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/zserv.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_STATIC
) || (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
)) {
108 static inline int zebra2proto(int proto
)
111 case ZEBRA_ROUTE_BABEL
:
112 proto
= RTPROT_BABEL
;
114 case ZEBRA_ROUTE_BGP
:
117 case ZEBRA_ROUTE_OSPF
:
118 case ZEBRA_ROUTE_OSPF6
:
121 case ZEBRA_ROUTE_STATIC
:
122 proto
= RTPROT_STATIC
;
124 case ZEBRA_ROUTE_ISIS
:
127 case ZEBRA_ROUTE_RIP
:
130 case ZEBRA_ROUTE_RIPNG
:
131 proto
= RTPROT_RIPNG
;
133 case ZEBRA_ROUTE_NHRP
:
136 case ZEBRA_ROUTE_EIGRP
:
137 proto
= RTPROT_EIGRP
;
139 case ZEBRA_ROUTE_LDP
:
142 case ZEBRA_ROUTE_SHARP
:
143 proto
= RTPROT_SHARP
;
146 proto
= RTPROT_ZEBRA
;
153 static inline int proto2zebra(int proto
, int family
)
157 proto
= ZEBRA_ROUTE_BABEL
;
160 proto
= ZEBRA_ROUTE_BGP
;
163 proto
= (family
== AFI_IP
) ?
164 ZEBRA_ROUTE_OSPF
: ZEBRA_ROUTE_OSPF6
;
167 proto
= ZEBRA_ROUTE_ISIS
;
170 proto
= ZEBRA_ROUTE_RIP
;
173 proto
= ZEBRA_ROUTE_RIPNG
;
176 proto
= ZEBRA_ROUTE_NHRP
;
179 proto
= ZEBRA_ROUTE_EIGRP
;
182 proto
= ZEBRA_ROUTE_LDP
;
185 proto
= ZEBRA_ROUTE_STATIC
;
188 proto
= ZEBRA_ROUTE_KERNEL
;
195 Pending: create an efficient table_id (in a tree/hash) based lookup)
197 static vrf_id_t
vrf_lookup_by_table(u_int32_t table_id
)
200 struct zebra_vrf
*zvrf
;
202 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
203 if ((zvrf
= vrf
->info
) == NULL
|| (zvrf
->table_id
!= table_id
))
206 return zvrf_id(zvrf
);
212 /* Looking up routing table by netlink interface. */
213 static int netlink_route_change_read_unicast(struct sockaddr_nl
*snl
,
214 struct nlmsghdr
*h
, ns_id_t ns_id
,
219 struct rtattr
*tb
[RTA_MAX
+ 1];
222 struct prefix_ipv6 src_p
= {};
223 vrf_id_t vrf_id
= VRF_DEFAULT
;
225 char anyaddr
[16] = {0};
227 int proto
= ZEBRA_ROUTE_KERNEL
;
232 uint8_t distance
= 0;
236 void *prefsrc
= NULL
; /* IPv4 preferred source host address */
237 void *src
= NULL
; /* IPv6 srcdest source prefix */
238 enum blackhole_type bh_type
= BLACKHOLE_UNSPEC
;
242 if (startup
&& h
->nlmsg_type
!= RTM_NEWROUTE
)
244 switch (rtm
->rtm_type
) {
248 bh_type
= BLACKHOLE_NULL
;
250 case RTN_UNREACHABLE
:
251 bh_type
= BLACKHOLE_REJECT
;
254 bh_type
= BLACKHOLE_ADMINPROHIB
;
260 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
264 memset(tb
, 0, sizeof tb
);
265 netlink_parse_rtattr(tb
, RTA_MAX
, RTM_RTA(rtm
), len
);
267 if (rtm
->rtm_flags
& RTM_F_CLONED
)
269 if (rtm
->rtm_protocol
== RTPROT_REDIRECT
)
271 if (rtm
->rtm_protocol
== RTPROT_KERNEL
)
274 if (!startup
&& is_selfroute(rtm
->rtm_protocol
)
275 && h
->nlmsg_type
== RTM_NEWROUTE
)
278 /* We don't care about change notifications for the MPLS table. */
279 /* TODO: Revisit this. */
280 if (rtm
->rtm_family
== AF_MPLS
)
283 /* Table corresponding to route. */
285 table
= *(int *)RTA_DATA(tb
[RTA_TABLE
]);
287 table
= rtm
->rtm_table
;
290 vrf_id
= vrf_lookup_by_table(table
);
291 if (vrf_id
== VRF_DEFAULT
) {
292 if (!is_zebra_valid_kernel_table(table
)
293 && !is_zebra_main_routing_table(table
))
297 /* Route which inserted by Zebra. */
298 if (is_selfroute(rtm
->rtm_protocol
)) {
299 flags
|= ZEBRA_FLAG_SELFROUTE
;
300 proto
= proto2zebra(rtm
->rtm_protocol
, rtm
->rtm_family
);
303 index
= *(int *)RTA_DATA(tb
[RTA_OIF
]);
306 dest
= RTA_DATA(tb
[RTA_DST
]);
311 src
= RTA_DATA(tb
[RTA_SRC
]);
316 prefsrc
= RTA_DATA(tb
[RTA_PREFSRC
]);
319 gate
= RTA_DATA(tb
[RTA_GATEWAY
]);
321 if (tb
[RTA_PRIORITY
])
322 metric
= *(int *)RTA_DATA(tb
[RTA_PRIORITY
]);
324 if (tb
[RTA_METRICS
]) {
325 struct rtattr
*mxrta
[RTAX_MAX
+ 1];
327 memset(mxrta
, 0, sizeof mxrta
);
328 netlink_parse_rtattr(mxrta
, RTAX_MAX
,
329 RTA_DATA(tb
[RTA_METRICS
]),
330 RTA_PAYLOAD(tb
[RTA_METRICS
]));
333 mtu
= *(u_int32_t
*)RTA_DATA(mxrta
[RTAX_MTU
]);
336 if (rtm
->rtm_family
== AF_INET
) {
338 memcpy(&p
.u
.prefix4
, dest
, 4);
339 p
.prefixlen
= rtm
->rtm_dst_len
;
342 0; // Forces debug below to not display anything
343 } else if (rtm
->rtm_family
== AF_INET6
) {
345 memcpy(&p
.u
.prefix6
, dest
, 16);
346 p
.prefixlen
= rtm
->rtm_dst_len
;
348 src_p
.family
= AF_INET6
;
349 memcpy(&src_p
.prefix
, src
, 16);
350 src_p
.prefixlen
= rtm
->rtm_src_len
;
353 if (rtm
->rtm_src_len
!= 0) {
354 char buf
[PREFIX_STRLEN
];
356 "unsupported IPv[4|6] sourcedest route (dest %s vrf %u)",
357 prefix2str(&p
, buf
, sizeof(buf
)), vrf_id
);
362 * For ZEBRA_ROUTE_KERNEL types:
364 * The metric/priority of the route received from the kernel
365 * is a 32 bit number. We are going to interpret the high
366 * order byte as the Admin Distance and the low order 3 bytes
369 * This will allow us to do two things:
370 * 1) Allow the creation of kernel routes that can be
371 * overridden by zebra.
372 * 2) Allow the old behavior for 'most' kernel route types
373 * if a user enters 'ip route ...' v4 routes get a metric
374 * of 0 and v6 routes get a metric of 1024. Both of these
375 * values will end up with a admin distance of 0, which
376 * will cause them to win for the purposes of zebra.
378 if (proto
== ZEBRA_ROUTE_KERNEL
) {
379 distance
= (metric
>> 24) & 0xFF;
380 metric
= (metric
& 0x00FFFFFF);
383 if (IS_ZEBRA_DEBUG_KERNEL
) {
384 char buf
[PREFIX_STRLEN
];
385 char buf2
[PREFIX_STRLEN
];
387 "%s %s%s%s vrf %u metric: %d Admin Distance: %d", nl_msg_type_to_str(h
->nlmsg_type
),
388 prefix2str(&p
, buf
, sizeof(buf
)),
389 src_p
.prefixlen
? " from " : "",
390 src_p
.prefixlen
? prefix2str(&src_p
, buf2
, sizeof(buf2
))
392 vrf_id
, metric
, distance
);
396 if (rtm
->rtm_family
== AF_INET6
)
399 if (h
->nlmsg_type
== RTM_NEWROUTE
) {
400 if (!tb
[RTA_MULTIPATH
]) {
402 size_t sz
= (afi
== AFI_IP
) ? 4 : 16;
404 memset(&nh
, 0, sizeof(nh
));
406 if (bh_type
== BLACKHOLE_UNSPEC
) {
408 nh
.type
= NEXTHOP_TYPE_IFINDEX
;
409 else if (index
&& gate
)
410 nh
.type
= (afi
== AFI_IP
)
411 ? NEXTHOP_TYPE_IPV4_IFINDEX
412 : NEXTHOP_TYPE_IPV6_IFINDEX
;
413 else if (!index
&& gate
)
414 nh
.type
= (afi
== AFI_IP
)
418 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
419 nh
.bh_type
= bh_type
;
422 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
423 nh
.bh_type
= bh_type
;
427 memcpy(&nh
.src
, prefsrc
, sz
);
429 memcpy(&nh
.gate
, gate
, sz
);
431 rib_add(afi
, SAFI_UNICAST
, vrf_id
, proto
,
432 0, flags
, &p
, NULL
, &nh
, table
, metric
, mtu
, distance
);
434 /* This is a multipath route */
436 struct route_entry
*re
;
437 struct rtnexthop
*rtnh
=
438 (struct rtnexthop
*)RTA_DATA(tb
[RTA_MULTIPATH
]);
440 len
= RTA_PAYLOAD(tb
[RTA_MULTIPATH
]);
442 re
= XCALLOC(MTYPE_RE
, sizeof(struct route_entry
));
444 re
->distance
= distance
;
451 re
->uptime
= time(NULL
);
454 if (len
< (int)sizeof(*rtnh
)
455 || rtnh
->rtnh_len
> len
)
458 index
= rtnh
->rtnh_ifindex
;
460 if (rtnh
->rtnh_len
> sizeof(*rtnh
)) {
461 memset(tb
, 0, sizeof(tb
));
462 netlink_parse_rtattr(
463 tb
, RTA_MAX
, RTNH_DATA(rtnh
),
464 rtnh
->rtnh_len
- sizeof(*rtnh
));
471 if (rtm
->rtm_family
== AF_INET
) {
473 route_entry_nexthop_ipv4_ifindex_add(
477 route_entry_nexthop_ipv4_add(
480 } else if (rtm
->rtm_family
483 route_entry_nexthop_ipv6_ifindex_add(
487 route_entry_nexthop_ipv6_add(
491 route_entry_nexthop_ifindex_add(re
,
494 len
-= NLMSG_ALIGN(rtnh
->rtnh_len
);
495 rtnh
= RTNH_NEXT(rtnh
);
498 zserv_nexthop_num_warn(__func__
,
499 (const struct prefix
*)&p
,
501 if (re
->nexthop_num
== 0)
504 rib_add_multipath(afi
, SAFI_UNICAST
, &p
,
508 if (!tb
[RTA_MULTIPATH
]) {
510 size_t sz
= (afi
== AFI_IP
) ? 4 : 16;
512 memset(&nh
, 0, sizeof(nh
));
513 if (bh_type
== BLACKHOLE_UNSPEC
) {
515 nh
.type
= NEXTHOP_TYPE_IFINDEX
;
516 else if (index
&& gate
)
519 ? NEXTHOP_TYPE_IPV4_IFINDEX
520 : NEXTHOP_TYPE_IPV6_IFINDEX
;
521 else if (!index
&& gate
)
522 nh
.type
= (afi
== AFI_IP
)
526 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
527 nh
.bh_type
= BLACKHOLE_UNSPEC
;
530 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
531 nh
.bh_type
= bh_type
;
535 memcpy(&nh
.gate
, gate
, sz
);
536 rib_delete(afi
, SAFI_UNICAST
, vrf_id
,
537 proto
, 0, flags
, &p
, NULL
, &nh
,
538 table
, metric
, true);
540 /* XXX: need to compare the entire list of nexthops
541 * here for NLM_F_APPEND stupidity */
542 rib_delete(afi
, SAFI_UNICAST
, vrf_id
,
543 proto
, 0, flags
, &p
, NULL
, NULL
,
544 table
, metric
, true);
551 static struct mcast_route_data
*mroute
= NULL
;
553 static int netlink_route_change_read_multicast(struct sockaddr_nl
*snl
,
555 ns_id_t ns_id
, int startup
)
559 struct rtattr
*tb
[RTA_MAX
+ 1];
560 struct mcast_route_data
*m
;
561 struct mcast_route_data mr
;
568 char oif_list
[256] = "\0";
569 vrf_id_t vrf
= ns_id
;
575 memset(&mr
, 0, sizeof(mr
));
581 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
583 memset(tb
, 0, sizeof tb
);
584 netlink_parse_rtattr(tb
, RTA_MAX
, RTM_RTA(rtm
), len
);
587 table
= *(int *)RTA_DATA(tb
[RTA_TABLE
]);
589 table
= rtm
->rtm_table
;
591 vrf
= vrf_lookup_by_table(table
);
594 iif
= *(int *)RTA_DATA(tb
[RTA_IIF
]);
597 m
->sg
.src
= *(struct in_addr
*)RTA_DATA(tb
[RTA_SRC
]);
600 m
->sg
.grp
= *(struct in_addr
*)RTA_DATA(tb
[RTA_DST
]);
602 if ((RTA_EXPIRES
<= RTA_MAX
) && tb
[RTA_EXPIRES
])
603 m
->lastused
= *(unsigned long long *)RTA_DATA(tb
[RTA_EXPIRES
]);
605 if (tb
[RTA_MULTIPATH
]) {
606 struct rtnexthop
*rtnh
=
607 (struct rtnexthop
*)RTA_DATA(tb
[RTA_MULTIPATH
]);
609 len
= RTA_PAYLOAD(tb
[RTA_MULTIPATH
]);
611 if (len
< (int)sizeof(*rtnh
) || rtnh
->rtnh_len
> len
)
614 oif
[oif_count
] = rtnh
->rtnh_ifindex
;
617 len
-= NLMSG_ALIGN(rtnh
->rtnh_len
);
618 rtnh
= RTNH_NEXT(rtnh
);
622 if (IS_ZEBRA_DEBUG_KERNEL
) {
623 struct interface
*ifp
;
624 strlcpy(sbuf
, inet_ntoa(m
->sg
.src
), sizeof(sbuf
));
625 strlcpy(gbuf
, inet_ntoa(m
->sg
.grp
), sizeof(gbuf
));
626 for (count
= 0; count
< oif_count
; count
++) {
627 ifp
= if_lookup_by_index(oif
[count
], vrf
);
630 sprintf(temp
, "%s ", ifp
->name
);
631 strcat(oif_list
, temp
);
633 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(vrf
);
634 ifp
= if_lookup_by_index(iif
, vrf
);
636 "MCAST VRF: %s(%d) %s (%s,%s) IIF: %s OIF: %s jiffies: %lld",
637 zvrf
->vrf
->name
, vrf
, nl_msg_type_to_str(h
->nlmsg_type
),
638 sbuf
, gbuf
, ifp
->name
, oif_list
, m
->lastused
);
643 int netlink_route_change(struct sockaddr_nl
*snl
, struct nlmsghdr
*h
,
644 ns_id_t ns_id
, int startup
)
647 vrf_id_t vrf_id
= ns_id
;
652 if (!(h
->nlmsg_type
== RTM_NEWROUTE
|| h
->nlmsg_type
== RTM_DELROUTE
)) {
653 /* If this is not route add/delete message print warning. */
654 zlog_warn("Kernel message: %d vrf %u\n", h
->nlmsg_type
, vrf_id
);
658 /* Connected route. */
659 if (IS_ZEBRA_DEBUG_KERNEL
)
660 zlog_debug("%s %s %s proto %s vrf %u",
661 nl_msg_type_to_str(h
->nlmsg_type
),
662 nl_family_to_str(rtm
->rtm_family
),
663 nl_rttype_to_str(rtm
->rtm_type
),
664 nl_rtproto_to_str(rtm
->rtm_protocol
), vrf_id
);
666 /* We don't care about change notifications for the MPLS table. */
667 /* TODO: Revisit this. */
668 if (rtm
->rtm_family
== AF_MPLS
)
671 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
675 if (rtm
->rtm_type
== RTN_MULTICAST
)
676 netlink_route_change_read_multicast(snl
, h
, ns_id
, startup
);
678 netlink_route_change_read_unicast(snl
, h
, ns_id
, startup
);
682 /* Request for specific route information from the kernel */
683 static int netlink_request_route(struct zebra_ns
*zns
, int family
, int type
)
690 /* Form the request, specifying filter (rtattr) if needed. */
691 memset(&req
, 0, sizeof(req
));
692 req
.n
.nlmsg_type
= type
;
693 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
694 req
.rtm
.rtm_family
= family
;
696 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
699 /* Routing table read function using netlink interface. Only called
701 int netlink_route_read(struct zebra_ns
*zns
)
705 /* Get IPv4 routing table. */
706 ret
= netlink_request_route(zns
, AF_INET
, RTM_GETROUTE
);
709 ret
= netlink_parse_info(netlink_route_change_read_unicast
,
710 &zns
->netlink_cmd
, zns
, 0, 1);
714 /* Get IPv6 routing table. */
715 ret
= netlink_request_route(zns
, AF_INET6
, RTM_GETROUTE
);
718 ret
= netlink_parse_info(netlink_route_change_read_unicast
,
719 &zns
->netlink_cmd
, zns
, 0, 1);
726 static void _netlink_route_nl_add_gateway_info(u_char route_family
,
728 struct nlmsghdr
*nlmsg
,
729 size_t req_size
, int bytelen
,
730 struct nexthop
*nexthop
)
732 if (route_family
== AF_MPLS
) {
733 struct gw_family_t gw_fam
;
735 gw_fam
.family
= gw_family
;
736 if (gw_family
== AF_INET
)
737 memcpy(&gw_fam
.gate
.ipv4
, &nexthop
->gate
.ipv4
, bytelen
);
739 memcpy(&gw_fam
.gate
.ipv6
, &nexthop
->gate
.ipv6
, bytelen
);
740 addattr_l(nlmsg
, req_size
, RTA_VIA
, &gw_fam
.family
,
743 if (gw_family
== AF_INET
)
744 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
,
745 &nexthop
->gate
.ipv4
, bytelen
);
747 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
,
748 &nexthop
->gate
.ipv6
, bytelen
);
752 static void _netlink_route_rta_add_gateway_info(u_char route_family
,
755 struct rtnexthop
*rtnh
,
756 size_t req_size
, int bytelen
,
757 struct nexthop
*nexthop
)
759 if (route_family
== AF_MPLS
) {
760 struct gw_family_t gw_fam
;
762 gw_fam
.family
= gw_family
;
763 if (gw_family
== AF_INET
)
764 memcpy(&gw_fam
.gate
.ipv4
, &nexthop
->gate
.ipv4
, bytelen
);
766 memcpy(&gw_fam
.gate
.ipv6
, &nexthop
->gate
.ipv6
, bytelen
);
767 rta_addattr_l(rta
, req_size
, RTA_VIA
, &gw_fam
.family
,
769 rtnh
->rtnh_len
+= RTA_LENGTH(bytelen
+ 2);
771 if (gw_family
== AF_INET
)
772 rta_addattr_l(rta
, req_size
, RTA_GATEWAY
,
773 &nexthop
->gate
.ipv4
, bytelen
);
775 rta_addattr_l(rta
, req_size
, RTA_GATEWAY
,
776 &nexthop
->gate
.ipv6
, bytelen
);
777 rtnh
->rtnh_len
+= sizeof(struct rtattr
) + bytelen
;
781 /* This function takes a nexthop as argument and adds
782 * the appropriate netlink attributes to an existing
785 * @param routedesc: Human readable description of route type
786 * (direct/recursive, single-/multipath)
787 * @param bytelen: Length of addresses in bytes.
788 * @param nexthop: Nexthop information
789 * @param nlmsg: nlmsghdr structure to fill in.
790 * @param req_size: The size allocated for the message.
792 static void _netlink_route_build_singlepath(const char *routedesc
, int bytelen
,
793 struct nexthop
*nexthop
,
794 struct nlmsghdr
*nlmsg
,
796 size_t req_size
, int cmd
)
798 struct nexthop_label
*nh_label
;
799 mpls_lse_t out_lse
[MPLS_MAX_LABELS
];
803 * label_buf is *only* currently used within debugging.
804 * As such when we assign it we are guarding it inside
805 * a debug test. If you want to change this make sure
806 * you fix this assumption
809 /* outgoing label - either as NEWDST (in the case of LSR) or as ENCAP
810 * (in the case of LER)
812 nh_label
= nexthop
->nh_label
;
813 if (rtmsg
->rtm_family
== AF_MPLS
) {
815 assert(nh_label
->num_labels
== 1);
818 if (nh_label
&& nh_label
->num_labels
) {
819 int i
, num_labels
= 0;
823 for (i
= 0; i
< nh_label
->num_labels
; i
++) {
824 if (nh_label
->label
[i
] != MPLS_IMP_NULL_LABEL
) {
825 bos
= ((i
== (nh_label
->num_labels
- 1)) ? 1
827 out_lse
[i
] = mpls_lse_encode(nh_label
->label
[i
],
829 if (IS_ZEBRA_DEBUG_KERNEL
) {
831 sprintf(label_buf
, "label %u",
834 sprintf(label_buf1
, "/%u",
836 strlcat(label_buf
, label_buf1
,
844 if (rtmsg
->rtm_family
== AF_MPLS
)
845 addattr_l(nlmsg
, req_size
, RTA_NEWDST
, &out_lse
,
846 num_labels
* sizeof(mpls_lse_t
));
849 u_int16_t encap
= LWTUNNEL_ENCAP_MPLS
;
851 addattr_l(nlmsg
, req_size
, RTA_ENCAP_TYPE
,
852 &encap
, sizeof(u_int16_t
));
853 nest
= addattr_nest(nlmsg
, req_size
, RTA_ENCAP
);
854 addattr_l(nlmsg
, req_size
, MPLS_IPTUNNEL_DST
,
856 num_labels
* sizeof(mpls_lse_t
));
857 addattr_nest_end(nlmsg
, nest
);
862 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
))
863 rtmsg
->rtm_flags
|= RTNH_F_ONLINK
;
865 if (rtmsg
->rtm_family
== AF_INET
866 && (nexthop
->type
== NEXTHOP_TYPE_IPV6
867 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
)) {
868 rtmsg
->rtm_flags
|= RTNH_F_ONLINK
;
869 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
, &ipv4_ll
, 4);
870 addattr32(nlmsg
, req_size
, RTA_OIF
, nexthop
->ifindex
);
872 if (nexthop
->rmap_src
.ipv4
.s_addr
&& (cmd
== RTM_NEWROUTE
))
873 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
874 &nexthop
->rmap_src
.ipv4
, bytelen
);
875 else if (nexthop
->src
.ipv4
.s_addr
&& (cmd
== RTM_NEWROUTE
))
876 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
877 &nexthop
->src
.ipv4
, bytelen
);
879 if (IS_ZEBRA_DEBUG_KERNEL
)
881 " 5549: _netlink_route_build_singlepath() (%s): "
882 "nexthop via %s %s if %u",
883 routedesc
, ipv4_ll_buf
, label_buf
,
888 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
889 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
890 /* Send deletes to the kernel without specifying the next-hop */
891 if (cmd
!= RTM_DELROUTE
)
892 _netlink_route_nl_add_gateway_info(
893 rtmsg
->rtm_family
, AF_INET
, nlmsg
, req_size
,
896 if (cmd
== RTM_NEWROUTE
) {
897 if (nexthop
->rmap_src
.ipv4
.s_addr
)
898 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
899 &nexthop
->rmap_src
.ipv4
, bytelen
);
900 else if (nexthop
->src
.ipv4
.s_addr
)
901 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
902 &nexthop
->src
.ipv4
, bytelen
);
905 if (IS_ZEBRA_DEBUG_KERNEL
)
907 "netlink_route_multipath() (%s): "
908 "nexthop via %s %s if %u",
909 routedesc
, inet_ntoa(nexthop
->gate
.ipv4
),
910 label_buf
, nexthop
->ifindex
);
913 if (nexthop
->type
== NEXTHOP_TYPE_IPV6
914 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
915 _netlink_route_nl_add_gateway_info(rtmsg
->rtm_family
, AF_INET6
,
916 nlmsg
, req_size
, bytelen
,
919 if (cmd
== RTM_NEWROUTE
) {
920 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
921 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
922 &nexthop
->rmap_src
.ipv6
, bytelen
);
923 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
924 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
925 &nexthop
->src
.ipv6
, bytelen
);
928 if (IS_ZEBRA_DEBUG_KERNEL
)
930 "netlink_route_multipath() (%s): "
931 "nexthop via %s %s if %u",
932 routedesc
, inet6_ntoa(nexthop
->gate
.ipv6
),
933 label_buf
, nexthop
->ifindex
);
935 if (nexthop
->type
== NEXTHOP_TYPE_IFINDEX
936 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
937 addattr32(nlmsg
, req_size
, RTA_OIF
, nexthop
->ifindex
);
939 if (cmd
== RTM_NEWROUTE
) {
940 if (nexthop
->rmap_src
.ipv4
.s_addr
)
941 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
942 &nexthop
->rmap_src
.ipv4
, bytelen
);
943 else if (nexthop
->src
.ipv4
.s_addr
)
944 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
945 &nexthop
->src
.ipv4
, bytelen
);
948 if (IS_ZEBRA_DEBUG_KERNEL
)
950 "netlink_route_multipath() (%s): "
952 routedesc
, nexthop
->ifindex
);
955 if (nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
956 addattr32(nlmsg
, req_size
, RTA_OIF
, nexthop
->ifindex
);
958 if (cmd
== RTM_NEWROUTE
) {
959 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
960 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
961 &nexthop
->rmap_src
.ipv6
, bytelen
);
962 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
963 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
964 &nexthop
->src
.ipv6
, bytelen
);
967 if (IS_ZEBRA_DEBUG_KERNEL
)
969 "netlink_route_multipath() (%s): "
971 routedesc
, nexthop
->ifindex
);
975 /* This function takes a nexthop as argument and
976 * appends to the given rtattr/rtnexthop pair the
977 * representation of the nexthop. If the nexthop
978 * defines a preferred source, the src parameter
979 * will be modified to point to that src, otherwise
980 * it will be kept unmodified.
982 * @param routedesc: Human readable description of route type
983 * (direct/recursive, single-/multipath)
984 * @param bytelen: Length of addresses in bytes.
985 * @param nexthop: Nexthop information
986 * @param rta: rtnetlink attribute structure
987 * @param rtnh: pointer to an rtnetlink nexthop structure
988 * @param src: pointer pointing to a location where
989 * the prefsrc should be stored.
991 static void _netlink_route_build_multipath(const char *routedesc
, int bytelen
,
992 struct nexthop
*nexthop
,
994 struct rtnexthop
*rtnh
,
998 struct nexthop_label
*nh_label
;
999 mpls_lse_t out_lse
[MPLS_MAX_LABELS
];
1000 char label_buf
[256];
1002 rtnh
->rtnh_len
= sizeof(*rtnh
);
1003 rtnh
->rtnh_flags
= 0;
1004 rtnh
->rtnh_hops
= 0;
1005 rta
->rta_len
+= rtnh
->rtnh_len
;
1008 * label_buf is *only* currently used within debugging.
1009 * As such when we assign it we are guarding it inside
1010 * a debug test. If you want to change this make sure
1011 * you fix this assumption
1013 label_buf
[0] = '\0';
1014 /* outgoing label - either as NEWDST (in the case of LSR) or as ENCAP
1015 * (in the case of LER)
1017 nh_label
= nexthop
->nh_label
;
1018 if (rtmsg
->rtm_family
== AF_MPLS
) {
1020 assert(nh_label
->num_labels
== 1);
1023 if (nh_label
&& nh_label
->num_labels
) {
1024 int i
, num_labels
= 0;
1026 char label_buf1
[20];
1028 for (i
= 0; i
< nh_label
->num_labels
; i
++) {
1029 if (nh_label
->label
[i
] != MPLS_IMP_NULL_LABEL
) {
1030 bos
= ((i
== (nh_label
->num_labels
- 1)) ? 1
1032 out_lse
[i
] = mpls_lse_encode(nh_label
->label
[i
],
1034 if (IS_ZEBRA_DEBUG_KERNEL
) {
1036 sprintf(label_buf
, "label %u",
1037 nh_label
->label
[i
]);
1039 sprintf(label_buf1
, "/%u",
1040 nh_label
->label
[i
]);
1041 strlcat(label_buf
, label_buf1
,
1049 if (rtmsg
->rtm_family
== AF_MPLS
) {
1050 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_NEWDST
,
1052 num_labels
* sizeof(mpls_lse_t
));
1053 rtnh
->rtnh_len
+= RTA_LENGTH(
1054 num_labels
* sizeof(mpls_lse_t
));
1056 struct rtattr
*nest
;
1057 u_int16_t encap
= LWTUNNEL_ENCAP_MPLS
;
1058 int len
= rta
->rta_len
;
1060 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
,
1061 RTA_ENCAP_TYPE
, &encap
,
1063 nest
= rta_nest(rta
, NL_PKT_BUF_SIZE
,
1065 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
,
1066 MPLS_IPTUNNEL_DST
, &out_lse
,
1067 num_labels
* sizeof(mpls_lse_t
));
1068 rta_nest_end(rta
, nest
);
1069 rtnh
->rtnh_len
+= rta
->rta_len
- len
;
1074 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
))
1075 rtnh
->rtnh_flags
|= RTNH_F_ONLINK
;
1077 if (rtmsg
->rtm_family
== AF_INET
1078 && (nexthop
->type
== NEXTHOP_TYPE_IPV6
1079 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
)) {
1081 rtnh
->rtnh_flags
|= RTNH_F_ONLINK
;
1082 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_GATEWAY
, &ipv4_ll
,
1084 rtnh
->rtnh_len
+= sizeof(struct rtattr
) + bytelen
;
1085 rtnh
->rtnh_ifindex
= nexthop
->ifindex
;
1087 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1088 *src
= &nexthop
->rmap_src
;
1089 else if (nexthop
->src
.ipv4
.s_addr
)
1090 *src
= &nexthop
->src
;
1092 if (IS_ZEBRA_DEBUG_KERNEL
)
1094 " 5549: netlink_route_build_multipath() (%s): "
1095 "nexthop via %s %s if %u",
1096 routedesc
, ipv4_ll_buf
, label_buf
,
1101 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
1102 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
1103 _netlink_route_rta_add_gateway_info(rtmsg
->rtm_family
, AF_INET
,
1104 rta
, rtnh
, NL_PKT_BUF_SIZE
,
1106 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1107 *src
= &nexthop
->rmap_src
;
1108 else if (nexthop
->src
.ipv4
.s_addr
)
1109 *src
= &nexthop
->src
;
1111 if (IS_ZEBRA_DEBUG_KERNEL
)
1113 "netlink_route_multipath() (%s): "
1114 "nexthop via %s %s if %u",
1115 routedesc
, inet_ntoa(nexthop
->gate
.ipv4
),
1116 label_buf
, nexthop
->ifindex
);
1118 if (nexthop
->type
== NEXTHOP_TYPE_IPV6
1119 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1120 _netlink_route_rta_add_gateway_info(rtmsg
->rtm_family
, AF_INET6
,
1121 rta
, rtnh
, NL_PKT_BUF_SIZE
,
1124 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1125 *src
= &nexthop
->rmap_src
;
1126 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1127 *src
= &nexthop
->src
;
1129 if (IS_ZEBRA_DEBUG_KERNEL
)
1131 "netlink_route_multipath() (%s): "
1132 "nexthop via %s %s if %u",
1133 routedesc
, inet6_ntoa(nexthop
->gate
.ipv6
),
1134 label_buf
, nexthop
->ifindex
);
1137 if (nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
1138 || nexthop
->type
== NEXTHOP_TYPE_IFINDEX
) {
1139 rtnh
->rtnh_ifindex
= nexthop
->ifindex
;
1141 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1142 *src
= &nexthop
->rmap_src
;
1143 else if (nexthop
->src
.ipv4
.s_addr
)
1144 *src
= &nexthop
->src
;
1146 if (IS_ZEBRA_DEBUG_KERNEL
)
1148 "netlink_route_multipath() (%s): "
1149 "nexthop via if %u",
1150 routedesc
, nexthop
->ifindex
);
1151 } else if (nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1152 rtnh
->rtnh_ifindex
= nexthop
->ifindex
;
1154 if (IS_ZEBRA_DEBUG_KERNEL
)
1156 "netlink_route_multipath() (%s): "
1157 "nexthop via if %u",
1158 routedesc
, nexthop
->ifindex
);
1160 rtnh
->rtnh_ifindex
= 0;
1164 static inline void _netlink_mpls_build_singlepath(const char *routedesc
,
1165 zebra_nhlfe_t
*nhlfe
,
1166 struct nlmsghdr
*nlmsg
,
1167 struct rtmsg
*rtmsg
,
1168 size_t req_size
, int cmd
)
1173 family
= NHLFE_FAMILY(nhlfe
);
1174 bytelen
= (family
== AF_INET
? 4 : 16);
1175 _netlink_route_build_singlepath(routedesc
, bytelen
, nhlfe
->nexthop
,
1176 nlmsg
, rtmsg
, req_size
, cmd
);
1181 _netlink_mpls_build_multipath(const char *routedesc
, zebra_nhlfe_t
*nhlfe
,
1182 struct rtattr
*rta
, struct rtnexthop
*rtnh
,
1183 struct rtmsg
*rtmsg
, union g_addr
**src
)
1188 family
= NHLFE_FAMILY(nhlfe
);
1189 bytelen
= (family
== AF_INET
? 4 : 16);
1190 _netlink_route_build_multipath(routedesc
, bytelen
, nhlfe
->nexthop
, rta
,
1195 /* Log debug information for netlink_route_multipath
1196 * if debug logging is enabled.
1198 * @param cmd: Netlink command which is to be processed
1199 * @param p: Prefix for which the change is due
1200 * @param nexthop: Nexthop which is currently processed
1201 * @param routedesc: Semantic annotation for nexthop
1202 * (recursive, multipath, etc.)
1203 * @param family: Address family which the change concerns
1205 static void _netlink_route_debug(int cmd
, struct prefix
*p
,
1206 struct nexthop
*nexthop
, const char *routedesc
,
1207 int family
, struct zebra_vrf
*zvrf
)
1209 if (IS_ZEBRA_DEBUG_KERNEL
) {
1210 char buf
[PREFIX_STRLEN
];
1212 "netlink_route_multipath() (%s): %s %s vrf %u type %s",
1213 routedesc
, nl_msg_type_to_str(cmd
),
1214 prefix2str(p
, buf
, sizeof(buf
)), zvrf_id(zvrf
),
1215 (nexthop
) ? nexthop_type_to_str(nexthop
->type
) : "UNK");
1219 static void _netlink_mpls_debug(int cmd
, u_int32_t label
, const char *routedesc
)
1221 if (IS_ZEBRA_DEBUG_KERNEL
)
1222 zlog_debug("netlink_mpls_multipath() (%s): %s %u/20", routedesc
,
1223 nl_msg_type_to_str(cmd
), label
);
1226 static int netlink_neigh_update(int cmd
, int ifindex
, uint32_t addr
, char *lla
,
1235 struct zebra_ns
*zns
= zebra_ns_lookup(NS_DEFAULT
);
1237 memset(&req
.n
, 0, sizeof(req
.n
));
1238 memset(&req
.ndm
, 0, sizeof(req
.ndm
));
1240 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1241 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
1242 req
.n
.nlmsg_type
= cmd
; // RTM_NEWNEIGH or RTM_DELNEIGH
1243 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1245 req
.ndm
.ndm_family
= AF_INET
;
1246 req
.ndm
.ndm_state
= NUD_PERMANENT
;
1247 req
.ndm
.ndm_ifindex
= ifindex
;
1248 req
.ndm
.ndm_type
= RTN_UNICAST
;
1250 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &addr
, 4);
1251 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, lla
, llalen
);
1253 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1257 /* Routing table change via netlink interface. */
1258 /* Update flag indicates whether this is a "replace" or not. */
1259 static int netlink_route_multipath(int cmd
, struct prefix
*p
,
1260 struct prefix
*src_p
, struct route_entry
*re
,
1264 struct sockaddr_nl snl
;
1265 struct nexthop
*nexthop
= NULL
;
1266 unsigned int nexthop_num
;
1268 int family
= PREFIX_FAMILY(p
);
1269 const char *routedesc
;
1276 char buf
[NL_PKT_BUF_SIZE
];
1279 struct zebra_ns
*zns
= zebra_ns_lookup(NS_DEFAULT
);
1280 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1282 memset(&req
, 0, sizeof req
- NL_PKT_BUF_SIZE
);
1284 bytelen
= (family
== AF_INET
? 4 : 16);
1286 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
1287 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
1288 if ((cmd
== RTM_NEWROUTE
) && update
)
1289 req
.n
.nlmsg_flags
|= NLM_F_REPLACE
;
1290 req
.n
.nlmsg_type
= cmd
;
1291 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1293 req
.r
.rtm_family
= family
;
1294 req
.r
.rtm_dst_len
= p
->prefixlen
;
1295 req
.r
.rtm_src_len
= src_p
? src_p
->prefixlen
: 0;
1296 req
.r
.rtm_protocol
= zebra2proto(re
->type
);
1297 req
.r
.rtm_scope
= RT_SCOPE_UNIVERSE
;
1298 req
.r
.rtm_type
= RTN_UNICAST
;
1300 addattr_l(&req
.n
, sizeof req
, RTA_DST
, &p
->u
.prefix
, bytelen
);
1302 addattr_l(&req
.n
, sizeof req
, RTA_SRC
, &src_p
->u
.prefix
,
1306 /* Hardcode the metric for all routes coming from zebra. Metric isn't
1308 * either by the kernel or by zebra. Its purely for calculating best
1310 * by the routing protocol and for communicating with protocol peers.
1312 addattr32(&req
.n
, sizeof req
, RTA_PRIORITY
, NL_DEFAULT_ROUTE_METRIC
);
1314 /* Table corresponding to this route. */
1315 if (re
->table
< 256)
1316 req
.r
.rtm_table
= re
->table
;
1318 req
.r
.rtm_table
= RT_TABLE_UNSPEC
;
1319 addattr32(&req
.n
, sizeof req
, RTA_TABLE
, re
->table
);
1325 if (re
->mtu
|| re
->nexthop_mtu
) {
1326 char buf
[NL_PKT_BUF_SIZE
];
1327 struct rtattr
*rta
= (void *)buf
;
1328 u_int32_t mtu
= re
->mtu
;
1329 if (!mtu
|| (re
->nexthop_mtu
&& re
->nexthop_mtu
< mtu
))
1330 mtu
= re
->nexthop_mtu
;
1331 rta
->rta_type
= RTA_METRICS
;
1332 rta
->rta_len
= RTA_LENGTH(0);
1333 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTAX_MTU
, &mtu
, sizeof mtu
);
1334 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_METRICS
, RTA_DATA(rta
),
1338 /* Count overall nexthops so we can decide whether to use singlepath
1339 * or multipath case. */
1341 for (ALL_NEXTHOPS(re
->nexthop
, nexthop
)) {
1342 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
))
1344 if (cmd
== RTM_NEWROUTE
1345 && !NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1347 if (cmd
== RTM_DELROUTE
1348 && !CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
1354 /* Singlepath case. */
1355 if (nexthop_num
== 1 || multipath_num
== 1) {
1357 for (ALL_NEXTHOPS(re
->nexthop
, nexthop
)) {
1359 * So we want to cover 2 types of blackhole
1361 * 1) A normal blackhole route( ala from a static
1363 * 2) A recursively resolved blackhole route
1365 if (nexthop
->type
== NEXTHOP_TYPE_BLACKHOLE
) {
1366 switch (nexthop
->bh_type
) {
1367 case BLACKHOLE_ADMINPROHIB
:
1368 req
.r
.rtm_type
= RTN_PROHIBIT
;
1370 case BLACKHOLE_REJECT
:
1371 req
.r
.rtm_type
= RTN_UNREACHABLE
;
1374 req
.r
.rtm_type
= RTN_BLACKHOLE
;
1379 if (CHECK_FLAG(nexthop
->flags
,
1380 NEXTHOP_FLAG_RECURSIVE
)) {
1382 if (family
== AF_INET
) {
1383 if (nexthop
->rmap_src
.ipv4
1390 } else if (nexthop
->src
.ipv4
1398 } else if (family
== AF_INET6
) {
1399 if (!IN6_IS_ADDR_UNSPECIFIED(
1407 !IN6_IS_ADDR_UNSPECIFIED(
1420 if ((cmd
== RTM_NEWROUTE
1421 && NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1422 || (cmd
== RTM_DELROUTE
1423 && CHECK_FLAG(nexthop
->flags
,
1424 NEXTHOP_FLAG_FIB
))) {
1425 routedesc
= nexthop
->rparent
1426 ? "recursive, single-path"
1429 _netlink_route_debug(cmd
, p
, nexthop
, routedesc
,
1431 _netlink_route_build_singlepath(
1432 routedesc
, bytelen
, nexthop
, &req
.n
,
1433 &req
.r
, sizeof req
, cmd
);
1438 if (setsrc
&& (cmd
== RTM_NEWROUTE
)) {
1439 if (family
== AF_INET
)
1440 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1441 &src
.ipv4
, bytelen
);
1442 else if (family
== AF_INET6
)
1443 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1444 &src
.ipv6
, bytelen
);
1447 char buf
[NL_PKT_BUF_SIZE
];
1448 struct rtattr
*rta
= (void *)buf
;
1449 struct rtnexthop
*rtnh
;
1450 union g_addr
*src1
= NULL
;
1452 rta
->rta_type
= RTA_MULTIPATH
;
1453 rta
->rta_len
= RTA_LENGTH(0);
1454 rtnh
= RTA_DATA(rta
);
1457 for (ALL_NEXTHOPS(re
->nexthop
, nexthop
)) {
1458 if (nexthop_num
>= multipath_num
)
1461 if (CHECK_FLAG(nexthop
->flags
,
1462 NEXTHOP_FLAG_RECURSIVE
)) {
1463 /* This only works for IPv4 now */
1465 if (family
== AF_INET
) {
1466 if (nexthop
->rmap_src
.ipv4
1473 } else if (nexthop
->src
.ipv4
1481 } else if (family
== AF_INET6
) {
1482 if (!IN6_IS_ADDR_UNSPECIFIED(
1490 !IN6_IS_ADDR_UNSPECIFIED(
1503 if ((cmd
== RTM_NEWROUTE
1504 && NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1505 || (cmd
== RTM_DELROUTE
1506 && CHECK_FLAG(nexthop
->flags
,
1507 NEXTHOP_FLAG_FIB
))) {
1508 routedesc
= nexthop
->rparent
1509 ? "recursive, multipath"
1513 _netlink_route_debug(cmd
, p
, nexthop
, routedesc
,
1515 _netlink_route_build_multipath(
1516 routedesc
, bytelen
, nexthop
, rta
, rtnh
,
1518 rtnh
= RTNH_NEXT(rtnh
);
1520 if (!setsrc
&& src1
) {
1521 if (family
== AF_INET
)
1522 src
.ipv4
= src1
->ipv4
;
1523 else if (family
== AF_INET6
)
1524 src
.ipv6
= src1
->ipv6
;
1530 if (setsrc
&& (cmd
== RTM_NEWROUTE
)) {
1531 if (family
== AF_INET
)
1532 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1533 &src
.ipv4
, bytelen
);
1534 else if (family
== AF_INET6
)
1535 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1536 &src
.ipv6
, bytelen
);
1537 if (IS_ZEBRA_DEBUG_KERNEL
)
1538 zlog_debug("Setting source");
1541 if (rta
->rta_len
> RTA_LENGTH(0))
1542 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_MULTIPATH
,
1543 RTA_DATA(rta
), RTA_PAYLOAD(rta
));
1546 /* If there is no useful nexthop then return. */
1547 if (nexthop_num
== 0) {
1548 if (IS_ZEBRA_DEBUG_KERNEL
)
1550 "netlink_route_multipath(): No useful nexthop.");
1556 /* Destination netlink address. */
1557 memset(&snl
, 0, sizeof snl
);
1558 snl
.nl_family
= AF_NETLINK
;
1560 /* Talk to netlink socket. */
1561 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1565 int kernel_get_ipmr_sg_stats(struct zebra_vrf
*zvrf
, void *in
)
1568 struct mcast_route_data
*mr
= (struct mcast_route_data
*)in
;
1576 struct zebra_ns
*zns
= zebra_ns_lookup(NS_DEFAULT
);
1578 memset(&req
.n
, 0, sizeof(req
.n
));
1579 memset(&req
.ndm
, 0, sizeof(req
.ndm
));
1581 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1582 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
1583 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1585 req
.ndm
.ndm_family
= RTNL_FAMILY_IPMR
;
1586 req
.n
.nlmsg_type
= RTM_GETROUTE
;
1588 addattr_l(&req
.n
, sizeof(req
), RTA_IIF
, &mroute
->ifindex
, 4);
1589 addattr_l(&req
.n
, sizeof(req
), RTA_OIF
, &mroute
->ifindex
, 4);
1590 addattr_l(&req
.n
, sizeof(req
), RTA_SRC
, &mroute
->sg
.src
.s_addr
, 4);
1591 addattr_l(&req
.n
, sizeof(req
), RTA_DST
, &mroute
->sg
.grp
.s_addr
, 4);
1592 addattr_l(&req
.n
, sizeof(req
), RTA_TABLE
, &zvrf
->table_id
, 4);
1594 suc
= netlink_talk(netlink_route_change_read_multicast
, &req
.n
,
1595 &zns
->netlink_cmd
, zns
, 0);
1601 void kernel_route_rib(struct prefix
*p
, struct prefix
*src_p
,
1602 struct route_entry
*old
, struct route_entry
*new)
1609 if (p
->family
== AF_INET
)
1610 ret
= netlink_route_multipath(RTM_NEWROUTE
, p
, src_p
,
1611 new, (old
) ? 1 : 0);
1614 * So v6 route replace semantics are not in
1615 * the kernel at this point as I understand it.
1616 * So let's do a delete than an add.
1617 * In the future once v6 route replace semantics
1618 * are in we can figure out what to do here to
1619 * allow working with old and new kernels.
1621 * I'm also intentionally ignoring the failure case
1622 * of the route delete. If that happens yeah we're
1626 netlink_route_multipath(RTM_DELROUTE
, p
,
1628 ret
= netlink_route_multipath(RTM_NEWROUTE
, p
,
1631 kernel_route_rib_pass_fail(p
, new,
1633 SOUTHBOUND_INSTALL_SUCCESS
:
1634 SOUTHBOUND_INSTALL_FAILURE
);
1639 ret
= netlink_route_multipath(RTM_DELROUTE
, p
, src_p
, old
, 0);
1641 kernel_route_rib_pass_fail(p
, old
,
1643 SOUTHBOUND_DELETE_SUCCESS
:
1644 SOUTHBOUND_DELETE_FAILURE
);
1648 int kernel_neigh_update(int add
, int ifindex
, uint32_t addr
, char *lla
,
1651 return netlink_neigh_update(add
? RTM_NEWNEIGH
: RTM_DELNEIGH
, ifindex
,
1656 * Add remote VTEP to the flood list for this VxLAN interface (VNI). This
1657 * is done by adding an FDB entry with a MAC of 00:00:00:00:00:00.
1659 static int netlink_vxlan_flood_list_update(struct interface
*ifp
,
1660 struct in_addr
*vtep_ip
, int cmd
)
1662 struct zebra_ns
*zns
= zebra_ns_lookup(NS_DEFAULT
);
1668 u_char dst_mac
[6] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0};
1670 memset(&req
.n
, 0, sizeof(req
.n
));
1671 memset(&req
.ndm
, 0, sizeof(req
.ndm
));
1673 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1674 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
1675 if (cmd
== RTM_NEWNEIGH
)
1676 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_APPEND
);
1677 req
.n
.nlmsg_type
= cmd
;
1678 req
.ndm
.ndm_family
= PF_BRIDGE
;
1679 req
.ndm
.ndm_state
= NUD_NOARP
| NUD_PERMANENT
;
1680 req
.ndm
.ndm_flags
|= NTF_SELF
; // Handle by "self", not "master"
1683 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, &dst_mac
, 6);
1684 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
1685 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &vtep_ip
->s_addr
, 4);
1687 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1692 * Add remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1694 * a "flood" MAC FDB entry.
1696 int kernel_add_vtep(vni_t vni
, struct interface
*ifp
, struct in_addr
*vtep_ip
)
1698 if (IS_ZEBRA_DEBUG_VXLAN
)
1699 zlog_debug("Install %s into flood list for VNI %u intf %s(%u)",
1700 inet_ntoa(*vtep_ip
), vni
, ifp
->name
, ifp
->ifindex
);
1702 return netlink_vxlan_flood_list_update(ifp
, vtep_ip
, RTM_NEWNEIGH
);
1706 * Remove remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1707 * deleting the "flood" MAC FDB entry.
1709 int kernel_del_vtep(vni_t vni
, struct interface
*ifp
, struct in_addr
*vtep_ip
)
1711 if (IS_ZEBRA_DEBUG_VXLAN
)
1713 "Uninstall %s from flood list for VNI %u intf %s(%u)",
1714 inet_ntoa(*vtep_ip
), vni
, ifp
->name
, ifp
->ifindex
);
1716 return netlink_vxlan_flood_list_update(ifp
, vtep_ip
, RTM_DELNEIGH
);
1720 #define NDA_RTA(r) \
1721 ((struct rtattr *)(((char *)(r)) + NLMSG_ALIGN(sizeof(struct ndmsg))))
1724 static int netlink_macfdb_change(struct sockaddr_nl
*snl
, struct nlmsghdr
*h
,
1728 struct interface
*ifp
;
1729 struct zebra_if
*zif
;
1730 struct rtattr
*tb
[NDA_MAX
+ 1];
1731 struct interface
*br_if
;
1734 struct prefix vtep_ip
;
1735 int vid_present
= 0, dst_present
= 0;
1736 char buf
[ETHER_ADDR_STRLEN
];
1741 ndm
= NLMSG_DATA(h
);
1743 /* We only process macfdb notifications if EVPN is enabled */
1744 if (!is_evpn_enabled())
1747 /* The interface should exist. */
1748 ifp
= if_lookup_by_index_per_ns(zebra_ns_lookup(NS_DEFAULT
),
1750 if (!ifp
|| !ifp
->info
)
1753 /* The interface should be something we're interested in. */
1754 if (!IS_ZEBRA_IF_BRIDGE_SLAVE(ifp
))
1757 /* Drop "permanent" entries. */
1758 if (ndm
->ndm_state
& NUD_PERMANENT
)
1761 zif
= (struct zebra_if
*)ifp
->info
;
1762 if ((br_if
= zif
->brslave_info
.br_if
) == NULL
) {
1763 zlog_warn("%s family %s IF %s(%u) brIF %u - no bridge master",
1764 nl_msg_type_to_str(h
->nlmsg_type
),
1765 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1766 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
);
1770 /* Parse attributes and extract fields of interest. */
1771 memset(tb
, 0, sizeof tb
);
1772 netlink_parse_rtattr(tb
, NDA_MAX
, NDA_RTA(ndm
), len
);
1774 if (!tb
[NDA_LLADDR
]) {
1775 zlog_warn("%s family %s IF %s(%u) brIF %u - no LLADDR",
1776 nl_msg_type_to_str(h
->nlmsg_type
),
1777 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1778 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
);
1782 if (RTA_PAYLOAD(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
1784 "%s family %s IF %s(%u) brIF %u - LLADDR is not MAC, len %lu",
1785 nl_msg_type_to_str(h
->nlmsg_type
),
1786 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1787 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
,
1788 (unsigned long)RTA_PAYLOAD(tb
[NDA_LLADDR
]));
1792 memcpy(&mac
, RTA_DATA(tb
[NDA_LLADDR
]), ETH_ALEN
);
1794 if ((NDA_VLAN
<= NDA_MAX
) && tb
[NDA_VLAN
]) {
1796 vid
= *(u_int16_t
*)RTA_DATA(tb
[NDA_VLAN
]);
1797 sprintf(vid_buf
, " VLAN %u", vid
);
1801 /* TODO: Only IPv4 supported now. */
1803 vtep_ip
.family
= AF_INET
;
1804 vtep_ip
.prefixlen
= IPV4_MAX_BITLEN
;
1805 memcpy(&(vtep_ip
.u
.prefix4
.s_addr
), RTA_DATA(tb
[NDA_DST
]),
1807 sprintf(dst_buf
, " dst %s", inet_ntoa(vtep_ip
.u
.prefix4
));
1810 sticky
= (ndm
->ndm_state
& NUD_NOARP
) ? 1 : 0;
1812 if (IS_ZEBRA_DEBUG_KERNEL
)
1813 zlog_debug("Rx %s family %s IF %s(%u)%s %sMAC %s%s",
1814 nl_msg_type_to_str(h
->nlmsg_type
),
1815 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1816 ndm
->ndm_ifindex
, vid_present
? vid_buf
: "",
1817 sticky
? "sticky " : "",
1818 prefix_mac2str(&mac
, buf
, sizeof(buf
)),
1819 dst_present
? dst_buf
: "");
1821 if (filter_vlan
&& vid
!= filter_vlan
)
1824 /* If add or update, do accordingly if learnt on a "local" interface; if
1825 * the notification is over VxLAN, this has to be related to
1827 * so perform an implicit delete of any local entry (if it exists).
1829 if (h
->nlmsg_type
== RTM_NEWNEIGH
) {
1830 /* Drop "permanent" entries. */
1831 if (ndm
->ndm_state
& NUD_PERMANENT
)
1834 if (IS_ZEBRA_IF_VXLAN(ifp
))
1835 return zebra_vxlan_check_del_local_mac(ifp
, br_if
, &mac
,
1838 return zebra_vxlan_local_mac_add_update(ifp
, br_if
, &mac
, vid
,
1842 /* This is a delete notification.
1843 * 1. For a MAC over VxLan, check if it needs to be refreshed(readded)
1844 * 2. For a MAC over "local" interface, delete the mac
1845 * Note: We will get notifications from both bridge driver and VxLAN
1847 * Ignore the notification from VxLan driver as it is also generated
1848 * when mac moves from remote to local.
1853 if (IS_ZEBRA_IF_VXLAN(ifp
))
1854 return zebra_vxlan_check_readd_remote_mac(ifp
, br_if
, &mac
,
1857 return zebra_vxlan_local_mac_del(ifp
, br_if
, &mac
, vid
);
1860 static int netlink_macfdb_table(struct sockaddr_nl
*snl
, struct nlmsghdr
*h
,
1861 ns_id_t ns_id
, int startup
)
1866 if (h
->nlmsg_type
!= RTM_NEWNEIGH
)
1869 /* Length validity. */
1870 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
1874 /* We are interested only in AF_BRIDGE notifications. */
1875 ndm
= NLMSG_DATA(h
);
1876 if (ndm
->ndm_family
!= AF_BRIDGE
)
1879 return netlink_macfdb_change(snl
, h
, len
);
1882 /* Request for MAC FDB information from the kernel */
1883 static int netlink_request_macs(struct zebra_ns
*zns
, int family
, int type
,
1884 ifindex_t master_ifindex
)
1888 struct ifinfomsg ifm
;
1892 /* Form the request, specifying filter (rtattr) if needed. */
1893 memset(&req
, 0, sizeof(req
));
1894 req
.n
.nlmsg_type
= type
;
1895 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ifinfomsg
));
1896 req
.ifm
.ifi_family
= family
;
1898 addattr32(&req
.n
, sizeof(req
), IFLA_MASTER
, master_ifindex
);
1900 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
1904 * MAC forwarding database read using netlink interface. This is invoked
1907 int netlink_macfdb_read(struct zebra_ns
*zns
)
1911 /* Get bridge FDB table. */
1912 ret
= netlink_request_macs(zns
, AF_BRIDGE
, RTM_GETNEIGH
, 0);
1915 /* We are reading entire table. */
1917 ret
= netlink_parse_info(netlink_macfdb_table
, &zns
->netlink_cmd
, zns
,
1924 * MAC forwarding database read using netlink interface. This is for a
1925 * specific bridge and matching specific access VLAN (if VLAN-aware bridge).
1927 int netlink_macfdb_read_for_bridge(struct zebra_ns
*zns
, struct interface
*ifp
,
1928 struct interface
*br_if
)
1930 struct zebra_if
*br_zif
;
1931 struct zebra_if
*zif
;
1932 struct zebra_l2info_vxlan
*vxl
;
1936 /* Save VLAN we're filtering on, if needed. */
1937 br_zif
= (struct zebra_if
*)br_if
->info
;
1938 zif
= (struct zebra_if
*)ifp
->info
;
1939 vxl
= &zif
->l2info
.vxl
;
1940 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif
))
1941 filter_vlan
= vxl
->access_vlan
;
1943 /* Get bridge FDB table for specific bridge - we do the VLAN filtering.
1945 ret
= netlink_request_macs(zns
, AF_BRIDGE
, RTM_GETNEIGH
,
1949 ret
= netlink_parse_info(netlink_macfdb_table
, &zns
->netlink_cmd
, zns
,
1952 /* Reset VLAN filter. */
1957 static int netlink_macfdb_update(struct interface
*ifp
, vlanid_t vid
,
1958 struct ethaddr
*mac
, struct in_addr vtep_ip
,
1959 int local
, int cmd
, u_char sticky
)
1961 struct zebra_ns
*zns
= zebra_ns_lookup(NS_DEFAULT
);
1968 struct zebra_if
*zif
;
1969 struct interface
*br_if
;
1970 struct zebra_if
*br_zif
;
1971 char buf
[ETHER_ADDR_STRLEN
];
1972 int vid_present
= 0, dst_present
= 0;
1977 if ((br_if
= zif
->brslave_info
.br_if
) == NULL
) {
1978 zlog_warn("MAC %s on IF %s(%u) - no mapping to bridge",
1979 (cmd
== RTM_NEWNEIGH
) ? "add" : "del", ifp
->name
,
1984 memset(&req
.n
, 0, sizeof(req
.n
));
1985 memset(&req
.ndm
, 0, sizeof(req
.ndm
));
1987 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1988 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
1989 if (cmd
== RTM_NEWNEIGH
)
1990 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_REPLACE
);
1991 req
.n
.nlmsg_type
= cmd
;
1992 req
.ndm
.ndm_family
= AF_BRIDGE
;
1993 req
.ndm
.ndm_flags
|= NTF_SELF
| NTF_MASTER
;
1994 req
.ndm
.ndm_state
= NUD_REACHABLE
;
1997 req
.ndm
.ndm_state
|= NUD_NOARP
;
1999 req
.ndm
.ndm_flags
|= NTF_EXT_LEARNED
;
2001 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, mac
, 6);
2002 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
2004 dst_alen
= 4; // TODO: hardcoded
2005 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &vtep_ip
, dst_alen
);
2007 sprintf(dst_buf
, " dst %s", inet_ntoa(vtep_ip
));
2009 br_zif
= (struct zebra_if
*)br_if
->info
;
2010 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif
) && vid
> 0) {
2011 addattr16(&req
.n
, sizeof(req
), NDA_VLAN
, vid
);
2013 sprintf(vid_buf
, " VLAN %u", vid
);
2015 addattr32(&req
.n
, sizeof(req
), NDA_MASTER
, br_if
->ifindex
);
2017 if (IS_ZEBRA_DEBUG_KERNEL
)
2018 zlog_debug("Tx %s family %s IF %s(%u)%s %sMAC %s%s",
2019 nl_msg_type_to_str(cmd
),
2020 nl_family_to_str(req
.ndm
.ndm_family
), ifp
->name
,
2021 ifp
->ifindex
, vid_present
? vid_buf
: "",
2022 sticky
? "sticky " : "",
2023 prefix_mac2str(mac
, buf
, sizeof(buf
)),
2024 dst_present
? dst_buf
: "");
2026 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2031 (NUD_PERMANENT | NUD_NOARP | NUD_REACHABLE | NUD_PROBE | NUD_STALE \
2034 static int netlink_ipneigh_change(struct sockaddr_nl
*snl
, struct nlmsghdr
*h
,
2038 struct interface
*ifp
;
2039 struct zebra_if
*zif
;
2040 struct rtattr
*tb
[NDA_MAX
+ 1];
2041 struct interface
*link_if
;
2044 char buf
[ETHER_ADDR_STRLEN
];
2045 char buf2
[INET6_ADDRSTRLEN
];
2046 int mac_present
= 0;
2049 ndm
= NLMSG_DATA(h
);
2051 /* We only process neigh notifications if EVPN is enabled */
2052 if (!is_evpn_enabled())
2055 /* The interface should exist. */
2056 ifp
= if_lookup_by_index_per_ns(zebra_ns_lookup(NS_DEFAULT
),
2058 if (!ifp
|| !ifp
->info
)
2061 /* Drop "permanent" entries. */
2062 if (ndm
->ndm_state
& NUD_PERMANENT
)
2065 zif
= (struct zebra_if
*)ifp
->info
;
2066 /* The neighbor is present on an SVI. From this, we locate the
2068 * bridge because we're only interested in neighbors on a VxLAN bridge.
2069 * The bridge is located based on the nature of the SVI:
2070 * (a) In the case of a VLAN-aware bridge, the SVI is a L3 VLAN
2072 * and is linked to the bridge
2073 * (b) In the case of a VLAN-unaware bridge, the SVI is the bridge
2077 if (IS_ZEBRA_IF_VLAN(ifp
)) {
2078 link_if
= if_lookup_by_index_per_ns(zebra_ns_lookup(NS_DEFAULT
),
2082 } else if (IS_ZEBRA_IF_BRIDGE(ifp
))
2087 /* Parse attributes and extract fields of interest. */
2088 memset(tb
, 0, sizeof tb
);
2089 netlink_parse_rtattr(tb
, NDA_MAX
, NDA_RTA(ndm
), len
);
2092 zlog_warn("%s family %s IF %s(%u) - no DST",
2093 nl_msg_type_to_str(h
->nlmsg_type
),
2094 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2098 memset(&mac
, 0, sizeof(struct ethaddr
));
2099 memset(&ip
, 0, sizeof(struct ipaddr
));
2100 ip
.ipa_type
= (ndm
->ndm_family
== AF_INET
) ? IPADDR_V4
: IPADDR_V6
;
2101 memcpy(&ip
.ip
.addr
, RTA_DATA(tb
[NDA_DST
]), RTA_PAYLOAD(tb
[NDA_DST
]));
2103 if (h
->nlmsg_type
== RTM_NEWNEIGH
) {
2104 if (tb
[NDA_LLADDR
]) {
2105 if (RTA_PAYLOAD(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
2107 "%s family %s IF %s(%u) - LLADDR is not MAC, len %lu",
2108 nl_msg_type_to_str(h
->nlmsg_type
),
2109 nl_family_to_str(ndm
->ndm_family
),
2110 ifp
->name
, ndm
->ndm_ifindex
,
2111 (unsigned long)RTA_PAYLOAD(tb
[NDA_LLADDR
]));
2116 memcpy(&mac
, RTA_DATA(tb
[NDA_LLADDR
]), ETH_ALEN
);
2119 ext_learned
= (ndm
->ndm_flags
& NTF_EXT_LEARNED
) ? 1 : 0;
2121 if (IS_ZEBRA_DEBUG_KERNEL
)
2123 "Rx %s family %s IF %s(%u) IP %s MAC %s state 0x%x flags 0x%x",
2124 nl_msg_type_to_str(h
->nlmsg_type
),
2125 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2127 ipaddr2str(&ip
, buf2
, sizeof(buf2
)),
2129 ? prefix_mac2str(&mac
, buf
, sizeof(buf
))
2131 ndm
->ndm_state
, ndm
->ndm_flags
);
2133 /* If the neighbor state is valid for use, process as an add or
2135 * else process as a delete. Note that the delete handling may
2137 * in re-adding the neighbor if it is a valid "remote" neighbor.
2139 if (ndm
->ndm_state
& NUD_VALID
)
2140 return zebra_vxlan_local_neigh_add_update(
2141 ifp
, link_if
, &ip
, &mac
, ndm
->ndm_state
,
2144 return zebra_vxlan_local_neigh_del(ifp
, link_if
, &ip
);
2147 if (IS_ZEBRA_DEBUG_KERNEL
)
2148 zlog_debug("Rx %s family %s IF %s(%u) IP %s",
2149 nl_msg_type_to_str(h
->nlmsg_type
),
2150 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2152 ipaddr2str(&ip
, buf2
, sizeof(buf2
)));
2154 /* Process the delete - it may result in re-adding the neighbor if it is
2155 * a valid "remote" neighbor.
2157 return zebra_vxlan_local_neigh_del(ifp
, link_if
, &ip
);
2160 static int netlink_neigh_table(struct sockaddr_nl
*snl
, struct nlmsghdr
*h
,
2161 ns_id_t ns_id
, int startup
)
2166 if (h
->nlmsg_type
!= RTM_NEWNEIGH
)
2169 /* Length validity. */
2170 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2174 /* We are interested only in AF_INET or AF_INET6 notifications. */
2175 ndm
= NLMSG_DATA(h
);
2176 if (ndm
->ndm_family
!= AF_INET
&& ndm
->ndm_family
!= AF_INET6
)
2179 return netlink_neigh_change(snl
, h
, len
);
2182 /* Request for IP neighbor information from the kernel */
2183 static int netlink_request_neigh(struct zebra_ns
*zns
, int family
, int type
,
2192 /* Form the request, specifying filter (rtattr) if needed. */
2193 memset(&req
, 0, sizeof(req
));
2194 req
.n
.nlmsg_type
= type
;
2195 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2196 req
.ndm
.ndm_family
= family
;
2198 addattr32(&req
.n
, sizeof(req
), NDA_IFINDEX
, ifindex
);
2200 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
2204 * IP Neighbor table read using netlink interface. This is invoked
2207 int netlink_neigh_read(struct zebra_ns
*zns
)
2211 /* Get IP neighbor table. */
2212 ret
= netlink_request_neigh(zns
, AF_UNSPEC
, RTM_GETNEIGH
, 0);
2215 ret
= netlink_parse_info(netlink_neigh_table
, &zns
->netlink_cmd
, zns
, 0,
2222 * IP Neighbor table read using netlink interface. This is for a specific
2225 int netlink_neigh_read_for_vlan(struct zebra_ns
*zns
, struct interface
*vlan_if
)
2229 ret
= netlink_request_neigh(zns
, AF_UNSPEC
, RTM_GETNEIGH
,
2233 ret
= netlink_parse_info(netlink_neigh_table
, &zns
->netlink_cmd
, zns
, 0,
2239 int netlink_neigh_change(struct sockaddr_nl
*snl
, struct nlmsghdr
*h
,
2245 if (!(h
->nlmsg_type
== RTM_NEWNEIGH
|| h
->nlmsg_type
== RTM_DELNEIGH
))
2248 /* Length validity. */
2249 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2253 /* Is this a notification for the MAC FDB or IP neighbor table? */
2254 ndm
= NLMSG_DATA(h
);
2255 if (ndm
->ndm_family
== AF_BRIDGE
)
2256 return netlink_macfdb_change(snl
, h
, len
);
2258 if (ndm
->ndm_type
!= RTN_UNICAST
)
2261 if (ndm
->ndm_family
== AF_INET
|| ndm
->ndm_family
== AF_INET6
)
2262 return netlink_ipneigh_change(snl
, h
, len
);
2267 static int netlink_neigh_update2(struct interface
*ifp
, struct ipaddr
*ip
,
2268 struct ethaddr
*mac
, u_int32_t flags
, int cmd
)
2277 struct zebra_ns
*zns
= zebra_ns_lookup(NS_DEFAULT
);
2278 char buf
[INET6_ADDRSTRLEN
];
2279 char buf2
[ETHER_ADDR_STRLEN
];
2281 memset(&req
.n
, 0, sizeof(req
.n
));
2282 memset(&req
.ndm
, 0, sizeof(req
.ndm
));
2284 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2285 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
2286 if (cmd
== RTM_NEWNEIGH
)
2287 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_REPLACE
);
2288 req
.n
.nlmsg_type
= cmd
; // RTM_NEWNEIGH or RTM_DELNEIGH
2289 req
.ndm
.ndm_family
= IS_IPADDR_V4(ip
) ? AF_INET
: AF_INET6
;
2290 req
.ndm
.ndm_state
= flags
;
2291 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
2292 req
.ndm
.ndm_type
= RTN_UNICAST
;
2293 req
.ndm
.ndm_flags
= NTF_EXT_LEARNED
;
2296 ipa_len
= IS_IPADDR_V4(ip
) ? IPV4_MAX_BYTELEN
: IPV6_MAX_BYTELEN
;
2297 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &ip
->ip
.addr
, ipa_len
);
2299 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, mac
, 6);
2301 if (IS_ZEBRA_DEBUG_KERNEL
)
2302 zlog_debug("Tx %s family %s IF %s(%u) Neigh %s MAC %s",
2303 nl_msg_type_to_str(cmd
),
2304 nl_family_to_str(req
.ndm
.ndm_family
), ifp
->name
,
2305 ifp
->ifindex
, ipaddr2str(ip
, buf
, sizeof(buf
)),
2306 mac
? prefix_mac2str(mac
, buf2
, sizeof(buf2
))
2309 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2313 int kernel_add_mac(struct interface
*ifp
, vlanid_t vid
, struct ethaddr
*mac
,
2314 struct in_addr vtep_ip
, u_char sticky
)
2316 return netlink_macfdb_update(ifp
, vid
, mac
, vtep_ip
, 0, RTM_NEWNEIGH
,
2320 int kernel_del_mac(struct interface
*ifp
, vlanid_t vid
, struct ethaddr
*mac
,
2321 struct in_addr vtep_ip
, int local
)
2323 return netlink_macfdb_update(ifp
, vid
, mac
, vtep_ip
, local
,
2327 int kernel_add_neigh(struct interface
*ifp
, struct ipaddr
*ip
,
2328 struct ethaddr
*mac
)
2330 return netlink_neigh_update2(ifp
, ip
, mac
, NUD_REACHABLE
, RTM_NEWNEIGH
);
2333 int kernel_del_neigh(struct interface
*ifp
, struct ipaddr
*ip
)
2335 return netlink_neigh_update2(ifp
, ip
, NULL
, 0, RTM_DELNEIGH
);
2339 * MPLS label forwarding table change via netlink interface.
2341 int netlink_mpls_multipath(int cmd
, zebra_lsp_t
*lsp
)
2344 zebra_nhlfe_t
*nhlfe
;
2345 struct nexthop
*nexthop
= NULL
;
2346 unsigned int nexthop_num
;
2347 const char *routedesc
;
2348 struct zebra_ns
*zns
= zebra_ns_lookup(NS_DEFAULT
);
2354 char buf
[NL_PKT_BUF_SIZE
];
2357 memset(&req
, 0, sizeof req
- NL_PKT_BUF_SIZE
);
2360 * Count # nexthops so we can decide whether to use singlepath
2361 * or multipath case.
2364 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2365 nexthop
= nhlfe
->nexthop
;
2368 if (cmd
== RTM_NEWROUTE
) {
2369 /* Count all selected NHLFEs */
2370 if (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2371 && CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
))
2375 /* Count all installed NHLFEs */
2376 if (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_INSTALLED
)
2377 && CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
2382 if ((nexthop_num
== 0) || (!lsp
->best_nhlfe
&& (cmd
!= RTM_DELROUTE
)))
2385 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
2386 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
2387 req
.n
.nlmsg_type
= cmd
;
2388 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
2390 req
.r
.rtm_family
= AF_MPLS
;
2391 req
.r
.rtm_table
= RT_TABLE_MAIN
;
2392 req
.r
.rtm_dst_len
= MPLS_LABEL_LEN_BITS
;
2393 req
.r
.rtm_scope
= RT_SCOPE_UNIVERSE
;
2394 req
.r
.rtm_type
= RTN_UNICAST
;
2396 if (cmd
== RTM_NEWROUTE
) {
2397 /* We do a replace to handle update. */
2398 req
.n
.nlmsg_flags
|= NLM_F_REPLACE
;
2400 /* set the protocol value if installing */
2401 route_type
= re_type_from_lsp_type(lsp
->best_nhlfe
->type
);
2402 req
.r
.rtm_protocol
= zebra2proto(route_type
);
2405 /* Fill destination */
2406 lse
= mpls_lse_encode(lsp
->ile
.in_label
, 0, 0, 1);
2407 addattr_l(&req
.n
, sizeof req
, RTA_DST
, &lse
, sizeof(mpls_lse_t
));
2409 /* Fill nexthops (paths) based on single-path or multipath. The paths
2410 * chosen depend on the operation.
2412 if (nexthop_num
== 1 || multipath_num
== 1) {
2413 routedesc
= "single-path";
2414 _netlink_mpls_debug(cmd
, lsp
->ile
.in_label
, routedesc
);
2417 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2418 nexthop
= nhlfe
->nexthop
;
2422 if ((cmd
== RTM_NEWROUTE
2423 && (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2424 && CHECK_FLAG(nexthop
->flags
,
2425 NEXTHOP_FLAG_ACTIVE
)))
2426 || (cmd
== RTM_DELROUTE
2427 && (CHECK_FLAG(nhlfe
->flags
,
2428 NHLFE_FLAG_INSTALLED
)
2429 && CHECK_FLAG(nexthop
->flags
,
2430 NEXTHOP_FLAG_FIB
)))) {
2431 /* Add the gateway */
2432 _netlink_mpls_build_singlepath(routedesc
, nhlfe
,
2439 } else /* Multipath case */
2441 char buf
[NL_PKT_BUF_SIZE
];
2442 struct rtattr
*rta
= (void *)buf
;
2443 struct rtnexthop
*rtnh
;
2444 union g_addr
*src1
= NULL
;
2446 rta
->rta_type
= RTA_MULTIPATH
;
2447 rta
->rta_len
= RTA_LENGTH(0);
2448 rtnh
= RTA_DATA(rta
);
2450 routedesc
= "multipath";
2451 _netlink_mpls_debug(cmd
, lsp
->ile
.in_label
, routedesc
);
2454 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2455 nexthop
= nhlfe
->nexthop
;
2459 if (nexthop_num
>= multipath_num
)
2462 if ((cmd
== RTM_NEWROUTE
2463 && (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2464 && CHECK_FLAG(nexthop
->flags
,
2465 NEXTHOP_FLAG_ACTIVE
)))
2466 || (cmd
== RTM_DELROUTE
2467 && (CHECK_FLAG(nhlfe
->flags
,
2468 NHLFE_FLAG_INSTALLED
)
2469 && CHECK_FLAG(nexthop
->flags
,
2470 NEXTHOP_FLAG_FIB
)))) {
2473 /* Build the multipath */
2474 _netlink_mpls_build_multipath(routedesc
, nhlfe
,
2477 rtnh
= RTNH_NEXT(rtnh
);
2481 /* Add the multipath */
2482 if (rta
->rta_len
> RTA_LENGTH(0))
2483 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_MULTIPATH
,
2484 RTA_DATA(rta
), RTA_PAYLOAD(rta
));
2487 /* Talk to netlink socket. */
2488 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2491 #endif /* HAVE_NETLINK */