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
));
299 zlog_err("%s: Message received from netlink is of a broken size %d %zu",
300 __PRETTY_FUNCTION__
, h
->nlmsg_len
,
301 (size_t)NLMSG_LENGTH(sizeof(struct rtmsg
)));
305 memset(tb
, 0, sizeof tb
);
306 netlink_parse_rtattr(tb
, RTA_MAX
, RTM_RTA(rtm
), len
);
308 if (rtm
->rtm_flags
& RTM_F_CLONED
)
310 if (rtm
->rtm_protocol
== RTPROT_REDIRECT
)
312 if (rtm
->rtm_protocol
== RTPROT_KERNEL
)
315 if (!startup
&& is_selfroute(rtm
->rtm_protocol
)
316 && h
->nlmsg_type
== RTM_NEWROUTE
) {
317 if (IS_ZEBRA_DEBUG_KERNEL
)
318 zlog_debug("Route type: %d Received that we think we have originated, ignoring",
323 /* We don't care about change notifications for the MPLS table. */
324 /* TODO: Revisit this. */
325 if (rtm
->rtm_family
== AF_MPLS
)
328 /* Table corresponding to route. */
330 table
= *(int *)RTA_DATA(tb
[RTA_TABLE
]);
332 table
= rtm
->rtm_table
;
335 vrf_id
= vrf_lookup_by_table(table
, ns_id
);
336 if (vrf_id
== VRF_DEFAULT
) {
337 if (!is_zebra_valid_kernel_table(table
)
338 && !is_zebra_main_routing_table(table
))
342 /* Route which inserted by Zebra. */
343 if (is_selfroute(rtm
->rtm_protocol
)) {
344 flags
|= ZEBRA_FLAG_SELFROUTE
;
345 proto
= proto2zebra(rtm
->rtm_protocol
, rtm
->rtm_family
);
348 index
= *(int *)RTA_DATA(tb
[RTA_OIF
]);
351 dest
= RTA_DATA(tb
[RTA_DST
]);
356 src
= RTA_DATA(tb
[RTA_SRC
]);
361 prefsrc
= RTA_DATA(tb
[RTA_PREFSRC
]);
364 gate
= RTA_DATA(tb
[RTA_GATEWAY
]);
366 if (tb
[RTA_PRIORITY
])
367 metric
= *(int *)RTA_DATA(tb
[RTA_PRIORITY
]);
369 #if defined(SUPPORT_REALMS)
371 tag
= *(uint32_t *)RTA_DATA(tb
[RTA_FLOW
]);
374 if (tb
[RTA_METRICS
]) {
375 struct rtattr
*mxrta
[RTAX_MAX
+ 1];
377 memset(mxrta
, 0, sizeof mxrta
);
378 netlink_parse_rtattr(mxrta
, RTAX_MAX
, RTA_DATA(tb
[RTA_METRICS
]),
379 RTA_PAYLOAD(tb
[RTA_METRICS
]));
382 mtu
= *(uint32_t *)RTA_DATA(mxrta
[RTAX_MTU
]);
385 if (rtm
->rtm_family
== AF_INET
) {
387 if (rtm
->rtm_dst_len
> IPV4_MAX_BITLEN
) {
389 "Invalid destination prefix length: %u received from kernel route change",
393 memcpy(&p
.u
.prefix4
, dest
, 4);
394 p
.prefixlen
= rtm
->rtm_dst_len
;
396 if (rtm
->rtm_src_len
!= 0) {
397 char buf
[PREFIX_STRLEN
];
398 zlog_warn("unsupported IPv4 sourcedest route (dest %s vrf %u)",
399 prefix2str(&p
, buf
, sizeof(buf
)), vrf_id
);
403 /* Force debug below to not display anything for source */
405 } else if (rtm
->rtm_family
== AF_INET6
) {
407 if (rtm
->rtm_dst_len
> IPV6_MAX_BITLEN
) {
409 "Invalid destination prefix length: %u received from kernel route change",
413 memcpy(&p
.u
.prefix6
, dest
, 16);
414 p
.prefixlen
= rtm
->rtm_dst_len
;
416 src_p
.family
= AF_INET6
;
417 if (rtm
->rtm_src_len
> IPV6_MAX_BITLEN
) {
419 "Invalid source prefix length: %u received from kernel route change",
423 memcpy(&src_p
.prefix
, src
, 16);
424 src_p
.prefixlen
= rtm
->rtm_src_len
;
428 * For ZEBRA_ROUTE_KERNEL types:
430 * The metric/priority of the route received from the kernel
431 * is a 32 bit number. We are going to interpret the high
432 * order byte as the Admin Distance and the low order 3 bytes
435 * This will allow us to do two things:
436 * 1) Allow the creation of kernel routes that can be
437 * overridden by zebra.
438 * 2) Allow the old behavior for 'most' kernel route types
439 * if a user enters 'ip route ...' v4 routes get a metric
440 * of 0 and v6 routes get a metric of 1024. Both of these
441 * values will end up with a admin distance of 0, which
442 * will cause them to win for the purposes of zebra.
444 if (proto
== ZEBRA_ROUTE_KERNEL
) {
445 distance
= (metric
>> 24) & 0xFF;
446 metric
= (metric
& 0x00FFFFFF);
449 if (IS_ZEBRA_DEBUG_KERNEL
) {
450 char buf
[PREFIX_STRLEN
];
451 char buf2
[PREFIX_STRLEN
];
452 zlog_debug("%s %s%s%s vrf %u(%u) metric: %d Admin Distance: %d",
453 nl_msg_type_to_str(h
->nlmsg_type
),
454 prefix2str(&p
, buf
, sizeof(buf
)),
455 src_p
.prefixlen
? " from " : "",
457 ? prefix2str(&src_p
, buf2
, sizeof(buf2
))
459 vrf_id
, table
, metric
, distance
);
463 if (rtm
->rtm_family
== AF_INET6
)
466 if (h
->nlmsg_type
== RTM_NEWROUTE
) {
467 struct interface
*ifp
;
468 vrf_id_t nh_vrf_id
= vrf_id
;
470 if (!tb
[RTA_MULTIPATH
]) {
472 size_t sz
= (afi
== AFI_IP
) ? 4 : 16;
474 memset(&nh
, 0, sizeof(nh
));
476 if (bh_type
== BLACKHOLE_UNSPEC
) {
478 nh
.type
= NEXTHOP_TYPE_IFINDEX
;
479 else if (index
&& gate
)
482 ? NEXTHOP_TYPE_IPV4_IFINDEX
483 : NEXTHOP_TYPE_IPV6_IFINDEX
;
484 else if (!index
&& gate
)
485 nh
.type
= (afi
== AFI_IP
)
489 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
490 nh
.bh_type
= bh_type
;
493 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
494 nh
.bh_type
= bh_type
;
498 memcpy(&nh
.src
, prefsrc
, sz
);
500 memcpy(&nh
.gate
, gate
, sz
);
503 ifp
= if_lookup_by_index_per_ns(
504 zebra_ns_lookup(ns_id
),
507 nh_vrf_id
= ifp
->vrf_id
;
509 nh
.vrf_id
= nh_vrf_id
;
511 rib_add(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
, &p
,
512 &src_p
, &nh
, table
, metric
, mtu
, distance
, tag
);
514 /* This is a multipath route */
516 struct route_entry
*re
;
517 struct rtnexthop
*rtnh
=
518 (struct rtnexthop
*)RTA_DATA(tb
[RTA_MULTIPATH
]);
520 len
= RTA_PAYLOAD(tb
[RTA_MULTIPATH
]);
522 re
= XCALLOC(MTYPE_RE
, sizeof(struct route_entry
));
524 re
->distance
= distance
;
531 re
->uptime
= time(NULL
);
536 if (len
< (int)sizeof(*rtnh
)
537 || rtnh
->rtnh_len
> len
)
540 index
= rtnh
->rtnh_ifindex
;
543 * Yes we are looking this up
544 * for every nexthop and just
545 * using the last one looked
548 ifp
= if_lookup_by_index_per_ns(
549 zebra_ns_lookup(ns_id
),
552 nh_vrf_id
= ifp
->vrf_id
;
555 "%s: Unknown interface %u specified, defaulting to VRF_DEFAULT",
558 nh_vrf_id
= VRF_DEFAULT
;
564 if (rtnh
->rtnh_len
> sizeof(*rtnh
)) {
565 memset(tb
, 0, sizeof(tb
));
566 netlink_parse_rtattr(
567 tb
, RTA_MAX
, RTNH_DATA(rtnh
),
568 rtnh
->rtnh_len
- sizeof(*rtnh
));
575 if (rtm
->rtm_family
== AF_INET
) {
577 route_entry_nexthop_ipv4_ifindex_add(
582 route_entry_nexthop_ipv4_add(
586 } else if (rtm
->rtm_family
589 route_entry_nexthop_ipv6_ifindex_add(
593 route_entry_nexthop_ipv6_add(
598 route_entry_nexthop_ifindex_add(
599 re
, index
, nh_vrf_id
);
601 if (rtnh
->rtnh_len
== 0)
604 len
-= NLMSG_ALIGN(rtnh
->rtnh_len
);
605 rtnh
= RTNH_NEXT(rtnh
);
608 zserv_nexthop_num_warn(__func__
,
609 (const struct prefix
*)&p
,
611 if (re
->nexthop_num
== 0)
614 rib_add_multipath(afi
, SAFI_UNICAST
, &p
,
618 if (!tb
[RTA_MULTIPATH
]) {
620 size_t sz
= (afi
== AFI_IP
) ? 4 : 16;
622 memset(&nh
, 0, sizeof(nh
));
623 if (bh_type
== BLACKHOLE_UNSPEC
) {
625 nh
.type
= NEXTHOP_TYPE_IFINDEX
;
626 else if (index
&& gate
)
629 ? NEXTHOP_TYPE_IPV4_IFINDEX
630 : NEXTHOP_TYPE_IPV6_IFINDEX
;
631 else if (!index
&& gate
)
632 nh
.type
= (afi
== AFI_IP
)
636 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
637 nh
.bh_type
= BLACKHOLE_UNSPEC
;
640 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
641 nh
.bh_type
= bh_type
;
645 memcpy(&nh
.gate
, gate
, sz
);
646 rib_delete(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
,
647 &p
, &src_p
, &nh
, table
, metric
, distance
,
650 /* XXX: need to compare the entire list of nexthops
651 * here for NLM_F_APPEND stupidity */
652 rib_delete(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
,
653 &p
, &src_p
, NULL
, table
, metric
, distance
,
661 static struct mcast_route_data
*mroute
= NULL
;
663 static int netlink_route_change_read_multicast(struct nlmsghdr
*h
,
664 ns_id_t ns_id
, int startup
)
668 struct rtattr
*tb
[RTA_MAX
+ 1];
669 struct mcast_route_data
*m
;
670 struct mcast_route_data mr
;
677 char oif_list
[256] = "\0";
684 memset(&mr
, 0, sizeof(mr
));
690 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
692 memset(tb
, 0, sizeof tb
);
693 netlink_parse_rtattr(tb
, RTA_MAX
, RTM_RTA(rtm
), len
);
696 table
= *(int *)RTA_DATA(tb
[RTA_TABLE
]);
698 table
= rtm
->rtm_table
;
700 vrf
= vrf_lookup_by_table(table
, ns_id
);
703 iif
= *(int *)RTA_DATA(tb
[RTA_IIF
]);
706 m
->sg
.src
= *(struct in_addr
*)RTA_DATA(tb
[RTA_SRC
]);
709 m
->sg
.grp
= *(struct in_addr
*)RTA_DATA(tb
[RTA_DST
]);
711 if ((RTA_EXPIRES
<= RTA_MAX
) && tb
[RTA_EXPIRES
])
712 m
->lastused
= *(unsigned long long *)RTA_DATA(tb
[RTA_EXPIRES
]);
714 if (tb
[RTA_MULTIPATH
]) {
715 struct rtnexthop
*rtnh
=
716 (struct rtnexthop
*)RTA_DATA(tb
[RTA_MULTIPATH
]);
718 len
= RTA_PAYLOAD(tb
[RTA_MULTIPATH
]);
720 if (len
< (int)sizeof(*rtnh
) || rtnh
->rtnh_len
> len
)
723 oif
[oif_count
] = rtnh
->rtnh_ifindex
;
726 if (rtnh
->rtnh_len
== 0)
729 len
-= NLMSG_ALIGN(rtnh
->rtnh_len
);
730 rtnh
= RTNH_NEXT(rtnh
);
734 if (IS_ZEBRA_DEBUG_KERNEL
) {
735 struct interface
*ifp
;
736 strlcpy(sbuf
, inet_ntoa(m
->sg
.src
), sizeof(sbuf
));
737 strlcpy(gbuf
, inet_ntoa(m
->sg
.grp
), sizeof(gbuf
));
738 for (count
= 0; count
< oif_count
; count
++) {
739 ifp
= if_lookup_by_index(oif
[count
], vrf
);
742 sprintf(temp
, "%s ", ifp
->name
);
743 strcat(oif_list
, temp
);
745 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(vrf
);
746 ifp
= if_lookup_by_index(iif
, vrf
);
748 "MCAST VRF: %s(%d) %s (%s,%s) IIF: %s OIF: %s jiffies: %lld",
749 zvrf
->vrf
->name
, vrf
, nl_msg_type_to_str(h
->nlmsg_type
),
750 sbuf
, gbuf
, ifp
->name
, oif_list
, m
->lastused
);
755 int netlink_route_change(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
762 if (!(h
->nlmsg_type
== RTM_NEWROUTE
|| h
->nlmsg_type
== RTM_DELROUTE
)) {
763 /* If this is not route add/delete message print warning. */
764 zlog_warn("Kernel message: %d NS %u\n", h
->nlmsg_type
, ns_id
);
768 if (!(rtm
->rtm_family
== AF_INET
|| rtm
->rtm_family
== AF_INET6
)) {
770 "Invalid address family: %u received from kernel route change: %u",
771 rtm
->rtm_family
, h
->nlmsg_type
);
775 /* Connected route. */
776 if (IS_ZEBRA_DEBUG_KERNEL
)
777 zlog_debug("%s %s %s proto %s NS %u",
778 nl_msg_type_to_str(h
->nlmsg_type
),
779 nl_family_to_str(rtm
->rtm_family
),
780 nl_rttype_to_str(rtm
->rtm_type
),
781 nl_rtproto_to_str(rtm
->rtm_protocol
), ns_id
);
784 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
786 zlog_err("%s: Message received from netlink is of a broken size: %d %zu",
789 (size_t)NLMSG_LENGTH(sizeof(struct rtmsg
)));
793 if (rtm
->rtm_type
== RTN_MULTICAST
)
794 netlink_route_change_read_multicast(h
, ns_id
, startup
);
796 netlink_route_change_read_unicast(h
, ns_id
, startup
);
800 /* Request for specific route information from the kernel */
801 static int netlink_request_route(struct zebra_ns
*zns
, int family
, int type
)
808 /* Form the request, specifying filter (rtattr) if needed. */
809 memset(&req
, 0, sizeof(req
));
810 req
.n
.nlmsg_type
= type
;
811 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
812 req
.rtm
.rtm_family
= family
;
814 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
817 /* Routing table read function using netlink interface. Only called
819 int netlink_route_read(struct zebra_ns
*zns
)
823 /* Get IPv4 routing table. */
824 ret
= netlink_request_route(zns
, AF_INET
, RTM_GETROUTE
);
827 ret
= netlink_parse_info(netlink_route_change_read_unicast
,
828 &zns
->netlink_cmd
, zns
, 0, 1);
832 /* Get IPv6 routing table. */
833 ret
= netlink_request_route(zns
, AF_INET6
, RTM_GETROUTE
);
836 ret
= netlink_parse_info(netlink_route_change_read_unicast
,
837 &zns
->netlink_cmd
, zns
, 0, 1);
844 static void _netlink_route_nl_add_gateway_info(uint8_t route_family
,
846 struct nlmsghdr
*nlmsg
,
847 size_t req_size
, int bytelen
,
848 struct nexthop
*nexthop
)
850 if (route_family
== AF_MPLS
) {
851 struct gw_family_t gw_fam
;
853 gw_fam
.family
= gw_family
;
854 if (gw_family
== AF_INET
)
855 memcpy(&gw_fam
.gate
.ipv4
, &nexthop
->gate
.ipv4
, bytelen
);
857 memcpy(&gw_fam
.gate
.ipv6
, &nexthop
->gate
.ipv6
, bytelen
);
858 addattr_l(nlmsg
, req_size
, RTA_VIA
, &gw_fam
.family
,
861 if (gw_family
== AF_INET
)
862 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
,
863 &nexthop
->gate
.ipv4
, bytelen
);
865 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
,
866 &nexthop
->gate
.ipv6
, bytelen
);
870 static void _netlink_route_rta_add_gateway_info(uint8_t route_family
,
873 struct rtnexthop
*rtnh
,
874 size_t req_size
, int bytelen
,
875 struct nexthop
*nexthop
)
877 if (route_family
== AF_MPLS
) {
878 struct gw_family_t gw_fam
;
880 gw_fam
.family
= gw_family
;
881 if (gw_family
== AF_INET
)
882 memcpy(&gw_fam
.gate
.ipv4
, &nexthop
->gate
.ipv4
, bytelen
);
884 memcpy(&gw_fam
.gate
.ipv6
, &nexthop
->gate
.ipv6
, bytelen
);
885 rta_addattr_l(rta
, req_size
, RTA_VIA
, &gw_fam
.family
,
887 rtnh
->rtnh_len
+= RTA_LENGTH(bytelen
+ 2);
889 if (gw_family
== AF_INET
)
890 rta_addattr_l(rta
, req_size
, RTA_GATEWAY
,
891 &nexthop
->gate
.ipv4
, bytelen
);
893 rta_addattr_l(rta
, req_size
, RTA_GATEWAY
,
894 &nexthop
->gate
.ipv6
, bytelen
);
895 rtnh
->rtnh_len
+= sizeof(struct rtattr
) + bytelen
;
899 /* This function takes a nexthop as argument and adds
900 * the appropriate netlink attributes to an existing
903 * @param routedesc: Human readable description of route type
904 * (direct/recursive, single-/multipath)
905 * @param bytelen: Length of addresses in bytes.
906 * @param nexthop: Nexthop information
907 * @param nlmsg: nlmsghdr structure to fill in.
908 * @param req_size: The size allocated for the message.
910 static void _netlink_route_build_singlepath(const char *routedesc
, int bytelen
,
911 struct nexthop
*nexthop
,
912 struct nlmsghdr
*nlmsg
,
914 size_t req_size
, int cmd
)
916 struct mpls_label_stack
*nh_label
;
917 mpls_lse_t out_lse
[MPLS_MAX_LABELS
];
922 * label_buf is *only* currently used within debugging.
923 * As such when we assign it we are guarding it inside
924 * a debug test. If you want to change this make sure
925 * you fix this assumption
930 for (struct nexthop
*nh
= nexthop
; nh
; nh
= nh
->rparent
) {
933 nh_label
= nh
->nh_label
;
934 if (!nh_label
|| !nh_label
->num_labels
)
937 for (int i
= 0; i
< nh_label
->num_labels
; i
++) {
938 if (nh_label
->label
[i
] == MPLS_LABEL_IMPLICIT_NULL
)
941 if (IS_ZEBRA_DEBUG_KERNEL
) {
943 sprintf(label_buf
, "label %u",
946 sprintf(label_buf1
, "/%u",
948 strlcat(label_buf
, label_buf1
,
953 out_lse
[num_labels
] =
954 mpls_lse_encode(nh_label
->label
[i
], 0, 0, 0);
960 /* Set the BoS bit */
961 out_lse
[num_labels
- 1] |= htonl(1 << MPLS_LS_S_SHIFT
);
963 if (rtmsg
->rtm_family
== AF_MPLS
)
964 addattr_l(nlmsg
, req_size
, RTA_NEWDST
, &out_lse
,
965 num_labels
* sizeof(mpls_lse_t
));
968 uint16_t encap
= LWTUNNEL_ENCAP_MPLS
;
970 addattr_l(nlmsg
, req_size
, RTA_ENCAP_TYPE
, &encap
,
972 nest
= addattr_nest(nlmsg
, req_size
, RTA_ENCAP
);
973 addattr_l(nlmsg
, req_size
, MPLS_IPTUNNEL_DST
, &out_lse
,
974 num_labels
* sizeof(mpls_lse_t
));
975 addattr_nest_end(nlmsg
, nest
);
979 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
))
980 rtmsg
->rtm_flags
|= RTNH_F_ONLINK
;
982 if (rtmsg
->rtm_family
== AF_INET
983 && (nexthop
->type
== NEXTHOP_TYPE_IPV6
984 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
)) {
985 rtmsg
->rtm_flags
|= RTNH_F_ONLINK
;
986 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
, &ipv4_ll
, 4);
987 addattr32(nlmsg
, req_size
, RTA_OIF
, nexthop
->ifindex
);
989 if (nexthop
->rmap_src
.ipv4
.s_addr
&& (cmd
== RTM_NEWROUTE
))
990 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
991 &nexthop
->rmap_src
.ipv4
, bytelen
);
992 else if (nexthop
->src
.ipv4
.s_addr
&& (cmd
== RTM_NEWROUTE
))
993 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
994 &nexthop
->src
.ipv4
, bytelen
);
996 if (IS_ZEBRA_DEBUG_KERNEL
)
998 " 5549: _netlink_route_build_singlepath() (%s): "
999 "nexthop via %s %s if %u(%u)",
1000 routedesc
, ipv4_ll_buf
, label_buf
,
1001 nexthop
->ifindex
, nexthop
->vrf_id
);
1005 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
1006 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
1007 /* Send deletes to the kernel without specifying the next-hop */
1008 if (cmd
!= RTM_DELROUTE
)
1009 _netlink_route_nl_add_gateway_info(
1010 rtmsg
->rtm_family
, AF_INET
, nlmsg
, req_size
,
1013 if (cmd
== RTM_NEWROUTE
) {
1014 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1015 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1016 &nexthop
->rmap_src
.ipv4
, bytelen
);
1017 else if (nexthop
->src
.ipv4
.s_addr
)
1018 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1019 &nexthop
->src
.ipv4
, bytelen
);
1022 if (IS_ZEBRA_DEBUG_KERNEL
)
1024 "netlink_route_multipath() (%s): "
1025 "nexthop via %s %s if %u(%u)",
1026 routedesc
, inet_ntoa(nexthop
->gate
.ipv4
),
1027 label_buf
, nexthop
->ifindex
, nexthop
->vrf_id
);
1030 if (nexthop
->type
== NEXTHOP_TYPE_IPV6
1031 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1032 _netlink_route_nl_add_gateway_info(rtmsg
->rtm_family
, AF_INET6
,
1033 nlmsg
, req_size
, bytelen
,
1036 if (cmd
== RTM_NEWROUTE
) {
1037 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1038 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1039 &nexthop
->rmap_src
.ipv6
, bytelen
);
1040 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1041 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1042 &nexthop
->src
.ipv6
, bytelen
);
1045 if (IS_ZEBRA_DEBUG_KERNEL
)
1047 "netlink_route_multipath() (%s): "
1048 "nexthop via %s %s if %u(%u)",
1049 routedesc
, inet6_ntoa(nexthop
->gate
.ipv6
),
1050 label_buf
, nexthop
->ifindex
, nexthop
->vrf_id
);
1054 * We have the ifindex so we should always send it
1055 * This is especially useful if we are doing route
1058 if (nexthop
->type
!= NEXTHOP_TYPE_BLACKHOLE
)
1059 addattr32(nlmsg
, req_size
, RTA_OIF
, nexthop
->ifindex
);
1061 if (nexthop
->type
== NEXTHOP_TYPE_IFINDEX
1062 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
1063 if (cmd
== RTM_NEWROUTE
) {
1064 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1065 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1066 &nexthop
->rmap_src
.ipv4
, bytelen
);
1067 else if (nexthop
->src
.ipv4
.s_addr
)
1068 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1069 &nexthop
->src
.ipv4
, bytelen
);
1072 if (IS_ZEBRA_DEBUG_KERNEL
)
1074 "netlink_route_multipath() (%s): "
1075 "nexthop via if %u(%u)",
1076 routedesc
, nexthop
->ifindex
, nexthop
->vrf_id
);
1079 if (nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1080 if (cmd
== RTM_NEWROUTE
) {
1081 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1082 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1083 &nexthop
->rmap_src
.ipv6
, bytelen
);
1084 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1085 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1086 &nexthop
->src
.ipv6
, bytelen
);
1089 if (IS_ZEBRA_DEBUG_KERNEL
)
1091 "netlink_route_multipath() (%s): "
1092 "nexthop via if %u(%u)",
1093 routedesc
, nexthop
->ifindex
, nexthop
->vrf_id
);
1097 /* This function takes a nexthop as argument and
1098 * appends to the given rtattr/rtnexthop pair the
1099 * representation of the nexthop. If the nexthop
1100 * defines a preferred source, the src parameter
1101 * will be modified to point to that src, otherwise
1102 * it will be kept unmodified.
1104 * @param routedesc: Human readable description of route type
1105 * (direct/recursive, single-/multipath)
1106 * @param bytelen: Length of addresses in bytes.
1107 * @param nexthop: Nexthop information
1108 * @param rta: rtnetlink attribute structure
1109 * @param rtnh: pointer to an rtnetlink nexthop structure
1110 * @param src: pointer pointing to a location where
1111 * the prefsrc should be stored.
1113 static void _netlink_route_build_multipath(const char *routedesc
, int bytelen
,
1114 struct nexthop
*nexthop
,
1116 struct rtnexthop
*rtnh
,
1117 struct rtmsg
*rtmsg
,
1120 struct mpls_label_stack
*nh_label
;
1121 mpls_lse_t out_lse
[MPLS_MAX_LABELS
];
1123 char label_buf
[256];
1125 rtnh
->rtnh_len
= sizeof(*rtnh
);
1126 rtnh
->rtnh_flags
= 0;
1127 rtnh
->rtnh_hops
= 0;
1128 rta
->rta_len
+= rtnh
->rtnh_len
;
1131 * label_buf is *only* currently used within debugging.
1132 * As such when we assign it we are guarding it inside
1133 * a debug test. If you want to change this make sure
1134 * you fix this assumption
1136 label_buf
[0] = '\0';
1139 for (struct nexthop
*nh
= nexthop
; nh
; nh
= nh
->rparent
) {
1140 char label_buf1
[20];
1142 nh_label
= nh
->nh_label
;
1143 if (!nh_label
|| !nh_label
->num_labels
)
1146 for (int i
= 0; i
< nh_label
->num_labels
; i
++) {
1147 if (nh_label
->label
[i
] == MPLS_LABEL_IMPLICIT_NULL
)
1150 if (IS_ZEBRA_DEBUG_KERNEL
) {
1152 sprintf(label_buf
, "label %u",
1153 nh_label
->label
[i
]);
1155 sprintf(label_buf1
, "/%u",
1156 nh_label
->label
[i
]);
1157 strlcat(label_buf
, label_buf1
,
1162 out_lse
[num_labels
] =
1163 mpls_lse_encode(nh_label
->label
[i
], 0, 0, 0);
1169 /* Set the BoS bit */
1170 out_lse
[num_labels
- 1] |= htonl(1 << MPLS_LS_S_SHIFT
);
1172 if (rtmsg
->rtm_family
== AF_MPLS
) {
1173 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_NEWDST
,
1175 num_labels
* sizeof(mpls_lse_t
));
1177 RTA_LENGTH(num_labels
* sizeof(mpls_lse_t
));
1179 struct rtattr
*nest
;
1180 uint16_t encap
= LWTUNNEL_ENCAP_MPLS
;
1181 int len
= rta
->rta_len
;
1183 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_ENCAP_TYPE
,
1184 &encap
, sizeof(uint16_t));
1185 nest
= rta_nest(rta
, NL_PKT_BUF_SIZE
, RTA_ENCAP
);
1186 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, MPLS_IPTUNNEL_DST
,
1188 num_labels
* sizeof(mpls_lse_t
));
1189 rta_nest_end(rta
, nest
);
1190 rtnh
->rtnh_len
+= rta
->rta_len
- len
;
1194 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
))
1195 rtnh
->rtnh_flags
|= RTNH_F_ONLINK
;
1197 if (rtmsg
->rtm_family
== AF_INET
1198 && (nexthop
->type
== NEXTHOP_TYPE_IPV6
1199 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
)) {
1201 rtnh
->rtnh_flags
|= RTNH_F_ONLINK
;
1202 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_GATEWAY
, &ipv4_ll
,
1204 rtnh
->rtnh_len
+= sizeof(struct rtattr
) + bytelen
;
1205 rtnh
->rtnh_ifindex
= nexthop
->ifindex
;
1207 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1208 *src
= &nexthop
->rmap_src
;
1209 else if (nexthop
->src
.ipv4
.s_addr
)
1210 *src
= &nexthop
->src
;
1212 if (IS_ZEBRA_DEBUG_KERNEL
)
1214 " 5549: netlink_route_build_multipath() (%s): "
1215 "nexthop via %s %s if %u",
1216 routedesc
, ipv4_ll_buf
, label_buf
,
1221 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
1222 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
1223 _netlink_route_rta_add_gateway_info(rtmsg
->rtm_family
, AF_INET
,
1224 rta
, rtnh
, NL_PKT_BUF_SIZE
,
1226 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1227 *src
= &nexthop
->rmap_src
;
1228 else if (nexthop
->src
.ipv4
.s_addr
)
1229 *src
= &nexthop
->src
;
1231 if (IS_ZEBRA_DEBUG_KERNEL
)
1233 "netlink_route_multipath() (%s): "
1234 "nexthop via %s %s if %u",
1235 routedesc
, inet_ntoa(nexthop
->gate
.ipv4
),
1236 label_buf
, nexthop
->ifindex
);
1238 if (nexthop
->type
== NEXTHOP_TYPE_IPV6
1239 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1240 _netlink_route_rta_add_gateway_info(rtmsg
->rtm_family
, AF_INET6
,
1241 rta
, rtnh
, NL_PKT_BUF_SIZE
,
1244 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1245 *src
= &nexthop
->rmap_src
;
1246 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1247 *src
= &nexthop
->src
;
1249 if (IS_ZEBRA_DEBUG_KERNEL
)
1251 "netlink_route_multipath() (%s): "
1252 "nexthop via %s %s if %u",
1253 routedesc
, inet6_ntoa(nexthop
->gate
.ipv6
),
1254 label_buf
, nexthop
->ifindex
);
1258 * We have figured out the ifindex so we should always send it
1259 * This is especially useful if we are doing route
1262 if (nexthop
->type
!= NEXTHOP_TYPE_BLACKHOLE
)
1263 rtnh
->rtnh_ifindex
= nexthop
->ifindex
;
1266 if (nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
1267 || nexthop
->type
== NEXTHOP_TYPE_IFINDEX
) {
1268 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1269 *src
= &nexthop
->rmap_src
;
1270 else if (nexthop
->src
.ipv4
.s_addr
)
1271 *src
= &nexthop
->src
;
1273 if (IS_ZEBRA_DEBUG_KERNEL
)
1275 "netlink_route_multipath() (%s): "
1276 "nexthop via if %u",
1277 routedesc
, nexthop
->ifindex
);
1281 static inline void _netlink_mpls_build_singlepath(const char *routedesc
,
1282 zebra_nhlfe_t
*nhlfe
,
1283 struct nlmsghdr
*nlmsg
,
1284 struct rtmsg
*rtmsg
,
1285 size_t req_size
, int cmd
)
1290 family
= NHLFE_FAMILY(nhlfe
);
1291 bytelen
= (family
== AF_INET
? 4 : 16);
1292 _netlink_route_build_singlepath(routedesc
, bytelen
, nhlfe
->nexthop
,
1293 nlmsg
, rtmsg
, req_size
, cmd
);
1298 _netlink_mpls_build_multipath(const char *routedesc
, zebra_nhlfe_t
*nhlfe
,
1299 struct rtattr
*rta
, struct rtnexthop
*rtnh
,
1300 struct rtmsg
*rtmsg
, union g_addr
**src
)
1305 family
= NHLFE_FAMILY(nhlfe
);
1306 bytelen
= (family
== AF_INET
? 4 : 16);
1307 _netlink_route_build_multipath(routedesc
, bytelen
, nhlfe
->nexthop
, rta
,
1312 /* Log debug information for netlink_route_multipath
1313 * if debug logging is enabled.
1315 * @param cmd: Netlink command which is to be processed
1316 * @param p: Prefix for which the change is due
1317 * @param family: Address family which the change concerns
1318 * @param zvrf: The vrf we are in
1319 * @param tableid: The table we are working on
1321 static void _netlink_route_debug(int cmd
, const struct prefix
*p
,
1322 int family
, vrf_id_t vrfid
,
1325 if (IS_ZEBRA_DEBUG_KERNEL
) {
1326 char buf
[PREFIX_STRLEN
];
1328 "netlink_route_multipath(): %s %s vrf %u(%u)",
1329 nl_msg_type_to_str(cmd
),
1330 prefix2str(p
, buf
, sizeof(buf
)),
1335 static void _netlink_mpls_debug(int cmd
, uint32_t label
, const char *routedesc
)
1337 if (IS_ZEBRA_DEBUG_KERNEL
)
1338 zlog_debug("netlink_mpls_multipath() (%s): %s %u/20", routedesc
,
1339 nl_msg_type_to_str(cmd
), label
);
1342 static int netlink_neigh_update(int cmd
, int ifindex
, uint32_t addr
, char *lla
,
1343 int llalen
, ns_id_t ns_id
)
1351 struct zebra_ns
*zns
= zebra_ns_lookup(ns_id
);
1353 memset(&req
, 0, sizeof(req
));
1355 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1356 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
1357 req
.n
.nlmsg_type
= cmd
; // RTM_NEWNEIGH or RTM_DELNEIGH
1358 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1360 req
.ndm
.ndm_family
= AF_INET
;
1361 req
.ndm
.ndm_state
= NUD_PERMANENT
;
1362 req
.ndm
.ndm_ifindex
= ifindex
;
1363 req
.ndm
.ndm_type
= RTN_UNICAST
;
1365 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &addr
, 4);
1366 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, lla
, llalen
);
1368 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1372 /* Routing table change via netlink interface. */
1373 /* Update flag indicates whether this is a "replace" or not. */
1374 static int netlink_route_multipath(int cmd
, const struct prefix
*p
,
1375 const struct prefix
*src_p
,
1376 struct route_entry
*re
,
1380 struct sockaddr_nl snl
;
1381 struct nexthop
*nexthop
= NULL
;
1382 unsigned int nexthop_num
;
1383 int family
= PREFIX_FAMILY(p
);
1384 const char *routedesc
;
1391 char buf
[NL_PKT_BUF_SIZE
];
1394 struct zebra_ns
*zns
;
1395 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1398 memset(&req
, 0, sizeof req
- NL_PKT_BUF_SIZE
);
1400 bytelen
= (family
== AF_INET
? 4 : 16);
1402 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
1403 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
1404 if ((cmd
== RTM_NEWROUTE
) && update
)
1405 req
.n
.nlmsg_flags
|= NLM_F_REPLACE
;
1406 req
.n
.nlmsg_type
= cmd
;
1407 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1409 req
.r
.rtm_family
= family
;
1410 req
.r
.rtm_dst_len
= p
->prefixlen
;
1411 req
.r
.rtm_src_len
= src_p
? src_p
->prefixlen
: 0;
1412 req
.r
.rtm_protocol
= zebra2proto(re
->type
);
1413 req
.r
.rtm_scope
= RT_SCOPE_UNIVERSE
;
1416 * blackhole routes are not RTN_UNICAST, they are
1417 * RTN_ BLACKHOLE|UNREACHABLE|PROHIBIT
1418 * so setting this value as a RTN_UNICAST would
1419 * cause the route lookup of just the prefix
1420 * to fail. So no need to specify this for
1421 * the RTM_DELROUTE case
1423 if (cmd
!= RTM_DELROUTE
)
1424 req
.r
.rtm_type
= RTN_UNICAST
;
1426 addattr_l(&req
.n
, sizeof req
, RTA_DST
, &p
->u
.prefix
, bytelen
);
1428 addattr_l(&req
.n
, sizeof req
, RTA_SRC
, &src_p
->u
.prefix
,
1432 /* Hardcode the metric for all routes coming from zebra. Metric isn't
1434 * either by the kernel or by zebra. Its purely for calculating best
1436 * by the routing protocol and for communicating with protocol peers.
1438 addattr32(&req
.n
, sizeof req
, RTA_PRIORITY
, NL_DEFAULT_ROUTE_METRIC
);
1439 #if defined(SUPPORT_REALMS)
1440 if (re
->tag
> 0 && re
->tag
<= 255)
1441 addattr32(&req
.n
, sizeof req
, RTA_FLOW
, re
->tag
);
1443 /* Table corresponding to this route. */
1444 if (re
->table
< 256)
1445 req
.r
.rtm_table
= re
->table
;
1447 req
.r
.rtm_table
= RT_TABLE_UNSPEC
;
1448 addattr32(&req
.n
, sizeof req
, RTA_TABLE
, re
->table
);
1451 _netlink_route_debug(cmd
, p
, family
, zvrf_id(zvrf
), re
->table
);
1454 * If we are not updating the route and we have received
1455 * a route delete, then all we need to fill in is the
1456 * prefix information to tell the kernel to schwack
1459 if (!update
&& cmd
== RTM_DELROUTE
)
1462 if (re
->mtu
|| re
->nexthop_mtu
) {
1463 char buf
[NL_PKT_BUF_SIZE
];
1464 struct rtattr
*rta
= (void *)buf
;
1465 uint32_t mtu
= re
->mtu
;
1466 if (!mtu
|| (re
->nexthop_mtu
&& re
->nexthop_mtu
< mtu
))
1467 mtu
= re
->nexthop_mtu
;
1468 rta
->rta_type
= RTA_METRICS
;
1469 rta
->rta_len
= RTA_LENGTH(0);
1470 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTAX_MTU
, &mtu
, sizeof mtu
);
1471 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_METRICS
, RTA_DATA(rta
),
1475 /* Count overall nexthops so we can decide whether to use singlepath
1476 * or multipath case. */
1478 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1479 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
))
1481 if (cmd
== RTM_NEWROUTE
&& !NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1483 if (cmd
== RTM_DELROUTE
1484 && !CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
1490 /* Singlepath case. */
1491 if (nexthop_num
== 1 || multipath_num
== 1) {
1493 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1495 * So we want to cover 2 types of blackhole
1497 * 1) A normal blackhole route( ala from a static
1499 * 2) A recursively resolved blackhole route
1501 if (nexthop
->type
== NEXTHOP_TYPE_BLACKHOLE
) {
1502 switch (nexthop
->bh_type
) {
1503 case BLACKHOLE_ADMINPROHIB
:
1504 req
.r
.rtm_type
= RTN_PROHIBIT
;
1506 case BLACKHOLE_REJECT
:
1507 req
.r
.rtm_type
= RTN_UNREACHABLE
;
1510 req
.r
.rtm_type
= RTN_BLACKHOLE
;
1515 if (CHECK_FLAG(nexthop
->flags
,
1516 NEXTHOP_FLAG_RECURSIVE
)) {
1518 if (family
== AF_INET
) {
1519 if (nexthop
->rmap_src
.ipv4
1526 } else if (nexthop
->src
.ipv4
1534 } else if (family
== AF_INET6
) {
1535 if (!IN6_IS_ADDR_UNSPECIFIED(
1543 !IN6_IS_ADDR_UNSPECIFIED(
1556 if ((cmd
== RTM_NEWROUTE
1557 && NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1558 || (cmd
== RTM_DELROUTE
1559 && CHECK_FLAG(nexthop
->flags
,
1560 NEXTHOP_FLAG_FIB
))) {
1561 routedesc
= nexthop
->rparent
1562 ? "recursive, single-path"
1565 _netlink_route_build_singlepath(
1566 routedesc
, bytelen
, nexthop
, &req
.n
,
1567 &req
.r
, sizeof req
, cmd
);
1572 if (setsrc
&& (cmd
== RTM_NEWROUTE
)) {
1573 if (family
== AF_INET
)
1574 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1575 &src
.ipv4
, bytelen
);
1576 else if (family
== AF_INET6
)
1577 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1578 &src
.ipv6
, bytelen
);
1581 char buf
[NL_PKT_BUF_SIZE
];
1582 struct rtattr
*rta
= (void *)buf
;
1583 struct rtnexthop
*rtnh
;
1584 union g_addr
*src1
= NULL
;
1586 rta
->rta_type
= RTA_MULTIPATH
;
1587 rta
->rta_len
= RTA_LENGTH(0);
1588 rtnh
= RTA_DATA(rta
);
1591 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1592 if (nexthop_num
>= multipath_num
)
1595 if (CHECK_FLAG(nexthop
->flags
,
1596 NEXTHOP_FLAG_RECURSIVE
)) {
1597 /* This only works for IPv4 now */
1599 if (family
== AF_INET
) {
1600 if (nexthop
->rmap_src
.ipv4
1607 } else if (nexthop
->src
.ipv4
1615 } else if (family
== AF_INET6
) {
1616 if (!IN6_IS_ADDR_UNSPECIFIED(
1624 !IN6_IS_ADDR_UNSPECIFIED(
1637 if ((cmd
== RTM_NEWROUTE
1638 && NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1639 || (cmd
== RTM_DELROUTE
1640 && CHECK_FLAG(nexthop
->flags
,
1641 NEXTHOP_FLAG_FIB
))) {
1642 routedesc
= nexthop
->rparent
1643 ? "recursive, multipath"
1647 _netlink_route_build_multipath(
1648 routedesc
, bytelen
, nexthop
, rta
, rtnh
,
1650 rtnh
= RTNH_NEXT(rtnh
);
1652 if (!setsrc
&& src1
) {
1653 if (family
== AF_INET
)
1654 src
.ipv4
= src1
->ipv4
;
1655 else if (family
== AF_INET6
)
1656 src
.ipv6
= src1
->ipv6
;
1662 if (setsrc
&& (cmd
== RTM_NEWROUTE
)) {
1663 if (family
== AF_INET
)
1664 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1665 &src
.ipv4
, bytelen
);
1666 else if (family
== AF_INET6
)
1667 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1668 &src
.ipv6
, bytelen
);
1669 if (IS_ZEBRA_DEBUG_KERNEL
)
1670 zlog_debug("Setting source");
1673 if (rta
->rta_len
> RTA_LENGTH(0))
1674 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_MULTIPATH
,
1675 RTA_DATA(rta
), RTA_PAYLOAD(rta
));
1678 /* If there is no useful nexthop then return. */
1679 if (nexthop_num
== 0) {
1680 if (IS_ZEBRA_DEBUG_KERNEL
)
1682 "netlink_route_multipath(): No useful nexthop.");
1688 /* Destination netlink address. */
1689 memset(&snl
, 0, sizeof snl
);
1690 snl
.nl_family
= AF_NETLINK
;
1692 /* Talk to netlink socket. */
1693 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1697 int kernel_get_ipmr_sg_stats(struct zebra_vrf
*zvrf
, void *in
)
1700 struct mcast_route_data
*mr
= (struct mcast_route_data
*)in
;
1708 struct zebra_ns
*zns
;
1711 memset(&req
, 0, sizeof(req
));
1713 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1714 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
1715 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1717 req
.ndm
.ndm_family
= RTNL_FAMILY_IPMR
;
1718 req
.n
.nlmsg_type
= RTM_GETROUTE
;
1720 addattr_l(&req
.n
, sizeof(req
), RTA_IIF
, &mroute
->ifindex
, 4);
1721 addattr_l(&req
.n
, sizeof(req
), RTA_OIF
, &mroute
->ifindex
, 4);
1722 addattr_l(&req
.n
, sizeof(req
), RTA_SRC
, &mroute
->sg
.src
.s_addr
, 4);
1723 addattr_l(&req
.n
, sizeof(req
), RTA_DST
, &mroute
->sg
.grp
.s_addr
, 4);
1724 addattr_l(&req
.n
, sizeof(req
), RTA_TABLE
, &zvrf
->table_id
, 4);
1726 suc
= netlink_talk(netlink_route_change_read_multicast
, &req
.n
,
1727 &zns
->netlink_cmd
, zns
, 0);
1733 enum dp_req_result
kernel_route_rib(struct route_node
*rn
,
1734 const struct prefix
*p
,
1735 const struct prefix
*src_p
,
1736 struct route_entry
*old
,
1737 struct route_entry
*new)
1744 if (p
->family
== AF_INET
|| v6_rr_semantics
)
1745 ret
= netlink_route_multipath(RTM_NEWROUTE
, p
, src_p
,
1746 new, (old
) ? 1 : 0);
1749 * So v6 route replace semantics are not in
1750 * the kernel at this point as I understand it.
1751 * So let's do a delete than an add.
1752 * In the future once v6 route replace semantics
1753 * are in we can figure out what to do here to
1754 * allow working with old and new kernels.
1756 * I'm also intentionally ignoring the failure case
1757 * of the route delete. If that happens yeah we're
1761 netlink_route_multipath(RTM_DELROUTE
, p
, src_p
,
1763 ret
= netlink_route_multipath(RTM_NEWROUTE
, p
, src_p
,
1766 kernel_route_rib_pass_fail(rn
, p
, new,
1767 (!ret
) ? DP_INSTALL_SUCCESS
1768 : DP_INSTALL_FAILURE
);
1769 return DP_REQUEST_SUCCESS
;
1773 ret
= netlink_route_multipath(RTM_DELROUTE
, p
, src_p
, old
, 0);
1775 kernel_route_rib_pass_fail(rn
, p
, old
,
1776 (!ret
) ? DP_DELETE_SUCCESS
1777 : DP_DELETE_FAILURE
);
1780 return DP_REQUEST_SUCCESS
;
1783 int kernel_neigh_update(int add
, int ifindex
, uint32_t addr
, char *lla
,
1784 int llalen
, ns_id_t ns_id
)
1786 return netlink_neigh_update(add
? RTM_NEWNEIGH
: RTM_DELNEIGH
, ifindex
,
1787 addr
, lla
, llalen
, ns_id
);
1791 * Add remote VTEP to the flood list for this VxLAN interface (VNI). This
1792 * is done by adding an FDB entry with a MAC of 00:00:00:00:00:00.
1794 static int netlink_vxlan_flood_list_update(struct interface
*ifp
,
1795 struct in_addr
*vtep_ip
, int cmd
)
1797 struct zebra_ns
*zns
;
1803 uint8_t dst_mac
[6] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0};
1804 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
1807 memset(&req
, 0, sizeof(req
));
1809 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1810 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
1811 if (cmd
== RTM_NEWNEIGH
)
1812 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_APPEND
);
1813 req
.n
.nlmsg_type
= cmd
;
1814 req
.ndm
.ndm_family
= PF_BRIDGE
;
1815 req
.ndm
.ndm_state
= NUD_NOARP
| NUD_PERMANENT
;
1816 req
.ndm
.ndm_flags
|= NTF_SELF
; // Handle by "self", not "master"
1819 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, &dst_mac
, 6);
1820 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
1821 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &vtep_ip
->s_addr
, 4);
1823 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1828 * Add remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1830 * a "flood" MAC FDB entry.
1832 int kernel_add_vtep(vni_t vni
, struct interface
*ifp
, struct in_addr
*vtep_ip
)
1834 if (IS_ZEBRA_DEBUG_VXLAN
)
1835 zlog_debug("Install %s into flood list for VNI %u intf %s(%u)",
1836 inet_ntoa(*vtep_ip
), vni
, ifp
->name
, ifp
->ifindex
);
1838 return netlink_vxlan_flood_list_update(ifp
, vtep_ip
, RTM_NEWNEIGH
);
1842 * Remove remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1843 * deleting the "flood" MAC FDB entry.
1845 int kernel_del_vtep(vni_t vni
, struct interface
*ifp
, struct in_addr
*vtep_ip
)
1847 if (IS_ZEBRA_DEBUG_VXLAN
)
1849 "Uninstall %s from flood list for VNI %u intf %s(%u)",
1850 inet_ntoa(*vtep_ip
), vni
, ifp
->name
, ifp
->ifindex
);
1852 return netlink_vxlan_flood_list_update(ifp
, vtep_ip
, RTM_DELNEIGH
);
1856 #define NDA_RTA(r) \
1857 ((struct rtattr *)(((char *)(r)) + NLMSG_ALIGN(sizeof(struct ndmsg))))
1860 static int netlink_macfdb_change(struct nlmsghdr
*h
, int len
, ns_id_t ns_id
)
1863 struct interface
*ifp
;
1864 struct zebra_if
*zif
;
1865 struct rtattr
*tb
[NDA_MAX
+ 1];
1866 struct interface
*br_if
;
1869 struct prefix vtep_ip
;
1870 int vid_present
= 0, dst_present
= 0;
1871 char buf
[ETHER_ADDR_STRLEN
];
1876 ndm
= NLMSG_DATA(h
);
1878 /* We only process macfdb notifications if EVPN is enabled */
1879 if (!is_evpn_enabled())
1882 /* The interface should exist. */
1883 ifp
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
1885 if (!ifp
|| !ifp
->info
)
1888 /* The interface should be something we're interested in. */
1889 if (!IS_ZEBRA_IF_BRIDGE_SLAVE(ifp
))
1892 /* Drop "permanent" entries. */
1893 if (ndm
->ndm_state
& NUD_PERMANENT
)
1896 zif
= (struct zebra_if
*)ifp
->info
;
1897 if ((br_if
= zif
->brslave_info
.br_if
) == NULL
) {
1898 zlog_warn("%s family %s IF %s(%u) brIF %u - no bridge master",
1899 nl_msg_type_to_str(h
->nlmsg_type
),
1900 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1901 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
);
1905 /* Parse attributes and extract fields of interest. */
1906 memset(tb
, 0, sizeof tb
);
1907 netlink_parse_rtattr(tb
, NDA_MAX
, NDA_RTA(ndm
), len
);
1909 if (!tb
[NDA_LLADDR
]) {
1910 zlog_warn("%s family %s IF %s(%u) brIF %u - no LLADDR",
1911 nl_msg_type_to_str(h
->nlmsg_type
),
1912 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1913 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
);
1917 if (RTA_PAYLOAD(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
1919 "%s family %s IF %s(%u) brIF %u - LLADDR is not MAC, len %lu",
1920 nl_msg_type_to_str(h
->nlmsg_type
),
1921 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1922 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
,
1923 (unsigned long)RTA_PAYLOAD(tb
[NDA_LLADDR
]));
1927 memcpy(&mac
, RTA_DATA(tb
[NDA_LLADDR
]), ETH_ALEN
);
1929 if ((NDA_VLAN
<= NDA_MAX
) && tb
[NDA_VLAN
]) {
1931 vid
= *(uint16_t *)RTA_DATA(tb
[NDA_VLAN
]);
1932 sprintf(vid_buf
, " VLAN %u", vid
);
1936 /* TODO: Only IPv4 supported now. */
1938 vtep_ip
.family
= AF_INET
;
1939 vtep_ip
.prefixlen
= IPV4_MAX_BITLEN
;
1940 memcpy(&(vtep_ip
.u
.prefix4
.s_addr
), RTA_DATA(tb
[NDA_DST
]),
1942 sprintf(dst_buf
, " dst %s", inet_ntoa(vtep_ip
.u
.prefix4
));
1945 sticky
= (ndm
->ndm_state
& NUD_NOARP
) ? 1 : 0;
1947 if (IS_ZEBRA_DEBUG_KERNEL
)
1948 zlog_debug("Rx %s family %s IF %s(%u)%s %sMAC %s%s",
1949 nl_msg_type_to_str(h
->nlmsg_type
),
1950 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1951 ndm
->ndm_ifindex
, vid_present
? vid_buf
: "",
1952 sticky
? "sticky " : "",
1953 prefix_mac2str(&mac
, buf
, sizeof(buf
)),
1954 dst_present
? dst_buf
: "");
1956 if (filter_vlan
&& vid
!= filter_vlan
)
1959 /* If add or update, do accordingly if learnt on a "local" interface; if
1960 * the notification is over VxLAN, this has to be related to
1962 * so perform an implicit delete of any local entry (if it exists).
1964 if (h
->nlmsg_type
== RTM_NEWNEIGH
) {
1965 /* Drop "permanent" entries. */
1966 if (ndm
->ndm_state
& NUD_PERMANENT
)
1969 if (IS_ZEBRA_IF_VXLAN(ifp
))
1970 return zebra_vxlan_check_del_local_mac(ifp
, br_if
, &mac
,
1973 return zebra_vxlan_local_mac_add_update(ifp
, br_if
, &mac
, vid
,
1977 /* This is a delete notification.
1978 * 1. For a MAC over VxLan, check if it needs to be refreshed(readded)
1979 * 2. For a MAC over "local" interface, delete the mac
1980 * Note: We will get notifications from both bridge driver and VxLAN
1982 * Ignore the notification from VxLan driver as it is also generated
1983 * when mac moves from remote to local.
1988 if (IS_ZEBRA_IF_VXLAN(ifp
))
1989 return zebra_vxlan_check_readd_remote_mac(ifp
, br_if
, &mac
,
1992 return zebra_vxlan_local_mac_del(ifp
, br_if
, &mac
, vid
);
1995 static int netlink_macfdb_table(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
2000 if (h
->nlmsg_type
!= RTM_NEWNEIGH
)
2003 /* Length validity. */
2004 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2008 /* We are interested only in AF_BRIDGE notifications. */
2009 ndm
= NLMSG_DATA(h
);
2010 if (ndm
->ndm_family
!= AF_BRIDGE
)
2013 return netlink_macfdb_change(h
, len
, ns_id
);
2016 /* Request for MAC FDB information from the kernel */
2017 static int netlink_request_macs(struct zebra_ns
*zns
, int family
, int type
,
2018 ifindex_t master_ifindex
)
2022 struct ifinfomsg ifm
;
2026 /* Form the request, specifying filter (rtattr) if needed. */
2027 memset(&req
, 0, sizeof(req
));
2028 req
.n
.nlmsg_type
= type
;
2029 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ifinfomsg
));
2030 req
.ifm
.ifi_family
= family
;
2032 addattr32(&req
.n
, sizeof(req
), IFLA_MASTER
, master_ifindex
);
2034 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
2038 * MAC forwarding database read using netlink interface. This is invoked
2041 int netlink_macfdb_read(struct zebra_ns
*zns
)
2045 /* Get bridge FDB table. */
2046 ret
= netlink_request_macs(zns
, AF_BRIDGE
, RTM_GETNEIGH
, 0);
2049 /* We are reading entire table. */
2051 ret
= netlink_parse_info(netlink_macfdb_table
, &zns
->netlink_cmd
, zns
,
2058 * MAC forwarding database read using netlink interface. This is for a
2059 * specific bridge and matching specific access VLAN (if VLAN-aware bridge).
2061 int netlink_macfdb_read_for_bridge(struct zebra_ns
*zns
, struct interface
*ifp
,
2062 struct interface
*br_if
)
2064 struct zebra_if
*br_zif
;
2065 struct zebra_if
*zif
;
2066 struct zebra_l2info_vxlan
*vxl
;
2070 /* Save VLAN we're filtering on, if needed. */
2071 br_zif
= (struct zebra_if
*)br_if
->info
;
2072 zif
= (struct zebra_if
*)ifp
->info
;
2073 vxl
= &zif
->l2info
.vxl
;
2074 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif
))
2075 filter_vlan
= vxl
->access_vlan
;
2077 /* Get bridge FDB table for specific bridge - we do the VLAN filtering.
2079 ret
= netlink_request_macs(zns
, AF_BRIDGE
, RTM_GETNEIGH
,
2083 ret
= netlink_parse_info(netlink_macfdb_table
, &zns
->netlink_cmd
, zns
,
2086 /* Reset VLAN filter. */
2091 static int netlink_macfdb_update(struct interface
*ifp
, vlanid_t vid
,
2092 struct ethaddr
*mac
, struct in_addr vtep_ip
,
2093 int local
, int cmd
, uint8_t sticky
)
2095 struct zebra_ns
*zns
;
2102 struct zebra_if
*zif
;
2103 struct interface
*br_if
;
2104 struct zebra_if
*br_zif
;
2105 char buf
[ETHER_ADDR_STRLEN
];
2106 int vid_present
= 0, dst_present
= 0;
2109 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
2113 if ((br_if
= zif
->brslave_info
.br_if
) == NULL
) {
2114 zlog_warn("MAC %s on IF %s(%u) - no mapping to bridge",
2115 (cmd
== RTM_NEWNEIGH
) ? "add" : "del", ifp
->name
,
2120 memset(&req
, 0, sizeof(req
));
2122 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2123 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
2124 if (cmd
== RTM_NEWNEIGH
)
2125 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_REPLACE
);
2126 req
.n
.nlmsg_type
= cmd
;
2127 req
.ndm
.ndm_family
= AF_BRIDGE
;
2128 req
.ndm
.ndm_flags
|= NTF_SELF
| NTF_MASTER
;
2129 req
.ndm
.ndm_state
= NUD_REACHABLE
;
2132 req
.ndm
.ndm_state
|= NUD_NOARP
;
2134 req
.ndm
.ndm_flags
|= NTF_EXT_LEARNED
;
2136 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, mac
, 6);
2137 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
2139 dst_alen
= 4; // TODO: hardcoded
2140 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &vtep_ip
, dst_alen
);
2142 sprintf(dst_buf
, " dst %s", inet_ntoa(vtep_ip
));
2144 br_zif
= (struct zebra_if
*)br_if
->info
;
2145 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif
) && vid
> 0) {
2146 addattr16(&req
.n
, sizeof(req
), NDA_VLAN
, vid
);
2148 sprintf(vid_buf
, " VLAN %u", vid
);
2150 addattr32(&req
.n
, sizeof(req
), NDA_MASTER
, br_if
->ifindex
);
2152 if (IS_ZEBRA_DEBUG_KERNEL
)
2153 zlog_debug("Tx %s family %s IF %s(%u)%s %sMAC %s%s",
2154 nl_msg_type_to_str(cmd
),
2155 nl_family_to_str(req
.ndm
.ndm_family
), ifp
->name
,
2156 ifp
->ifindex
, vid_present
? vid_buf
: "",
2157 sticky
? "sticky " : "",
2158 prefix_mac2str(mac
, buf
, sizeof(buf
)),
2159 dst_present
? dst_buf
: "");
2161 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2166 (NUD_PERMANENT | NUD_NOARP | NUD_REACHABLE | NUD_PROBE | NUD_STALE \
2169 static int netlink_ipneigh_change(struct nlmsghdr
*h
, int len
, ns_id_t ns_id
)
2172 struct interface
*ifp
;
2173 struct zebra_if
*zif
;
2174 struct rtattr
*tb
[NDA_MAX
+ 1];
2175 struct interface
*link_if
;
2178 char buf
[ETHER_ADDR_STRLEN
];
2179 char buf2
[INET6_ADDRSTRLEN
];
2180 int mac_present
= 0;
2181 uint8_t ext_learned
;
2182 uint8_t router_flag
;
2184 ndm
= NLMSG_DATA(h
);
2186 /* The interface should exist. */
2187 ifp
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
2189 if (!ifp
|| !ifp
->info
)
2192 zif
= (struct zebra_if
*)ifp
->info
;
2194 /* Parse attributes and extract fields of interest. */
2195 memset(tb
, 0, sizeof tb
);
2196 netlink_parse_rtattr(tb
, NDA_MAX
, NDA_RTA(ndm
), len
);
2199 zlog_warn("%s family %s IF %s(%u) - no DST",
2200 nl_msg_type_to_str(h
->nlmsg_type
),
2201 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2206 memset(&ip
, 0, sizeof(struct ipaddr
));
2207 ip
.ipa_type
= (ndm
->ndm_family
== AF_INET
) ? IPADDR_V4
: IPADDR_V6
;
2208 memcpy(&ip
.ip
.addr
, RTA_DATA(tb
[NDA_DST
]), RTA_PAYLOAD(tb
[NDA_DST
]));
2210 /* Drop some "permanent" entries. */
2211 if (ndm
->ndm_state
& NUD_PERMANENT
) {
2212 char buf
[16] = "169.254.0.1";
2213 struct in_addr ipv4_ll
;
2215 if (ndm
->ndm_family
!= AF_INET
)
2218 if (!zif
->v6_2_v4_ll_neigh_entry
)
2221 if (h
->nlmsg_type
!= RTM_DELNEIGH
)
2224 inet_pton(AF_INET
, buf
, &ipv4_ll
);
2225 if (ipv4_ll
.s_addr
!= ip
.ip
._v4_addr
.s_addr
)
2228 if_nbr_ipv6ll_to_ipv4ll_neigh_update(
2229 ifp
, &zif
->v6_2_v4_ll_addr6
, true);
2233 /* The neighbor is present on an SVI. From this, we locate the
2235 * bridge because we're only interested in neighbors on a VxLAN bridge.
2236 * The bridge is located based on the nature of the SVI:
2237 * (a) In the case of a VLAN-aware bridge, the SVI is a L3 VLAN
2239 * and is linked to the bridge
2240 * (b) In the case of a VLAN-unaware bridge, the SVI is the bridge
2244 if (IS_ZEBRA_IF_VLAN(ifp
)) {
2245 link_if
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
2249 } else if (IS_ZEBRA_IF_BRIDGE(ifp
))
2254 memset(&mac
, 0, sizeof(struct ethaddr
));
2255 if (h
->nlmsg_type
== RTM_NEWNEIGH
) {
2256 if (tb
[NDA_LLADDR
]) {
2257 if (RTA_PAYLOAD(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
2259 "%s family %s IF %s(%u) - LLADDR is not MAC, len %lu",
2260 nl_msg_type_to_str(h
->nlmsg_type
),
2261 nl_family_to_str(ndm
->ndm_family
),
2262 ifp
->name
, ndm
->ndm_ifindex
,
2263 (unsigned long)RTA_PAYLOAD(
2269 memcpy(&mac
, RTA_DATA(tb
[NDA_LLADDR
]), ETH_ALEN
);
2272 ext_learned
= (ndm
->ndm_flags
& NTF_EXT_LEARNED
) ? 1 : 0;
2273 router_flag
= (ndm
->ndm_flags
& NTF_ROUTER
) ? 1 : 0;
2275 if (IS_ZEBRA_DEBUG_KERNEL
)
2277 "Rx %s family %s IF %s(%u) IP %s MAC %s state 0x%x flags 0x%x",
2278 nl_msg_type_to_str(h
->nlmsg_type
),
2279 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2281 ipaddr2str(&ip
, buf2
, sizeof(buf2
)),
2283 ? prefix_mac2str(&mac
, buf
, sizeof(buf
))
2285 ndm
->ndm_state
, ndm
->ndm_flags
);
2287 /* If the neighbor state is valid for use, process as an add or
2289 * else process as a delete. Note that the delete handling may
2291 * in re-adding the neighbor if it is a valid "remote" neighbor.
2293 if (ndm
->ndm_state
& NUD_VALID
)
2294 return zebra_vxlan_handle_kernel_neigh_update(
2295 ifp
, link_if
, &ip
, &mac
, ndm
->ndm_state
,
2296 ext_learned
, router_flag
);
2298 return zebra_vxlan_handle_kernel_neigh_del(ifp
, link_if
, &ip
);
2301 if (IS_ZEBRA_DEBUG_KERNEL
)
2302 zlog_debug("Rx %s family %s IF %s(%u) IP %s",
2303 nl_msg_type_to_str(h
->nlmsg_type
),
2304 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2306 ipaddr2str(&ip
, buf2
, sizeof(buf2
)));
2308 /* Process the delete - it may result in re-adding the neighbor if it is
2309 * a valid "remote" neighbor.
2311 return zebra_vxlan_handle_kernel_neigh_del(ifp
, link_if
, &ip
);
2314 static int netlink_neigh_table(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
2319 if (h
->nlmsg_type
!= RTM_NEWNEIGH
)
2322 /* Length validity. */
2323 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2327 /* We are interested only in AF_INET or AF_INET6 notifications. */
2328 ndm
= NLMSG_DATA(h
);
2329 if (ndm
->ndm_family
!= AF_INET
&& ndm
->ndm_family
!= AF_INET6
)
2332 return netlink_neigh_change(h
, len
);
2335 /* Request for IP neighbor information from the kernel */
2336 static int netlink_request_neigh(struct zebra_ns
*zns
, int family
, int type
,
2345 /* Form the request, specifying filter (rtattr) if needed. */
2346 memset(&req
, 0, sizeof(req
));
2347 req
.n
.nlmsg_type
= type
;
2348 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2349 req
.ndm
.ndm_family
= family
;
2351 addattr32(&req
.n
, sizeof(req
), NDA_IFINDEX
, ifindex
);
2353 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
2357 * IP Neighbor table read using netlink interface. This is invoked
2360 int netlink_neigh_read(struct zebra_ns
*zns
)
2364 /* Get IP neighbor table. */
2365 ret
= netlink_request_neigh(zns
, AF_UNSPEC
, RTM_GETNEIGH
, 0);
2368 ret
= netlink_parse_info(netlink_neigh_table
, &zns
->netlink_cmd
, zns
, 0,
2375 * IP Neighbor table read using netlink interface. This is for a specific
2378 int netlink_neigh_read_for_vlan(struct zebra_ns
*zns
, struct interface
*vlan_if
)
2382 ret
= netlink_request_neigh(zns
, AF_UNSPEC
, RTM_GETNEIGH
,
2386 ret
= netlink_parse_info(netlink_neigh_table
, &zns
->netlink_cmd
, zns
, 0,
2392 int netlink_neigh_change(struct nlmsghdr
*h
, ns_id_t ns_id
)
2397 if (!(h
->nlmsg_type
== RTM_NEWNEIGH
|| h
->nlmsg_type
== RTM_DELNEIGH
))
2400 /* Length validity. */
2401 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2403 zlog_err("%s: Message received from netlink is of a broken size %d %zu",
2404 __PRETTY_FUNCTION__
, h
->nlmsg_len
,
2405 (size_t)NLMSG_LENGTH(sizeof(struct ndmsg
)));
2409 /* Is this a notification for the MAC FDB or IP neighbor table? */
2410 ndm
= NLMSG_DATA(h
);
2411 if (ndm
->ndm_family
== AF_BRIDGE
)
2412 return netlink_macfdb_change(h
, len
, ns_id
);
2414 if (ndm
->ndm_type
!= RTN_UNICAST
)
2417 if (ndm
->ndm_family
== AF_INET
|| ndm
->ndm_family
== AF_INET6
)
2418 return netlink_ipneigh_change(h
, len
, ns_id
);
2421 "Invalid address family: %u received from kernel neighbor change: %u",
2422 ndm
->ndm_family
, h
->nlmsg_type
);
2429 static int netlink_neigh_update2(struct interface
*ifp
, struct ipaddr
*ip
,
2430 struct ethaddr
*mac
, uint8_t flags
,
2431 uint16_t state
, int cmd
)
2440 struct zebra_ns
*zns
;
2441 char buf
[INET6_ADDRSTRLEN
];
2442 char buf2
[ETHER_ADDR_STRLEN
];
2443 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
2446 memset(&req
, 0, sizeof(req
));
2448 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2449 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
2450 if (cmd
== RTM_NEWNEIGH
)
2451 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_REPLACE
);
2452 req
.n
.nlmsg_type
= cmd
; // RTM_NEWNEIGH or RTM_DELNEIGH
2453 req
.ndm
.ndm_family
= IS_IPADDR_V4(ip
) ? AF_INET
: AF_INET6
;
2454 req
.ndm
.ndm_state
= state
;
2455 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
2456 req
.ndm
.ndm_type
= RTN_UNICAST
;
2457 req
.ndm
.ndm_flags
= flags
;
2459 ipa_len
= IS_IPADDR_V4(ip
) ? IPV4_MAX_BYTELEN
: IPV6_MAX_BYTELEN
;
2460 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &ip
->ip
.addr
, ipa_len
);
2462 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, mac
, 6);
2464 if (IS_ZEBRA_DEBUG_KERNEL
)
2465 zlog_debug("Tx %s family %s IF %s(%u) Neigh %s MAC %s flags 0x%x",
2466 nl_msg_type_to_str(cmd
),
2467 nl_family_to_str(req
.ndm
.ndm_family
), ifp
->name
,
2468 ifp
->ifindex
, ipaddr2str(ip
, buf
, sizeof(buf
)),
2469 mac
? prefix_mac2str(mac
, buf2
, sizeof(buf2
))
2472 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2476 int kernel_add_mac(struct interface
*ifp
, vlanid_t vid
, struct ethaddr
*mac
,
2477 struct in_addr vtep_ip
, uint8_t sticky
)
2479 return netlink_macfdb_update(ifp
, vid
, mac
, vtep_ip
, 0, RTM_NEWNEIGH
,
2483 int kernel_del_mac(struct interface
*ifp
, vlanid_t vid
, struct ethaddr
*mac
,
2484 struct in_addr vtep_ip
, int local
)
2486 return netlink_macfdb_update(ifp
, vid
, mac
, vtep_ip
, local
,
2490 int kernel_add_neigh(struct interface
*ifp
, struct ipaddr
*ip
,
2491 struct ethaddr
*mac
, uint8_t flags
)
2493 return netlink_neigh_update2(ifp
, ip
, mac
, flags
,
2494 NUD_NOARP
, RTM_NEWNEIGH
);
2497 int kernel_del_neigh(struct interface
*ifp
, struct ipaddr
*ip
)
2499 return netlink_neigh_update2(ifp
, ip
, NULL
, 0, 0, RTM_DELNEIGH
);
2503 * MPLS label forwarding table change via netlink interface.
2505 int netlink_mpls_multipath(int cmd
, zebra_lsp_t
*lsp
)
2508 zebra_nhlfe_t
*nhlfe
;
2509 struct nexthop
*nexthop
= NULL
;
2510 unsigned int nexthop_num
;
2511 const char *routedesc
;
2512 struct zebra_ns
*zns
= zebra_ns_lookup(NS_DEFAULT
);
2518 char buf
[NL_PKT_BUF_SIZE
];
2521 memset(&req
, 0, sizeof req
- NL_PKT_BUF_SIZE
);
2524 * Count # nexthops so we can decide whether to use singlepath
2525 * or multipath case.
2528 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2529 nexthop
= nhlfe
->nexthop
;
2532 if (cmd
== RTM_NEWROUTE
) {
2533 /* Count all selected NHLFEs */
2534 if (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2535 && CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
))
2539 /* Count all installed NHLFEs */
2540 if (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_INSTALLED
)
2541 && CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
2546 if ((nexthop_num
== 0) || (!lsp
->best_nhlfe
&& (cmd
!= RTM_DELROUTE
)))
2549 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
2550 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
2551 req
.n
.nlmsg_type
= cmd
;
2552 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
2554 req
.r
.rtm_family
= AF_MPLS
;
2555 req
.r
.rtm_table
= RT_TABLE_MAIN
;
2556 req
.r
.rtm_dst_len
= MPLS_LABEL_LEN_BITS
;
2557 req
.r
.rtm_scope
= RT_SCOPE_UNIVERSE
;
2558 req
.r
.rtm_type
= RTN_UNICAST
;
2560 if (cmd
== RTM_NEWROUTE
) {
2561 /* We do a replace to handle update. */
2562 req
.n
.nlmsg_flags
|= NLM_F_REPLACE
;
2564 /* set the protocol value if installing */
2565 route_type
= re_type_from_lsp_type(lsp
->best_nhlfe
->type
);
2566 req
.r
.rtm_protocol
= zebra2proto(route_type
);
2569 /* Fill destination */
2570 lse
= mpls_lse_encode(lsp
->ile
.in_label
, 0, 0, 1);
2571 addattr_l(&req
.n
, sizeof req
, RTA_DST
, &lse
, sizeof(mpls_lse_t
));
2573 /* Fill nexthops (paths) based on single-path or multipath. The paths
2574 * chosen depend on the operation.
2576 if (nexthop_num
== 1 || multipath_num
== 1) {
2577 routedesc
= "single-path";
2578 _netlink_mpls_debug(cmd
, lsp
->ile
.in_label
, routedesc
);
2581 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2582 nexthop
= nhlfe
->nexthop
;
2586 if ((cmd
== RTM_NEWROUTE
2587 && (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2588 && CHECK_FLAG(nexthop
->flags
,
2589 NEXTHOP_FLAG_ACTIVE
)))
2590 || (cmd
== RTM_DELROUTE
2591 && (CHECK_FLAG(nhlfe
->flags
,
2592 NHLFE_FLAG_INSTALLED
)
2593 && CHECK_FLAG(nexthop
->flags
,
2594 NEXTHOP_FLAG_FIB
)))) {
2595 /* Add the gateway */
2596 _netlink_mpls_build_singlepath(routedesc
, nhlfe
,
2603 } else /* Multipath case */
2605 char buf
[NL_PKT_BUF_SIZE
];
2606 struct rtattr
*rta
= (void *)buf
;
2607 struct rtnexthop
*rtnh
;
2608 union g_addr
*src1
= NULL
;
2610 rta
->rta_type
= RTA_MULTIPATH
;
2611 rta
->rta_len
= RTA_LENGTH(0);
2612 rtnh
= RTA_DATA(rta
);
2614 routedesc
= "multipath";
2615 _netlink_mpls_debug(cmd
, lsp
->ile
.in_label
, routedesc
);
2618 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2619 nexthop
= nhlfe
->nexthop
;
2623 if (nexthop_num
>= multipath_num
)
2626 if ((cmd
== RTM_NEWROUTE
2627 && (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2628 && CHECK_FLAG(nexthop
->flags
,
2629 NEXTHOP_FLAG_ACTIVE
)))
2630 || (cmd
== RTM_DELROUTE
2631 && (CHECK_FLAG(nhlfe
->flags
,
2632 NHLFE_FLAG_INSTALLED
)
2633 && CHECK_FLAG(nexthop
->flags
,
2634 NEXTHOP_FLAG_FIB
)))) {
2637 /* Build the multipath */
2638 _netlink_mpls_build_multipath(routedesc
, nhlfe
,
2641 rtnh
= RTNH_NEXT(rtnh
);
2645 /* Add the multipath */
2646 if (rta
->rta_len
> RTA_LENGTH(0))
2647 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_MULTIPATH
,
2648 RTA_DATA(rta
), RTA_PAYLOAD(rta
));
2651 /* Talk to netlink socket. */
2652 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2655 #endif /* HAVE_NETLINK */