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 memcpy(&p
.u
.prefix4
, dest
, 4);
388 p
.prefixlen
= rtm
->rtm_dst_len
;
390 if (rtm
->rtm_src_len
!= 0) {
391 char buf
[PREFIX_STRLEN
];
392 zlog_warn("unsupported IPv4 sourcedest route (dest %s vrf %u)",
393 prefix2str(&p
, buf
, sizeof(buf
)), vrf_id
);
397 /* Force debug below to not display anything for source */
399 } else if (rtm
->rtm_family
== AF_INET6
) {
401 memcpy(&p
.u
.prefix6
, dest
, 16);
402 p
.prefixlen
= rtm
->rtm_dst_len
;
404 src_p
.family
= AF_INET6
;
405 memcpy(&src_p
.prefix
, src
, 16);
406 src_p
.prefixlen
= rtm
->rtm_src_len
;
410 * For ZEBRA_ROUTE_KERNEL types:
412 * The metric/priority of the route received from the kernel
413 * is a 32 bit number. We are going to interpret the high
414 * order byte as the Admin Distance and the low order 3 bytes
417 * This will allow us to do two things:
418 * 1) Allow the creation of kernel routes that can be
419 * overridden by zebra.
420 * 2) Allow the old behavior for 'most' kernel route types
421 * if a user enters 'ip route ...' v4 routes get a metric
422 * of 0 and v6 routes get a metric of 1024. Both of these
423 * values will end up with a admin distance of 0, which
424 * will cause them to win for the purposes of zebra.
426 if (proto
== ZEBRA_ROUTE_KERNEL
) {
427 distance
= (metric
>> 24) & 0xFF;
428 metric
= (metric
& 0x00FFFFFF);
431 if (IS_ZEBRA_DEBUG_KERNEL
) {
432 char buf
[PREFIX_STRLEN
];
433 char buf2
[PREFIX_STRLEN
];
434 zlog_debug("%s %s%s%s vrf %u(%u) metric: %d Admin Distance: %d",
435 nl_msg_type_to_str(h
->nlmsg_type
),
436 prefix2str(&p
, buf
, sizeof(buf
)),
437 src_p
.prefixlen
? " from " : "",
439 ? prefix2str(&src_p
, buf2
, sizeof(buf2
))
441 vrf_id
, table
, metric
, distance
);
445 if (rtm
->rtm_family
== AF_INET6
)
448 if (h
->nlmsg_type
== RTM_NEWROUTE
) {
449 struct interface
*ifp
;
450 vrf_id_t nh_vrf_id
= vrf_id
;
452 if (!tb
[RTA_MULTIPATH
]) {
454 size_t sz
= (afi
== AFI_IP
) ? 4 : 16;
456 memset(&nh
, 0, sizeof(nh
));
458 if (bh_type
== BLACKHOLE_UNSPEC
) {
460 nh
.type
= NEXTHOP_TYPE_IFINDEX
;
461 else if (index
&& gate
)
464 ? NEXTHOP_TYPE_IPV4_IFINDEX
465 : NEXTHOP_TYPE_IPV6_IFINDEX
;
466 else if (!index
&& gate
)
467 nh
.type
= (afi
== AFI_IP
)
471 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
472 nh
.bh_type
= bh_type
;
475 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
476 nh
.bh_type
= bh_type
;
480 memcpy(&nh
.src
, prefsrc
, sz
);
482 memcpy(&nh
.gate
, gate
, sz
);
485 ifp
= if_lookup_by_index_per_ns(
486 zebra_ns_lookup(ns_id
),
489 nh_vrf_id
= ifp
->vrf_id
;
491 nh
.vrf_id
= nh_vrf_id
;
493 rib_add(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
, &p
,
494 &src_p
, &nh
, table
, metric
, mtu
, distance
, tag
);
496 /* This is a multipath route */
498 struct route_entry
*re
;
499 struct rtnexthop
*rtnh
=
500 (struct rtnexthop
*)RTA_DATA(tb
[RTA_MULTIPATH
]);
502 len
= RTA_PAYLOAD(tb
[RTA_MULTIPATH
]);
504 re
= XCALLOC(MTYPE_RE
, sizeof(struct route_entry
));
506 re
->distance
= distance
;
513 re
->uptime
= time(NULL
);
518 if (len
< (int)sizeof(*rtnh
)
519 || rtnh
->rtnh_len
> len
)
522 index
= rtnh
->rtnh_ifindex
;
525 * Yes we are looking this up
526 * for every nexthop and just
527 * using the last one looked
530 ifp
= if_lookup_by_index_per_ns(
531 zebra_ns_lookup(ns_id
),
534 nh_vrf_id
= ifp
->vrf_id
;
537 "%s: Unknown interface %u specified, defaulting to VRF_DEFAULT",
540 nh_vrf_id
= VRF_DEFAULT
;
546 if (rtnh
->rtnh_len
> sizeof(*rtnh
)) {
547 memset(tb
, 0, sizeof(tb
));
548 netlink_parse_rtattr(
549 tb
, RTA_MAX
, RTNH_DATA(rtnh
),
550 rtnh
->rtnh_len
- sizeof(*rtnh
));
557 if (rtm
->rtm_family
== AF_INET
) {
559 route_entry_nexthop_ipv4_ifindex_add(
564 route_entry_nexthop_ipv4_add(
568 } else if (rtm
->rtm_family
571 route_entry_nexthop_ipv6_ifindex_add(
575 route_entry_nexthop_ipv6_add(
580 route_entry_nexthop_ifindex_add(
581 re
, index
, nh_vrf_id
);
583 len
-= NLMSG_ALIGN(rtnh
->rtnh_len
);
584 rtnh
= RTNH_NEXT(rtnh
);
587 zserv_nexthop_num_warn(__func__
,
588 (const struct prefix
*)&p
,
590 if (re
->nexthop_num
== 0)
593 rib_add_multipath(afi
, SAFI_UNICAST
, &p
,
597 if (!tb
[RTA_MULTIPATH
]) {
599 size_t sz
= (afi
== AFI_IP
) ? 4 : 16;
601 memset(&nh
, 0, sizeof(nh
));
602 if (bh_type
== BLACKHOLE_UNSPEC
) {
604 nh
.type
= NEXTHOP_TYPE_IFINDEX
;
605 else if (index
&& gate
)
608 ? NEXTHOP_TYPE_IPV4_IFINDEX
609 : NEXTHOP_TYPE_IPV6_IFINDEX
;
610 else if (!index
&& gate
)
611 nh
.type
= (afi
== AFI_IP
)
615 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
616 nh
.bh_type
= BLACKHOLE_UNSPEC
;
619 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
620 nh
.bh_type
= bh_type
;
624 memcpy(&nh
.gate
, gate
, sz
);
625 rib_delete(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
,
626 &p
, &src_p
, &nh
, table
, metric
, true);
628 /* XXX: need to compare the entire list of nexthops
629 * here for NLM_F_APPEND stupidity */
630 rib_delete(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
,
631 &p
, &src_p
, NULL
, table
, metric
, true);
638 static struct mcast_route_data
*mroute
= NULL
;
640 static int netlink_route_change_read_multicast(struct nlmsghdr
*h
,
641 ns_id_t ns_id
, int startup
)
645 struct rtattr
*tb
[RTA_MAX
+ 1];
646 struct mcast_route_data
*m
;
647 struct mcast_route_data mr
;
654 char oif_list
[256] = "\0";
661 memset(&mr
, 0, sizeof(mr
));
667 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
669 memset(tb
, 0, sizeof tb
);
670 netlink_parse_rtattr(tb
, RTA_MAX
, RTM_RTA(rtm
), len
);
673 table
= *(int *)RTA_DATA(tb
[RTA_TABLE
]);
675 table
= rtm
->rtm_table
;
677 vrf
= vrf_lookup_by_table(table
, ns_id
);
680 iif
= *(int *)RTA_DATA(tb
[RTA_IIF
]);
683 m
->sg
.src
= *(struct in_addr
*)RTA_DATA(tb
[RTA_SRC
]);
686 m
->sg
.grp
= *(struct in_addr
*)RTA_DATA(tb
[RTA_DST
]);
688 if ((RTA_EXPIRES
<= RTA_MAX
) && tb
[RTA_EXPIRES
])
689 m
->lastused
= *(unsigned long long *)RTA_DATA(tb
[RTA_EXPIRES
]);
691 if (tb
[RTA_MULTIPATH
]) {
692 struct rtnexthop
*rtnh
=
693 (struct rtnexthop
*)RTA_DATA(tb
[RTA_MULTIPATH
]);
695 len
= RTA_PAYLOAD(tb
[RTA_MULTIPATH
]);
697 if (len
< (int)sizeof(*rtnh
) || rtnh
->rtnh_len
> len
)
700 oif
[oif_count
] = rtnh
->rtnh_ifindex
;
703 len
-= NLMSG_ALIGN(rtnh
->rtnh_len
);
704 rtnh
= RTNH_NEXT(rtnh
);
708 if (IS_ZEBRA_DEBUG_KERNEL
) {
709 struct interface
*ifp
;
710 strlcpy(sbuf
, inet_ntoa(m
->sg
.src
), sizeof(sbuf
));
711 strlcpy(gbuf
, inet_ntoa(m
->sg
.grp
), sizeof(gbuf
));
712 for (count
= 0; count
< oif_count
; count
++) {
713 ifp
= if_lookup_by_index(oif
[count
], vrf
);
716 sprintf(temp
, "%s ", ifp
->name
);
717 strcat(oif_list
, temp
);
719 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(vrf
);
720 ifp
= if_lookup_by_index(iif
, vrf
);
722 "MCAST VRF: %s(%d) %s (%s,%s) IIF: %s OIF: %s jiffies: %lld",
723 zvrf
->vrf
->name
, vrf
, nl_msg_type_to_str(h
->nlmsg_type
),
724 sbuf
, gbuf
, ifp
->name
, oif_list
, m
->lastused
);
729 int netlink_route_change(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
736 if (!(h
->nlmsg_type
== RTM_NEWROUTE
|| h
->nlmsg_type
== RTM_DELROUTE
)) {
737 /* If this is not route add/delete message print warning. */
738 zlog_warn("Kernel message: %d NS %u\n", h
->nlmsg_type
, ns_id
);
742 if (!(rtm
->rtm_family
== AF_INET
|| rtm
->rtm_family
== AF_INET6
743 || rtm
->rtm_family
== AF_ETHERNET
744 || rtm
->rtm_family
== AF_MPLS
)) {
746 "Invalid address family: %d received from kernel route change: %d",
747 rtm
->rtm_family
, h
->nlmsg_type
);
751 /* Connected route. */
752 if (IS_ZEBRA_DEBUG_KERNEL
)
753 zlog_debug("%s %s %s proto %s NS %u",
754 nl_msg_type_to_str(h
->nlmsg_type
),
755 nl_family_to_str(rtm
->rtm_family
),
756 nl_rttype_to_str(rtm
->rtm_type
),
757 nl_rtproto_to_str(rtm
->rtm_protocol
), ns_id
);
759 /* We don't care about change notifications for the MPLS table. */
760 /* TODO: Revisit this. */
761 if (rtm
->rtm_family
== AF_MPLS
)
764 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
766 zlog_err("%s: Message received from netlink is of a broken size: %d %zu",
769 (size_t)NLMSG_LENGTH(sizeof(struct rtmsg
)));
773 if (rtm
->rtm_type
== RTN_MULTICAST
)
774 netlink_route_change_read_multicast(h
, ns_id
, startup
);
776 netlink_route_change_read_unicast(h
, ns_id
, startup
);
780 /* Request for specific route information from the kernel */
781 static int netlink_request_route(struct zebra_ns
*zns
, int family
, int type
)
788 /* Form the request, specifying filter (rtattr) if needed. */
789 memset(&req
, 0, sizeof(req
));
790 req
.n
.nlmsg_type
= type
;
791 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
792 req
.rtm
.rtm_family
= family
;
794 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
797 /* Routing table read function using netlink interface. Only called
799 int netlink_route_read(struct zebra_ns
*zns
)
803 /* Get IPv4 routing table. */
804 ret
= netlink_request_route(zns
, AF_INET
, RTM_GETROUTE
);
807 ret
= netlink_parse_info(netlink_route_change_read_unicast
,
808 &zns
->netlink_cmd
, zns
, 0, 1);
812 /* Get IPv6 routing table. */
813 ret
= netlink_request_route(zns
, AF_INET6
, RTM_GETROUTE
);
816 ret
= netlink_parse_info(netlink_route_change_read_unicast
,
817 &zns
->netlink_cmd
, zns
, 0, 1);
824 static void _netlink_route_nl_add_gateway_info(uint8_t route_family
,
826 struct nlmsghdr
*nlmsg
,
827 size_t req_size
, int bytelen
,
828 struct nexthop
*nexthop
)
830 if (route_family
== AF_MPLS
) {
831 struct gw_family_t gw_fam
;
833 gw_fam
.family
= gw_family
;
834 if (gw_family
== AF_INET
)
835 memcpy(&gw_fam
.gate
.ipv4
, &nexthop
->gate
.ipv4
, bytelen
);
837 memcpy(&gw_fam
.gate
.ipv6
, &nexthop
->gate
.ipv6
, bytelen
);
838 addattr_l(nlmsg
, req_size
, RTA_VIA
, &gw_fam
.family
,
841 if (gw_family
== AF_INET
)
842 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
,
843 &nexthop
->gate
.ipv4
, bytelen
);
845 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
,
846 &nexthop
->gate
.ipv6
, bytelen
);
850 static void _netlink_route_rta_add_gateway_info(uint8_t route_family
,
853 struct rtnexthop
*rtnh
,
854 size_t req_size
, int bytelen
,
855 struct nexthop
*nexthop
)
857 if (route_family
== AF_MPLS
) {
858 struct gw_family_t gw_fam
;
860 gw_fam
.family
= gw_family
;
861 if (gw_family
== AF_INET
)
862 memcpy(&gw_fam
.gate
.ipv4
, &nexthop
->gate
.ipv4
, bytelen
);
864 memcpy(&gw_fam
.gate
.ipv6
, &nexthop
->gate
.ipv6
, bytelen
);
865 rta_addattr_l(rta
, req_size
, RTA_VIA
, &gw_fam
.family
,
867 rtnh
->rtnh_len
+= RTA_LENGTH(bytelen
+ 2);
869 if (gw_family
== AF_INET
)
870 rta_addattr_l(rta
, req_size
, RTA_GATEWAY
,
871 &nexthop
->gate
.ipv4
, bytelen
);
873 rta_addattr_l(rta
, req_size
, RTA_GATEWAY
,
874 &nexthop
->gate
.ipv6
, bytelen
);
875 rtnh
->rtnh_len
+= sizeof(struct rtattr
) + bytelen
;
879 /* This function takes a nexthop as argument and adds
880 * the appropriate netlink attributes to an existing
883 * @param routedesc: Human readable description of route type
884 * (direct/recursive, single-/multipath)
885 * @param bytelen: Length of addresses in bytes.
886 * @param nexthop: Nexthop information
887 * @param nlmsg: nlmsghdr structure to fill in.
888 * @param req_size: The size allocated for the message.
890 static void _netlink_route_build_singlepath(const char *routedesc
, int bytelen
,
891 struct nexthop
*nexthop
,
892 struct nlmsghdr
*nlmsg
,
894 size_t req_size
, int cmd
)
896 struct mpls_label_stack
*nh_label
;
897 mpls_lse_t out_lse
[MPLS_MAX_LABELS
];
902 * label_buf is *only* currently used within debugging.
903 * As such when we assign it we are guarding it inside
904 * a debug test. If you want to change this make sure
905 * you fix this assumption
910 for (struct nexthop
*nh
= nexthop
; nh
; nh
= nh
->rparent
) {
913 nh_label
= nh
->nh_label
;
914 if (!nh_label
|| !nh_label
->num_labels
)
917 for (int i
= 0; i
< nh_label
->num_labels
; i
++) {
918 if (nh_label
->label
[i
] == MPLS_LABEL_IMPLICIT_NULL
)
921 if (IS_ZEBRA_DEBUG_KERNEL
) {
923 sprintf(label_buf
, "label %u",
926 sprintf(label_buf1
, "/%u",
928 strlcat(label_buf
, label_buf1
,
933 out_lse
[num_labels
] =
934 mpls_lse_encode(nh_label
->label
[i
], 0, 0, 0);
940 /* Set the BoS bit */
941 out_lse
[num_labels
- 1] |= htonl(1 << MPLS_LS_S_SHIFT
);
943 if (rtmsg
->rtm_family
== AF_MPLS
)
944 addattr_l(nlmsg
, req_size
, RTA_NEWDST
, &out_lse
,
945 num_labels
* sizeof(mpls_lse_t
));
948 uint16_t encap
= LWTUNNEL_ENCAP_MPLS
;
950 addattr_l(nlmsg
, req_size
, RTA_ENCAP_TYPE
, &encap
,
952 nest
= addattr_nest(nlmsg
, req_size
, RTA_ENCAP
);
953 addattr_l(nlmsg
, req_size
, MPLS_IPTUNNEL_DST
, &out_lse
,
954 num_labels
* sizeof(mpls_lse_t
));
955 addattr_nest_end(nlmsg
, nest
);
959 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
))
960 rtmsg
->rtm_flags
|= RTNH_F_ONLINK
;
962 if (rtmsg
->rtm_family
== AF_INET
963 && (nexthop
->type
== NEXTHOP_TYPE_IPV6
964 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
)) {
965 rtmsg
->rtm_flags
|= RTNH_F_ONLINK
;
966 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
, &ipv4_ll
, 4);
967 addattr32(nlmsg
, req_size
, RTA_OIF
, nexthop
->ifindex
);
969 if (nexthop
->rmap_src
.ipv4
.s_addr
&& (cmd
== RTM_NEWROUTE
))
970 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
971 &nexthop
->rmap_src
.ipv4
, bytelen
);
972 else if (nexthop
->src
.ipv4
.s_addr
&& (cmd
== RTM_NEWROUTE
))
973 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
974 &nexthop
->src
.ipv4
, bytelen
);
976 if (IS_ZEBRA_DEBUG_KERNEL
)
978 " 5549: _netlink_route_build_singlepath() (%s): "
979 "nexthop via %s %s if %u(%u)",
980 routedesc
, ipv4_ll_buf
, label_buf
,
981 nexthop
->ifindex
, nexthop
->vrf_id
);
985 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
986 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
987 /* Send deletes to the kernel without specifying the next-hop */
988 if (cmd
!= RTM_DELROUTE
)
989 _netlink_route_nl_add_gateway_info(
990 rtmsg
->rtm_family
, AF_INET
, nlmsg
, req_size
,
993 if (cmd
== RTM_NEWROUTE
) {
994 if (nexthop
->rmap_src
.ipv4
.s_addr
)
995 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
996 &nexthop
->rmap_src
.ipv4
, bytelen
);
997 else if (nexthop
->src
.ipv4
.s_addr
)
998 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
999 &nexthop
->src
.ipv4
, bytelen
);
1002 if (IS_ZEBRA_DEBUG_KERNEL
)
1004 "netlink_route_multipath() (%s): "
1005 "nexthop via %s %s if %u(%u)",
1006 routedesc
, inet_ntoa(nexthop
->gate
.ipv4
),
1007 label_buf
, nexthop
->ifindex
, nexthop
->vrf_id
);
1010 if (nexthop
->type
== NEXTHOP_TYPE_IPV6
1011 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1012 _netlink_route_nl_add_gateway_info(rtmsg
->rtm_family
, AF_INET6
,
1013 nlmsg
, req_size
, bytelen
,
1016 if (cmd
== RTM_NEWROUTE
) {
1017 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1018 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1019 &nexthop
->rmap_src
.ipv6
, bytelen
);
1020 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1021 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1022 &nexthop
->src
.ipv6
, bytelen
);
1025 if (IS_ZEBRA_DEBUG_KERNEL
)
1027 "netlink_route_multipath() (%s): "
1028 "nexthop via %s %s if %u(%u)",
1029 routedesc
, inet6_ntoa(nexthop
->gate
.ipv6
),
1030 label_buf
, nexthop
->ifindex
, nexthop
->vrf_id
);
1034 * We have the ifindex so we should always send it
1035 * This is especially useful if we are doing route
1038 if (nexthop
->type
!= NEXTHOP_TYPE_BLACKHOLE
)
1039 addattr32(nlmsg
, req_size
, RTA_OIF
, nexthop
->ifindex
);
1041 if (nexthop
->type
== NEXTHOP_TYPE_IFINDEX
1042 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
1043 if (cmd
== RTM_NEWROUTE
) {
1044 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1045 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1046 &nexthop
->rmap_src
.ipv4
, bytelen
);
1047 else if (nexthop
->src
.ipv4
.s_addr
)
1048 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1049 &nexthop
->src
.ipv4
, bytelen
);
1052 if (IS_ZEBRA_DEBUG_KERNEL
)
1054 "netlink_route_multipath() (%s): "
1055 "nexthop via if %u(%u)",
1056 routedesc
, nexthop
->ifindex
, nexthop
->vrf_id
);
1059 if (nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1060 if (cmd
== RTM_NEWROUTE
) {
1061 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1062 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1063 &nexthop
->rmap_src
.ipv6
, bytelen
);
1064 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1065 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1066 &nexthop
->src
.ipv6
, bytelen
);
1069 if (IS_ZEBRA_DEBUG_KERNEL
)
1071 "netlink_route_multipath() (%s): "
1072 "nexthop via if %u(%u)",
1073 routedesc
, nexthop
->ifindex
, nexthop
->vrf_id
);
1077 /* This function takes a nexthop as argument and
1078 * appends to the given rtattr/rtnexthop pair the
1079 * representation of the nexthop. If the nexthop
1080 * defines a preferred source, the src parameter
1081 * will be modified to point to that src, otherwise
1082 * it will be kept unmodified.
1084 * @param routedesc: Human readable description of route type
1085 * (direct/recursive, single-/multipath)
1086 * @param bytelen: Length of addresses in bytes.
1087 * @param nexthop: Nexthop information
1088 * @param rta: rtnetlink attribute structure
1089 * @param rtnh: pointer to an rtnetlink nexthop structure
1090 * @param src: pointer pointing to a location where
1091 * the prefsrc should be stored.
1093 static void _netlink_route_build_multipath(const char *routedesc
, int bytelen
,
1094 struct nexthop
*nexthop
,
1096 struct rtnexthop
*rtnh
,
1097 struct rtmsg
*rtmsg
,
1100 struct mpls_label_stack
*nh_label
;
1101 mpls_lse_t out_lse
[MPLS_MAX_LABELS
];
1103 char label_buf
[256];
1105 rtnh
->rtnh_len
= sizeof(*rtnh
);
1106 rtnh
->rtnh_flags
= 0;
1107 rtnh
->rtnh_hops
= 0;
1108 rta
->rta_len
+= rtnh
->rtnh_len
;
1111 * label_buf is *only* currently used within debugging.
1112 * As such when we assign it we are guarding it inside
1113 * a debug test. If you want to change this make sure
1114 * you fix this assumption
1116 label_buf
[0] = '\0';
1119 for (struct nexthop
*nh
= nexthop
; nh
; nh
= nh
->rparent
) {
1120 char label_buf1
[20];
1122 nh_label
= nh
->nh_label
;
1123 if (!nh_label
|| !nh_label
->num_labels
)
1126 for (int i
= 0; i
< nh_label
->num_labels
; i
++) {
1127 if (nh_label
->label
[i
] == MPLS_LABEL_IMPLICIT_NULL
)
1130 if (IS_ZEBRA_DEBUG_KERNEL
) {
1132 sprintf(label_buf
, "label %u",
1133 nh_label
->label
[i
]);
1135 sprintf(label_buf1
, "/%u",
1136 nh_label
->label
[i
]);
1137 strlcat(label_buf
, label_buf1
,
1142 out_lse
[num_labels
] =
1143 mpls_lse_encode(nh_label
->label
[i
], 0, 0, 0);
1149 /* Set the BoS bit */
1150 out_lse
[num_labels
- 1] |= htonl(1 << MPLS_LS_S_SHIFT
);
1152 if (rtmsg
->rtm_family
== AF_MPLS
) {
1153 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_NEWDST
,
1155 num_labels
* sizeof(mpls_lse_t
));
1157 RTA_LENGTH(num_labels
* sizeof(mpls_lse_t
));
1159 struct rtattr
*nest
;
1160 uint16_t encap
= LWTUNNEL_ENCAP_MPLS
;
1161 int len
= rta
->rta_len
;
1163 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_ENCAP_TYPE
,
1164 &encap
, sizeof(uint16_t));
1165 nest
= rta_nest(rta
, NL_PKT_BUF_SIZE
, RTA_ENCAP
);
1166 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, MPLS_IPTUNNEL_DST
,
1168 num_labels
* sizeof(mpls_lse_t
));
1169 rta_nest_end(rta
, nest
);
1170 rtnh
->rtnh_len
+= rta
->rta_len
- len
;
1174 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
))
1175 rtnh
->rtnh_flags
|= RTNH_F_ONLINK
;
1177 if (rtmsg
->rtm_family
== AF_INET
1178 && (nexthop
->type
== NEXTHOP_TYPE_IPV6
1179 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
)) {
1181 rtnh
->rtnh_flags
|= RTNH_F_ONLINK
;
1182 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_GATEWAY
, &ipv4_ll
,
1184 rtnh
->rtnh_len
+= sizeof(struct rtattr
) + bytelen
;
1185 rtnh
->rtnh_ifindex
= nexthop
->ifindex
;
1187 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1188 *src
= &nexthop
->rmap_src
;
1189 else if (nexthop
->src
.ipv4
.s_addr
)
1190 *src
= &nexthop
->src
;
1192 if (IS_ZEBRA_DEBUG_KERNEL
)
1194 " 5549: netlink_route_build_multipath() (%s): "
1195 "nexthop via %s %s if %u",
1196 routedesc
, ipv4_ll_buf
, label_buf
,
1201 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
1202 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
1203 _netlink_route_rta_add_gateway_info(rtmsg
->rtm_family
, AF_INET
,
1204 rta
, rtnh
, NL_PKT_BUF_SIZE
,
1206 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1207 *src
= &nexthop
->rmap_src
;
1208 else if (nexthop
->src
.ipv4
.s_addr
)
1209 *src
= &nexthop
->src
;
1211 if (IS_ZEBRA_DEBUG_KERNEL
)
1213 "netlink_route_multipath() (%s): "
1214 "nexthop via %s %s if %u",
1215 routedesc
, inet_ntoa(nexthop
->gate
.ipv4
),
1216 label_buf
, nexthop
->ifindex
);
1218 if (nexthop
->type
== NEXTHOP_TYPE_IPV6
1219 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1220 _netlink_route_rta_add_gateway_info(rtmsg
->rtm_family
, AF_INET6
,
1221 rta
, rtnh
, NL_PKT_BUF_SIZE
,
1224 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1225 *src
= &nexthop
->rmap_src
;
1226 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1227 *src
= &nexthop
->src
;
1229 if (IS_ZEBRA_DEBUG_KERNEL
)
1231 "netlink_route_multipath() (%s): "
1232 "nexthop via %s %s if %u",
1233 routedesc
, inet6_ntoa(nexthop
->gate
.ipv6
),
1234 label_buf
, nexthop
->ifindex
);
1238 * We have figured out the ifindex so we should always send it
1239 * This is especially useful if we are doing route
1242 if (nexthop
->type
!= NEXTHOP_TYPE_BLACKHOLE
)
1243 rtnh
->rtnh_ifindex
= nexthop
->ifindex
;
1246 if (nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
1247 || nexthop
->type
== NEXTHOP_TYPE_IFINDEX
) {
1248 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1249 *src
= &nexthop
->rmap_src
;
1250 else if (nexthop
->src
.ipv4
.s_addr
)
1251 *src
= &nexthop
->src
;
1253 if (IS_ZEBRA_DEBUG_KERNEL
)
1255 "netlink_route_multipath() (%s): "
1256 "nexthop via if %u",
1257 routedesc
, nexthop
->ifindex
);
1261 static inline void _netlink_mpls_build_singlepath(const char *routedesc
,
1262 zebra_nhlfe_t
*nhlfe
,
1263 struct nlmsghdr
*nlmsg
,
1264 struct rtmsg
*rtmsg
,
1265 size_t req_size
, int cmd
)
1270 family
= NHLFE_FAMILY(nhlfe
);
1271 bytelen
= (family
== AF_INET
? 4 : 16);
1272 _netlink_route_build_singlepath(routedesc
, bytelen
, nhlfe
->nexthop
,
1273 nlmsg
, rtmsg
, req_size
, cmd
);
1278 _netlink_mpls_build_multipath(const char *routedesc
, zebra_nhlfe_t
*nhlfe
,
1279 struct rtattr
*rta
, struct rtnexthop
*rtnh
,
1280 struct rtmsg
*rtmsg
, union g_addr
**src
)
1285 family
= NHLFE_FAMILY(nhlfe
);
1286 bytelen
= (family
== AF_INET
? 4 : 16);
1287 _netlink_route_build_multipath(routedesc
, bytelen
, nhlfe
->nexthop
, rta
,
1292 /* Log debug information for netlink_route_multipath
1293 * if debug logging is enabled.
1295 * @param cmd: Netlink command which is to be processed
1296 * @param p: Prefix for which the change is due
1297 * @param family: Address family which the change concerns
1298 * @param zvrf: The vrf we are in
1299 * @param tableid: The table we are working on
1301 static void _netlink_route_debug(int cmd
, const struct prefix
*p
,
1302 int family
, vrf_id_t vrfid
,
1305 if (IS_ZEBRA_DEBUG_KERNEL
) {
1306 char buf
[PREFIX_STRLEN
];
1308 "netlink_route_multipath(): %s %s vrf %u(%u)",
1309 nl_msg_type_to_str(cmd
),
1310 prefix2str(p
, buf
, sizeof(buf
)),
1315 static void _netlink_mpls_debug(int cmd
, uint32_t label
, const char *routedesc
)
1317 if (IS_ZEBRA_DEBUG_KERNEL
)
1318 zlog_debug("netlink_mpls_multipath() (%s): %s %u/20", routedesc
,
1319 nl_msg_type_to_str(cmd
), label
);
1322 static int netlink_neigh_update(int cmd
, int ifindex
, uint32_t addr
, char *lla
,
1323 int llalen
, ns_id_t ns_id
)
1331 struct zebra_ns
*zns
= zebra_ns_lookup(ns_id
);
1333 memset(&req
, 0, sizeof(req
));
1335 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1336 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
1337 req
.n
.nlmsg_type
= cmd
; // RTM_NEWNEIGH or RTM_DELNEIGH
1338 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1340 req
.ndm
.ndm_family
= AF_INET
;
1341 req
.ndm
.ndm_state
= NUD_PERMANENT
;
1342 req
.ndm
.ndm_ifindex
= ifindex
;
1343 req
.ndm
.ndm_type
= RTN_UNICAST
;
1345 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &addr
, 4);
1346 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, lla
, llalen
);
1348 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1352 /* Routing table change via netlink interface. */
1353 /* Update flag indicates whether this is a "replace" or not. */
1354 static int netlink_route_multipath(int cmd
, const struct prefix
*p
,
1355 const struct prefix
*src_p
,
1356 struct route_entry
*re
,
1360 struct sockaddr_nl snl
;
1361 struct nexthop
*nexthop
= NULL
;
1362 unsigned int nexthop_num
;
1363 int family
= PREFIX_FAMILY(p
);
1364 const char *routedesc
;
1371 char buf
[NL_PKT_BUF_SIZE
];
1374 struct zebra_ns
*zns
;
1375 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1378 memset(&req
, 0, sizeof req
- NL_PKT_BUF_SIZE
);
1380 bytelen
= (family
== AF_INET
? 4 : 16);
1382 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
1383 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
1384 if ((cmd
== RTM_NEWROUTE
) && update
)
1385 req
.n
.nlmsg_flags
|= NLM_F_REPLACE
;
1386 req
.n
.nlmsg_type
= cmd
;
1387 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1389 req
.r
.rtm_family
= family
;
1390 req
.r
.rtm_dst_len
= p
->prefixlen
;
1391 req
.r
.rtm_src_len
= src_p
? src_p
->prefixlen
: 0;
1392 req
.r
.rtm_protocol
= zebra2proto(re
->type
);
1393 req
.r
.rtm_scope
= RT_SCOPE_UNIVERSE
;
1396 * blackhole routes are not RTN_UNICAST, they are
1397 * RTN_ BLACKHOLE|UNREACHABLE|PROHIBIT
1398 * so setting this value as a RTN_UNICAST would
1399 * cause the route lookup of just the prefix
1400 * to fail. So no need to specify this for
1401 * the RTM_DELROUTE case
1403 if (cmd
!= RTM_DELROUTE
)
1404 req
.r
.rtm_type
= RTN_UNICAST
;
1406 addattr_l(&req
.n
, sizeof req
, RTA_DST
, &p
->u
.prefix
, bytelen
);
1408 addattr_l(&req
.n
, sizeof req
, RTA_SRC
, &src_p
->u
.prefix
,
1412 /* Hardcode the metric for all routes coming from zebra. Metric isn't
1414 * either by the kernel or by zebra. Its purely for calculating best
1416 * by the routing protocol and for communicating with protocol peers.
1418 addattr32(&req
.n
, sizeof req
, RTA_PRIORITY
, NL_DEFAULT_ROUTE_METRIC
);
1419 #if defined(SUPPORT_REALMS)
1420 if (re
->tag
> 0 && re
->tag
<= 255)
1421 addattr32(&req
.n
, sizeof req
, RTA_FLOW
, re
->tag
);
1423 /* Table corresponding to this route. */
1424 if (re
->table
< 256)
1425 req
.r
.rtm_table
= re
->table
;
1427 req
.r
.rtm_table
= RT_TABLE_UNSPEC
;
1428 addattr32(&req
.n
, sizeof req
, RTA_TABLE
, re
->table
);
1431 _netlink_route_debug(cmd
, p
, family
, zvrf_id(zvrf
), re
->table
);
1434 * If we are not updating the route and we have received
1435 * a route delete, then all we need to fill in is the
1436 * prefix information to tell the kernel to schwack
1439 if (!update
&& cmd
== RTM_DELROUTE
)
1442 if (re
->mtu
|| re
->nexthop_mtu
) {
1443 char buf
[NL_PKT_BUF_SIZE
];
1444 struct rtattr
*rta
= (void *)buf
;
1445 uint32_t mtu
= re
->mtu
;
1446 if (!mtu
|| (re
->nexthop_mtu
&& re
->nexthop_mtu
< mtu
))
1447 mtu
= re
->nexthop_mtu
;
1448 rta
->rta_type
= RTA_METRICS
;
1449 rta
->rta_len
= RTA_LENGTH(0);
1450 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTAX_MTU
, &mtu
, sizeof mtu
);
1451 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_METRICS
, RTA_DATA(rta
),
1455 /* Count overall nexthops so we can decide whether to use singlepath
1456 * or multipath case. */
1458 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1459 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
))
1461 if (cmd
== RTM_NEWROUTE
&& !NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1463 if (cmd
== RTM_DELROUTE
1464 && !CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
1470 /* Singlepath case. */
1471 if (nexthop_num
== 1 || multipath_num
== 1) {
1473 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1475 * So we want to cover 2 types of blackhole
1477 * 1) A normal blackhole route( ala from a static
1479 * 2) A recursively resolved blackhole route
1481 if (nexthop
->type
== NEXTHOP_TYPE_BLACKHOLE
) {
1482 switch (nexthop
->bh_type
) {
1483 case BLACKHOLE_ADMINPROHIB
:
1484 req
.r
.rtm_type
= RTN_PROHIBIT
;
1486 case BLACKHOLE_REJECT
:
1487 req
.r
.rtm_type
= RTN_UNREACHABLE
;
1490 req
.r
.rtm_type
= RTN_BLACKHOLE
;
1495 if (CHECK_FLAG(nexthop
->flags
,
1496 NEXTHOP_FLAG_RECURSIVE
)) {
1498 if (family
== AF_INET
) {
1499 if (nexthop
->rmap_src
.ipv4
1506 } else if (nexthop
->src
.ipv4
1514 } else if (family
== AF_INET6
) {
1515 if (!IN6_IS_ADDR_UNSPECIFIED(
1523 !IN6_IS_ADDR_UNSPECIFIED(
1536 if ((cmd
== RTM_NEWROUTE
1537 && NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1538 || (cmd
== RTM_DELROUTE
1539 && CHECK_FLAG(nexthop
->flags
,
1540 NEXTHOP_FLAG_FIB
))) {
1541 routedesc
= nexthop
->rparent
1542 ? "recursive, single-path"
1545 _netlink_route_build_singlepath(
1546 routedesc
, bytelen
, nexthop
, &req
.n
,
1547 &req
.r
, sizeof req
, cmd
);
1552 if (setsrc
&& (cmd
== RTM_NEWROUTE
)) {
1553 if (family
== AF_INET
)
1554 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1555 &src
.ipv4
, bytelen
);
1556 else if (family
== AF_INET6
)
1557 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1558 &src
.ipv6
, bytelen
);
1561 char buf
[NL_PKT_BUF_SIZE
];
1562 struct rtattr
*rta
= (void *)buf
;
1563 struct rtnexthop
*rtnh
;
1564 union g_addr
*src1
= NULL
;
1566 rta
->rta_type
= RTA_MULTIPATH
;
1567 rta
->rta_len
= RTA_LENGTH(0);
1568 rtnh
= RTA_DATA(rta
);
1571 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1572 if (nexthop_num
>= multipath_num
)
1575 if (CHECK_FLAG(nexthop
->flags
,
1576 NEXTHOP_FLAG_RECURSIVE
)) {
1577 /* This only works for IPv4 now */
1579 if (family
== AF_INET
) {
1580 if (nexthop
->rmap_src
.ipv4
1587 } else if (nexthop
->src
.ipv4
1595 } else if (family
== AF_INET6
) {
1596 if (!IN6_IS_ADDR_UNSPECIFIED(
1604 !IN6_IS_ADDR_UNSPECIFIED(
1617 if ((cmd
== RTM_NEWROUTE
1618 && NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1619 || (cmd
== RTM_DELROUTE
1620 && CHECK_FLAG(nexthop
->flags
,
1621 NEXTHOP_FLAG_FIB
))) {
1622 routedesc
= nexthop
->rparent
1623 ? "recursive, multipath"
1627 _netlink_route_build_multipath(
1628 routedesc
, bytelen
, nexthop
, rta
, rtnh
,
1630 rtnh
= RTNH_NEXT(rtnh
);
1632 if (!setsrc
&& src1
) {
1633 if (family
== AF_INET
)
1634 src
.ipv4
= src1
->ipv4
;
1635 else if (family
== AF_INET6
)
1636 src
.ipv6
= src1
->ipv6
;
1642 if (setsrc
&& (cmd
== RTM_NEWROUTE
)) {
1643 if (family
== AF_INET
)
1644 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1645 &src
.ipv4
, bytelen
);
1646 else if (family
== AF_INET6
)
1647 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1648 &src
.ipv6
, bytelen
);
1649 if (IS_ZEBRA_DEBUG_KERNEL
)
1650 zlog_debug("Setting source");
1653 if (rta
->rta_len
> RTA_LENGTH(0))
1654 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_MULTIPATH
,
1655 RTA_DATA(rta
), RTA_PAYLOAD(rta
));
1658 /* If there is no useful nexthop then return. */
1659 if (nexthop_num
== 0) {
1660 if (IS_ZEBRA_DEBUG_KERNEL
)
1662 "netlink_route_multipath(): No useful nexthop.");
1668 /* Destination netlink address. */
1669 memset(&snl
, 0, sizeof snl
);
1670 snl
.nl_family
= AF_NETLINK
;
1672 /* Talk to netlink socket. */
1673 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1677 int kernel_get_ipmr_sg_stats(struct zebra_vrf
*zvrf
, void *in
)
1680 struct mcast_route_data
*mr
= (struct mcast_route_data
*)in
;
1688 struct zebra_ns
*zns
;
1691 memset(&req
, 0, sizeof(req
));
1693 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1694 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
1695 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1697 req
.ndm
.ndm_family
= RTNL_FAMILY_IPMR
;
1698 req
.n
.nlmsg_type
= RTM_GETROUTE
;
1700 addattr_l(&req
.n
, sizeof(req
), RTA_IIF
, &mroute
->ifindex
, 4);
1701 addattr_l(&req
.n
, sizeof(req
), RTA_OIF
, &mroute
->ifindex
, 4);
1702 addattr_l(&req
.n
, sizeof(req
), RTA_SRC
, &mroute
->sg
.src
.s_addr
, 4);
1703 addattr_l(&req
.n
, sizeof(req
), RTA_DST
, &mroute
->sg
.grp
.s_addr
, 4);
1704 addattr_l(&req
.n
, sizeof(req
), RTA_TABLE
, &zvrf
->table_id
, 4);
1706 suc
= netlink_talk(netlink_route_change_read_multicast
, &req
.n
,
1707 &zns
->netlink_cmd
, zns
, 0);
1713 enum dp_req_result
kernel_route_rib(struct route_node
*rn
,
1714 const struct prefix
*p
,
1715 const struct prefix
*src_p
,
1716 struct route_entry
*old
,
1717 struct route_entry
*new)
1724 if (p
->family
== AF_INET
|| v6_rr_semantics
)
1725 ret
= netlink_route_multipath(RTM_NEWROUTE
, p
, src_p
,
1726 new, (old
) ? 1 : 0);
1729 * So v6 route replace semantics are not in
1730 * the kernel at this point as I understand it.
1731 * So let's do a delete than an add.
1732 * In the future once v6 route replace semantics
1733 * are in we can figure out what to do here to
1734 * allow working with old and new kernels.
1736 * I'm also intentionally ignoring the failure case
1737 * of the route delete. If that happens yeah we're
1741 netlink_route_multipath(RTM_DELROUTE
, p
, src_p
,
1743 ret
= netlink_route_multipath(RTM_NEWROUTE
, p
, src_p
,
1746 kernel_route_rib_pass_fail(rn
, p
, new,
1747 (!ret
) ? DP_INSTALL_SUCCESS
1748 : DP_INSTALL_FAILURE
);
1749 return DP_REQUEST_SUCCESS
;
1753 ret
= netlink_route_multipath(RTM_DELROUTE
, p
, src_p
, old
, 0);
1755 kernel_route_rib_pass_fail(rn
, p
, old
,
1756 (!ret
) ? DP_DELETE_SUCCESS
1757 : DP_DELETE_FAILURE
);
1760 return DP_REQUEST_SUCCESS
;
1763 int kernel_neigh_update(int add
, int ifindex
, uint32_t addr
, char *lla
,
1764 int llalen
, ns_id_t ns_id
)
1766 return netlink_neigh_update(add
? RTM_NEWNEIGH
: RTM_DELNEIGH
, ifindex
,
1767 addr
, lla
, llalen
, ns_id
);
1771 * Add remote VTEP to the flood list for this VxLAN interface (VNI). This
1772 * is done by adding an FDB entry with a MAC of 00:00:00:00:00:00.
1774 static int netlink_vxlan_flood_list_update(struct interface
*ifp
,
1775 struct in_addr
*vtep_ip
, int cmd
)
1777 struct zebra_ns
*zns
;
1783 uint8_t dst_mac
[6] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0};
1784 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
1787 memset(&req
, 0, sizeof(req
));
1789 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1790 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
1791 if (cmd
== RTM_NEWNEIGH
)
1792 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_APPEND
);
1793 req
.n
.nlmsg_type
= cmd
;
1794 req
.ndm
.ndm_family
= PF_BRIDGE
;
1795 req
.ndm
.ndm_state
= NUD_NOARP
| NUD_PERMANENT
;
1796 req
.ndm
.ndm_flags
|= NTF_SELF
; // Handle by "self", not "master"
1799 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, &dst_mac
, 6);
1800 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
1801 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &vtep_ip
->s_addr
, 4);
1803 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1808 * Add remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1810 * a "flood" MAC FDB entry.
1812 int kernel_add_vtep(vni_t vni
, struct interface
*ifp
, struct in_addr
*vtep_ip
)
1814 if (IS_ZEBRA_DEBUG_VXLAN
)
1815 zlog_debug("Install %s into flood list for VNI %u intf %s(%u)",
1816 inet_ntoa(*vtep_ip
), vni
, ifp
->name
, ifp
->ifindex
);
1818 return netlink_vxlan_flood_list_update(ifp
, vtep_ip
, RTM_NEWNEIGH
);
1822 * Remove remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1823 * deleting the "flood" MAC FDB entry.
1825 int kernel_del_vtep(vni_t vni
, struct interface
*ifp
, struct in_addr
*vtep_ip
)
1827 if (IS_ZEBRA_DEBUG_VXLAN
)
1829 "Uninstall %s from flood list for VNI %u intf %s(%u)",
1830 inet_ntoa(*vtep_ip
), vni
, ifp
->name
, ifp
->ifindex
);
1832 return netlink_vxlan_flood_list_update(ifp
, vtep_ip
, RTM_DELNEIGH
);
1836 #define NDA_RTA(r) \
1837 ((struct rtattr *)(((char *)(r)) + NLMSG_ALIGN(sizeof(struct ndmsg))))
1840 static int netlink_macfdb_change(struct nlmsghdr
*h
, int len
, ns_id_t ns_id
)
1843 struct interface
*ifp
;
1844 struct zebra_if
*zif
;
1845 struct rtattr
*tb
[NDA_MAX
+ 1];
1846 struct interface
*br_if
;
1849 struct prefix vtep_ip
;
1850 int vid_present
= 0, dst_present
= 0;
1851 char buf
[ETHER_ADDR_STRLEN
];
1856 ndm
= NLMSG_DATA(h
);
1858 /* We only process macfdb notifications if EVPN is enabled */
1859 if (!is_evpn_enabled())
1862 /* The interface should exist. */
1863 ifp
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
1865 if (!ifp
|| !ifp
->info
)
1868 /* The interface should be something we're interested in. */
1869 if (!IS_ZEBRA_IF_BRIDGE_SLAVE(ifp
))
1872 /* Drop "permanent" entries. */
1873 if (ndm
->ndm_state
& NUD_PERMANENT
)
1876 zif
= (struct zebra_if
*)ifp
->info
;
1877 if ((br_if
= zif
->brslave_info
.br_if
) == NULL
) {
1878 zlog_warn("%s family %s IF %s(%u) brIF %u - no bridge master",
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
);
1885 /* Parse attributes and extract fields of interest. */
1886 memset(tb
, 0, sizeof tb
);
1887 netlink_parse_rtattr(tb
, NDA_MAX
, NDA_RTA(ndm
), len
);
1889 if (!tb
[NDA_LLADDR
]) {
1890 zlog_warn("%s family %s IF %s(%u) brIF %u - no LLADDR",
1891 nl_msg_type_to_str(h
->nlmsg_type
),
1892 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1893 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
);
1897 if (RTA_PAYLOAD(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
1899 "%s family %s IF %s(%u) brIF %u - LLADDR is not MAC, len %lu",
1900 nl_msg_type_to_str(h
->nlmsg_type
),
1901 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1902 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
,
1903 (unsigned long)RTA_PAYLOAD(tb
[NDA_LLADDR
]));
1907 memcpy(&mac
, RTA_DATA(tb
[NDA_LLADDR
]), ETH_ALEN
);
1909 if ((NDA_VLAN
<= NDA_MAX
) && tb
[NDA_VLAN
]) {
1911 vid
= *(uint16_t *)RTA_DATA(tb
[NDA_VLAN
]);
1912 sprintf(vid_buf
, " VLAN %u", vid
);
1916 /* TODO: Only IPv4 supported now. */
1918 vtep_ip
.family
= AF_INET
;
1919 vtep_ip
.prefixlen
= IPV4_MAX_BITLEN
;
1920 memcpy(&(vtep_ip
.u
.prefix4
.s_addr
), RTA_DATA(tb
[NDA_DST
]),
1922 sprintf(dst_buf
, " dst %s", inet_ntoa(vtep_ip
.u
.prefix4
));
1925 sticky
= (ndm
->ndm_state
& NUD_NOARP
) ? 1 : 0;
1927 if (IS_ZEBRA_DEBUG_KERNEL
)
1928 zlog_debug("Rx %s family %s IF %s(%u)%s %sMAC %s%s",
1929 nl_msg_type_to_str(h
->nlmsg_type
),
1930 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1931 ndm
->ndm_ifindex
, vid_present
? vid_buf
: "",
1932 sticky
? "sticky " : "",
1933 prefix_mac2str(&mac
, buf
, sizeof(buf
)),
1934 dst_present
? dst_buf
: "");
1936 if (filter_vlan
&& vid
!= filter_vlan
)
1939 /* If add or update, do accordingly if learnt on a "local" interface; if
1940 * the notification is over VxLAN, this has to be related to
1942 * so perform an implicit delete of any local entry (if it exists).
1944 if (h
->nlmsg_type
== RTM_NEWNEIGH
) {
1945 /* Drop "permanent" entries. */
1946 if (ndm
->ndm_state
& NUD_PERMANENT
)
1949 if (IS_ZEBRA_IF_VXLAN(ifp
))
1950 return zebra_vxlan_check_del_local_mac(ifp
, br_if
, &mac
,
1953 return zebra_vxlan_local_mac_add_update(ifp
, br_if
, &mac
, vid
,
1957 /* This is a delete notification.
1958 * 1. For a MAC over VxLan, check if it needs to be refreshed(readded)
1959 * 2. For a MAC over "local" interface, delete the mac
1960 * Note: We will get notifications from both bridge driver and VxLAN
1962 * Ignore the notification from VxLan driver as it is also generated
1963 * when mac moves from remote to local.
1968 if (IS_ZEBRA_IF_VXLAN(ifp
))
1969 return zebra_vxlan_check_readd_remote_mac(ifp
, br_if
, &mac
,
1972 return zebra_vxlan_local_mac_del(ifp
, br_if
, &mac
, vid
);
1975 static int netlink_macfdb_table(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
1980 if (h
->nlmsg_type
!= RTM_NEWNEIGH
)
1983 /* Length validity. */
1984 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
1988 /* We are interested only in AF_BRIDGE notifications. */
1989 ndm
= NLMSG_DATA(h
);
1990 if (ndm
->ndm_family
!= AF_BRIDGE
)
1993 return netlink_macfdb_change(h
, len
, ns_id
);
1996 /* Request for MAC FDB information from the kernel */
1997 static int netlink_request_macs(struct zebra_ns
*zns
, int family
, int type
,
1998 ifindex_t master_ifindex
)
2002 struct ifinfomsg ifm
;
2006 /* Form the request, specifying filter (rtattr) if needed. */
2007 memset(&req
, 0, sizeof(req
));
2008 req
.n
.nlmsg_type
= type
;
2009 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ifinfomsg
));
2010 req
.ifm
.ifi_family
= family
;
2012 addattr32(&req
.n
, sizeof(req
), IFLA_MASTER
, master_ifindex
);
2014 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
2018 * MAC forwarding database read using netlink interface. This is invoked
2021 int netlink_macfdb_read(struct zebra_ns
*zns
)
2025 /* Get bridge FDB table. */
2026 ret
= netlink_request_macs(zns
, AF_BRIDGE
, RTM_GETNEIGH
, 0);
2029 /* We are reading entire table. */
2031 ret
= netlink_parse_info(netlink_macfdb_table
, &zns
->netlink_cmd
, zns
,
2038 * MAC forwarding database read using netlink interface. This is for a
2039 * specific bridge and matching specific access VLAN (if VLAN-aware bridge).
2041 int netlink_macfdb_read_for_bridge(struct zebra_ns
*zns
, struct interface
*ifp
,
2042 struct interface
*br_if
)
2044 struct zebra_if
*br_zif
;
2045 struct zebra_if
*zif
;
2046 struct zebra_l2info_vxlan
*vxl
;
2050 /* Save VLAN we're filtering on, if needed. */
2051 br_zif
= (struct zebra_if
*)br_if
->info
;
2052 zif
= (struct zebra_if
*)ifp
->info
;
2053 vxl
= &zif
->l2info
.vxl
;
2054 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif
))
2055 filter_vlan
= vxl
->access_vlan
;
2057 /* Get bridge FDB table for specific bridge - we do the VLAN filtering.
2059 ret
= netlink_request_macs(zns
, AF_BRIDGE
, RTM_GETNEIGH
,
2063 ret
= netlink_parse_info(netlink_macfdb_table
, &zns
->netlink_cmd
, zns
,
2066 /* Reset VLAN filter. */
2071 static int netlink_macfdb_update(struct interface
*ifp
, vlanid_t vid
,
2072 struct ethaddr
*mac
, struct in_addr vtep_ip
,
2073 int local
, int cmd
, uint8_t sticky
)
2075 struct zebra_ns
*zns
;
2082 struct zebra_if
*zif
;
2083 struct interface
*br_if
;
2084 struct zebra_if
*br_zif
;
2085 char buf
[ETHER_ADDR_STRLEN
];
2086 int vid_present
= 0, dst_present
= 0;
2089 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
2093 if ((br_if
= zif
->brslave_info
.br_if
) == NULL
) {
2094 zlog_warn("MAC %s on IF %s(%u) - no mapping to bridge",
2095 (cmd
== RTM_NEWNEIGH
) ? "add" : "del", ifp
->name
,
2100 memset(&req
, 0, sizeof(req
));
2102 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2103 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
2104 if (cmd
== RTM_NEWNEIGH
)
2105 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_REPLACE
);
2106 req
.n
.nlmsg_type
= cmd
;
2107 req
.ndm
.ndm_family
= AF_BRIDGE
;
2108 req
.ndm
.ndm_flags
|= NTF_SELF
| NTF_MASTER
;
2109 req
.ndm
.ndm_state
= NUD_REACHABLE
;
2112 req
.ndm
.ndm_state
|= NUD_NOARP
;
2114 req
.ndm
.ndm_flags
|= NTF_EXT_LEARNED
;
2116 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, mac
, 6);
2117 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
2119 dst_alen
= 4; // TODO: hardcoded
2120 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &vtep_ip
, dst_alen
);
2122 sprintf(dst_buf
, " dst %s", inet_ntoa(vtep_ip
));
2124 br_zif
= (struct zebra_if
*)br_if
->info
;
2125 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif
) && vid
> 0) {
2126 addattr16(&req
.n
, sizeof(req
), NDA_VLAN
, vid
);
2128 sprintf(vid_buf
, " VLAN %u", vid
);
2130 addattr32(&req
.n
, sizeof(req
), NDA_MASTER
, br_if
->ifindex
);
2132 if (IS_ZEBRA_DEBUG_KERNEL
)
2133 zlog_debug("Tx %s family %s IF %s(%u)%s %sMAC %s%s",
2134 nl_msg_type_to_str(cmd
),
2135 nl_family_to_str(req
.ndm
.ndm_family
), ifp
->name
,
2136 ifp
->ifindex
, vid_present
? vid_buf
: "",
2137 sticky
? "sticky " : "",
2138 prefix_mac2str(mac
, buf
, sizeof(buf
)),
2139 dst_present
? dst_buf
: "");
2141 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2146 (NUD_PERMANENT | NUD_NOARP | NUD_REACHABLE | NUD_PROBE | NUD_STALE \
2149 static int netlink_ipneigh_change(struct nlmsghdr
*h
, int len
, ns_id_t ns_id
)
2152 struct interface
*ifp
;
2153 struct zebra_if
*zif
;
2154 struct rtattr
*tb
[NDA_MAX
+ 1];
2155 struct interface
*link_if
;
2158 char buf
[ETHER_ADDR_STRLEN
];
2159 char buf2
[INET6_ADDRSTRLEN
];
2160 int mac_present
= 0;
2161 uint8_t ext_learned
;
2163 ndm
= NLMSG_DATA(h
);
2165 /* The interface should exist. */
2166 ifp
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
2168 if (!ifp
|| !ifp
->info
)
2171 zif
= (struct zebra_if
*)ifp
->info
;
2173 /* Parse attributes and extract fields of interest. */
2174 memset(tb
, 0, sizeof tb
);
2175 netlink_parse_rtattr(tb
, NDA_MAX
, NDA_RTA(ndm
), len
);
2178 zlog_warn("%s family %s IF %s(%u) - no DST",
2179 nl_msg_type_to_str(h
->nlmsg_type
),
2180 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2185 memset(&ip
, 0, sizeof(struct ipaddr
));
2186 ip
.ipa_type
= (ndm
->ndm_family
== AF_INET
) ? IPADDR_V4
: IPADDR_V6
;
2187 memcpy(&ip
.ip
.addr
, RTA_DATA(tb
[NDA_DST
]), RTA_PAYLOAD(tb
[NDA_DST
]));
2189 /* Drop some "permanent" entries. */
2190 if (ndm
->ndm_state
& NUD_PERMANENT
) {
2191 char buf
[16] = "169.254.0.1";
2192 struct in_addr ipv4_ll
;
2194 if (ndm
->ndm_family
!= AF_INET
)
2197 if (!zif
->v6_2_v4_ll_neigh_entry
)
2200 if (h
->nlmsg_type
!= RTM_DELNEIGH
)
2203 inet_pton(AF_INET
, buf
, &ipv4_ll
);
2204 if (ipv4_ll
.s_addr
!= ip
.ip
._v4_addr
.s_addr
)
2207 if_nbr_ipv6ll_to_ipv4ll_neigh_update(
2208 ifp
, &zif
->v6_2_v4_ll_addr6
, true);
2212 /* The neighbor is present on an SVI. From this, we locate the
2214 * bridge because we're only interested in neighbors on a VxLAN bridge.
2215 * The bridge is located based on the nature of the SVI:
2216 * (a) In the case of a VLAN-aware bridge, the SVI is a L3 VLAN
2218 * and is linked to the bridge
2219 * (b) In the case of a VLAN-unaware bridge, the SVI is the bridge
2223 if (IS_ZEBRA_IF_VLAN(ifp
)) {
2224 link_if
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
2228 } else if (IS_ZEBRA_IF_BRIDGE(ifp
))
2233 memset(&mac
, 0, sizeof(struct ethaddr
));
2234 if (h
->nlmsg_type
== RTM_NEWNEIGH
) {
2235 if (tb
[NDA_LLADDR
]) {
2236 if (RTA_PAYLOAD(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
2238 "%s family %s IF %s(%u) - LLADDR is not MAC, len %lu",
2239 nl_msg_type_to_str(h
->nlmsg_type
),
2240 nl_family_to_str(ndm
->ndm_family
),
2241 ifp
->name
, ndm
->ndm_ifindex
,
2242 (unsigned long)RTA_PAYLOAD(
2248 memcpy(&mac
, RTA_DATA(tb
[NDA_LLADDR
]), ETH_ALEN
);
2251 ext_learned
= (ndm
->ndm_flags
& NTF_EXT_LEARNED
) ? 1 : 0;
2253 if (IS_ZEBRA_DEBUG_KERNEL
)
2255 "Rx %s family %s IF %s(%u) IP %s MAC %s state 0x%x flags 0x%x",
2256 nl_msg_type_to_str(h
->nlmsg_type
),
2257 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2259 ipaddr2str(&ip
, buf2
, sizeof(buf2
)),
2261 ? prefix_mac2str(&mac
, buf
, sizeof(buf
))
2263 ndm
->ndm_state
, ndm
->ndm_flags
);
2265 /* If the neighbor state is valid for use, process as an add or
2267 * else process as a delete. Note that the delete handling may
2269 * in re-adding the neighbor if it is a valid "remote" neighbor.
2271 if (ndm
->ndm_state
& NUD_VALID
)
2272 return zebra_vxlan_handle_kernel_neigh_update(
2273 ifp
, link_if
, &ip
, &mac
, ndm
->ndm_state
,
2276 return zebra_vxlan_handle_kernel_neigh_del(ifp
, link_if
, &ip
);
2279 if (IS_ZEBRA_DEBUG_KERNEL
)
2280 zlog_debug("Rx %s family %s IF %s(%u) IP %s",
2281 nl_msg_type_to_str(h
->nlmsg_type
),
2282 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2284 ipaddr2str(&ip
, buf2
, sizeof(buf2
)));
2286 /* Process the delete - it may result in re-adding the neighbor if it is
2287 * a valid "remote" neighbor.
2289 return zebra_vxlan_handle_kernel_neigh_del(ifp
, link_if
, &ip
);
2292 static int netlink_neigh_table(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
2297 if (h
->nlmsg_type
!= RTM_NEWNEIGH
)
2300 /* Length validity. */
2301 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2305 /* We are interested only in AF_INET or AF_INET6 notifications. */
2306 ndm
= NLMSG_DATA(h
);
2307 if (ndm
->ndm_family
!= AF_INET
&& ndm
->ndm_family
!= AF_INET6
)
2310 return netlink_neigh_change(h
, len
);
2313 /* Request for IP neighbor information from the kernel */
2314 static int netlink_request_neigh(struct zebra_ns
*zns
, int family
, int type
,
2323 /* Form the request, specifying filter (rtattr) if needed. */
2324 memset(&req
, 0, sizeof(req
));
2325 req
.n
.nlmsg_type
= type
;
2326 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2327 req
.ndm
.ndm_family
= family
;
2329 addattr32(&req
.n
, sizeof(req
), NDA_IFINDEX
, ifindex
);
2331 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
2335 * IP Neighbor table read using netlink interface. This is invoked
2338 int netlink_neigh_read(struct zebra_ns
*zns
)
2342 /* Get IP neighbor table. */
2343 ret
= netlink_request_neigh(zns
, AF_UNSPEC
, RTM_GETNEIGH
, 0);
2346 ret
= netlink_parse_info(netlink_neigh_table
, &zns
->netlink_cmd
, zns
, 0,
2353 * IP Neighbor table read using netlink interface. This is for a specific
2356 int netlink_neigh_read_for_vlan(struct zebra_ns
*zns
, struct interface
*vlan_if
)
2360 ret
= netlink_request_neigh(zns
, AF_UNSPEC
, RTM_GETNEIGH
,
2364 ret
= netlink_parse_info(netlink_neigh_table
, &zns
->netlink_cmd
, zns
, 0,
2370 int netlink_neigh_change(struct nlmsghdr
*h
, ns_id_t ns_id
)
2375 if (!(h
->nlmsg_type
== RTM_NEWNEIGH
|| h
->nlmsg_type
== RTM_DELNEIGH
))
2378 /* Length validity. */
2379 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2381 zlog_err("%s: Message received from netlink is of a broken size %d %zu",
2382 __PRETTY_FUNCTION__
, h
->nlmsg_len
,
2383 (size_t)NLMSG_LENGTH(sizeof(struct ndmsg
)));
2387 /* Is this a notification for the MAC FDB or IP neighbor table? */
2388 ndm
= NLMSG_DATA(h
);
2389 if (ndm
->ndm_family
== AF_BRIDGE
)
2390 return netlink_macfdb_change(h
, len
, ns_id
);
2392 if (ndm
->ndm_type
!= RTN_UNICAST
)
2395 if (ndm
->ndm_family
== AF_INET
|| ndm
->ndm_family
== AF_INET6
)
2396 return netlink_ipneigh_change(h
, len
, ns_id
);
2399 "Invalid address family: %d received from kernel neighbor change: %d",
2400 ndm
->ndm_family
, h
->nlmsg_type
);
2407 static int netlink_neigh_update2(struct interface
*ifp
, struct ipaddr
*ip
,
2408 struct ethaddr
*mac
, uint32_t flags
, int cmd
)
2417 struct zebra_ns
*zns
;
2418 char buf
[INET6_ADDRSTRLEN
];
2419 char buf2
[ETHER_ADDR_STRLEN
];
2420 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
2423 memset(&req
, 0, sizeof(req
));
2425 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2426 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
2427 if (cmd
== RTM_NEWNEIGH
)
2428 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_REPLACE
);
2429 req
.n
.nlmsg_type
= cmd
; // RTM_NEWNEIGH or RTM_DELNEIGH
2430 req
.ndm
.ndm_family
= IS_IPADDR_V4(ip
) ? AF_INET
: AF_INET6
;
2431 req
.ndm
.ndm_state
= flags
;
2432 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
2433 req
.ndm
.ndm_type
= RTN_UNICAST
;
2434 req
.ndm
.ndm_flags
= NTF_EXT_LEARNED
;
2437 ipa_len
= IS_IPADDR_V4(ip
) ? IPV4_MAX_BYTELEN
: IPV6_MAX_BYTELEN
;
2438 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &ip
->ip
.addr
, ipa_len
);
2440 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, mac
, 6);
2442 if (IS_ZEBRA_DEBUG_KERNEL
)
2443 zlog_debug("Tx %s family %s IF %s(%u) Neigh %s MAC %s",
2444 nl_msg_type_to_str(cmd
),
2445 nl_family_to_str(req
.ndm
.ndm_family
), ifp
->name
,
2446 ifp
->ifindex
, ipaddr2str(ip
, buf
, sizeof(buf
)),
2447 mac
? prefix_mac2str(mac
, buf2
, sizeof(buf2
))
2450 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2454 int kernel_add_mac(struct interface
*ifp
, vlanid_t vid
, struct ethaddr
*mac
,
2455 struct in_addr vtep_ip
, uint8_t sticky
)
2457 return netlink_macfdb_update(ifp
, vid
, mac
, vtep_ip
, 0, RTM_NEWNEIGH
,
2461 int kernel_del_mac(struct interface
*ifp
, vlanid_t vid
, struct ethaddr
*mac
,
2462 struct in_addr vtep_ip
, int local
)
2464 return netlink_macfdb_update(ifp
, vid
, mac
, vtep_ip
, local
,
2468 int kernel_add_neigh(struct interface
*ifp
, struct ipaddr
*ip
,
2469 struct ethaddr
*mac
)
2471 return netlink_neigh_update2(ifp
, ip
, mac
, NUD_NOARP
, RTM_NEWNEIGH
);
2474 int kernel_del_neigh(struct interface
*ifp
, struct ipaddr
*ip
)
2476 return netlink_neigh_update2(ifp
, ip
, NULL
, 0, RTM_DELNEIGH
);
2480 * MPLS label forwarding table change via netlink interface.
2482 int netlink_mpls_multipath(int cmd
, zebra_lsp_t
*lsp
)
2485 zebra_nhlfe_t
*nhlfe
;
2486 struct nexthop
*nexthop
= NULL
;
2487 unsigned int nexthop_num
;
2488 const char *routedesc
;
2489 struct zebra_ns
*zns
= zebra_ns_lookup(NS_DEFAULT
);
2495 char buf
[NL_PKT_BUF_SIZE
];
2498 memset(&req
, 0, sizeof req
- NL_PKT_BUF_SIZE
);
2501 * Count # nexthops so we can decide whether to use singlepath
2502 * or multipath case.
2505 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2506 nexthop
= nhlfe
->nexthop
;
2509 if (cmd
== RTM_NEWROUTE
) {
2510 /* Count all selected NHLFEs */
2511 if (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2512 && CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
))
2516 /* Count all installed NHLFEs */
2517 if (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_INSTALLED
)
2518 && CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
2523 if ((nexthop_num
== 0) || (!lsp
->best_nhlfe
&& (cmd
!= RTM_DELROUTE
)))
2526 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
2527 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
2528 req
.n
.nlmsg_type
= cmd
;
2529 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
2531 req
.r
.rtm_family
= AF_MPLS
;
2532 req
.r
.rtm_table
= RT_TABLE_MAIN
;
2533 req
.r
.rtm_dst_len
= MPLS_LABEL_LEN_BITS
;
2534 req
.r
.rtm_scope
= RT_SCOPE_UNIVERSE
;
2535 req
.r
.rtm_type
= RTN_UNICAST
;
2537 if (cmd
== RTM_NEWROUTE
) {
2538 /* We do a replace to handle update. */
2539 req
.n
.nlmsg_flags
|= NLM_F_REPLACE
;
2541 /* set the protocol value if installing */
2542 route_type
= re_type_from_lsp_type(lsp
->best_nhlfe
->type
);
2543 req
.r
.rtm_protocol
= zebra2proto(route_type
);
2546 /* Fill destination */
2547 lse
= mpls_lse_encode(lsp
->ile
.in_label
, 0, 0, 1);
2548 addattr_l(&req
.n
, sizeof req
, RTA_DST
, &lse
, sizeof(mpls_lse_t
));
2550 /* Fill nexthops (paths) based on single-path or multipath. The paths
2551 * chosen depend on the operation.
2553 if (nexthop_num
== 1 || multipath_num
== 1) {
2554 routedesc
= "single-path";
2555 _netlink_mpls_debug(cmd
, lsp
->ile
.in_label
, routedesc
);
2558 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2559 nexthop
= nhlfe
->nexthop
;
2563 if ((cmd
== RTM_NEWROUTE
2564 && (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2565 && CHECK_FLAG(nexthop
->flags
,
2566 NEXTHOP_FLAG_ACTIVE
)))
2567 || (cmd
== RTM_DELROUTE
2568 && (CHECK_FLAG(nhlfe
->flags
,
2569 NHLFE_FLAG_INSTALLED
)
2570 && CHECK_FLAG(nexthop
->flags
,
2571 NEXTHOP_FLAG_FIB
)))) {
2572 /* Add the gateway */
2573 _netlink_mpls_build_singlepath(routedesc
, nhlfe
,
2580 } else /* Multipath case */
2582 char buf
[NL_PKT_BUF_SIZE
];
2583 struct rtattr
*rta
= (void *)buf
;
2584 struct rtnexthop
*rtnh
;
2585 union g_addr
*src1
= NULL
;
2587 rta
->rta_type
= RTA_MULTIPATH
;
2588 rta
->rta_len
= RTA_LENGTH(0);
2589 rtnh
= RTA_DATA(rta
);
2591 routedesc
= "multipath";
2592 _netlink_mpls_debug(cmd
, lsp
->ile
.in_label
, routedesc
);
2595 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2596 nexthop
= nhlfe
->nexthop
;
2600 if (nexthop_num
>= multipath_num
)
2603 if ((cmd
== RTM_NEWROUTE
2604 && (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2605 && CHECK_FLAG(nexthop
->flags
,
2606 NEXTHOP_FLAG_ACTIVE
)))
2607 || (cmd
== RTM_DELROUTE
2608 && (CHECK_FLAG(nhlfe
->flags
,
2609 NHLFE_FLAG_INSTALLED
)
2610 && CHECK_FLAG(nexthop
->flags
,
2611 NEXTHOP_FLAG_FIB
)))) {
2614 /* Build the multipath */
2615 _netlink_mpls_build_multipath(routedesc
, nhlfe
,
2618 rtnh
= RTNH_NEXT(rtnh
);
2622 /* Add the multipath */
2623 if (rta
->rta_len
> RTA_LENGTH(0))
2624 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_MULTIPATH
,
2625 RTA_DATA(rta
), RTA_PAYLOAD(rta
));
2628 /* Talk to netlink socket. */
2629 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2632 #endif /* HAVE_NETLINK */