1 /* Kernel routing table updates using netlink over GNU/Linux system.
2 * Copyright (C) 1997, 98, 99 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
25 #include <net/if_arp.h>
26 #include <linux/lwtunnel.h>
27 #include <linux/mpls_iptunnel.h>
28 #include <linux/neighbour.h>
29 #include <linux/rtnetlink.h>
31 /* Hack for GNU libc version 2. */
33 #define MSG_TRUNC 0x20
34 #endif /* MSG_TRUNC */
40 #include "connected.h"
43 #include "zebra_memory.h"
53 #include "zebra/zapi_msg.h"
54 #include "zebra/zebra_ns.h"
55 #include "zebra/zebra_vrf.h"
57 #include "zebra/redistribute.h"
58 #include "zebra/interface.h"
59 #include "zebra/debug.h"
60 #include "zebra/rtadv.h"
61 #include "zebra/zebra_ptm.h"
62 #include "zebra/zebra_mpls.h"
63 #include "zebra/kernel_netlink.h"
64 #include "zebra/rt_netlink.h"
65 #include "zebra/zebra_mroute.h"
66 #include "zebra/zebra_vxlan.h"
72 static vlanid_t filter_vlan
= 0;
80 char ipv4_ll_buf
[16] = "169.254.0.1";
81 struct in_addr ipv4_ll
;
84 * The ipv4_ll data structure is used for all 5549
85 * additions to the kernel. Let's figure out the
86 * correct value one time instead for every
87 * install/remove of a 5549 type route
89 void rt_netlink_init(void)
91 inet_pton(AF_INET
, ipv4_ll_buf
, &ipv4_ll
);
94 static inline int is_selfroute(int proto
)
96 if ((proto
== RTPROT_BGP
) || (proto
== RTPROT_OSPF
)
97 || (proto
== RTPROT_ZSTATIC
) || (proto
== RTPROT_ZEBRA
)
98 || (proto
== RTPROT_ISIS
) || (proto
== RTPROT_RIPNG
)
99 || (proto
== RTPROT_NHRP
) || (proto
== RTPROT_EIGRP
)
100 || (proto
== RTPROT_LDP
) || (proto
== RTPROT_BABEL
)
101 || (proto
== RTPROT_RIP
) || (proto
== RTPROT_SHARP
)
102 || (proto
== RTPROT_PBR
)) {
109 static inline int zebra2proto(int proto
)
112 case ZEBRA_ROUTE_BABEL
:
113 proto
= RTPROT_BABEL
;
115 case ZEBRA_ROUTE_BGP
:
118 case ZEBRA_ROUTE_OSPF
:
119 case ZEBRA_ROUTE_OSPF6
:
122 case ZEBRA_ROUTE_STATIC
:
123 proto
= RTPROT_ZSTATIC
;
125 case ZEBRA_ROUTE_ISIS
:
128 case ZEBRA_ROUTE_RIP
:
131 case ZEBRA_ROUTE_RIPNG
:
132 proto
= RTPROT_RIPNG
;
134 case ZEBRA_ROUTE_NHRP
:
137 case ZEBRA_ROUTE_EIGRP
:
138 proto
= RTPROT_EIGRP
;
140 case ZEBRA_ROUTE_LDP
:
143 case ZEBRA_ROUTE_SHARP
:
144 proto
= RTPROT_SHARP
;
146 case ZEBRA_ROUTE_PBR
:
151 * When a user adds a new protocol this will show up
152 * to let them know to do something about it. This
153 * is intentionally a warn because we should see
154 * this as part of development of a new protocol
156 zlog_warn("%s: Please add this protocol(%d) to proper rt_netlink.c handling",
157 __PRETTY_FUNCTION__
, proto
);
158 proto
= RTPROT_ZEBRA
;
165 static inline int proto2zebra(int proto
, int family
)
169 proto
= ZEBRA_ROUTE_BABEL
;
172 proto
= ZEBRA_ROUTE_BGP
;
175 proto
= (family
== AFI_IP
) ? ZEBRA_ROUTE_OSPF
179 proto
= ZEBRA_ROUTE_ISIS
;
182 proto
= ZEBRA_ROUTE_RIP
;
185 proto
= ZEBRA_ROUTE_RIPNG
;
188 proto
= ZEBRA_ROUTE_NHRP
;
191 proto
= ZEBRA_ROUTE_EIGRP
;
194 proto
= ZEBRA_ROUTE_LDP
;
198 proto
= ZEBRA_ROUTE_STATIC
;
201 proto
= ZEBRA_ROUTE_SHARP
;
204 proto
= ZEBRA_ROUTE_PBR
;
208 * When a user adds a new protocol this will show up
209 * to let them know to do something about it. This
210 * is intentionally a warn because we should see
211 * this as part of development of a new protocol
213 zlog_warn("%s: Please add this protocol(%d) to proper rt_netlink.c handling",
216 proto
= ZEBRA_ROUTE_KERNEL
;
223 Pending: create an efficient table_id (in a tree/hash) based lookup)
225 static vrf_id_t
vrf_lookup_by_table(uint32_t table_id
, ns_id_t ns_id
)
228 struct zebra_vrf
*zvrf
;
230 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
234 /* case vrf with netns : match the netnsid */
235 if (vrf_is_backend_netns()) {
236 if (ns_id
== zvrf_id(zvrf
))
237 return zvrf_id(zvrf
);
239 /* VRF is VRF_BACKEND_VRF_LITE */
240 if (zvrf
->table_id
!= table_id
)
242 return zvrf_id(zvrf
);
249 /* Looking up routing table by netlink interface. */
250 static int netlink_route_change_read_unicast(struct nlmsghdr
*h
, ns_id_t ns_id
,
255 struct rtattr
*tb
[RTA_MAX
+ 1];
258 struct prefix_ipv6 src_p
= {};
261 char anyaddr
[16] = {0};
263 int proto
= ZEBRA_ROUTE_KERNEL
;
268 uint8_t distance
= 0;
273 void *prefsrc
= NULL
; /* IPv4 preferred source host address */
274 void *src
= NULL
; /* IPv6 srcdest source prefix */
275 enum blackhole_type bh_type
= BLACKHOLE_UNSPEC
;
279 if (startup
&& h
->nlmsg_type
!= RTM_NEWROUTE
)
281 switch (rtm
->rtm_type
) {
285 bh_type
= BLACKHOLE_NULL
;
287 case RTN_UNREACHABLE
:
288 bh_type
= BLACKHOLE_REJECT
;
291 bh_type
= BLACKHOLE_ADMINPROHIB
;
297 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
301 memset(tb
, 0, sizeof tb
);
302 netlink_parse_rtattr(tb
, RTA_MAX
, RTM_RTA(rtm
), len
);
304 if (rtm
->rtm_flags
& RTM_F_CLONED
)
306 if (rtm
->rtm_protocol
== RTPROT_REDIRECT
)
308 if (rtm
->rtm_protocol
== RTPROT_KERNEL
)
311 if (!startup
&& is_selfroute(rtm
->rtm_protocol
)
312 && h
->nlmsg_type
== RTM_NEWROUTE
) {
313 if (IS_ZEBRA_DEBUG_KERNEL
)
314 zlog_debug("Route type: %d Received that we think we have originated, ignoring",
319 /* We don't care about change notifications for the MPLS table. */
320 /* TODO: Revisit this. */
321 if (rtm
->rtm_family
== AF_MPLS
)
324 /* Table corresponding to route. */
326 table
= *(int *)RTA_DATA(tb
[RTA_TABLE
]);
328 table
= rtm
->rtm_table
;
331 vrf_id
= vrf_lookup_by_table(table
, ns_id
);
332 if (vrf_id
== VRF_DEFAULT
) {
333 if (!is_zebra_valid_kernel_table(table
)
334 && !is_zebra_main_routing_table(table
))
338 /* Route which inserted by Zebra. */
339 if (is_selfroute(rtm
->rtm_protocol
)) {
340 flags
|= ZEBRA_FLAG_SELFROUTE
;
341 proto
= proto2zebra(rtm
->rtm_protocol
, rtm
->rtm_family
);
344 index
= *(int *)RTA_DATA(tb
[RTA_OIF
]);
347 dest
= RTA_DATA(tb
[RTA_DST
]);
352 src
= RTA_DATA(tb
[RTA_SRC
]);
357 prefsrc
= RTA_DATA(tb
[RTA_PREFSRC
]);
360 gate
= RTA_DATA(tb
[RTA_GATEWAY
]);
362 if (tb
[RTA_PRIORITY
])
363 metric
= *(int *)RTA_DATA(tb
[RTA_PRIORITY
]);
365 #if defined(SUPPORT_REALMS)
367 tag
= *(uint32_t *)RTA_DATA(tb
[RTA_FLOW
]);
370 if (tb
[RTA_METRICS
]) {
371 struct rtattr
*mxrta
[RTAX_MAX
+ 1];
373 memset(mxrta
, 0, sizeof mxrta
);
374 netlink_parse_rtattr(mxrta
, RTAX_MAX
, RTA_DATA(tb
[RTA_METRICS
]),
375 RTA_PAYLOAD(tb
[RTA_METRICS
]));
378 mtu
= *(uint32_t *)RTA_DATA(mxrta
[RTAX_MTU
]);
381 if (rtm
->rtm_family
== AF_INET
) {
383 memcpy(&p
.u
.prefix4
, dest
, 4);
384 p
.prefixlen
= rtm
->rtm_dst_len
;
387 0; // Forces debug below to not display anything
388 } else if (rtm
->rtm_family
== AF_INET6
) {
390 memcpy(&p
.u
.prefix6
, dest
, 16);
391 p
.prefixlen
= rtm
->rtm_dst_len
;
393 src_p
.family
= AF_INET6
;
394 memcpy(&src_p
.prefix
, src
, 16);
395 src_p
.prefixlen
= rtm
->rtm_src_len
;
398 if (rtm
->rtm_src_len
!= 0) {
399 char buf
[PREFIX_STRLEN
];
401 "unsupported IPv[4|6] sourcedest route (dest %s vrf %u)",
402 prefix2str(&p
, buf
, sizeof(buf
)), vrf_id
);
407 * For ZEBRA_ROUTE_KERNEL types:
409 * The metric/priority of the route received from the kernel
410 * is a 32 bit number. We are going to interpret the high
411 * order byte as the Admin Distance and the low order 3 bytes
414 * This will allow us to do two things:
415 * 1) Allow the creation of kernel routes that can be
416 * overridden by zebra.
417 * 2) Allow the old behavior for 'most' kernel route types
418 * if a user enters 'ip route ...' v4 routes get a metric
419 * of 0 and v6 routes get a metric of 1024. Both of these
420 * values will end up with a admin distance of 0, which
421 * will cause them to win for the purposes of zebra.
423 if (proto
== ZEBRA_ROUTE_KERNEL
) {
424 distance
= (metric
>> 24) & 0xFF;
425 metric
= (metric
& 0x00FFFFFF);
428 if (IS_ZEBRA_DEBUG_KERNEL
) {
429 char buf
[PREFIX_STRLEN
];
430 char buf2
[PREFIX_STRLEN
];
431 zlog_debug("%s %s%s%s vrf %u(%u) metric: %d Admin Distance: %d",
432 nl_msg_type_to_str(h
->nlmsg_type
),
433 prefix2str(&p
, buf
, sizeof(buf
)),
434 src_p
.prefixlen
? " from " : "",
436 ? prefix2str(&src_p
, buf2
, sizeof(buf2
))
438 vrf_id
, table
, metric
, distance
);
442 if (rtm
->rtm_family
== AF_INET6
)
445 if (h
->nlmsg_type
== RTM_NEWROUTE
) {
446 struct interface
*ifp
;
447 vrf_id_t nh_vrf_id
= vrf_id
;
449 if (!tb
[RTA_MULTIPATH
]) {
451 size_t sz
= (afi
== AFI_IP
) ? 4 : 16;
453 memset(&nh
, 0, sizeof(nh
));
455 if (bh_type
== BLACKHOLE_UNSPEC
) {
457 nh
.type
= NEXTHOP_TYPE_IFINDEX
;
458 else if (index
&& gate
)
461 ? NEXTHOP_TYPE_IPV4_IFINDEX
462 : NEXTHOP_TYPE_IPV6_IFINDEX
;
463 else if (!index
&& gate
)
464 nh
.type
= (afi
== AFI_IP
)
468 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
469 nh
.bh_type
= bh_type
;
472 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
473 nh
.bh_type
= bh_type
;
477 memcpy(&nh
.src
, prefsrc
, sz
);
479 memcpy(&nh
.gate
, gate
, sz
);
482 ifp
= if_lookup_by_index(index
, VRF_UNKNOWN
);
484 nh_vrf_id
= ifp
->vrf_id
;
486 nh
.vrf_id
= nh_vrf_id
;
488 rib_add(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
, &p
,
489 NULL
, &nh
, table
, metric
, mtu
, distance
, tag
);
491 /* This is a multipath route */
493 struct route_entry
*re
;
494 struct rtnexthop
*rtnh
=
495 (struct rtnexthop
*)RTA_DATA(tb
[RTA_MULTIPATH
]);
497 len
= RTA_PAYLOAD(tb
[RTA_MULTIPATH
]);
499 re
= XCALLOC(MTYPE_RE
, sizeof(struct route_entry
));
501 re
->distance
= distance
;
508 re
->uptime
= time(NULL
);
513 if (len
< (int)sizeof(*rtnh
)
514 || rtnh
->rtnh_len
> len
)
517 index
= rtnh
->rtnh_ifindex
;
520 * Yes we are looking this up
521 * for every nexthop and just
522 * using the last one looked
525 ifp
= if_lookup_by_index(index
,
528 nh_vrf_id
= ifp
->vrf_id
;
531 "%s: Unknown interface %u specified, defaulting to VRF_DEFAULT",
534 nh_vrf_id
= VRF_DEFAULT
;
540 if (rtnh
->rtnh_len
> sizeof(*rtnh
)) {
541 memset(tb
, 0, sizeof(tb
));
542 netlink_parse_rtattr(
543 tb
, RTA_MAX
, RTNH_DATA(rtnh
),
544 rtnh
->rtnh_len
- sizeof(*rtnh
));
551 if (rtm
->rtm_family
== AF_INET
) {
553 route_entry_nexthop_ipv4_ifindex_add(
558 route_entry_nexthop_ipv4_add(
562 } else if (rtm
->rtm_family
565 route_entry_nexthop_ipv6_ifindex_add(
569 route_entry_nexthop_ipv6_add(
574 route_entry_nexthop_ifindex_add(
575 re
, index
, nh_vrf_id
);
577 len
-= NLMSG_ALIGN(rtnh
->rtnh_len
);
578 rtnh
= RTNH_NEXT(rtnh
);
581 zserv_nexthop_num_warn(__func__
,
582 (const struct prefix
*)&p
,
584 if (re
->nexthop_num
== 0)
587 rib_add_multipath(afi
, SAFI_UNICAST
, &p
, NULL
,
591 if (!tb
[RTA_MULTIPATH
]) {
593 size_t sz
= (afi
== AFI_IP
) ? 4 : 16;
595 memset(&nh
, 0, sizeof(nh
));
596 if (bh_type
== BLACKHOLE_UNSPEC
) {
598 nh
.type
= NEXTHOP_TYPE_IFINDEX
;
599 else if (index
&& gate
)
602 ? NEXTHOP_TYPE_IPV4_IFINDEX
603 : NEXTHOP_TYPE_IPV6_IFINDEX
;
604 else if (!index
&& gate
)
605 nh
.type
= (afi
== AFI_IP
)
609 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
610 nh
.bh_type
= BLACKHOLE_UNSPEC
;
613 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
614 nh
.bh_type
= bh_type
;
618 memcpy(&nh
.gate
, gate
, sz
);
619 rib_delete(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
,
620 &p
, NULL
, &nh
, table
, metric
, true);
622 /* XXX: need to compare the entire list of nexthops
623 * here for NLM_F_APPEND stupidity */
624 rib_delete(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
,
625 &p
, NULL
, NULL
, table
, metric
, true);
632 static struct mcast_route_data
*mroute
= NULL
;
634 static int netlink_route_change_read_multicast(struct nlmsghdr
*h
,
635 ns_id_t ns_id
, int startup
)
639 struct rtattr
*tb
[RTA_MAX
+ 1];
640 struct mcast_route_data
*m
;
641 struct mcast_route_data mr
;
648 char oif_list
[256] = "\0";
655 memset(&mr
, 0, sizeof(mr
));
661 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
663 memset(tb
, 0, sizeof tb
);
664 netlink_parse_rtattr(tb
, RTA_MAX
, RTM_RTA(rtm
), len
);
667 table
= *(int *)RTA_DATA(tb
[RTA_TABLE
]);
669 table
= rtm
->rtm_table
;
671 vrf
= vrf_lookup_by_table(table
, ns_id
);
674 iif
= *(int *)RTA_DATA(tb
[RTA_IIF
]);
677 m
->sg
.src
= *(struct in_addr
*)RTA_DATA(tb
[RTA_SRC
]);
680 m
->sg
.grp
= *(struct in_addr
*)RTA_DATA(tb
[RTA_DST
]);
682 if ((RTA_EXPIRES
<= RTA_MAX
) && tb
[RTA_EXPIRES
])
683 m
->lastused
= *(unsigned long long *)RTA_DATA(tb
[RTA_EXPIRES
]);
685 if (tb
[RTA_MULTIPATH
]) {
686 struct rtnexthop
*rtnh
=
687 (struct rtnexthop
*)RTA_DATA(tb
[RTA_MULTIPATH
]);
689 len
= RTA_PAYLOAD(tb
[RTA_MULTIPATH
]);
691 if (len
< (int)sizeof(*rtnh
) || rtnh
->rtnh_len
> len
)
694 oif
[oif_count
] = rtnh
->rtnh_ifindex
;
697 len
-= NLMSG_ALIGN(rtnh
->rtnh_len
);
698 rtnh
= RTNH_NEXT(rtnh
);
702 if (IS_ZEBRA_DEBUG_KERNEL
) {
703 struct interface
*ifp
;
704 strlcpy(sbuf
, inet_ntoa(m
->sg
.src
), sizeof(sbuf
));
705 strlcpy(gbuf
, inet_ntoa(m
->sg
.grp
), sizeof(gbuf
));
706 for (count
= 0; count
< oif_count
; count
++) {
707 ifp
= if_lookup_by_index(oif
[count
], vrf
);
710 sprintf(temp
, "%s ", ifp
->name
);
711 strcat(oif_list
, temp
);
713 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(vrf
);
714 ifp
= if_lookup_by_index(iif
, vrf
);
716 "MCAST VRF: %s(%d) %s (%s,%s) IIF: %s OIF: %s jiffies: %lld",
717 zvrf
->vrf
->name
, vrf
, nl_msg_type_to_str(h
->nlmsg_type
),
718 sbuf
, gbuf
, ifp
->name
, oif_list
, m
->lastused
);
723 int netlink_route_change(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
730 if (!(h
->nlmsg_type
== RTM_NEWROUTE
|| h
->nlmsg_type
== RTM_DELROUTE
)) {
731 /* If this is not route add/delete message print warning. */
732 zlog_warn("Kernel message: %d NS %u\n", h
->nlmsg_type
, ns_id
);
736 /* Connected route. */
737 if (IS_ZEBRA_DEBUG_KERNEL
)
738 zlog_debug("%s %s %s proto %s NS %u",
739 nl_msg_type_to_str(h
->nlmsg_type
),
740 nl_family_to_str(rtm
->rtm_family
),
741 nl_rttype_to_str(rtm
->rtm_type
),
742 nl_rtproto_to_str(rtm
->rtm_protocol
), ns_id
);
744 /* We don't care about change notifications for the MPLS table. */
745 /* TODO: Revisit this. */
746 if (rtm
->rtm_family
== AF_MPLS
)
749 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
753 if (rtm
->rtm_type
== RTN_MULTICAST
)
754 netlink_route_change_read_multicast(h
, ns_id
, startup
);
756 netlink_route_change_read_unicast(h
, ns_id
, startup
);
760 /* Request for specific route information from the kernel */
761 static int netlink_request_route(struct zebra_ns
*zns
, int family
, int type
)
768 /* Form the request, specifying filter (rtattr) if needed. */
769 memset(&req
, 0, sizeof(req
));
770 req
.n
.nlmsg_type
= type
;
771 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
772 req
.rtm
.rtm_family
= family
;
774 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
777 /* Routing table read function using netlink interface. Only called
779 int netlink_route_read(struct zebra_ns
*zns
)
783 /* Get IPv4 routing table. */
784 ret
= netlink_request_route(zns
, AF_INET
, RTM_GETROUTE
);
787 ret
= netlink_parse_info(netlink_route_change_read_unicast
,
788 &zns
->netlink_cmd
, zns
, 0, 1);
792 /* Get IPv6 routing table. */
793 ret
= netlink_request_route(zns
, AF_INET6
, RTM_GETROUTE
);
796 ret
= netlink_parse_info(netlink_route_change_read_unicast
,
797 &zns
->netlink_cmd
, zns
, 0, 1);
804 static void _netlink_route_nl_add_gateway_info(uint8_t route_family
,
806 struct nlmsghdr
*nlmsg
,
807 size_t req_size
, int bytelen
,
808 struct nexthop
*nexthop
)
810 if (route_family
== AF_MPLS
) {
811 struct gw_family_t gw_fam
;
813 gw_fam
.family
= gw_family
;
814 if (gw_family
== AF_INET
)
815 memcpy(&gw_fam
.gate
.ipv4
, &nexthop
->gate
.ipv4
, bytelen
);
817 memcpy(&gw_fam
.gate
.ipv6
, &nexthop
->gate
.ipv6
, bytelen
);
818 addattr_l(nlmsg
, req_size
, RTA_VIA
, &gw_fam
.family
,
821 if (gw_family
== AF_INET
)
822 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
,
823 &nexthop
->gate
.ipv4
, bytelen
);
825 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
,
826 &nexthop
->gate
.ipv6
, bytelen
);
830 static void _netlink_route_rta_add_gateway_info(uint8_t route_family
,
833 struct rtnexthop
*rtnh
,
834 size_t req_size
, int bytelen
,
835 struct nexthop
*nexthop
)
837 if (route_family
== AF_MPLS
) {
838 struct gw_family_t gw_fam
;
840 gw_fam
.family
= gw_family
;
841 if (gw_family
== AF_INET
)
842 memcpy(&gw_fam
.gate
.ipv4
, &nexthop
->gate
.ipv4
, bytelen
);
844 memcpy(&gw_fam
.gate
.ipv6
, &nexthop
->gate
.ipv6
, bytelen
);
845 rta_addattr_l(rta
, req_size
, RTA_VIA
, &gw_fam
.family
,
847 rtnh
->rtnh_len
+= RTA_LENGTH(bytelen
+ 2);
849 if (gw_family
== AF_INET
)
850 rta_addattr_l(rta
, req_size
, RTA_GATEWAY
,
851 &nexthop
->gate
.ipv4
, bytelen
);
853 rta_addattr_l(rta
, req_size
, RTA_GATEWAY
,
854 &nexthop
->gate
.ipv6
, bytelen
);
855 rtnh
->rtnh_len
+= sizeof(struct rtattr
) + bytelen
;
859 /* This function takes a nexthop as argument and adds
860 * the appropriate netlink attributes to an existing
863 * @param routedesc: Human readable description of route type
864 * (direct/recursive, single-/multipath)
865 * @param bytelen: Length of addresses in bytes.
866 * @param nexthop: Nexthop information
867 * @param nlmsg: nlmsghdr structure to fill in.
868 * @param req_size: The size allocated for the message.
870 static void _netlink_route_build_singlepath(const char *routedesc
, int bytelen
,
871 struct nexthop
*nexthop
,
872 struct nlmsghdr
*nlmsg
,
874 size_t req_size
, int cmd
)
876 struct mpls_label_stack
*nh_label
;
877 mpls_lse_t out_lse
[MPLS_MAX_LABELS
];
882 * label_buf is *only* currently used within debugging.
883 * As such when we assign it we are guarding it inside
884 * a debug test. If you want to change this make sure
885 * you fix this assumption
890 for (struct nexthop
*nh
= nexthop
; nh
; nh
= nh
->rparent
) {
893 nh_label
= nh
->nh_label
;
894 if (!nh_label
|| !nh_label
->num_labels
)
897 for (int i
= 0; i
< nh_label
->num_labels
; i
++) {
898 if (nh_label
->label
[i
] == MPLS_LABEL_IMPLICIT_NULL
)
901 if (IS_ZEBRA_DEBUG_KERNEL
) {
903 sprintf(label_buf
, "label %u",
906 sprintf(label_buf1
, "/%u",
908 strlcat(label_buf
, label_buf1
,
913 out_lse
[num_labels
] =
914 mpls_lse_encode(nh_label
->label
[i
], 0, 0, 0);
920 /* Set the BoS bit */
921 out_lse
[num_labels
- 1] |= htonl(1 << MPLS_LS_S_SHIFT
);
923 if (rtmsg
->rtm_family
== AF_MPLS
)
924 addattr_l(nlmsg
, req_size
, RTA_NEWDST
, &out_lse
,
925 num_labels
* sizeof(mpls_lse_t
));
928 uint16_t encap
= LWTUNNEL_ENCAP_MPLS
;
930 addattr_l(nlmsg
, req_size
, RTA_ENCAP_TYPE
, &encap
,
932 nest
= addattr_nest(nlmsg
, req_size
, RTA_ENCAP
);
933 addattr_l(nlmsg
, req_size
, MPLS_IPTUNNEL_DST
, &out_lse
,
934 num_labels
* sizeof(mpls_lse_t
));
935 addattr_nest_end(nlmsg
, nest
);
939 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
))
940 rtmsg
->rtm_flags
|= RTNH_F_ONLINK
;
942 if (rtmsg
->rtm_family
== AF_INET
943 && (nexthop
->type
== NEXTHOP_TYPE_IPV6
944 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
)) {
945 rtmsg
->rtm_flags
|= RTNH_F_ONLINK
;
946 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
, &ipv4_ll
, 4);
947 addattr32(nlmsg
, req_size
, RTA_OIF
, nexthop
->ifindex
);
949 if (nexthop
->rmap_src
.ipv4
.s_addr
&& (cmd
== RTM_NEWROUTE
))
950 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
951 &nexthop
->rmap_src
.ipv4
, bytelen
);
952 else if (nexthop
->src
.ipv4
.s_addr
&& (cmd
== RTM_NEWROUTE
))
953 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
954 &nexthop
->src
.ipv4
, bytelen
);
956 if (IS_ZEBRA_DEBUG_KERNEL
)
958 " 5549: _netlink_route_build_singlepath() (%s): "
959 "nexthop via %s %s if %u(%u)",
960 routedesc
, ipv4_ll_buf
, label_buf
,
961 nexthop
->ifindex
, nexthop
->vrf_id
);
965 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
966 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
967 /* Send deletes to the kernel without specifying the next-hop */
968 if (cmd
!= RTM_DELROUTE
)
969 _netlink_route_nl_add_gateway_info(
970 rtmsg
->rtm_family
, AF_INET
, nlmsg
, req_size
,
973 if (cmd
== RTM_NEWROUTE
) {
974 if (nexthop
->rmap_src
.ipv4
.s_addr
)
975 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
976 &nexthop
->rmap_src
.ipv4
, bytelen
);
977 else if (nexthop
->src
.ipv4
.s_addr
)
978 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
979 &nexthop
->src
.ipv4
, bytelen
);
982 if (IS_ZEBRA_DEBUG_KERNEL
)
984 "netlink_route_multipath() (%s): "
985 "nexthop via %s %s if %u(%u)",
986 routedesc
, inet_ntoa(nexthop
->gate
.ipv4
),
987 label_buf
, nexthop
->ifindex
, nexthop
->vrf_id
);
990 if (nexthop
->type
== NEXTHOP_TYPE_IPV6
991 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
992 _netlink_route_nl_add_gateway_info(rtmsg
->rtm_family
, AF_INET6
,
993 nlmsg
, req_size
, bytelen
,
996 if (cmd
== RTM_NEWROUTE
) {
997 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
998 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
999 &nexthop
->rmap_src
.ipv6
, bytelen
);
1000 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1001 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1002 &nexthop
->src
.ipv6
, bytelen
);
1005 if (IS_ZEBRA_DEBUG_KERNEL
)
1007 "netlink_route_multipath() (%s): "
1008 "nexthop via %s %s if %u(%u)",
1009 routedesc
, inet6_ntoa(nexthop
->gate
.ipv6
),
1010 label_buf
, nexthop
->ifindex
, nexthop
->vrf_id
);
1014 * We have the ifindex so we should always send it
1015 * This is especially useful if we are doing route
1018 if (nexthop
->type
!= NEXTHOP_TYPE_BLACKHOLE
)
1019 addattr32(nlmsg
, req_size
, RTA_OIF
, nexthop
->ifindex
);
1021 if (nexthop
->type
== NEXTHOP_TYPE_IFINDEX
1022 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
1023 if (cmd
== RTM_NEWROUTE
) {
1024 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1025 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1026 &nexthop
->rmap_src
.ipv4
, bytelen
);
1027 else if (nexthop
->src
.ipv4
.s_addr
)
1028 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1029 &nexthop
->src
.ipv4
, bytelen
);
1032 if (IS_ZEBRA_DEBUG_KERNEL
)
1034 "netlink_route_multipath() (%s): "
1035 "nexthop via if %u(%u)",
1036 routedesc
, nexthop
->ifindex
, nexthop
->vrf_id
);
1039 if (nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1040 if (cmd
== RTM_NEWROUTE
) {
1041 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1042 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1043 &nexthop
->rmap_src
.ipv6
, bytelen
);
1044 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1045 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1046 &nexthop
->src
.ipv6
, bytelen
);
1049 if (IS_ZEBRA_DEBUG_KERNEL
)
1051 "netlink_route_multipath() (%s): "
1052 "nexthop via if %u(%u)",
1053 routedesc
, nexthop
->ifindex
, nexthop
->vrf_id
);
1057 /* This function takes a nexthop as argument and
1058 * appends to the given rtattr/rtnexthop pair the
1059 * representation of the nexthop. If the nexthop
1060 * defines a preferred source, the src parameter
1061 * will be modified to point to that src, otherwise
1062 * it will be kept unmodified.
1064 * @param routedesc: Human readable description of route type
1065 * (direct/recursive, single-/multipath)
1066 * @param bytelen: Length of addresses in bytes.
1067 * @param nexthop: Nexthop information
1068 * @param rta: rtnetlink attribute structure
1069 * @param rtnh: pointer to an rtnetlink nexthop structure
1070 * @param src: pointer pointing to a location where
1071 * the prefsrc should be stored.
1073 static void _netlink_route_build_multipath(const char *routedesc
, int bytelen
,
1074 struct nexthop
*nexthop
,
1076 struct rtnexthop
*rtnh
,
1077 struct rtmsg
*rtmsg
,
1080 struct mpls_label_stack
*nh_label
;
1081 mpls_lse_t out_lse
[MPLS_MAX_LABELS
];
1083 char label_buf
[256];
1085 rtnh
->rtnh_len
= sizeof(*rtnh
);
1086 rtnh
->rtnh_flags
= 0;
1087 rtnh
->rtnh_hops
= 0;
1088 rta
->rta_len
+= rtnh
->rtnh_len
;
1091 * label_buf is *only* currently used within debugging.
1092 * As such when we assign it we are guarding it inside
1093 * a debug test. If you want to change this make sure
1094 * you fix this assumption
1096 label_buf
[0] = '\0';
1099 for (struct nexthop
*nh
= nexthop
; nh
; nh
= nh
->rparent
) {
1100 char label_buf1
[20];
1102 nh_label
= nh
->nh_label
;
1103 if (!nh_label
|| !nh_label
->num_labels
)
1106 for (int i
= 0; i
< nh_label
->num_labels
; i
++) {
1107 if (nh_label
->label
[i
] == MPLS_LABEL_IMPLICIT_NULL
)
1110 if (IS_ZEBRA_DEBUG_KERNEL
) {
1112 sprintf(label_buf
, "label %u",
1113 nh_label
->label
[i
]);
1115 sprintf(label_buf1
, "/%u",
1116 nh_label
->label
[i
]);
1117 strlcat(label_buf
, label_buf1
,
1122 out_lse
[num_labels
] =
1123 mpls_lse_encode(nh_label
->label
[i
], 0, 0, 0);
1129 /* Set the BoS bit */
1130 out_lse
[num_labels
- 1] |= htonl(1 << MPLS_LS_S_SHIFT
);
1132 if (rtmsg
->rtm_family
== AF_MPLS
) {
1133 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_NEWDST
,
1135 num_labels
* sizeof(mpls_lse_t
));
1137 RTA_LENGTH(num_labels
* sizeof(mpls_lse_t
));
1139 struct rtattr
*nest
;
1140 uint16_t encap
= LWTUNNEL_ENCAP_MPLS
;
1141 int len
= rta
->rta_len
;
1143 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_ENCAP_TYPE
,
1144 &encap
, sizeof(uint16_t));
1145 nest
= rta_nest(rta
, NL_PKT_BUF_SIZE
, RTA_ENCAP
);
1146 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, MPLS_IPTUNNEL_DST
,
1148 num_labels
* sizeof(mpls_lse_t
));
1149 rta_nest_end(rta
, nest
);
1150 rtnh
->rtnh_len
+= rta
->rta_len
- len
;
1154 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
))
1155 rtnh
->rtnh_flags
|= RTNH_F_ONLINK
;
1157 if (rtmsg
->rtm_family
== AF_INET
1158 && (nexthop
->type
== NEXTHOP_TYPE_IPV6
1159 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
)) {
1161 rtnh
->rtnh_flags
|= RTNH_F_ONLINK
;
1162 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_GATEWAY
, &ipv4_ll
,
1164 rtnh
->rtnh_len
+= sizeof(struct rtattr
) + bytelen
;
1165 rtnh
->rtnh_ifindex
= nexthop
->ifindex
;
1167 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1168 *src
= &nexthop
->rmap_src
;
1169 else if (nexthop
->src
.ipv4
.s_addr
)
1170 *src
= &nexthop
->src
;
1172 if (IS_ZEBRA_DEBUG_KERNEL
)
1174 " 5549: netlink_route_build_multipath() (%s): "
1175 "nexthop via %s %s if %u",
1176 routedesc
, ipv4_ll_buf
, label_buf
,
1181 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
1182 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
1183 _netlink_route_rta_add_gateway_info(rtmsg
->rtm_family
, AF_INET
,
1184 rta
, rtnh
, NL_PKT_BUF_SIZE
,
1186 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1187 *src
= &nexthop
->rmap_src
;
1188 else if (nexthop
->src
.ipv4
.s_addr
)
1189 *src
= &nexthop
->src
;
1191 if (IS_ZEBRA_DEBUG_KERNEL
)
1193 "netlink_route_multipath() (%s): "
1194 "nexthop via %s %s if %u",
1195 routedesc
, inet_ntoa(nexthop
->gate
.ipv4
),
1196 label_buf
, nexthop
->ifindex
);
1198 if (nexthop
->type
== NEXTHOP_TYPE_IPV6
1199 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1200 _netlink_route_rta_add_gateway_info(rtmsg
->rtm_family
, AF_INET6
,
1201 rta
, rtnh
, NL_PKT_BUF_SIZE
,
1204 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1205 *src
= &nexthop
->rmap_src
;
1206 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1207 *src
= &nexthop
->src
;
1209 if (IS_ZEBRA_DEBUG_KERNEL
)
1211 "netlink_route_multipath() (%s): "
1212 "nexthop via %s %s if %u",
1213 routedesc
, inet6_ntoa(nexthop
->gate
.ipv6
),
1214 label_buf
, nexthop
->ifindex
);
1218 * We have figured out the ifindex so we should always send it
1219 * This is especially useful if we are doing route
1222 if (nexthop
->type
!= NEXTHOP_TYPE_BLACKHOLE
)
1223 rtnh
->rtnh_ifindex
= nexthop
->ifindex
;
1226 if (nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
1227 || nexthop
->type
== NEXTHOP_TYPE_IFINDEX
) {
1228 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1229 *src
= &nexthop
->rmap_src
;
1230 else if (nexthop
->src
.ipv4
.s_addr
)
1231 *src
= &nexthop
->src
;
1233 if (IS_ZEBRA_DEBUG_KERNEL
)
1235 "netlink_route_multipath() (%s): "
1236 "nexthop via if %u",
1237 routedesc
, nexthop
->ifindex
);
1241 static inline void _netlink_mpls_build_singlepath(const char *routedesc
,
1242 zebra_nhlfe_t
*nhlfe
,
1243 struct nlmsghdr
*nlmsg
,
1244 struct rtmsg
*rtmsg
,
1245 size_t req_size
, int cmd
)
1250 family
= NHLFE_FAMILY(nhlfe
);
1251 bytelen
= (family
== AF_INET
? 4 : 16);
1252 _netlink_route_build_singlepath(routedesc
, bytelen
, nhlfe
->nexthop
,
1253 nlmsg
, rtmsg
, req_size
, cmd
);
1258 _netlink_mpls_build_multipath(const char *routedesc
, zebra_nhlfe_t
*nhlfe
,
1259 struct rtattr
*rta
, struct rtnexthop
*rtnh
,
1260 struct rtmsg
*rtmsg
, union g_addr
**src
)
1265 family
= NHLFE_FAMILY(nhlfe
);
1266 bytelen
= (family
== AF_INET
? 4 : 16);
1267 _netlink_route_build_multipath(routedesc
, bytelen
, nhlfe
->nexthop
, rta
,
1272 /* Log debug information for netlink_route_multipath
1273 * if debug logging is enabled.
1275 * @param cmd: Netlink command which is to be processed
1276 * @param p: Prefix for which the change is due
1277 * @param family: Address family which the change concerns
1278 * @param zvrf: The vrf we are in
1279 * @param tableid: The table we are working on
1281 static void _netlink_route_debug(int cmd
, struct prefix
*p
,
1282 int family
, struct zebra_vrf
*zvrf
,
1285 if (IS_ZEBRA_DEBUG_KERNEL
) {
1286 char buf
[PREFIX_STRLEN
];
1288 "netlink_route_multipath(): %s %s vrf %u(%u)",
1289 nl_msg_type_to_str(cmd
),
1290 prefix2str(p
, buf
, sizeof(buf
)),
1291 zvrf_id(zvrf
), tableid
);
1295 static void _netlink_mpls_debug(int cmd
, uint32_t label
, const char *routedesc
)
1297 if (IS_ZEBRA_DEBUG_KERNEL
)
1298 zlog_debug("netlink_mpls_multipath() (%s): %s %u/20", routedesc
,
1299 nl_msg_type_to_str(cmd
), label
);
1302 static int netlink_neigh_update(int cmd
, int ifindex
, uint32_t addr
, char *lla
,
1303 int llalen
, ns_id_t ns_id
)
1311 struct zebra_ns
*zns
= zebra_ns_lookup(ns_id
);
1313 memset(&req
, 0, sizeof(req
));
1315 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1316 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
1317 req
.n
.nlmsg_type
= cmd
; // RTM_NEWNEIGH or RTM_DELNEIGH
1318 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1320 req
.ndm
.ndm_family
= AF_INET
;
1321 req
.ndm
.ndm_state
= NUD_PERMANENT
;
1322 req
.ndm
.ndm_ifindex
= ifindex
;
1323 req
.ndm
.ndm_type
= RTN_UNICAST
;
1325 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &addr
, 4);
1326 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, lla
, llalen
);
1328 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1332 /* Routing table change via netlink interface. */
1333 /* Update flag indicates whether this is a "replace" or not. */
1334 static int netlink_route_multipath(int cmd
, struct prefix
*p
,
1335 struct prefix
*src_p
, struct route_entry
*re
,
1339 struct sockaddr_nl snl
;
1340 struct nexthop
*nexthop
= NULL
;
1341 unsigned int nexthop_num
;
1342 int family
= PREFIX_FAMILY(p
);
1343 const char *routedesc
;
1350 char buf
[NL_PKT_BUF_SIZE
];
1353 struct zebra_ns
*zns
;
1354 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1357 memset(&req
, 0, sizeof req
- NL_PKT_BUF_SIZE
);
1359 bytelen
= (family
== AF_INET
? 4 : 16);
1361 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
1362 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
1363 if ((cmd
== RTM_NEWROUTE
) && update
)
1364 req
.n
.nlmsg_flags
|= NLM_F_REPLACE
;
1365 req
.n
.nlmsg_type
= cmd
;
1366 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1368 req
.r
.rtm_family
= family
;
1369 req
.r
.rtm_dst_len
= p
->prefixlen
;
1370 req
.r
.rtm_src_len
= src_p
? src_p
->prefixlen
: 0;
1371 req
.r
.rtm_protocol
= zebra2proto(re
->type
);
1372 req
.r
.rtm_scope
= RT_SCOPE_UNIVERSE
;
1375 * blackhole routes are not RTN_UNICAST, they are
1376 * RTN_ BLACKHOLE|UNREACHABLE|PROHIBIT
1377 * so setting this value as a RTN_UNICAST would
1378 * cause the route lookup of just the prefix
1379 * to fail. So no need to specify this for
1380 * the RTM_DELROUTE case
1382 if (cmd
!= RTM_DELROUTE
)
1383 req
.r
.rtm_type
= RTN_UNICAST
;
1385 addattr_l(&req
.n
, sizeof req
, RTA_DST
, &p
->u
.prefix
, bytelen
);
1387 addattr_l(&req
.n
, sizeof req
, RTA_SRC
, &src_p
->u
.prefix
,
1391 /* Hardcode the metric for all routes coming from zebra. Metric isn't
1393 * either by the kernel or by zebra. Its purely for calculating best
1395 * by the routing protocol and for communicating with protocol peers.
1397 addattr32(&req
.n
, sizeof req
, RTA_PRIORITY
, NL_DEFAULT_ROUTE_METRIC
);
1398 #if defined(SUPPORT_REALMS)
1399 if (re
->tag
> 0 && re
->tag
<= 255)
1400 addattr32(&req
.n
, sizeof req
, RTA_FLOW
, re
->tag
);
1402 /* Table corresponding to this route. */
1403 if (re
->table
< 256)
1404 req
.r
.rtm_table
= re
->table
;
1406 req
.r
.rtm_table
= RT_TABLE_UNSPEC
;
1407 addattr32(&req
.n
, sizeof req
, RTA_TABLE
, re
->table
);
1410 _netlink_route_debug(cmd
, p
, family
, zvrf
, re
->table
);
1413 * If we are not updating the route and we have received
1414 * a route delete, then all we need to fill in is the
1415 * prefix information to tell the kernel to schwack
1418 if (!update
&& cmd
== RTM_DELROUTE
)
1421 if (re
->mtu
|| re
->nexthop_mtu
) {
1422 char buf
[NL_PKT_BUF_SIZE
];
1423 struct rtattr
*rta
= (void *)buf
;
1424 uint32_t mtu
= re
->mtu
;
1425 if (!mtu
|| (re
->nexthop_mtu
&& re
->nexthop_mtu
< mtu
))
1426 mtu
= re
->nexthop_mtu
;
1427 rta
->rta_type
= RTA_METRICS
;
1428 rta
->rta_len
= RTA_LENGTH(0);
1429 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTAX_MTU
, &mtu
, sizeof mtu
);
1430 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_METRICS
, RTA_DATA(rta
),
1434 /* Count overall nexthops so we can decide whether to use singlepath
1435 * or multipath case. */
1437 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1438 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
))
1440 if (cmd
== RTM_NEWROUTE
&& !NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1442 if (cmd
== RTM_DELROUTE
1443 && !CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
1449 /* Singlepath case. */
1450 if (nexthop_num
== 1 || multipath_num
== 1) {
1452 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1454 * So we want to cover 2 types of blackhole
1456 * 1) A normal blackhole route( ala from a static
1458 * 2) A recursively resolved blackhole route
1460 if (nexthop
->type
== NEXTHOP_TYPE_BLACKHOLE
) {
1461 switch (nexthop
->bh_type
) {
1462 case BLACKHOLE_ADMINPROHIB
:
1463 req
.r
.rtm_type
= RTN_PROHIBIT
;
1465 case BLACKHOLE_REJECT
:
1466 req
.r
.rtm_type
= RTN_UNREACHABLE
;
1469 req
.r
.rtm_type
= RTN_BLACKHOLE
;
1474 if (CHECK_FLAG(nexthop
->flags
,
1475 NEXTHOP_FLAG_RECURSIVE
)) {
1477 if (family
== AF_INET
) {
1478 if (nexthop
->rmap_src
.ipv4
1485 } else if (nexthop
->src
.ipv4
1493 } else if (family
== AF_INET6
) {
1494 if (!IN6_IS_ADDR_UNSPECIFIED(
1502 !IN6_IS_ADDR_UNSPECIFIED(
1515 if ((cmd
== RTM_NEWROUTE
1516 && NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1517 || (cmd
== RTM_DELROUTE
1518 && CHECK_FLAG(nexthop
->flags
,
1519 NEXTHOP_FLAG_FIB
))) {
1520 routedesc
= nexthop
->rparent
1521 ? "recursive, single-path"
1524 _netlink_route_build_singlepath(
1525 routedesc
, bytelen
, nexthop
, &req
.n
,
1526 &req
.r
, sizeof req
, cmd
);
1531 if (setsrc
&& (cmd
== RTM_NEWROUTE
)) {
1532 if (family
== AF_INET
)
1533 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1534 &src
.ipv4
, bytelen
);
1535 else if (family
== AF_INET6
)
1536 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1537 &src
.ipv6
, bytelen
);
1540 char buf
[NL_PKT_BUF_SIZE
];
1541 struct rtattr
*rta
= (void *)buf
;
1542 struct rtnexthop
*rtnh
;
1543 union g_addr
*src1
= NULL
;
1545 rta
->rta_type
= RTA_MULTIPATH
;
1546 rta
->rta_len
= RTA_LENGTH(0);
1547 rtnh
= RTA_DATA(rta
);
1550 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1551 if (nexthop_num
>= multipath_num
)
1554 if (CHECK_FLAG(nexthop
->flags
,
1555 NEXTHOP_FLAG_RECURSIVE
)) {
1556 /* This only works for IPv4 now */
1558 if (family
== AF_INET
) {
1559 if (nexthop
->rmap_src
.ipv4
1566 } else if (nexthop
->src
.ipv4
1574 } else if (family
== AF_INET6
) {
1575 if (!IN6_IS_ADDR_UNSPECIFIED(
1583 !IN6_IS_ADDR_UNSPECIFIED(
1596 if ((cmd
== RTM_NEWROUTE
1597 && NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1598 || (cmd
== RTM_DELROUTE
1599 && CHECK_FLAG(nexthop
->flags
,
1600 NEXTHOP_FLAG_FIB
))) {
1601 routedesc
= nexthop
->rparent
1602 ? "recursive, multipath"
1606 _netlink_route_build_multipath(
1607 routedesc
, bytelen
, nexthop
, rta
, rtnh
,
1609 rtnh
= RTNH_NEXT(rtnh
);
1611 if (!setsrc
&& src1
) {
1612 if (family
== AF_INET
)
1613 src
.ipv4
= src1
->ipv4
;
1614 else if (family
== AF_INET6
)
1615 src
.ipv6
= src1
->ipv6
;
1621 if (setsrc
&& (cmd
== RTM_NEWROUTE
)) {
1622 if (family
== AF_INET
)
1623 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1624 &src
.ipv4
, bytelen
);
1625 else if (family
== AF_INET6
)
1626 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1627 &src
.ipv6
, bytelen
);
1628 if (IS_ZEBRA_DEBUG_KERNEL
)
1629 zlog_debug("Setting source");
1632 if (rta
->rta_len
> RTA_LENGTH(0))
1633 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_MULTIPATH
,
1634 RTA_DATA(rta
), RTA_PAYLOAD(rta
));
1637 /* If there is no useful nexthop then return. */
1638 if (nexthop_num
== 0) {
1639 if (IS_ZEBRA_DEBUG_KERNEL
)
1641 "netlink_route_multipath(): No useful nexthop.");
1647 /* Destination netlink address. */
1648 memset(&snl
, 0, sizeof snl
);
1649 snl
.nl_family
= AF_NETLINK
;
1651 /* Talk to netlink socket. */
1652 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1656 int kernel_get_ipmr_sg_stats(struct zebra_vrf
*zvrf
, void *in
)
1659 struct mcast_route_data
*mr
= (struct mcast_route_data
*)in
;
1667 struct zebra_ns
*zns
;
1670 memset(&req
, 0, sizeof(req
));
1672 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1673 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
1674 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1676 req
.ndm
.ndm_family
= RTNL_FAMILY_IPMR
;
1677 req
.n
.nlmsg_type
= RTM_GETROUTE
;
1679 addattr_l(&req
.n
, sizeof(req
), RTA_IIF
, &mroute
->ifindex
, 4);
1680 addattr_l(&req
.n
, sizeof(req
), RTA_OIF
, &mroute
->ifindex
, 4);
1681 addattr_l(&req
.n
, sizeof(req
), RTA_SRC
, &mroute
->sg
.src
.s_addr
, 4);
1682 addattr_l(&req
.n
, sizeof(req
), RTA_DST
, &mroute
->sg
.grp
.s_addr
, 4);
1683 addattr_l(&req
.n
, sizeof(req
), RTA_TABLE
, &zvrf
->table_id
, 4);
1685 suc
= netlink_talk(netlink_route_change_read_multicast
, &req
.n
,
1686 &zns
->netlink_cmd
, zns
, 0);
1692 enum dp_req_result
kernel_route_rib(struct route_node
*rn
,
1694 struct prefix
*src_p
,
1695 struct route_entry
*old
,
1696 struct route_entry
*new)
1703 if (p
->family
== AF_INET
|| v6_rr_semantics
)
1704 ret
= netlink_route_multipath(RTM_NEWROUTE
, p
, src_p
,
1705 new, (old
) ? 1 : 0);
1708 * So v6 route replace semantics are not in
1709 * the kernel at this point as I understand it.
1710 * So let's do a delete than an add.
1711 * In the future once v6 route replace semantics
1712 * are in we can figure out what to do here to
1713 * allow working with old and new kernels.
1715 * I'm also intentionally ignoring the failure case
1716 * of the route delete. If that happens yeah we're
1720 netlink_route_multipath(RTM_DELROUTE
, p
, src_p
,
1722 ret
= netlink_route_multipath(RTM_NEWROUTE
, p
, src_p
,
1725 kernel_route_rib_pass_fail(rn
, p
, new,
1726 (!ret
) ? DP_INSTALL_SUCCESS
1727 : DP_INSTALL_FAILURE
);
1728 return DP_REQUEST_SUCCESS
;
1732 ret
= netlink_route_multipath(RTM_DELROUTE
, p
, src_p
, old
, 0);
1734 kernel_route_rib_pass_fail(rn
, p
, old
,
1735 (!ret
) ? DP_DELETE_SUCCESS
1736 : DP_DELETE_FAILURE
);
1739 return DP_REQUEST_SUCCESS
;
1742 int kernel_neigh_update(int add
, int ifindex
, uint32_t addr
, char *lla
,
1743 int llalen
, ns_id_t ns_id
)
1745 return netlink_neigh_update(add
? RTM_NEWNEIGH
: RTM_DELNEIGH
, ifindex
,
1746 addr
, lla
, llalen
, ns_id
);
1750 * Add remote VTEP to the flood list for this VxLAN interface (VNI). This
1751 * is done by adding an FDB entry with a MAC of 00:00:00:00:00:00.
1753 static int netlink_vxlan_flood_list_update(struct interface
*ifp
,
1754 struct in_addr
*vtep_ip
, int cmd
)
1756 struct zebra_ns
*zns
;
1762 uint8_t dst_mac
[6] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0};
1763 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
1766 memset(&req
, 0, sizeof(req
));
1768 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1769 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
1770 if (cmd
== RTM_NEWNEIGH
)
1771 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_APPEND
);
1772 req
.n
.nlmsg_type
= cmd
;
1773 req
.ndm
.ndm_family
= PF_BRIDGE
;
1774 req
.ndm
.ndm_state
= NUD_NOARP
| NUD_PERMANENT
;
1775 req
.ndm
.ndm_flags
|= NTF_SELF
; // Handle by "self", not "master"
1778 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, &dst_mac
, 6);
1779 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
1780 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &vtep_ip
->s_addr
, 4);
1782 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1787 * Add remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1789 * a "flood" MAC FDB entry.
1791 int kernel_add_vtep(vni_t vni
, struct interface
*ifp
, struct in_addr
*vtep_ip
)
1793 if (IS_ZEBRA_DEBUG_VXLAN
)
1794 zlog_debug("Install %s into flood list for VNI %u intf %s(%u)",
1795 inet_ntoa(*vtep_ip
), vni
, ifp
->name
, ifp
->ifindex
);
1797 return netlink_vxlan_flood_list_update(ifp
, vtep_ip
, RTM_NEWNEIGH
);
1801 * Remove remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1802 * deleting the "flood" MAC FDB entry.
1804 int kernel_del_vtep(vni_t vni
, struct interface
*ifp
, struct in_addr
*vtep_ip
)
1806 if (IS_ZEBRA_DEBUG_VXLAN
)
1808 "Uninstall %s from flood list for VNI %u intf %s(%u)",
1809 inet_ntoa(*vtep_ip
), vni
, ifp
->name
, ifp
->ifindex
);
1811 return netlink_vxlan_flood_list_update(ifp
, vtep_ip
, RTM_DELNEIGH
);
1815 #define NDA_RTA(r) \
1816 ((struct rtattr *)(((char *)(r)) + NLMSG_ALIGN(sizeof(struct ndmsg))))
1819 static int netlink_macfdb_change(struct nlmsghdr
*h
, int len
, ns_id_t ns_id
)
1822 struct interface
*ifp
;
1823 struct zebra_if
*zif
;
1824 struct rtattr
*tb
[NDA_MAX
+ 1];
1825 struct interface
*br_if
;
1828 struct prefix vtep_ip
;
1829 int vid_present
= 0, dst_present
= 0;
1830 char buf
[ETHER_ADDR_STRLEN
];
1835 ndm
= NLMSG_DATA(h
);
1837 /* We only process macfdb notifications if EVPN is enabled */
1838 if (!is_evpn_enabled())
1841 /* The interface should exist. */
1842 ifp
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
1844 if (!ifp
|| !ifp
->info
)
1847 /* The interface should be something we're interested in. */
1848 if (!IS_ZEBRA_IF_BRIDGE_SLAVE(ifp
))
1851 /* Drop "permanent" entries. */
1852 if (ndm
->ndm_state
& NUD_PERMANENT
)
1855 zif
= (struct zebra_if
*)ifp
->info
;
1856 if ((br_if
= zif
->brslave_info
.br_if
) == NULL
) {
1857 zlog_warn("%s family %s IF %s(%u) brIF %u - no bridge master",
1858 nl_msg_type_to_str(h
->nlmsg_type
),
1859 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1860 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
);
1864 /* Parse attributes and extract fields of interest. */
1865 memset(tb
, 0, sizeof tb
);
1866 netlink_parse_rtattr(tb
, NDA_MAX
, NDA_RTA(ndm
), len
);
1868 if (!tb
[NDA_LLADDR
]) {
1869 zlog_warn("%s family %s IF %s(%u) brIF %u - no LLADDR",
1870 nl_msg_type_to_str(h
->nlmsg_type
),
1871 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1872 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
);
1876 if (RTA_PAYLOAD(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
1878 "%s family %s IF %s(%u) brIF %u - LLADDR is not MAC, len %lu",
1879 nl_msg_type_to_str(h
->nlmsg_type
),
1880 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1881 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
,
1882 (unsigned long)RTA_PAYLOAD(tb
[NDA_LLADDR
]));
1886 memcpy(&mac
, RTA_DATA(tb
[NDA_LLADDR
]), ETH_ALEN
);
1888 if ((NDA_VLAN
<= NDA_MAX
) && tb
[NDA_VLAN
]) {
1890 vid
= *(uint16_t *)RTA_DATA(tb
[NDA_VLAN
]);
1891 sprintf(vid_buf
, " VLAN %u", vid
);
1895 /* TODO: Only IPv4 supported now. */
1897 vtep_ip
.family
= AF_INET
;
1898 vtep_ip
.prefixlen
= IPV4_MAX_BITLEN
;
1899 memcpy(&(vtep_ip
.u
.prefix4
.s_addr
), RTA_DATA(tb
[NDA_DST
]),
1901 sprintf(dst_buf
, " dst %s", inet_ntoa(vtep_ip
.u
.prefix4
));
1904 sticky
= (ndm
->ndm_state
& NUD_NOARP
) ? 1 : 0;
1906 if (IS_ZEBRA_DEBUG_KERNEL
)
1907 zlog_debug("Rx %s family %s IF %s(%u)%s %sMAC %s%s",
1908 nl_msg_type_to_str(h
->nlmsg_type
),
1909 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1910 ndm
->ndm_ifindex
, vid_present
? vid_buf
: "",
1911 sticky
? "sticky " : "",
1912 prefix_mac2str(&mac
, buf
, sizeof(buf
)),
1913 dst_present
? dst_buf
: "");
1915 if (filter_vlan
&& vid
!= filter_vlan
)
1918 /* If add or update, do accordingly if learnt on a "local" interface; if
1919 * the notification is over VxLAN, this has to be related to
1921 * so perform an implicit delete of any local entry (if it exists).
1923 if (h
->nlmsg_type
== RTM_NEWNEIGH
) {
1924 /* Drop "permanent" entries. */
1925 if (ndm
->ndm_state
& NUD_PERMANENT
)
1928 if (IS_ZEBRA_IF_VXLAN(ifp
))
1929 return zebra_vxlan_check_del_local_mac(ifp
, br_if
, &mac
,
1932 return zebra_vxlan_local_mac_add_update(ifp
, br_if
, &mac
, vid
,
1936 /* This is a delete notification.
1937 * 1. For a MAC over VxLan, check if it needs to be refreshed(readded)
1938 * 2. For a MAC over "local" interface, delete the mac
1939 * Note: We will get notifications from both bridge driver and VxLAN
1941 * Ignore the notification from VxLan driver as it is also generated
1942 * when mac moves from remote to local.
1947 if (IS_ZEBRA_IF_VXLAN(ifp
))
1948 return zebra_vxlan_check_readd_remote_mac(ifp
, br_if
, &mac
,
1951 return zebra_vxlan_local_mac_del(ifp
, br_if
, &mac
, vid
);
1954 static int netlink_macfdb_table(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
1959 if (h
->nlmsg_type
!= RTM_NEWNEIGH
)
1962 /* Length validity. */
1963 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
1967 /* We are interested only in AF_BRIDGE notifications. */
1968 ndm
= NLMSG_DATA(h
);
1969 if (ndm
->ndm_family
!= AF_BRIDGE
)
1972 return netlink_macfdb_change(h
, len
, ns_id
);
1975 /* Request for MAC FDB information from the kernel */
1976 static int netlink_request_macs(struct zebra_ns
*zns
, int family
, int type
,
1977 ifindex_t master_ifindex
)
1981 struct ifinfomsg ifm
;
1985 /* Form the request, specifying filter (rtattr) if needed. */
1986 memset(&req
, 0, sizeof(req
));
1987 req
.n
.nlmsg_type
= type
;
1988 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ifinfomsg
));
1989 req
.ifm
.ifi_family
= family
;
1991 addattr32(&req
.n
, sizeof(req
), IFLA_MASTER
, master_ifindex
);
1993 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
1997 * MAC forwarding database read using netlink interface. This is invoked
2000 int netlink_macfdb_read(struct zebra_ns
*zns
)
2004 /* Get bridge FDB table. */
2005 ret
= netlink_request_macs(zns
, AF_BRIDGE
, RTM_GETNEIGH
, 0);
2008 /* We are reading entire table. */
2010 ret
= netlink_parse_info(netlink_macfdb_table
, &zns
->netlink_cmd
, zns
,
2017 * MAC forwarding database read using netlink interface. This is for a
2018 * specific bridge and matching specific access VLAN (if VLAN-aware bridge).
2020 int netlink_macfdb_read_for_bridge(struct zebra_ns
*zns
, struct interface
*ifp
,
2021 struct interface
*br_if
)
2023 struct zebra_if
*br_zif
;
2024 struct zebra_if
*zif
;
2025 struct zebra_l2info_vxlan
*vxl
;
2029 /* Save VLAN we're filtering on, if needed. */
2030 br_zif
= (struct zebra_if
*)br_if
->info
;
2031 zif
= (struct zebra_if
*)ifp
->info
;
2032 vxl
= &zif
->l2info
.vxl
;
2033 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif
))
2034 filter_vlan
= vxl
->access_vlan
;
2036 /* Get bridge FDB table for specific bridge - we do the VLAN filtering.
2038 ret
= netlink_request_macs(zns
, AF_BRIDGE
, RTM_GETNEIGH
,
2042 ret
= netlink_parse_info(netlink_macfdb_table
, &zns
->netlink_cmd
, zns
,
2045 /* Reset VLAN filter. */
2050 static int netlink_macfdb_update(struct interface
*ifp
, vlanid_t vid
,
2051 struct ethaddr
*mac
, struct in_addr vtep_ip
,
2052 int local
, int cmd
, uint8_t sticky
)
2054 struct zebra_ns
*zns
;
2061 struct zebra_if
*zif
;
2062 struct interface
*br_if
;
2063 struct zebra_if
*br_zif
;
2064 char buf
[ETHER_ADDR_STRLEN
];
2065 int vid_present
= 0, dst_present
= 0;
2068 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
2072 if ((br_if
= zif
->brslave_info
.br_if
) == NULL
) {
2073 zlog_warn("MAC %s on IF %s(%u) - no mapping to bridge",
2074 (cmd
== RTM_NEWNEIGH
) ? "add" : "del", ifp
->name
,
2079 memset(&req
, 0, sizeof(req
));
2081 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2082 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
2083 if (cmd
== RTM_NEWNEIGH
)
2084 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_REPLACE
);
2085 req
.n
.nlmsg_type
= cmd
;
2086 req
.ndm
.ndm_family
= AF_BRIDGE
;
2087 req
.ndm
.ndm_flags
|= NTF_SELF
| NTF_MASTER
;
2088 req
.ndm
.ndm_state
= NUD_REACHABLE
;
2091 req
.ndm
.ndm_state
|= NUD_NOARP
;
2093 req
.ndm
.ndm_flags
|= NTF_EXT_LEARNED
;
2095 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, mac
, 6);
2096 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
2098 dst_alen
= 4; // TODO: hardcoded
2099 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &vtep_ip
, dst_alen
);
2101 sprintf(dst_buf
, " dst %s", inet_ntoa(vtep_ip
));
2103 br_zif
= (struct zebra_if
*)br_if
->info
;
2104 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif
) && vid
> 0) {
2105 addattr16(&req
.n
, sizeof(req
), NDA_VLAN
, vid
);
2107 sprintf(vid_buf
, " VLAN %u", vid
);
2109 addattr32(&req
.n
, sizeof(req
), NDA_MASTER
, br_if
->ifindex
);
2111 if (IS_ZEBRA_DEBUG_KERNEL
)
2112 zlog_debug("Tx %s family %s IF %s(%u)%s %sMAC %s%s",
2113 nl_msg_type_to_str(cmd
),
2114 nl_family_to_str(req
.ndm
.ndm_family
), ifp
->name
,
2115 ifp
->ifindex
, vid_present
? vid_buf
: "",
2116 sticky
? "sticky " : "",
2117 prefix_mac2str(mac
, buf
, sizeof(buf
)),
2118 dst_present
? dst_buf
: "");
2120 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2125 (NUD_PERMANENT | NUD_NOARP | NUD_REACHABLE | NUD_PROBE | NUD_STALE \
2128 static int netlink_ipneigh_change(struct nlmsghdr
*h
, int len
, ns_id_t ns_id
)
2131 struct interface
*ifp
;
2132 struct zebra_if
*zif
;
2133 struct rtattr
*tb
[NDA_MAX
+ 1];
2134 struct interface
*link_if
;
2137 char buf
[ETHER_ADDR_STRLEN
];
2138 char buf2
[INET6_ADDRSTRLEN
];
2139 int mac_present
= 0;
2140 uint8_t ext_learned
;
2142 ndm
= NLMSG_DATA(h
);
2144 /* The interface should exist. */
2145 ifp
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
2147 if (!ifp
|| !ifp
->info
)
2150 zif
= (struct zebra_if
*)ifp
->info
;
2152 /* Parse attributes and extract fields of interest. */
2153 memset(tb
, 0, sizeof tb
);
2154 netlink_parse_rtattr(tb
, NDA_MAX
, NDA_RTA(ndm
), len
);
2157 zlog_warn("%s family %s IF %s(%u) - no DST",
2158 nl_msg_type_to_str(h
->nlmsg_type
),
2159 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2164 memset(&ip
, 0, sizeof(struct ipaddr
));
2165 ip
.ipa_type
= (ndm
->ndm_family
== AF_INET
) ? IPADDR_V4
: IPADDR_V6
;
2166 memcpy(&ip
.ip
.addr
, RTA_DATA(tb
[NDA_DST
]), RTA_PAYLOAD(tb
[NDA_DST
]));
2168 /* Drop some "permanent" entries. */
2169 if (ndm
->ndm_state
& NUD_PERMANENT
) {
2170 char buf
[16] = "169.254.0.1";
2171 struct in_addr ipv4_ll
;
2173 if (ndm
->ndm_family
!= AF_INET
)
2176 if (!zif
->v6_2_v4_ll_neigh_entry
)
2179 if (h
->nlmsg_type
!= RTM_DELNEIGH
)
2182 inet_pton(AF_INET
, buf
, &ipv4_ll
);
2183 if (ipv4_ll
.s_addr
!= ip
.ip
._v4_addr
.s_addr
)
2186 if_nbr_ipv6ll_to_ipv4ll_neigh_update(
2187 ifp
, &zif
->v6_2_v4_ll_addr6
, true);
2191 /* The neighbor is present on an SVI. From this, we locate the
2193 * bridge because we're only interested in neighbors on a VxLAN bridge.
2194 * The bridge is located based on the nature of the SVI:
2195 * (a) In the case of a VLAN-aware bridge, the SVI is a L3 VLAN
2197 * and is linked to the bridge
2198 * (b) In the case of a VLAN-unaware bridge, the SVI is the bridge
2202 if (IS_ZEBRA_IF_VLAN(ifp
)) {
2203 link_if
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
2207 } else if (IS_ZEBRA_IF_BRIDGE(ifp
))
2212 memset(&mac
, 0, sizeof(struct ethaddr
));
2213 if (h
->nlmsg_type
== RTM_NEWNEIGH
) {
2214 if (tb
[NDA_LLADDR
]) {
2215 if (RTA_PAYLOAD(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
2217 "%s family %s IF %s(%u) - LLADDR is not MAC, len %lu",
2218 nl_msg_type_to_str(h
->nlmsg_type
),
2219 nl_family_to_str(ndm
->ndm_family
),
2220 ifp
->name
, ndm
->ndm_ifindex
,
2221 (unsigned long)RTA_PAYLOAD(
2227 memcpy(&mac
, RTA_DATA(tb
[NDA_LLADDR
]), ETH_ALEN
);
2230 ext_learned
= (ndm
->ndm_flags
& NTF_EXT_LEARNED
) ? 1 : 0;
2232 if (IS_ZEBRA_DEBUG_KERNEL
)
2234 "Rx %s family %s IF %s(%u) IP %s MAC %s state 0x%x flags 0x%x",
2235 nl_msg_type_to_str(h
->nlmsg_type
),
2236 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2238 ipaddr2str(&ip
, buf2
, sizeof(buf2
)),
2240 ? prefix_mac2str(&mac
, buf
, sizeof(buf
))
2242 ndm
->ndm_state
, ndm
->ndm_flags
);
2244 /* If the neighbor state is valid for use, process as an add or
2246 * else process as a delete. Note that the delete handling may
2248 * in re-adding the neighbor if it is a valid "remote" neighbor.
2250 if (ndm
->ndm_state
& NUD_VALID
)
2251 return zebra_vxlan_handle_kernel_neigh_update(
2252 ifp
, link_if
, &ip
, &mac
, ndm
->ndm_state
,
2255 return zebra_vxlan_handle_kernel_neigh_del(ifp
, link_if
, &ip
);
2258 if (IS_ZEBRA_DEBUG_KERNEL
)
2259 zlog_debug("Rx %s family %s IF %s(%u) IP %s",
2260 nl_msg_type_to_str(h
->nlmsg_type
),
2261 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2263 ipaddr2str(&ip
, buf2
, sizeof(buf2
)));
2265 /* Process the delete - it may result in re-adding the neighbor if it is
2266 * a valid "remote" neighbor.
2268 return zebra_vxlan_handle_kernel_neigh_del(ifp
, link_if
, &ip
);
2271 static int netlink_neigh_table(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
2276 if (h
->nlmsg_type
!= RTM_NEWNEIGH
)
2279 /* Length validity. */
2280 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2284 /* We are interested only in AF_INET or AF_INET6 notifications. */
2285 ndm
= NLMSG_DATA(h
);
2286 if (ndm
->ndm_family
!= AF_INET
&& ndm
->ndm_family
!= AF_INET6
)
2289 return netlink_neigh_change(h
, len
);
2292 /* Request for IP neighbor information from the kernel */
2293 static int netlink_request_neigh(struct zebra_ns
*zns
, int family
, int type
,
2302 /* Form the request, specifying filter (rtattr) if needed. */
2303 memset(&req
, 0, sizeof(req
));
2304 req
.n
.nlmsg_type
= type
;
2305 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2306 req
.ndm
.ndm_family
= family
;
2308 addattr32(&req
.n
, sizeof(req
), NDA_IFINDEX
, ifindex
);
2310 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
2314 * IP Neighbor table read using netlink interface. This is invoked
2317 int netlink_neigh_read(struct zebra_ns
*zns
)
2321 /* Get IP neighbor table. */
2322 ret
= netlink_request_neigh(zns
, AF_UNSPEC
, RTM_GETNEIGH
, 0);
2325 ret
= netlink_parse_info(netlink_neigh_table
, &zns
->netlink_cmd
, zns
, 0,
2332 * IP Neighbor table read using netlink interface. This is for a specific
2335 int netlink_neigh_read_for_vlan(struct zebra_ns
*zns
, struct interface
*vlan_if
)
2339 ret
= netlink_request_neigh(zns
, AF_UNSPEC
, RTM_GETNEIGH
,
2343 ret
= netlink_parse_info(netlink_neigh_table
, &zns
->netlink_cmd
, zns
, 0,
2349 int netlink_neigh_change(struct nlmsghdr
*h
, ns_id_t ns_id
)
2354 if (!(h
->nlmsg_type
== RTM_NEWNEIGH
|| h
->nlmsg_type
== RTM_DELNEIGH
))
2357 /* Length validity. */
2358 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2362 /* Is this a notification for the MAC FDB or IP neighbor table? */
2363 ndm
= NLMSG_DATA(h
);
2364 if (ndm
->ndm_family
== AF_BRIDGE
)
2365 return netlink_macfdb_change(h
, len
, ns_id
);
2367 if (ndm
->ndm_type
!= RTN_UNICAST
)
2370 if (ndm
->ndm_family
== AF_INET
|| ndm
->ndm_family
== AF_INET6
)
2371 return netlink_ipneigh_change(h
, len
, ns_id
);
2376 static int netlink_neigh_update2(struct interface
*ifp
, struct ipaddr
*ip
,
2377 struct ethaddr
*mac
, uint32_t flags
, int cmd
)
2386 struct zebra_ns
*zns
;
2387 char buf
[INET6_ADDRSTRLEN
];
2388 char buf2
[ETHER_ADDR_STRLEN
];
2389 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
2392 memset(&req
, 0, sizeof(req
));
2394 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2395 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
2396 if (cmd
== RTM_NEWNEIGH
)
2397 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_REPLACE
);
2398 req
.n
.nlmsg_type
= cmd
; // RTM_NEWNEIGH or RTM_DELNEIGH
2399 req
.ndm
.ndm_family
= IS_IPADDR_V4(ip
) ? AF_INET
: AF_INET6
;
2400 req
.ndm
.ndm_state
= flags
;
2401 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
2402 req
.ndm
.ndm_type
= RTN_UNICAST
;
2403 req
.ndm
.ndm_flags
= NTF_EXT_LEARNED
;
2406 ipa_len
= IS_IPADDR_V4(ip
) ? IPV4_MAX_BYTELEN
: IPV6_MAX_BYTELEN
;
2407 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &ip
->ip
.addr
, ipa_len
);
2409 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, mac
, 6);
2411 if (IS_ZEBRA_DEBUG_KERNEL
)
2412 zlog_debug("Tx %s family %s IF %s(%u) Neigh %s MAC %s",
2413 nl_msg_type_to_str(cmd
),
2414 nl_family_to_str(req
.ndm
.ndm_family
), ifp
->name
,
2415 ifp
->ifindex
, ipaddr2str(ip
, buf
, sizeof(buf
)),
2416 mac
? prefix_mac2str(mac
, buf2
, sizeof(buf2
))
2419 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2423 int kernel_add_mac(struct interface
*ifp
, vlanid_t vid
, struct ethaddr
*mac
,
2424 struct in_addr vtep_ip
, uint8_t sticky
)
2426 return netlink_macfdb_update(ifp
, vid
, mac
, vtep_ip
, 0, RTM_NEWNEIGH
,
2430 int kernel_del_mac(struct interface
*ifp
, vlanid_t vid
, struct ethaddr
*mac
,
2431 struct in_addr vtep_ip
, int local
)
2433 return netlink_macfdb_update(ifp
, vid
, mac
, vtep_ip
, local
,
2437 int kernel_add_neigh(struct interface
*ifp
, struct ipaddr
*ip
,
2438 struct ethaddr
*mac
)
2440 return netlink_neigh_update2(ifp
, ip
, mac
, NUD_NOARP
, RTM_NEWNEIGH
);
2443 int kernel_del_neigh(struct interface
*ifp
, struct ipaddr
*ip
)
2445 return netlink_neigh_update2(ifp
, ip
, NULL
, 0, RTM_DELNEIGH
);
2449 * MPLS label forwarding table change via netlink interface.
2451 int netlink_mpls_multipath(int cmd
, zebra_lsp_t
*lsp
)
2454 zebra_nhlfe_t
*nhlfe
;
2455 struct nexthop
*nexthop
= NULL
;
2456 unsigned int nexthop_num
;
2457 const char *routedesc
;
2458 struct zebra_ns
*zns
= zebra_ns_lookup(NS_DEFAULT
);
2464 char buf
[NL_PKT_BUF_SIZE
];
2467 memset(&req
, 0, sizeof req
- NL_PKT_BUF_SIZE
);
2470 * Count # nexthops so we can decide whether to use singlepath
2471 * or multipath case.
2474 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2475 nexthop
= nhlfe
->nexthop
;
2478 if (cmd
== RTM_NEWROUTE
) {
2479 /* Count all selected NHLFEs */
2480 if (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2481 && CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
))
2485 /* Count all installed NHLFEs */
2486 if (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_INSTALLED
)
2487 && CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
2492 if ((nexthop_num
== 0) || (!lsp
->best_nhlfe
&& (cmd
!= RTM_DELROUTE
)))
2495 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
2496 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
2497 req
.n
.nlmsg_type
= cmd
;
2498 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
2500 req
.r
.rtm_family
= AF_MPLS
;
2501 req
.r
.rtm_table
= RT_TABLE_MAIN
;
2502 req
.r
.rtm_dst_len
= MPLS_LABEL_LEN_BITS
;
2503 req
.r
.rtm_scope
= RT_SCOPE_UNIVERSE
;
2504 req
.r
.rtm_type
= RTN_UNICAST
;
2506 if (cmd
== RTM_NEWROUTE
) {
2507 /* We do a replace to handle update. */
2508 req
.n
.nlmsg_flags
|= NLM_F_REPLACE
;
2510 /* set the protocol value if installing */
2511 route_type
= re_type_from_lsp_type(lsp
->best_nhlfe
->type
);
2512 req
.r
.rtm_protocol
= zebra2proto(route_type
);
2515 /* Fill destination */
2516 lse
= mpls_lse_encode(lsp
->ile
.in_label
, 0, 0, 1);
2517 addattr_l(&req
.n
, sizeof req
, RTA_DST
, &lse
, sizeof(mpls_lse_t
));
2519 /* Fill nexthops (paths) based on single-path or multipath. The paths
2520 * chosen depend on the operation.
2522 if (nexthop_num
== 1 || multipath_num
== 1) {
2523 routedesc
= "single-path";
2524 _netlink_mpls_debug(cmd
, lsp
->ile
.in_label
, routedesc
);
2527 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2528 nexthop
= nhlfe
->nexthop
;
2532 if ((cmd
== RTM_NEWROUTE
2533 && (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2534 && CHECK_FLAG(nexthop
->flags
,
2535 NEXTHOP_FLAG_ACTIVE
)))
2536 || (cmd
== RTM_DELROUTE
2537 && (CHECK_FLAG(nhlfe
->flags
,
2538 NHLFE_FLAG_INSTALLED
)
2539 && CHECK_FLAG(nexthop
->flags
,
2540 NEXTHOP_FLAG_FIB
)))) {
2541 /* Add the gateway */
2542 _netlink_mpls_build_singlepath(routedesc
, nhlfe
,
2549 } else /* Multipath case */
2551 char buf
[NL_PKT_BUF_SIZE
];
2552 struct rtattr
*rta
= (void *)buf
;
2553 struct rtnexthop
*rtnh
;
2554 union g_addr
*src1
= NULL
;
2556 rta
->rta_type
= RTA_MULTIPATH
;
2557 rta
->rta_len
= RTA_LENGTH(0);
2558 rtnh
= RTA_DATA(rta
);
2560 routedesc
= "multipath";
2561 _netlink_mpls_debug(cmd
, lsp
->ile
.in_label
, routedesc
);
2564 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2565 nexthop
= nhlfe
->nexthop
;
2569 if (nexthop_num
>= multipath_num
)
2572 if ((cmd
== RTM_NEWROUTE
2573 && (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2574 && CHECK_FLAG(nexthop
->flags
,
2575 NEXTHOP_FLAG_ACTIVE
)))
2576 || (cmd
== RTM_DELROUTE
2577 && (CHECK_FLAG(nhlfe
->flags
,
2578 NHLFE_FLAG_INSTALLED
)
2579 && CHECK_FLAG(nexthop
->flags
,
2580 NEXTHOP_FLAG_FIB
)))) {
2583 /* Build the multipath */
2584 _netlink_mpls_build_multipath(routedesc
, nhlfe
,
2587 rtnh
= RTNH_NEXT(rtnh
);
2591 /* Add the multipath */
2592 if (rta
->rta_len
> RTA_LENGTH(0))
2593 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_MULTIPATH
,
2594 RTA_DATA(rta
), RTA_PAYLOAD(rta
));
2597 /* Talk to netlink socket. */
2598 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2601 #endif /* HAVE_NETLINK */