre
= rib_match(AFI_IP
, SAFI_MULTICAST
, vrf_id
, &gaddr
, &m_rn
);
732 ure
= rib_match(AFI_IP
, SAFI_UNICAST
, vrf_id
, &gaddr
, &u_rn
);
734 re
= u_rn
->p
.prefixlen
> m_rn
->p
.prefixlen
? ure
: mre
;
743 *rn_out
= (re
== mre
) ? m_rn
: u_rn
;
745 if (IS_ZEBRA_DEBUG_RIB
) {
747 inet_ntop(AF_INET
, &addr
, buf
, BUFSIZ
);
749 zlog_debug("%s: %s: vrf: %u found %s, using %s",
750 __func__
, buf
, vrf_id
,
751 mre
? (ure
? "MRIB+URIB" : "MRIB")
752 : ure
? "URIB" : "nothing",
753 re
== ure
? "URIB" : re
== mre
? "MRIB" : "none");
758 void multicast_mode_ipv4_set(enum multicast_mode mode
)
760 if (IS_ZEBRA_DEBUG_RIB
)
761 zlog_debug("%s: multicast lookup mode set (%d)", __func__
,
763 ipv4_multicast_mode
= mode
;
766 enum multicast_mode
multicast_mode_ipv4_get(void)
768 return ipv4_multicast_mode
;
771 struct route_entry
*rib_lookup_ipv4(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
773 struct route_table
*table
;
774 struct route_node
*rn
;
775 struct route_entry
*match
= NULL
;
776 struct nexthop
*nexthop
;
780 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
784 rn
= route_node_lookup(table
, (struct prefix
*)p
);
786 /* No route for this prefix. */
791 route_unlock_node(rn
);
792 dest
= rib_dest_from_rnode(rn
);
794 if (dest
&& dest
->selected_fib
795 && !CHECK_FLAG(dest
->selected_fib
->status
, ROUTE_ENTRY_REMOVED
))
796 match
= dest
->selected_fib
;
801 if (match
->type
== ZEBRA_ROUTE_CONNECT
)
804 for (ALL_NEXTHOPS(match
->ng
, nexthop
))
805 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
811 #define RIB_SYSTEM_ROUTE(R) \
812 ((R)->type == ZEBRA_ROUTE_KERNEL || (R)->type == ZEBRA_ROUTE_CONNECT)
814 #define RIB_KERNEL_ROUTE(R) \
815 ((R)->type == ZEBRA_ROUTE_KERNEL)
817 /* This function verifies reachability of one given nexthop, which can be
818 * numbered or unnumbered, IPv4 or IPv6. The result is unconditionally stored
819 * in nexthop->flags field. If the 4th parameter, 'set', is non-zero,
820 * nexthop->ifindex will be updated appropriately as well.
821 * An existing route map can turn (otherwise active) nexthop into inactive, but
824 * The return value is the final value of 'ACTIVE' flag.
827 static unsigned nexthop_active_check(struct route_node
*rn
,
828 struct route_entry
*re
,
829 struct nexthop
*nexthop
, bool set
)
831 struct interface
*ifp
;
832 route_map_result_t ret
= RMAP_MATCH
;
834 char buf
[SRCDEST2STR_BUFFER
];
835 const struct prefix
*p
, *src_p
;
836 struct zebra_vrf
*zvrf
;
838 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
840 if (rn
->p
.family
== AF_INET
)
842 else if (rn
->p
.family
== AF_INET6
)
846 switch (nexthop
->type
) {
847 case NEXTHOP_TYPE_IFINDEX
:
848 ifp
= if_lookup_by_index(nexthop
->ifindex
, nexthop
->vrf_id
);
849 if (ifp
&& if_is_operative(ifp
))
850 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
852 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
854 case NEXTHOP_TYPE_IPV4
:
855 case NEXTHOP_TYPE_IPV4_IFINDEX
:
857 if (nexthop_active(AFI_IP
, re
, nexthop
, set
, rn
))
858 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
860 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
862 case NEXTHOP_TYPE_IPV6
:
864 if (nexthop_active(AFI_IP6
, re
, nexthop
, set
, rn
))
865 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
867 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
869 case NEXTHOP_TYPE_IPV6_IFINDEX
:
870 /* RFC 5549, v4 prefix with v6 NH */
871 if (rn
->p
.family
!= AF_INET
)
873 if (IN6_IS_ADDR_LINKLOCAL(&nexthop
->gate
.ipv6
)) {
874 ifp
= if_lookup_by_index(nexthop
->ifindex
,
876 if (ifp
&& if_is_operative(ifp
))
877 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
879 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
881 if (nexthop_active(AFI_IP6
, re
, nexthop
, set
, rn
))
882 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
884 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
887 case NEXTHOP_TYPE_BLACKHOLE
:
888 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
893 if (!CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
)) {
894 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
895 zlog_debug("\t%s: Unable to find a active nexthop",
896 __PRETTY_FUNCTION__
);
900 /* XXX: What exactly do those checks do? Do we support
901 * e.g. IPv4 routes with IPv6 nexthops or vice versa?
903 if (RIB_SYSTEM_ROUTE(re
) || (family
== AFI_IP
&& p
->family
!= AF_INET
)
904 || (family
== AFI_IP6
&& p
->family
!= AF_INET6
))
905 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
907 /* The original code didn't determine the family correctly
908 * e.g. for NEXTHOP_TYPE_IFINDEX. Retrieve the correct afi
909 * from the rib_table_info in those cases.
910 * Possibly it may be better to use only the rib_table_info
914 rib_table_info_t
*info
;
916 info
= srcdest_rnode_table_info(rn
);
920 memset(&nexthop
->rmap_src
.ipv6
, 0, sizeof(union g_addr
));
922 zvrf
= zebra_vrf_lookup_by_id(nexthop
->vrf_id
);
924 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
925 zlog_debug("\t%s: zvrf is NULL", __PRETTY_FUNCTION__
);
926 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
929 /* It'll get set if required inside */
930 ret
= zebra_route_map_check(family
, re
->type
, re
->instance
, p
,
931 nexthop
, zvrf
, re
->tag
);
932 if (ret
== RMAP_DENYMATCH
) {
933 if (IS_ZEBRA_DEBUG_RIB
) {
934 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
936 "%u:%s: Filtering out with NH out %s due to route map",
938 ifindex2ifname(nexthop
->ifindex
,
941 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
943 return CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
946 /* Iterate over all nexthops of the given RIB entry and refresh their
947 * ACTIVE flag. re->nexthop_active_num is updated accordingly. If any
948 * nexthop is found to toggle the ACTIVE flag, the whole re structure
949 * is flagged with ROUTE_ENTRY_CHANGED. The 4th 'set' argument is
950 * transparently passed to nexthop_active_check().
952 * Return value is the new number of active nexthops.
955 static int nexthop_active_update(struct route_node
*rn
, struct route_entry
*re
,
958 struct nexthop
*nexthop
;
959 union g_addr prev_src
;
960 unsigned int prev_active
, new_active
, old_num_nh
;
961 ifindex_t prev_index
;
963 old_num_nh
= re
->nexthop_active_num
;
965 re
->nexthop_active_num
= 0;
966 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
968 for (nexthop
= re
->ng
.nexthop
; nexthop
; nexthop
= nexthop
->next
) {
969 /* No protocol daemon provides src and so we're skipping
971 prev_src
= nexthop
->rmap_src
;
972 prev_active
= CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
973 prev_index
= nexthop
->ifindex
;
975 * We need to respect the multipath_num here
976 * as that what we should be able to install from
977 * a multipath perpsective should not be a data plane
980 new_active
= nexthop_active_check(rn
, re
, nexthop
, set
);
981 if (new_active
&& re
->nexthop_active_num
>= multipath_num
) {
982 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
);
986 re
->nexthop_active_num
++;
987 /* Don't allow src setting on IPv6 addr for now */
988 if (prev_active
!= new_active
|| prev_index
!= nexthop
->ifindex
989 || ((nexthop
->type
>= NEXTHOP_TYPE_IFINDEX
990 && nexthop
->type
< NEXTHOP_TYPE_IPV6
)
991 && prev_src
.ipv4
.s_addr
992 != nexthop
->rmap_src
.ipv4
.s_addr
)
993 || ((nexthop
->type
>= NEXTHOP_TYPE_IPV6
994 && nexthop
->type
< NEXTHOP_TYPE_BLACKHOLE
)
995 && !(IPV6_ADDR_SAME(&prev_src
.ipv6
,
996 &nexthop
->rmap_src
.ipv6
)))) {
997 SET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
998 SET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1002 if (old_num_nh
!= re
->nexthop_active_num
)
1003 SET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1005 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
)) {
1006 SET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1009 return re
->nexthop_active_num
;
1013 * Is this RIB labeled-unicast? It must be of type BGP and all paths
1014 * (nexthops) must have a label.
1016 int zebra_rib_labeled_unicast(struct route_entry
*re
)
1018 struct nexthop
*nexthop
= NULL
;
1020 if (re
->type
!= ZEBRA_ROUTE_BGP
)
1023 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1024 if (!nexthop
->nh_label
|| !nexthop
->nh_label
->num_labels
)
1030 /* Update flag indicates whether this is a "replace" or not. Currently, this
1031 * is only used for IPv4.
1033 void rib_install_kernel(struct route_node
*rn
, struct route_entry
*re
,
1034 struct route_entry
*old
)
1036 struct nexthop
*nexthop
;
1037 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1038 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1039 const struct prefix
*p
, *src_p
;
1040 enum zebra_dplane_result ret
;
1042 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1044 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1046 if (info
->safi
!= SAFI_UNICAST
) {
1047 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1048 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1051 struct nexthop
*prev
;
1053 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1054 UNSET_FLAG (nexthop
->flags
, NEXTHOP_FLAG_DUPLICATE
);
1055 for (ALL_NEXTHOPS(re
->ng
, prev
)) {
1056 if (prev
== nexthop
)
1058 if (nexthop_same_firsthop(nexthop
, prev
)) {
1059 SET_FLAG(nexthop
->flags
,
1060 NEXTHOP_FLAG_DUPLICATE
);
1068 * If this is a replace to a new RE let the originator of the RE
1069 * know that they've lost
1071 if (old
&& (old
!= re
) && (old
->type
!= re
->type
))
1072 zsend_route_notify_owner(old
, p
, ZAPI_ROUTE_BETTER_ADMIN_WON
);
1074 /* Update fib selection */
1075 dest
->selected_fib
= re
;
1078 * Make sure we update the FPM any time we send new information to
1081 hook_call(rib_update
, rn
, "installing in kernel");
1083 /* Send add or update */
1084 if (old
&& (old
!= re
))
1085 ret
= dplane_route_update(rn
, re
, old
);
1087 ret
= dplane_route_add(rn
, re
);
1090 case ZEBRA_DPLANE_REQUEST_QUEUED
:
1092 zvrf
->installs_queued
++;
1094 case ZEBRA_DPLANE_REQUEST_FAILURE
:
1096 char str
[SRCDEST2STR_BUFFER
];
1098 srcdest_rnode2str(rn
, str
, sizeof(str
));
1099 flog_err(EC_ZEBRA_DP_INSTALL_FAIL
,
1100 "%u:%s: Failed to enqueue dataplane install",
1104 case ZEBRA_DPLANE_REQUEST_SUCCESS
:
1113 /* Uninstall the route from kernel. */
1114 void rib_uninstall_kernel(struct route_node
*rn
, struct route_entry
*re
)
1116 struct nexthop
*nexthop
;
1117 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1118 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1120 if (info
->safi
!= SAFI_UNICAST
) {
1121 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1122 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1127 * Make sure we update the FPM any time we send new information to
1130 hook_call(rib_update
, rn
, "uninstalling from kernel");
1132 switch (dplane_route_delete(rn
, re
)) {
1133 case ZEBRA_DPLANE_REQUEST_QUEUED
:
1135 zvrf
->removals_queued
++;
1137 case ZEBRA_DPLANE_REQUEST_FAILURE
:
1139 char str
[SRCDEST2STR_BUFFER
];
1141 srcdest_rnode2str(rn
, str
, sizeof(str
));
1142 flog_err(EC_ZEBRA_DP_INSTALL_FAIL
,
1143 "%u:%s: Failed to enqueue dataplane uninstall",
1147 case ZEBRA_DPLANE_REQUEST_SUCCESS
:
1156 /* Uninstall the route from kernel. */
1157 static void rib_uninstall(struct route_node
*rn
, struct route_entry
*re
)
1159 rib_table_info_t
*info
= srcdest_rnode_table_info(rn
);
1160 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1161 struct nexthop
*nexthop
;
1163 if (dest
&& dest
->selected_fib
== re
) {
1164 if (info
->safi
== SAFI_UNICAST
)
1165 hook_call(rib_update
, rn
, "rib_uninstall");
1167 /* If labeled-unicast route, uninstall transit LSP. */
1168 if (zebra_rib_labeled_unicast(re
))
1169 zebra_mpls_lsp_uninstall(info
->zvrf
, rn
, re
);
1171 if (!RIB_SYSTEM_ROUTE(re
))
1172 rib_uninstall_kernel(rn
, re
);
1174 dest
->selected_fib
= NULL
;
1176 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
1177 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1180 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)) {
1181 const struct prefix
*p
, *src_p
;
1183 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1185 redistribute_delete(p
, src_p
, re
);
1186 UNSET_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
);
1191 * rib_can_delete_dest
1193 * Returns TRUE if the given dest can be deleted from the table.
1195 static int rib_can_delete_dest(rib_dest_t
*dest
)
1202 * Don't delete the dest if we have to update the FPM about this
1205 if (CHECK_FLAG(dest
->flags
, RIB_DEST_UPDATE_FPM
)
1206 || CHECK_FLAG(dest
->flags
, RIB_DEST_SENT_TO_FPM
))
1215 * Garbage collect the rib dest corresponding to the given route node
1218 * Returns TRUE if the dest was deleted, FALSE otherwise.
1220 int rib_gc_dest(struct route_node
*rn
)
1224 dest
= rib_dest_from_rnode(rn
);
1228 if (!rib_can_delete_dest(dest
))
1231 if (IS_ZEBRA_DEBUG_RIB
) {
1232 struct zebra_vrf
*zvrf
;
1234 zvrf
= rib_dest_vrf(dest
);
1235 rnode_debug(rn
, zvrf_id(zvrf
), "removing dest from table");
1239 XFREE(MTYPE_RIB_DEST
, dest
);
1243 * Release the one reference that we keep on the route node.
1245 route_unlock_node(rn
);
1249 static void rib_process_add_fib(struct zebra_vrf
*zvrf
, struct route_node
*rn
,
1250 struct route_entry
*new)
1252 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1254 hook_call(rib_update
, rn
, "new route selected");
1256 /* Update real nexthop. This may actually determine if nexthop is active
1258 if (!nexthop_active_update(rn
, new, true)) {
1259 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1263 if (IS_ZEBRA_DEBUG_RIB
) {
1264 char buf
[SRCDEST2STR_BUFFER
];
1265 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1266 zlog_debug("%u:%s: Adding route rn %p, re %p (type %d)",
1267 zvrf_id(zvrf
), buf
, rn
, new, new->type
);
1270 /* If labeled-unicast route, install transit LSP. */
1271 if (zebra_rib_labeled_unicast(new))
1272 zebra_mpls_lsp_install(zvrf
, rn
, new);
1274 if (!RIB_SYSTEM_ROUTE(new))
1275 rib_install_kernel(rn
, new, NULL
);
1277 dest
->selected_fib
= new;
1279 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1282 static void rib_process_del_fib(struct zebra_vrf
*zvrf
, struct route_node
*rn
,
1283 struct route_entry
*old
)
1285 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1286 hook_call(rib_update
, rn
, "removing existing route");
1288 /* Uninstall from kernel. */
1289 if (IS_ZEBRA_DEBUG_RIB
) {
1290 char buf
[SRCDEST2STR_BUFFER
];
1291 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1292 zlog_debug("%u:%s: Deleting route rn %p, re %p (type %d)",
1293 zvrf_id(zvrf
), buf
, rn
, old
, old
->type
);
1296 /* If labeled-unicast route, uninstall transit LSP. */
1297 if (zebra_rib_labeled_unicast(old
))
1298 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1300 if (!RIB_SYSTEM_ROUTE(old
))
1301 rib_uninstall_kernel(rn
, old
);
1304 * We are setting this to NULL here
1305 * because that is what we traditionally
1306 * have been doing. I am not positive
1307 * that this is the right thing to do
1308 * but let's leave the code alone
1309 * for the RIB_SYSTEM_ROUTE case
1311 dest
->selected_fib
= NULL
;
1314 /* Update nexthop for route, reset changed flag. */
1315 /* Note: this code also handles the Linux case when an interface goes
1316 * down, causing the kernel to delete routes without sending DELROUTE
1319 if (!nexthop_active_update(rn
, old
, true) &&
1320 (RIB_KERNEL_ROUTE(old
)))
1321 SET_FLAG(old
->status
, ROUTE_ENTRY_REMOVED
);
1323 UNSET_FLAG(old
->status
, ROUTE_ENTRY_CHANGED
);
1326 static void rib_process_update_fib(struct zebra_vrf
*zvrf
,
1327 struct route_node
*rn
,
1328 struct route_entry
*old
,
1329 struct route_entry
*new)
1331 struct nexthop
*nexthop
= NULL
;
1333 rib_dest_t
*dest
= rib_dest_from_rnode(rn
);
1336 * We have to install or update if a new route has been selected or
1337 * something has changed.
1339 if (new != old
|| CHECK_FLAG(new->status
, ROUTE_ENTRY_CHANGED
)) {
1340 hook_call(rib_update
, rn
, "updating existing route");
1342 /* Update the nexthop; we could determine here that nexthop is
1344 if (nexthop_active_update(rn
, new, true))
1347 /* If nexthop is active, install the selected route, if
1349 * the install succeeds, cleanup flags for prior route, if
1354 if (IS_ZEBRA_DEBUG_RIB
) {
1355 char buf
[SRCDEST2STR_BUFFER
];
1356 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1359 "%u:%s: Updating route rn %p, re %p (type %d) "
1361 zvrf_id(zvrf
), buf
, rn
, new,
1362 new->type
, old
, old
->type
);
1365 "%u:%s: Updating route rn %p, re %p (type %d)",
1366 zvrf_id(zvrf
), buf
, rn
, new,
1370 /* If labeled-unicast route, uninstall transit LSP. */
1371 if (zebra_rib_labeled_unicast(old
))
1372 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1374 /* Non-system route should be installed. */
1375 if (!RIB_SYSTEM_ROUTE(new)) {
1376 /* If labeled-unicast route, install transit
1378 if (zebra_rib_labeled_unicast(new))
1379 zebra_mpls_lsp_install(zvrf
, rn
, new);
1381 rib_install_kernel(rn
, new, old
);
1384 * We do not need to install the
1385 * selected route because it
1386 * is already isntalled by
1387 * the system( ie not us )
1388 * so just mark it as winning
1389 * we do need to ensure that
1390 * if we uninstall a route
1391 * from ourselves we don't
1392 * over write this pointer
1394 dest
->selected_fib
= NULL
;
1396 /* If install succeeded or system route, cleanup flags
1397 * for prior route. */
1399 if (RIB_SYSTEM_ROUTE(new)) {
1400 if (!RIB_SYSTEM_ROUTE(old
))
1401 rib_uninstall_kernel(rn
, old
);
1403 for (nexthop
= old
->ng
.nexthop
; nexthop
;
1404 nexthop
= nexthop
->next
)
1405 UNSET_FLAG(nexthop
->flags
,
1412 * If nexthop for selected route is not active or install
1414 * may need to uninstall and delete for redistribution.
1417 if (IS_ZEBRA_DEBUG_RIB
) {
1418 char buf
[SRCDEST2STR_BUFFER
];
1419 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1422 "%u:%s: Deleting route rn %p, re %p (type %d) "
1423 "old %p (type %d) - nexthop inactive",
1424 zvrf_id(zvrf
), buf
, rn
, new,
1425 new->type
, old
, old
->type
);
1428 "%u:%s: Deleting route rn %p, re %p (type %d) - nexthop inactive",
1429 zvrf_id(zvrf
), buf
, rn
, new,
1433 /* If labeled-unicast route, uninstall transit LSP. */
1434 if (zebra_rib_labeled_unicast(old
))
1435 zebra_mpls_lsp_uninstall(zvrf
, rn
, old
);
1437 if (!RIB_SYSTEM_ROUTE(old
))
1438 rib_uninstall_kernel(rn
, old
);
1440 dest
->selected_fib
= NULL
;
1444 * Same route selected; check if in the FIB and if not,
1446 * is housekeeping code to deal with race conditions in kernel
1448 * netlink reporting interface up before IPv4 or IPv6 protocol
1452 if (!RIB_SYSTEM_ROUTE(new)) {
1453 bool in_fib
= false;
1455 for (ALL_NEXTHOPS(new->ng
, nexthop
))
1456 if (CHECK_FLAG(nexthop
->flags
,
1457 NEXTHOP_FLAG_FIB
)) {
1462 rib_install_kernel(rn
, new, NULL
);
1466 /* Update prior route. */
1468 /* Set real nexthop. */
1469 nexthop_active_update(rn
, old
, true);
1470 UNSET_FLAG(old
->status
, ROUTE_ENTRY_CHANGED
);
1473 /* Clear changed flag. */
1474 UNSET_FLAG(new->status
, ROUTE_ENTRY_CHANGED
);
1477 /* Check if 'alternate' RIB entry is better than 'current'. */
1478 static struct route_entry
*rib_choose_best(struct route_entry
*current
,
1479 struct route_entry
*alternate
)
1481 if (current
== NULL
)
1484 /* filter route selection in following order:
1485 * - connected beats other types
1486 * - if both connected, loopback or vrf wins
1487 * - lower distance beats higher
1488 * - lower metric beats higher for equal distance
1489 * - last, hence oldest, route wins tie break.
1492 /* Connected routes. Check to see if either are a vrf
1493 * or loopback interface. If not, pick the last connected
1494 * route of the set of lowest metric connected routes.
1496 if (alternate
->type
== ZEBRA_ROUTE_CONNECT
) {
1497 if (current
->type
!= ZEBRA_ROUTE_CONNECT
)
1500 /* both are connected. are either loop or vrf? */
1501 struct nexthop
*nexthop
= NULL
;
1503 for (ALL_NEXTHOPS(alternate
->ng
, nexthop
)) {
1504 if (if_is_loopback_or_vrf(if_lookup_by_index(
1505 nexthop
->ifindex
, alternate
->vrf_id
)))
1509 for (ALL_NEXTHOPS(current
->ng
, nexthop
)) {
1510 if (if_is_loopback_or_vrf(if_lookup_by_index(
1511 nexthop
->ifindex
, current
->vrf_id
)))
1515 /* Neither are loop or vrf so pick best metric */
1516 if (alternate
->metric
<= current
->metric
)
1522 if (current
->type
== ZEBRA_ROUTE_CONNECT
)
1525 /* higher distance loses */
1526 if (alternate
->distance
< current
->distance
)
1528 if (current
->distance
< alternate
->distance
)
1531 /* metric tie-breaks equal distance */
1532 if (alternate
->metric
<= current
->metric
)
1538 /* Core function for processing routing information base. */
1539 static void rib_process(struct route_node
*rn
)
1541 struct route_entry
*re
;
1542 struct route_entry
*next
;
1543 struct route_entry
*old_selected
= NULL
;
1544 struct route_entry
*new_selected
= NULL
;
1545 struct route_entry
*old_fib
= NULL
;
1546 struct route_entry
*new_fib
= NULL
;
1547 struct route_entry
*best
= NULL
;
1548 char buf
[SRCDEST2STR_BUFFER
];
1550 struct zebra_vrf
*zvrf
= NULL
;
1551 const struct prefix
*p
, *src_p
;
1553 srcdest_rnode_prefixes(rn
, &p
, &src_p
);
1554 vrf_id_t vrf_id
= VRF_UNKNOWN
;
1558 dest
= rib_dest_from_rnode(rn
);
1560 zvrf
= rib_dest_vrf(dest
);
1561 vrf_id
= zvrf_id(zvrf
);
1564 if (IS_ZEBRA_DEBUG_RIB
)
1565 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
1567 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1568 zlog_debug("%u:%s: Processing rn %p", vrf_id
, buf
, rn
);
1571 * we can have rn's that have a NULL info pointer
1572 * (dest). As such let's not let the deref happen
1573 * additionally we know RNODE_FOREACH_RE_SAFE
1574 * will not iterate so we are ok.
1577 old_fib
= dest
->selected_fib
;
1579 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
1580 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
1582 "%u:%s: Examine re %p (type %d) status %x flags %x "
1583 "dist %d metric %d",
1584 vrf_id
, buf
, re
, re
->type
, re
->status
,
1585 re
->flags
, re
->distance
, re
->metric
);
1587 UNSET_FLAG(re
->status
, ROUTE_ENTRY_NEXTHOPS_CHANGED
);
1589 /* Currently selected re. */
1590 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)) {
1591 assert(old_selected
== NULL
);
1595 /* Skip deleted entries from selection */
1596 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
1599 /* Skip unreachable nexthop. */
1600 /* This first call to nexthop_active_update is merely to
1602 * there's any change to nexthops associated with this RIB
1604 * rib_process() can be invoked due to an external event such as
1606 * down or due to next-hop-tracking evaluation. In the latter
1608 * a decision has already been made that the NHs have changed.
1610 * need to invoke a potentially expensive call again. Further,
1612 * the change might be in a recursive NH which is not caught in
1613 * the nexthop_active_update() code. Thus, we might miss changes
1617 if (!CHECK_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
)
1618 && !nexthop_active_update(rn
, re
, false)) {
1619 if (re
->type
== ZEBRA_ROUTE_TABLE
) {
1620 /* XXX: HERE BE DRAGONS!!!!!
1621 * In all honesty, I have not yet figured out
1623 * does or why the ROUTE_ENTRY_CHANGED test
1625 * or why we need to delete a route here, and
1627 * this concerns both selected and fib route, or
1630 /* This entry was denied by the 'ip protocol
1631 * table' route-map, we
1632 * need to delete it */
1633 if (re
!= old_selected
) {
1634 if (IS_ZEBRA_DEBUG_RIB
)
1636 "%s: %u:%s: imported via import-table but denied "
1637 "by the ip protocol table route-map",
1638 __func__
, vrf_id
, buf
);
1641 SET_FLAG(re
->status
,
1642 ROUTE_ENTRY_REMOVED
);
1648 /* Infinite distance. */
1649 if (re
->distance
== DISTANCE_INFINITY
) {
1650 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1654 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_FIB_OVERRIDE
)) {
1655 best
= rib_choose_best(new_fib
, re
);
1656 if (new_fib
&& best
!= new_fib
)
1657 UNSET_FLAG(new_fib
->status
,
1658 ROUTE_ENTRY_CHANGED
);
1661 best
= rib_choose_best(new_selected
, re
);
1662 if (new_selected
&& best
!= new_selected
)
1663 UNSET_FLAG(new_selected
->status
,
1664 ROUTE_ENTRY_CHANGED
);
1665 new_selected
= best
;
1668 UNSET_FLAG(re
->status
, ROUTE_ENTRY_CHANGED
);
1669 } /* RNODE_FOREACH_RE */
1671 /* If no FIB override route, use the selected route also for FIB */
1672 if (new_fib
== NULL
)
1673 new_fib
= new_selected
;
1675 /* After the cycle is finished, the following pointers will be set:
1676 * old_selected --- RE entry currently having SELECTED
1677 * new_selected --- RE entry that is newly SELECTED
1678 * old_fib --- RE entry currently in kernel FIB
1679 * new_fib --- RE entry that is newly to be in kernel FIB
1681 * new_selected will get SELECTED flag, and is going to be redistributed
1682 * the zclients. new_fib (which can be new_selected) will be installed
1686 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
1688 "%u:%s: After processing: old_selected %p new_selected %p old_fib %p new_fib %p",
1689 vrf_id
, buf
, (void *)old_selected
, (void *)new_selected
,
1690 (void *)old_fib
, (void *)new_fib
);
1693 /* Buffer ROUTE_ENTRY_CHANGED here, because it will get cleared if
1694 * fib == selected */
1695 bool selected_changed
= new_selected
&& CHECK_FLAG(new_selected
->status
,
1696 ROUTE_ENTRY_CHANGED
);
1698 /* Update fib according to selection results */
1699 if (new_fib
&& old_fib
)
1700 rib_process_update_fib(zvrf
, rn
, old_fib
, new_fib
);
1702 rib_process_add_fib(zvrf
, rn
, new_fib
);
1704 rib_process_del_fib(zvrf
, rn
, old_fib
);
1706 /* Update SELECTED entry */
1707 if (old_selected
!= new_selected
|| selected_changed
) {
1709 if (new_selected
&& new_selected
!= new_fib
) {
1710 nexthop_active_update(rn
, new_selected
, true);
1711 UNSET_FLAG(new_selected
->status
, ROUTE_ENTRY_CHANGED
);
1715 SET_FLAG(new_selected
->flags
, ZEBRA_FLAG_SELECTED
);
1717 /* Special case: new route is system route, so
1718 * dataplane update will not be done - ensure we
1719 * redistribute the route.
1721 if (RIB_SYSTEM_ROUTE(new_selected
))
1722 redistribute_update(p
, src_p
, new_selected
,
1728 redistribute_delete(p
, src_p
, old_selected
);
1729 if (old_selected
!= new_selected
)
1730 UNSET_FLAG(old_selected
->flags
,
1731 ZEBRA_FLAG_SELECTED
);
1735 /* Remove all RE entries queued for removal */
1736 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
1737 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
1738 if (IS_ZEBRA_DEBUG_RIB
) {
1739 rnode_debug(rn
, vrf_id
, "rn %p, removing re %p",
1740 (void *)rn
, (void *)re
);
1747 * Check if the dest can be deleted now.
1753 * Utility to match route with dplane context data
1755 static bool rib_route_match_ctx(const struct route_entry
*re
,
1756 const struct zebra_dplane_ctx
*ctx
,
1759 bool result
= false;
1763 * In 'update' case, we test info about the 'previous' or
1766 if ((re
->type
== dplane_ctx_get_old_type(ctx
)) &&
1767 (re
->instance
== dplane_ctx_get_old_instance(ctx
))) {
1770 /* TODO -- we're using this extra test, but it's not
1771 * exactly clear why.
1773 if (re
->type
== ZEBRA_ROUTE_STATIC
&&
1774 (re
->distance
!= dplane_ctx_get_old_distance(ctx
) ||
1775 re
->tag
!= dplane_ctx_get_old_tag(ctx
))) {
1782 * Ordinary, single-route case using primary context info
1784 if ((dplane_ctx_get_op(ctx
) != DPLANE_OP_ROUTE_DELETE
) &&
1785 CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
1786 /* Skip route that's been deleted */
1790 if ((re
->type
== dplane_ctx_get_type(ctx
)) &&
1791 (re
->instance
== dplane_ctx_get_instance(ctx
))) {
1794 /* TODO -- we're using this extra test, but it's not
1795 * exactly clear why.
1797 if (re
->type
== ZEBRA_ROUTE_STATIC
&&
1798 (re
->distance
!= dplane_ctx_get_distance(ctx
) ||
1799 re
->tag
!= dplane_ctx_get_tag(ctx
))) {
1811 * Route-update results processing after async dataplane update.
1813 static void rib_process_after(struct zebra_dplane_ctx
*ctx
)
1815 struct route_table
*table
= NULL
;
1816 struct zebra_vrf
*zvrf
= NULL
;
1817 struct route_node
*rn
= NULL
;
1818 struct route_entry
*re
= NULL
, *old_re
= NULL
, *rib
;
1819 bool is_update
= false;
1820 struct nexthop
*nexthop
, *ctx_nexthop
;
1821 char dest_str
[PREFIX_STRLEN
] = "";
1822 enum dplane_op_e op
;
1823 enum zebra_dplane_result status
;
1824 const struct prefix
*dest_pfx
, *src_pfx
;
1826 /* Locate rn and re(s) from ctx */
1828 table
= zebra_vrf_table_with_table_id(dplane_ctx_get_afi(ctx
),
1829 dplane_ctx_get_safi(ctx
),
1830 dplane_ctx_get_vrf(ctx
),
1831 dplane_ctx_get_table(ctx
));
1832 if (table
== NULL
) {
1833 if (IS_ZEBRA_DEBUG_DPLANE
) {
1834 zlog_debug("Failed to process dplane results: no table for afi %d, safi %d, vrf %u",
1835 dplane_ctx_get_afi(ctx
),
1836 dplane_ctx_get_safi(ctx
),
1837 dplane_ctx_get_vrf(ctx
));
1842 zvrf
= vrf_info_lookup(dplane_ctx_get_vrf(ctx
));
1844 dest_pfx
= dplane_ctx_get_dest(ctx
);
1846 /* Note well: only capturing the prefix string if debug is enabled here;
1847 * unconditional log messages will have to generate the string.
1849 if (IS_ZEBRA_DEBUG_DPLANE
)
1850 prefix2str(dest_pfx
, dest_str
, sizeof(dest_str
));
1852 src_pfx
= dplane_ctx_get_src(ctx
);
1853 rn
= srcdest_rnode_get(table
, dplane_ctx_get_dest(ctx
),
1854 src_pfx
? (struct prefix_ipv6
*)src_pfx
: NULL
);
1856 if (IS_ZEBRA_DEBUG_DPLANE
) {
1857 zlog_debug("Failed to process dplane results: no route for %u:%s",
1858 dplane_ctx_get_vrf(ctx
), dest_str
);
1863 srcdest_rnode_prefixes(rn
, &dest_pfx
, &src_pfx
);
1865 op
= dplane_ctx_get_op(ctx
);
1866 status
= dplane_ctx_get_status(ctx
);
1868 if (IS_ZEBRA_DEBUG_DPLANE_DETAIL
)
1869 zlog_debug("%u:%s Processing dplane ctx %p, op %s result %s",
1870 dplane_ctx_get_vrf(ctx
), dest_str
, ctx
,
1871 dplane_op2str(op
), dplane_res2str(status
));
1873 if (op
== DPLANE_OP_ROUTE_DELETE
) {
1875 * In the delete case, the zebra core datastructs were
1876 * updated (or removed) at the time the delete was issued,
1877 * so we're just notifying the route owner.
1879 if (status
== ZEBRA_DPLANE_REQUEST_SUCCESS
) {
1880 zsend_route_notify_owner_ctx(ctx
, ZAPI_ROUTE_REMOVED
);
1885 zsend_route_notify_owner_ctx(ctx
,
1886 ZAPI_ROUTE_REMOVE_FAIL
);
1888 zlog_warn("%u:%s: Route Deletion failure",
1889 dplane_ctx_get_vrf(ctx
),
1890 prefix2str(dest_pfx
,
1891 dest_str
, sizeof(dest_str
)));
1894 /* Nothing more to do in delete case */
1899 * Update is a bit of a special case, where we may have both old and new
1900 * routes to post-process.
1902 is_update
= dplane_ctx_is_update(ctx
);
1905 * Take a pass through the routes, look for matches with the context
1908 RNODE_FOREACH_RE(rn
, rib
) {
1911 if (rib_route_match_ctx(rib
, ctx
, false))
1915 /* Check for old route match */
1916 if (is_update
&& (old_re
== NULL
)) {
1917 if (rib_route_match_ctx(rib
, ctx
, true /*is_update*/))
1921 /* Have we found the routes we need to work on? */
1922 if (re
&& ((!is_update
|| old_re
)))
1927 * Check sequence number(s) to detect stale results before continuing
1929 if (re
&& (re
->dplane_sequence
!= dplane_ctx_get_seq(ctx
))) {
1930 if (IS_ZEBRA_DEBUG_DPLANE_DETAIL
) {
1931 zlog_debug("%u:%s Stale dplane result for re %p",
1932 dplane_ctx_get_vrf(ctx
), dest_str
, re
);
1938 (old_re
->dplane_sequence
!= dplane_ctx_get_old_seq(ctx
))) {
1939 if (IS_ZEBRA_DEBUG_DPLANE_DETAIL
) {
1940 zlog_debug("%u:%s Stale dplane result for old_re %p",
1941 dplane_ctx_get_vrf(ctx
), dest_str
, old_re
);
1947 * Here's sort of a tough one: the route update result is stale.
1948 * Is it better to use the context block info to generate
1949 * redist and owner notification, or is it better to wait
1950 * for the up-to-date result to arrive?
1953 /* TODO -- for now, only expose up-to-date results */
1957 if (status
== ZEBRA_DPLANE_REQUEST_SUCCESS
) {
1958 /* Update zebra nexthop FIB flag for each
1959 * nexthop that was installed.
1961 for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx
), ctx_nexthop
)) {
1963 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1964 if (nexthop_same(ctx_nexthop
, nexthop
))
1968 if (nexthop
== NULL
)
1971 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
))
1974 if (CHECK_FLAG(ctx_nexthop
->flags
,
1976 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1978 UNSET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
1983 /* Set flag for nexthop tracking processing */
1984 zvrf
->flags
|= ZEBRA_VRF_RIB_SCHEDULED
;
1988 /* TODO -- still calling the redist api using the route_entries,
1989 * and there's a corner-case here: if there's no client
1990 * for the 'new' route, a redist deleting the 'old' route
1991 * will be sent. But if the 'old' context info was stale,
1992 * 'old_re' will be NULL here and that delete will not be sent.
1994 redistribute_update(dest_pfx
, src_pfx
, re
, old_re
);
1996 /* Notify route owner */
1997 zsend_route_notify_owner(re
,
1998 dest_pfx
, ZAPI_ROUTE_INSTALLED
);
2001 zsend_route_notify_owner(re
, dest_pfx
,
2002 ZAPI_ROUTE_FAIL_INSTALL
);
2004 zlog_warn("%u:%s: Route install failed",
2005 dplane_ctx_get_vrf(ctx
),
2006 prefix2str(dest_pfx
,
2007 dest_str
, sizeof(dest_str
)));
2012 /* Return context to dataplane module */
2013 dplane_ctx_fini(&ctx
);
2016 /* Take a list of route_node structs and return 1, if there was a record
2017 * picked from it and processed by rib_process(). Don't process more,
2018 * than one RN record; operate only in the specified sub-queue.
2020 static unsigned int process_subq(struct list
*subq
, uint8_t qindex
)
2022 struct listnode
*lnode
= listhead(subq
);
2023 struct route_node
*rnode
;
2025 struct zebra_vrf
*zvrf
= NULL
;
2030 rnode
= listgetdata(lnode
);
2031 dest
= rib_dest_from_rnode(rnode
);
2033 zvrf
= rib_dest_vrf(dest
);
2037 if (IS_ZEBRA_DEBUG_RIB_DETAILED
) {
2038 char buf
[SRCDEST2STR_BUFFER
];
2039 srcdest_rnode2str(rnode
, buf
, sizeof(buf
));
2040 zlog_debug("%u:%s: rn %p dequeued from sub-queue %u",
2041 zvrf
? zvrf_id(zvrf
) : 0, buf
, rnode
, qindex
);
2045 UNSET_FLAG(rib_dest_from_rnode(rnode
)->flags
,
2046 RIB_ROUTE_QUEUED(qindex
));
2051 zlog_debug ("%s: called for route_node (%p, %d) with no ribs",
2052 __func__
, rnode
, rnode
->lock
);
2053 zlog_backtrace(LOG_DEBUG
);
2056 route_unlock_node(rnode
);
2057 list_delete_node(subq
, lnode
);
2062 * Perform next-hop tracking processing after RIB updates.
2064 static void do_nht_processing(void)
2067 struct zebra_vrf
*zvrf
;
2069 /* Evaluate nexthops for those VRFs which underwent route processing.
2071 * should limit the evaluation to the necessary VRFs in most common
2074 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
2076 if (zvrf
== NULL
|| !(zvrf
->flags
& ZEBRA_VRF_RIB_SCHEDULED
))
2079 if (IS_ZEBRA_DEBUG_RIB_DETAILED
|| IS_ZEBRA_DEBUG_NHT
)
2080 zlog_debug("NHT processing check for zvrf %s",
2083 zvrf
->flags
&= ~ZEBRA_VRF_RIB_SCHEDULED
;
2084 zebra_evaluate_rnh(zvrf
, AF_INET
, 0, RNH_NEXTHOP_TYPE
, NULL
);
2085 zebra_evaluate_rnh(zvrf
, AF_INET
, 0, RNH_IMPORT_CHECK_TYPE
,
2087 zebra_evaluate_rnh(zvrf
, AF_INET6
, 0, RNH_NEXTHOP_TYPE
, NULL
);
2088 zebra_evaluate_rnh(zvrf
, AF_INET6
, 0, RNH_IMPORT_CHECK_TYPE
,
2092 /* Schedule LSPs for processing, if needed. */
2093 zvrf
= vrf_info_lookup(VRF_DEFAULT
);
2094 if (mpls_should_lsps_be_processed(zvrf
)) {
2095 if (IS_ZEBRA_DEBUG_MPLS
)
2097 "%u: Scheduling all LSPs upon RIB completion",
2099 zebra_mpls_lsp_schedule(zvrf
);
2100 mpls_unmark_lsps_for_processing(zvrf
);
2105 * All meta queues have been processed. Trigger next-hop evaluation.
2107 static void meta_queue_process_complete(struct work_queue
*dummy
)
2109 do_nht_processing();
2112 /* Dispatch the meta queue by picking, processing and unlocking the next RN from
2113 * a non-empty sub-queue with lowest priority. wq is equal to zebra->ribq and
2115 * is pointed to the meta queue structure.
2117 static wq_item_status
meta_queue_process(struct work_queue
*dummy
, void *data
)
2119 struct meta_queue
*mq
= data
;
2121 uint32_t queue_len
, queue_limit
;
2123 /* Ensure there's room for more dataplane updates */
2124 queue_limit
= dplane_get_in_queue_limit();
2125 queue_len
= dplane_get_in_queue_len();
2126 if (queue_len
> queue_limit
) {
2127 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2128 zlog_debug("rib queue: dplane queue len %u, limit %u, retrying",
2129 queue_len
, queue_limit
);
2131 /* Ensure that the meta-queue is actually enqueued */
2132 if (work_queue_empty(zebrad
.ribq
))
2133 work_queue_add(zebrad
.ribq
, zebrad
.mq
);
2135 return WQ_QUEUE_BLOCKED
;
2138 for (i
= 0; i
< MQ_SIZE
; i
++)
2139 if (process_subq(mq
->subq
[i
], i
)) {
2143 return mq
->size
? WQ_REQUEUE
: WQ_SUCCESS
;
2147 * Map from rib types to queue type (priority) in meta queue
2149 static const uint8_t meta_queue_map
[ZEBRA_ROUTE_MAX
] = {
2150 [ZEBRA_ROUTE_SYSTEM
] = 4,
2151 [ZEBRA_ROUTE_KERNEL
] = 0,
2152 [ZEBRA_ROUTE_CONNECT
] = 0,
2153 [ZEBRA_ROUTE_STATIC
] = 1,
2154 [ZEBRA_ROUTE_RIP
] = 2,
2155 [ZEBRA_ROUTE_RIPNG
] = 2,
2156 [ZEBRA_ROUTE_OSPF
] = 2,
2157 [ZEBRA_ROUTE_OSPF6
] = 2,
2158 [ZEBRA_ROUTE_ISIS
] = 2,
2159 [ZEBRA_ROUTE_BGP
] = 3,
2160 [ZEBRA_ROUTE_PIM
] = 4, // Shouldn't happen but for safety
2161 [ZEBRA_ROUTE_EIGRP
] = 2,
2162 [ZEBRA_ROUTE_NHRP
] = 2,
2163 [ZEBRA_ROUTE_HSLS
] = 4,
2164 [ZEBRA_ROUTE_OLSR
] = 4,
2165 [ZEBRA_ROUTE_TABLE
] = 1,
2166 [ZEBRA_ROUTE_LDP
] = 4,
2167 [ZEBRA_ROUTE_VNC
] = 3,
2168 [ZEBRA_ROUTE_VNC_DIRECT
] = 3,
2169 [ZEBRA_ROUTE_VNC_DIRECT_RH
] = 3,
2170 [ZEBRA_ROUTE_BGP_DIRECT
] = 3,
2171 [ZEBRA_ROUTE_BGP_DIRECT_EXT
] = 3,
2172 [ZEBRA_ROUTE_BABEL
] = 2,
2173 [ZEBRA_ROUTE_ALL
] = 4, // Shouldn't happen but for safety
2176 /* Look into the RN and queue it into one or more priority queues,
2177 * increasing the size for each data push done.
2179 static void rib_meta_queue_add(struct meta_queue
*mq
, struct route_node
*rn
)
2181 struct route_entry
*re
;
2183 RNODE_FOREACH_RE (rn
, re
) {
2184 uint8_t qindex
= meta_queue_map
[re
->type
];
2185 struct zebra_vrf
*zvrf
;
2187 /* Invariant: at this point we always have rn->info set. */
2188 if (CHECK_FLAG(rib_dest_from_rnode(rn
)->flags
,
2189 RIB_ROUTE_QUEUED(qindex
))) {
2190 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2193 "rn %p is already queued in sub-queue %u",
2194 (void *)rn
, qindex
);
2198 SET_FLAG(rib_dest_from_rnode(rn
)->flags
,
2199 RIB_ROUTE_QUEUED(qindex
));
2200 listnode_add(mq
->subq
[qindex
], rn
);
2201 route_lock_node(rn
);
2204 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2205 rnode_debug(rn
, re
->vrf_id
,
2206 "queued rn %p into sub-queue %u",
2207 (void *)rn
, qindex
);
2209 zvrf
= zebra_vrf_lookup_by_id(re
->vrf_id
);
2211 zvrf
->flags
|= ZEBRA_VRF_RIB_SCHEDULED
;
2215 /* Add route_node to work queue and schedule processing */
2216 void rib_queue_add(struct route_node
*rn
)
2220 /* Pointless to queue a route_node with no RIB entries to add or remove
2222 if (!rnode_to_ribs(rn
)) {
2223 zlog_debug("%s: called for route_node (%p, %d) with no ribs",
2224 __func__
, (void *)rn
, rn
->lock
);
2225 zlog_backtrace(LOG_DEBUG
);
2229 if (zebrad
.ribq
== NULL
) {
2230 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2231 "%s: work_queue does not exist!", __func__
);
2236 * The RIB queue should normally be either empty or holding the only
2237 * work_queue_item element. In the latter case this element would
2238 * hold a pointer to the meta queue structure, which must be used to
2239 * actually queue the route nodes to process. So create the MQ
2240 * holder, if necessary, then push the work into it in any case.
2241 * This semantics was introduced after 0.99.9 release.
2243 if (work_queue_empty(zebrad
.ribq
))
2244 work_queue_add(zebrad
.ribq
, zebrad
.mq
);
2246 rib_meta_queue_add(zebrad
.mq
, rn
);
2251 /* Create new meta queue.
2252 A destructor function doesn't seem to be necessary here.
2254 static struct meta_queue
*meta_queue_new(void)
2256 struct meta_queue
*new;
2259 new = XCALLOC(MTYPE_WORK_QUEUE
, sizeof(struct meta_queue
));
2261 for (i
= 0; i
< MQ_SIZE
; i
++) {
2262 new->subq
[i
] = list_new();
2263 assert(new->subq
[i
]);
2269 void meta_queue_free(struct meta_queue
*mq
)
2273 for (i
= 0; i
< MQ_SIZE
; i
++)
2274 list_delete(&mq
->subq
[i
]);
2276 XFREE(MTYPE_WORK_QUEUE
, mq
);
2279 /* initialise zebra rib work queue */
2280 static void rib_queue_init(struct zebra_t
*zebra
)
2285 work_queue_new(zebra
->master
, "route_node processing"))) {
2286 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2287 "%s: could not initialise work queue!", __func__
);
2291 /* fill in the work queue spec */
2292 zebra
->ribq
->spec
.workfunc
= &meta_queue_process
;
2293 zebra
->ribq
->spec
.errorfunc
= NULL
;
2294 zebra
->ribq
->spec
.completion_func
= &meta_queue_process_complete
;
2295 /* XXX: TODO: These should be runtime configurable via vty */
2296 zebra
->ribq
->spec
.max_retries
= 3;
2297 zebra
->ribq
->spec
.hold
= ZEBRA_RIB_PROCESS_HOLD_TIME
;
2298 zebra
->ribq
->spec
.retry
= ZEBRA_RIB_PROCESS_RETRY_TIME
;
2300 if (!(zebra
->mq
= meta_queue_new())) {
2301 flog_err(EC_ZEBRA_WQ_NONEXISTENT
,
2302 "%s: could not initialise meta queue!", __func__
);
2308 /* RIB updates are processed via a queue of pointers to route_nodes.
2310 * The queue length is bounded by the maximal size of the routing table,
2311 * as a route_node will not be requeued, if already queued.
2313 * REs are submitted via rib_addnode or rib_delnode which set minimal
2314 * state, or static_install_route (when an existing RE is updated)
2315 * and then submit route_node to queue for best-path selection later.
2316 * Order of add/delete state changes are preserved for any given RE.
2318 * Deleted REs are reaped during best-path selection.
2321 * |-> rib_link or unset ROUTE_ENTRY_REMOVE |->Update kernel with
2322 * |-------->| | best RE, if required
2324 * static_install->|->rib_addqueue...... -> rib_process
2326 * |-------->| |-> rib_unlink
2327 * |-> set ROUTE_ENTRY_REMOVE |
2328 * rib_delnode (RE freed)
2330 * The 'info' pointer of a route_node points to a rib_dest_t
2331 * ('dest'). Queueing state for a route_node is kept on the dest. The
2332 * dest is created on-demand by rib_link() and is kept around at least
2333 * as long as there are ribs hanging off it (@see rib_gc_dest()).
2335 * Refcounting (aka "locking" throughout the GNU Zebra and Quagga code):
2337 * - route_nodes: refcounted by:
2338 * - dest attached to route_node:
2339 * - managed by: rib_link/rib_gc_dest
2340 * - route_node processing queue
2341 * - managed by: rib_addqueue, rib_process.
2345 /* Add RE to head of the route node. */
2346 static void rib_link(struct route_node
*rn
, struct route_entry
*re
, int process
)
2348 struct route_entry
*head
;
2351 const char *rmap_name
;
2355 dest
= rib_dest_from_rnode(rn
);
2357 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2358 rnode_debug(rn
, re
->vrf_id
, "rn %p adding dest", rn
);
2360 dest
= XCALLOC(MTYPE_RIB_DEST
, sizeof(rib_dest_t
));
2361 route_lock_node(rn
); /* rn route table reference */
2366 head
= dest
->routes
;
2373 afi
= (rn
->p
.family
== AF_INET
)
2375 : (rn
->p
.family
== AF_INET6
) ? AFI_IP6
: AFI_MAX
;
2376 if (is_zebra_import_table_enabled(afi
, re
->table
)) {
2377 rmap_name
= zebra_get_import_table_route_map(afi
, re
->table
);
2378 zebra_add_import_table_entry(rn
, re
, rmap_name
);
2383 static void rib_addnode(struct route_node
*rn
,
2384 struct route_entry
*re
, int process
)
2386 /* RE node has been un-removed before route-node is processed.
2387 * route_node must hence already be on the queue for processing..
2389 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)) {
2390 if (IS_ZEBRA_DEBUG_RIB
)
2391 rnode_debug(rn
, re
->vrf_id
, "rn %p, un-removed re %p",
2392 (void *)rn
, (void *)re
);
2394 UNSET_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
);
2397 rib_link(rn
, re
, process
);
2403 * Detach a rib structure from a route_node.
2405 * Note that a call to rib_unlink() should be followed by a call to
2406 * rib_gc_dest() at some point. This allows a rib_dest_t that is no
2407 * longer required to be deleted.
2409 void rib_unlink(struct route_node
*rn
, struct route_entry
*re
)
2415 if (IS_ZEBRA_DEBUG_RIB
)
2416 rnode_debug(rn
, re
->vrf_id
, "rn %p, re %p", (void *)rn
,
2419 dest
= rib_dest_from_rnode(rn
);
2422 re
->next
->prev
= re
->prev
;
2425 re
->prev
->next
= re
->next
;
2427 dest
->routes
= re
->next
;
2430 if (dest
->selected_fib
== re
)
2431 dest
->selected_fib
= NULL
;
2433 nexthops_free(re
->ng
.nexthop
);
2434 XFREE(MTYPE_RE
, re
);
2437 void rib_delnode(struct route_node
*rn
, struct route_entry
*re
)
2441 if (IS_ZEBRA_DEBUG_RIB
)
2442 rnode_debug(rn
, re
->vrf_id
, "rn %p, re %p, removing",
2443 (void *)rn
, (void *)re
);
2444 SET_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
);
2446 afi
= (rn
->p
.family
== AF_INET
)
2448 : (rn
->p
.family
== AF_INET6
) ? AFI_IP6
: AFI_MAX
;
2449 if (is_zebra_import_table_enabled(afi
, re
->table
)) {
2450 zebra_del_import_table_entry(rn
, re
);
2451 /* Just clean up if non main table */
2452 if (IS_ZEBRA_DEBUG_RIB
) {
2453 char buf
[SRCDEST2STR_BUFFER
];
2454 srcdest_rnode2str(rn
, buf
, sizeof(buf
));
2456 "%u:%s: Freeing route rn %p, re %p (type %d)",
2457 re
->vrf_id
, buf
, rn
, re
, re
->type
);
2466 /* This function dumps the contents of a given RE entry into
2467 * standard debug log. Calling function name and IP prefix in
2468 * question are passed as 1st and 2nd arguments.
2471 void _route_entry_dump(const char *func
, union prefixconstptr pp
,
2472 union prefixconstptr src_pp
,
2473 const struct route_entry
*re
)
2475 const struct prefix
*src_p
= src_pp
.p
;
2476 bool is_srcdst
= src_p
&& src_p
->prefixlen
;
2477 char straddr
[PREFIX_STRLEN
];
2478 char srcaddr
[PREFIX_STRLEN
];
2479 struct nexthop
*nexthop
;
2481 zlog_debug("%s: dumping RE entry %p for %s%s%s vrf %u", func
,
2482 (const void *)re
, prefix2str(pp
, straddr
, sizeof(straddr
)),
2483 is_srcdst
? " from " : "",
2484 is_srcdst
? prefix2str(src_pp
, srcaddr
, sizeof(srcaddr
))
2487 zlog_debug("%s: uptime == %lu, type == %u, instance == %d, table == %d",
2488 func
, (unsigned long)re
->uptime
, re
->type
, re
->instance
,
2491 "%s: metric == %u, mtu == %u, distance == %u, flags == %u, status == %u",
2492 func
, re
->metric
, re
->mtu
, re
->distance
, re
->flags
, re
->status
);
2493 zlog_debug("%s: nexthop_num == %u, nexthop_active_num == %u", func
,
2494 re
->nexthop_num
, re
->nexthop_active_num
);
2496 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
2497 struct interface
*ifp
;
2498 struct vrf
*vrf
= vrf_lookup_by_id(nexthop
->vrf_id
);
2500 switch (nexthop
->type
) {
2501 case NEXTHOP_TYPE_BLACKHOLE
:
2502 sprintf(straddr
, "Blackhole");
2504 case NEXTHOP_TYPE_IFINDEX
:
2505 ifp
= if_lookup_by_index(nexthop
->ifindex
,
2507 sprintf(straddr
, "%s", ifp
? ifp
->name
: "Unknown");
2509 case NEXTHOP_TYPE_IPV4
:
2511 case NEXTHOP_TYPE_IPV4_IFINDEX
:
2512 inet_ntop(AF_INET
, &nexthop
->gate
, straddr
,
2515 case NEXTHOP_TYPE_IPV6
:
2516 case NEXTHOP_TYPE_IPV6_IFINDEX
:
2517 inet_ntop(AF_INET6
, &nexthop
->gate
, straddr
,
2521 zlog_debug("%s: %s %s[%u] vrf %s(%u) with flags %s%s%s", func
,
2522 (nexthop
->rparent
? " NH" : "NH"), straddr
,
2523 nexthop
->ifindex
, vrf
? vrf
->name
: "Unknown",
2525 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
)
2528 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
)
2531 (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
)
2535 zlog_debug("%s: dump complete", func
);
2538 /* This is an exported helper to rtm_read() to dump the strange
2539 * RE entry found by rib_lookup_ipv4_route()
2542 void rib_lookup_and_dump(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
2544 struct route_table
*table
;
2545 struct route_node
*rn
;
2546 struct route_entry
*re
;
2547 char prefix_buf
[INET_ADDRSTRLEN
];
2550 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
2552 flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED
,
2553 "%s:%u zebra_vrf_table() returned NULL", __func__
,
2558 /* Scan the RIB table for exactly matching RE entry. */
2559 rn
= route_node_lookup(table
, (struct prefix
*)p
);
2561 /* No route for this prefix. */
2563 zlog_debug("%s:%u lookup failed for %s", __func__
, vrf_id
,
2564 prefix2str((struct prefix
*)p
, prefix_buf
,
2565 sizeof(prefix_buf
)));
2570 route_unlock_node(rn
);
2573 RNODE_FOREACH_RE (rn
, re
) {
2574 zlog_debug("%s:%u rn %p, re %p: %s, %s",
2576 (void *)rn
, (void *)re
,
2577 (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
)
2580 (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELECTED
)
2583 route_entry_dump(p
, NULL
, re
);
2587 /* Check if requested address assignment will fail due to another
2588 * route being installed by zebra in FIB already. Take necessary
2589 * actions, if needed: remove such a route from FIB and deSELECT
2590 * corresponding RE entry. Then put affected RN into RIBQ head.
2592 void rib_lookup_and_pushup(struct prefix_ipv4
*p
, vrf_id_t vrf_id
)
2594 struct route_table
*table
;
2595 struct route_node
*rn
;
2596 unsigned changed
= 0;
2599 if (NULL
== (table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
))) {
2600 flog_err(EC_ZEBRA_TABLE_LOOKUP_FAILED
,
2601 "%s:%u zebra_vrf_table() returned NULL", __func__
,
2606 /* No matches would be the simplest case. */
2607 if (NULL
== (rn
= route_node_lookup(table
, (struct prefix
*)p
)))
2611 route_unlock_node(rn
);
2613 dest
= rib_dest_from_rnode(rn
);
2614 /* Check all RE entries. In case any changes have to be done, requeue
2615 * the RN into RIBQ head. If the routing message about the new connected
2616 * route (generated by the IP address we are going to assign very soon)
2617 * comes before the RIBQ is processed, the new RE entry will join
2618 * RIBQ record already on head. This is necessary for proper
2620 * of the rest of the RE.
2622 if (dest
->selected_fib
&& !RIB_SYSTEM_ROUTE(dest
->selected_fib
)) {
2624 if (IS_ZEBRA_DEBUG_RIB
) {
2625 char buf
[PREFIX_STRLEN
];
2627 zlog_debug("%u:%s: freeing way for connected prefix",
2628 dest
->selected_fib
->vrf_id
,
2629 prefix2str(&rn
->p
, buf
, sizeof(buf
)));
2630 route_entry_dump(&rn
->p
, NULL
, dest
->selected_fib
);
2632 rib_uninstall(rn
, dest
->selected_fib
);
2638 int rib_add_multipath(afi_t afi
, safi_t safi
, struct prefix
*p
,
2639 struct prefix_ipv6
*src_p
, struct route_entry
*re
)
2641 struct route_table
*table
;
2642 struct route_node
*rn
;
2643 struct route_entry
*same
= NULL
;
2644 struct nexthop
*nexthop
;
2650 assert(!src_p
|| !src_p
->prefixlen
|| afi
== AFI_IP6
);
2653 table
= zebra_vrf_table_with_table_id(afi
, safi
, re
->vrf_id
, re
->table
);
2655 XFREE(MTYPE_RE
, re
);
2659 /* Make it sure prefixlen is applied to the prefix. */
2662 apply_mask_ipv6(src_p
);
2664 /* Set default distance by route type. */
2665 if (re
->distance
== 0) {
2666 re
->distance
= route_distance(re
->type
);
2668 /* iBGP distance is 200. */
2669 if (re
->type
== ZEBRA_ROUTE_BGP
2670 && CHECK_FLAG(re
->flags
, ZEBRA_FLAG_IBGP
))
2674 /* Lookup route node.*/
2675 rn
= srcdest_rnode_get(table
, p
, src_p
);
2678 * If same type of route are installed, treat it as a implicit
2680 * If the user has specified the No route replace semantics
2681 * for the install don't do a route replace.
2683 RNODE_FOREACH_RE (rn
, same
) {
2684 if (CHECK_FLAG(same
->status
, ROUTE_ENTRY_REMOVED
))
2687 if (same
->type
!= re
->type
)
2689 if (same
->instance
!= re
->instance
)
2691 if (same
->type
== ZEBRA_ROUTE_KERNEL
2692 && same
->metric
!= re
->metric
)
2695 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_RR_USE_DISTANCE
) &&
2696 same
->distance
!= re
->distance
)
2700 * We should allow duplicate connected routes
2701 * because of IPv6 link-local routes and unnumbered
2702 * interfaces on Linux.
2704 if (same
->type
!= ZEBRA_ROUTE_CONNECT
)
2708 /* If this route is kernel route, set FIB flag to the route. */
2709 if (RIB_SYSTEM_ROUTE(re
))
2710 for (nexthop
= re
->ng
.nexthop
; nexthop
; nexthop
= nexthop
->next
)
2711 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
2713 /* Link new re to node.*/
2714 if (IS_ZEBRA_DEBUG_RIB
) {
2717 "Inserting route rn %p, re %p (type %d) existing %p",
2718 (void *)rn
, (void *)re
, re
->type
, (void *)same
);
2720 if (IS_ZEBRA_DEBUG_RIB_DETAILED
)
2721 route_entry_dump(p
, src_p
, re
);
2723 rib_addnode(rn
, re
, 1);
2726 /* Free implicit route.*/
2728 rib_delnode(rn
, same
);
2732 route_unlock_node(rn
);
2736 void rib_delete(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
, int type
,
2737 unsigned short instance
, int flags
, struct prefix
*p
,
2738 struct prefix_ipv6
*src_p
, const struct nexthop
*nh
,
2739 uint32_t table_id
, uint32_t metric
, uint8_t distance
,
2742 struct route_table
*table
;
2743 struct route_node
*rn
;
2744 struct route_entry
*re
;
2745 struct route_entry
*fib
= NULL
;
2746 struct route_entry
*same
= NULL
;
2747 struct nexthop
*rtnh
;
2748 char buf2
[INET6_ADDRSTRLEN
];
2751 assert(!src_p
|| !src_p
->prefixlen
|| afi
== AFI_IP6
);
2754 table
= zebra_vrf_table_with_table_id(afi
, safi
, vrf_id
, table_id
);
2761 apply_mask_ipv6(src_p
);
2763 /* Lookup route node. */
2764 rn
= srcdest_rnode_lookup(table
, p
, src_p
);
2766 char dst_buf
[PREFIX_STRLEN
], src_buf
[PREFIX_STRLEN
];
2768 prefix2str(p
, dst_buf
, sizeof(dst_buf
));
2769 if (src_p
&& src_p
->prefixlen
)
2770 prefix2str(src_p
, src_buf
, sizeof(src_buf
));
2774 if (IS_ZEBRA_DEBUG_RIB
)
2775 zlog_debug("%u:%s%s%s doesn't exist in rib", vrf_id
,
2777 (src_buf
[0] != '\0') ? " from " : "",
2782 dest
= rib_dest_from_rnode(rn
);
2783 fib
= dest
->selected_fib
;
2785 /* Lookup same type route. */
2786 RNODE_FOREACH_RE (rn
, re
) {
2787 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
2790 if (re
->type
!= type
)
2792 if (re
->instance
!= instance
)
2794 if (CHECK_FLAG(re
->flags
, ZEBRA_FLAG_RR_USE_DISTANCE
) &&
2795 distance
!= re
->distance
)
2798 if (re
->type
== ZEBRA_ROUTE_KERNEL
&& re
->metric
!= metric
)
2800 if (re
->type
== ZEBRA_ROUTE_CONNECT
&& (rtnh
= re
->ng
.nexthop
)
2801 && rtnh
->type
== NEXTHOP_TYPE_IFINDEX
&& nh
) {
2802 if (rtnh
->ifindex
!= nh
->ifindex
)
2807 /* Make sure that the route found has the same gateway. */
2813 for (ALL_NEXTHOPS(re
->ng
, rtnh
))
2814 if (nexthop_same_no_recurse(rtnh
, nh
)) {
2822 /* If same type of route can't be found and this message is from
2826 * In the past(HA!) we could get here because
2827 * we were receiving a route delete from the
2828 * kernel and we're not marking the proto
2829 * as coming from it's appropriate originator.
2830 * Now that we are properly noticing the fact
2831 * that the kernel has deleted our route we
2832 * are not going to get called in this path
2833 * I am going to leave this here because
2834 * this might still work this way on non-linux
2835 * platforms as well as some weird state I have
2836 * not properly thought of yet.
2837 * If we can show that this code path is
2838 * dead then we can remove it.
2840 if (fib
&& CHECK_FLAG(flags
, ZEBRA_FLAG_SELFROUTE
)) {
2841 if (IS_ZEBRA_DEBUG_RIB
) {
2844 "rn %p, re %p (type %d) was deleted from kernel, adding",
2845 rn
, fib
, fib
->type
);
2849 for (rtnh
= fib
->ng
.nexthop
; rtnh
;
2851 UNSET_FLAG(rtnh
->flags
,
2855 * This is a non FRR route
2856 * as such we should mark
2859 dest
->selected_fib
= NULL
;
2861 /* This means someone else, other than Zebra,
2863 * a Zebra router from the kernel. We will add
2865 rib_install_kernel(rn
, fib
, NULL
);
2868 if (IS_ZEBRA_DEBUG_RIB
) {
2872 "via %s ifindex %d type %d "
2873 "doesn't exist in rib",
2874 inet_ntop(afi2family(afi
),
2881 "type %d doesn't exist in rib",
2884 route_unlock_node(rn
);
2890 if (fromkernel
&& CHECK_FLAG(flags
, ZEBRA_FLAG_SELFROUTE
)
2892 rib_install_kernel(rn
, same
, NULL
);
2893 route_unlock_node(rn
);
2898 if (CHECK_FLAG(flags
, ZEBRA_FLAG_EVPN_ROUTE
)) {
2899 struct nexthop
*tmp_nh
;
2901 for (ALL_NEXTHOPS(re
->ng
, tmp_nh
)) {
2902 struct ipaddr vtep_ip
;
2904 memset(&vtep_ip
, 0, sizeof(struct ipaddr
));
2905 if (afi
== AFI_IP
) {
2906 vtep_ip
.ipa_type
= IPADDR_V4
;
2907 memcpy(&(vtep_ip
.ipaddr_v4
),
2908 &(tmp_nh
->gate
.ipv4
),
2909 sizeof(struct in_addr
));
2911 vtep_ip
.ipa_type
= IPADDR_V6
;
2912 memcpy(&(vtep_ip
.ipaddr_v6
),
2913 &(tmp_nh
->gate
.ipv6
),
2914 sizeof(struct in6_addr
));
2916 zebra_vxlan_evpn_vrf_route_del(re
->vrf_id
,
2920 rib_delnode(rn
, same
);
2923 route_unlock_node(rn
);
2928 int rib_add(afi_t afi
, safi_t safi
, vrf_id_t vrf_id
, int type
,
2929 unsigned short instance
, int flags
, struct prefix
*p
,
2930 struct prefix_ipv6
*src_p
, const struct nexthop
*nh
,
2931 uint32_t table_id
, uint32_t metric
, uint32_t mtu
, uint8_t distance
,
2934 struct route_entry
*re
;
2935 struct nexthop
*nexthop
;
2937 /* Allocate new route_entry structure. */
2938 re
= XCALLOC(MTYPE_RE
, sizeof(struct route_entry
));
2940 re
->instance
= instance
;
2941 re
->distance
= distance
;
2943 re
->metric
= metric
;
2945 re
->table
= table_id
;
2946 re
->vrf_id
= vrf_id
;
2947 re
->nexthop_num
= 0;
2948 re
->uptime
= time(NULL
);
2952 nexthop
= nexthop_new();
2954 route_entry_nexthop_add(re
, nexthop
);
2956 return rib_add_multipath(afi
, safi
, p
, src_p
, re
);
2959 /* Schedule routes of a particular table (address-family) based on event. */
2960 void rib_update_table(struct route_table
*table
, rib_update_event_t event
)
2962 struct route_node
*rn
;
2963 struct route_entry
*re
, *next
;
2965 /* Walk all routes and queue for processing, if appropriate for
2966 * the trigger event.
2968 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
2970 * If we are looking at a route node and the node
2971 * has already been queued we don't
2972 * need to queue it up again
2974 if (rn
->info
&& CHECK_FLAG(rib_dest_from_rnode(rn
)->flags
,
2975 RIB_ROUTE_ANY_QUEUED
))
2978 case RIB_UPDATE_IF_CHANGE
:
2979 /* Examine all routes that won't get processed by the
2981 * triggered by nexthop evaluation (NHT). This would be
2983 * kernel and certain static routes. Note that NHT will
2985 * triggered upon an interface event as connected routes
2987 * get queued for processing.
2989 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
2992 if (re
->type
!= ZEBRA_ROUTE_SYSTEM
2993 && re
->type
!= ZEBRA_ROUTE_KERNEL
2994 && re
->type
!= ZEBRA_ROUTE_CONNECT
2995 && re
->type
!= ZEBRA_ROUTE_STATIC
)
2998 if (re
->type
!= ZEBRA_ROUTE_STATIC
) {
3003 for (nh
= re
->ng
.nexthop
; nh
; nh
= nh
->next
)
3004 if (!(nh
->type
== NEXTHOP_TYPE_IPV4
3005 || nh
->type
== NEXTHOP_TYPE_IPV6
))
3008 /* If we only have nexthops to a
3017 case RIB_UPDATE_RMAP_CHANGE
:
3018 case RIB_UPDATE_OTHER
:
3019 /* Right now, examine all routes. Can restrict to a
3021 * some cases (TODO).
3023 if (rnode_to_ribs(rn
))
3033 /* RIB update function. */
3034 void rib_update(vrf_id_t vrf_id
, rib_update_event_t event
)
3036 struct route_table
*table
;
3038 /* Process routes of interested address-families. */
3039 table
= zebra_vrf_table(AFI_IP
, SAFI_UNICAST
, vrf_id
);
3041 if (IS_ZEBRA_DEBUG_EVENT
)
3042 zlog_debug("%s : AFI_IP event %d", __func__
, event
);
3043 rib_update_table(table
, event
);
3046 table
= zebra_vrf_table(AFI_IP6
, SAFI_UNICAST
, vrf_id
);
3048 if (IS_ZEBRA_DEBUG_EVENT
)
3049 zlog_debug("%s : AFI_IP6 event %d", __func__
, event
);
3050 rib_update_table(table
, event
);
3054 /* Delete self installed routes after zebra is relaunched. */
3055 void rib_sweep_table(struct route_table
*table
)
3057 struct route_node
*rn
;
3058 struct route_entry
*re
;
3059 struct route_entry
*next
;
3060 struct nexthop
*nexthop
;
3065 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
3066 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
3067 if (IS_ZEBRA_DEBUG_RIB
)
3068 route_entry_dump(&rn
->p
, NULL
, re
);
3070 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
3073 if (!CHECK_FLAG(re
->flags
, ZEBRA_FLAG_SELFROUTE
))
3077 * So we are starting up and have received
3078 * routes from the kernel that we have installed
3079 * from a previous run of zebra but not cleaned
3080 * up ( say a kill -9 )
3081 * But since we haven't actually installed
3082 * them yet( we received them from the kernel )
3083 * we don't think they are active.
3084 * So let's pretend they are active to actually
3086 * In all honesty I'm not sure if we should
3087 * mark them as active when we receive them
3088 * This is startup only so probably ok.
3090 * If we ever decide to move rib_sweep_table
3091 * to a different spot (ie startup )
3092 * this decision needs to be revisited
3094 for (ALL_NEXTHOPS(re
->ng
, nexthop
))
3095 SET_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
);
3097 rib_uninstall_kernel(rn
, re
);
3098 rib_delnode(rn
, re
);
3103 /* Sweep all RIB tables. */
3104 void rib_sweep_route(void)
3107 struct zebra_vrf
*zvrf
;
3109 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
3110 if ((zvrf
= vrf
->info
) == NULL
)
3113 rib_sweep_table(zvrf
->table
[AFI_IP
][SAFI_UNICAST
]);
3114 rib_sweep_table(zvrf
->table
[AFI_IP6
][SAFI_UNICAST
]);
3117 zebra_router_sweep_route();
3120 /* Remove specific by protocol routes from 'table'. */
3121 unsigned long rib_score_proto_table(uint8_t proto
, unsigned short instance
,
3122 struct route_table
*table
)
3124 struct route_node
*rn
;
3125 struct route_entry
*re
;
3126 struct route_entry
*next
;
3127 unsigned long n
= 0;
3130 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
))
3131 RNODE_FOREACH_RE_SAFE (rn
, re
, next
) {
3132 if (CHECK_FLAG(re
->status
, ROUTE_ENTRY_REMOVED
))
3134 if (re
->type
== proto
3135 && re
->instance
== instance
) {
3136 rib_delnode(rn
, re
);
3143 /* Remove specific by protocol routes. */
3144 unsigned long rib_score_proto(uint8_t proto
, unsigned short instance
)
3147 struct zebra_vrf
*zvrf
;
3148 unsigned long cnt
= 0;
3150 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
)
3151 if ((zvrf
= vrf
->info
) != NULL
)
3152 cnt
+= rib_score_proto_table(
3154 zvrf
->table
[AFI_IP
][SAFI_UNICAST
])
3155 + rib_score_proto_table(
3157 zvrf
->table
[AFI_IP6
][SAFI_UNICAST
]);
3159 cnt
+= zebra_router_score_proto(proto
, instance
);
3164 /* Close RIB and clean up kernel routes. */
3165 void rib_close_table(struct route_table
*table
)
3167 struct route_node
*rn
;
3168 rib_table_info_t
*info
;
3174 info
= route_table_get_info(table
);
3176 for (rn
= route_top(table
); rn
; rn
= srcdest_route_next(rn
)) {
3177 dest
= rib_dest_from_rnode(rn
);
3179 if (dest
&& dest
->selected_fib
) {
3180 if (info
->safi
== SAFI_UNICAST
)
3181 hook_call(rib_update
, rn
, NULL
);
3183 if (!RIB_SYSTEM_ROUTE(dest
->selected_fib
)) {
3184 rib_uninstall_kernel(rn
, dest
->selected_fib
);
3185 dest
->selected_fib
= NULL
;
3194 static int rib_process_dplane_results(struct thread
*thread
)
3196 struct zebra_dplane_ctx
*ctx
;
3199 /* Take lock controlling queue of results */
3200 pthread_mutex_lock(&dplane_mutex
);
3202 /* Dequeue context block */
3203 ctx
= dplane_ctx_dequeue(&rib_dplane_q
);
3205 pthread_mutex_unlock(&dplane_mutex
);
3208 rib_process_after(ctx
);
3214 /* Check for nexthop tracking processing after finishing with results */
3215 do_nht_processing();
3221 * Results are returned from the dataplane subsystem, in the context of
3222 * the dataplane pthread. We enqueue the results here for processing by
3223 * the main thread later.
3225 static int rib_dplane_results(struct zebra_dplane_ctx
*ctx
)
3227 /* Take lock controlling queue of results */
3228 pthread_mutex_lock(&dplane_mutex
);
3230 /* Enqueue context block */
3231 dplane_ctx_enqueue_tail(&rib_dplane_q
, ctx
);
3233 pthread_mutex_unlock(&dplane_mutex
);
3235 /* Ensure event is signalled to zebra main thread */
3236 thread_add_event(zebrad
.master
, rib_process_dplane_results
, NULL
, 0,
3242 /* Routing information base initialize. */
3245 rib_queue_init(&zebrad
);
3247 /* Init dataplane, and register for results */
3248 pthread_mutex_init(&dplane_mutex
, NULL
);
3249 TAILQ_INIT(&rib_dplane_q
);
3250 zebra_dplane_init();
3251 dplane_results_register(rib_dplane_results
);
3257 * Get the first vrf id that is greater than the given vrf id if any.
3259 * Returns TRUE if a vrf id was found, FALSE otherwise.
3261 static inline int vrf_id_get_next(vrf_id_t vrf_id
, vrf_id_t
*next_id_p
)
3265 vrf
= vrf_lookup_by_id(vrf_id
);
3267 vrf
= RB_NEXT(vrf_id_head
, vrf
);
3269 *next_id_p
= vrf
->vrf_id
;
3278 * rib_tables_iter_next
3280 * Returns the next table in the iteration.
3282 struct route_table
*rib_tables_iter_next(rib_tables_iter_t
*iter
)
3284 struct route_table
*table
;
3287 * Array that helps us go over all AFI/SAFI combinations via one
3294 {AFI_IP
, SAFI_UNICAST
}, {AFI_IP
, SAFI_MULTICAST
},
3295 {AFI_IP
, SAFI_LABELED_UNICAST
}, {AFI_IP6
, SAFI_UNICAST
},
3296 {AFI_IP6
, SAFI_MULTICAST
}, {AFI_IP6
, SAFI_LABELED_UNICAST
},
3301 switch (iter
->state
) {
3303 case RIB_TABLES_ITER_S_INIT
:
3304 iter
->vrf_id
= VRF_DEFAULT
;
3305 iter
->afi_safi_ix
= -1;
3309 case RIB_TABLES_ITER_S_ITERATING
:
3310 iter
->afi_safi_ix
++;
3313 while (iter
->afi_safi_ix
3314 < (int)ZEBRA_NUM_OF(afi_safis
)) {
3315 table
= zebra_vrf_table(
3316 afi_safis
[iter
->afi_safi_ix
].afi
,
3317 afi_safis
[iter
->afi_safi_ix
].safi
,
3322 iter
->afi_safi_ix
++;
3326 * Found another table in this vrf.
3332 * Done with all tables in the current vrf, go to the
3336 if (!vrf_id_get_next(iter
->vrf_id
, &iter
->vrf_id
))
3339 iter
->afi_safi_ix
= 0;
3344 case RIB_TABLES_ITER_S_DONE
:
3349 iter
->state
= RIB_TABLES_ITER_S_ITERATING
;
3351 iter
->state
= RIB_TABLES_ITER_S_DONE
;