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"
67 #include "zebra/zebra_errors.h"
73 static vlanid_t filter_vlan
= 0;
81 char ipv4_ll_buf
[16] = "169.254.0.1";
82 struct in_addr ipv4_ll
;
85 * The ipv4_ll data structure is used for all 5549
86 * additions to the kernel. Let's figure out the
87 * correct value one time instead for every
88 * install/remove of a 5549 type route
90 void rt_netlink_init(void)
92 inet_pton(AF_INET
, ipv4_ll_buf
, &ipv4_ll
);
95 static inline int is_selfroute(int proto
)
97 if ((proto
== RTPROT_BGP
) || (proto
== RTPROT_OSPF
)
98 || (proto
== RTPROT_ZSTATIC
) || (proto
== RTPROT_ZEBRA
)
99 || (proto
== RTPROT_ISIS
) || (proto
== RTPROT_RIPNG
)
100 || (proto
== RTPROT_NHRP
) || (proto
== RTPROT_EIGRP
)
101 || (proto
== RTPROT_LDP
) || (proto
== RTPROT_BABEL
)
102 || (proto
== RTPROT_RIP
) || (proto
== RTPROT_SHARP
)
103 || (proto
== RTPROT_PBR
) || (proto
== RTPROT_OPENFABRIC
)) {
110 static inline int zebra2proto(int proto
)
113 case ZEBRA_ROUTE_BABEL
:
114 proto
= RTPROT_BABEL
;
116 case ZEBRA_ROUTE_BGP
:
119 case ZEBRA_ROUTE_OSPF
:
120 case ZEBRA_ROUTE_OSPF6
:
123 case ZEBRA_ROUTE_STATIC
:
124 proto
= RTPROT_ZSTATIC
;
126 case ZEBRA_ROUTE_ISIS
:
129 case ZEBRA_ROUTE_RIP
:
132 case ZEBRA_ROUTE_RIPNG
:
133 proto
= RTPROT_RIPNG
;
135 case ZEBRA_ROUTE_NHRP
:
138 case ZEBRA_ROUTE_EIGRP
:
139 proto
= RTPROT_EIGRP
;
141 case ZEBRA_ROUTE_LDP
:
144 case ZEBRA_ROUTE_SHARP
:
145 proto
= RTPROT_SHARP
;
147 case ZEBRA_ROUTE_PBR
:
150 case ZEBRA_ROUTE_OPENFABRIC
:
151 proto
= RTPROT_OPENFABRIC
;
155 * When a user adds a new protocol this will show up
156 * to let them know to do something about it. This
157 * is intentionally a warn because we should see
158 * this as part of development of a new protocol
161 "%s: Please add this protocol(%d) to proper rt_netlink.c handling",
162 __PRETTY_FUNCTION__
, proto
);
163 proto
= RTPROT_ZEBRA
;
170 static inline int proto2zebra(int proto
, int family
)
174 proto
= ZEBRA_ROUTE_BABEL
;
177 proto
= ZEBRA_ROUTE_BGP
;
180 proto
= (family
== AFI_IP
) ? ZEBRA_ROUTE_OSPF
184 proto
= ZEBRA_ROUTE_ISIS
;
187 proto
= ZEBRA_ROUTE_RIP
;
190 proto
= ZEBRA_ROUTE_RIPNG
;
193 proto
= ZEBRA_ROUTE_NHRP
;
196 proto
= ZEBRA_ROUTE_EIGRP
;
199 proto
= ZEBRA_ROUTE_LDP
;
203 proto
= ZEBRA_ROUTE_STATIC
;
206 proto
= ZEBRA_ROUTE_SHARP
;
209 proto
= ZEBRA_ROUTE_PBR
;
211 case RTPROT_OPENFABRIC
:
212 proto
= ZEBRA_ROUTE_OPENFABRIC
;
216 * When a user adds a new protocol this will show up
217 * to let them know to do something about it. This
218 * is intentionally a warn because we should see
219 * this as part of development of a new protocol
222 "%s: Please add this protocol(%d) to proper rt_netlink.c handling",
223 __PRETTY_FUNCTION__
, proto
);
224 proto
= ZEBRA_ROUTE_KERNEL
;
231 Pending: create an efficient table_id (in a tree/hash) based lookup)
233 static vrf_id_t
vrf_lookup_by_table(uint32_t table_id
, ns_id_t ns_id
)
236 struct zebra_vrf
*zvrf
;
238 RB_FOREACH (vrf
, vrf_id_head
, &vrfs_by_id
) {
242 /* case vrf with netns : match the netnsid */
243 if (vrf_is_backend_netns()) {
244 if (ns_id
== zvrf_id(zvrf
))
245 return zvrf_id(zvrf
);
247 /* VRF is VRF_BACKEND_VRF_LITE */
248 if (zvrf
->table_id
!= table_id
)
250 return zvrf_id(zvrf
);
258 * @parse_encap_mpls() - Parses encapsulated mpls attributes
259 * @tb: Pointer to rtattr to look for nested items in.
260 * @labels: Pointer to store labels in.
262 * Return: Number of mpls labels found.
264 static int parse_encap_mpls(struct rtattr
*tb
, mpls_label_t
*labels
)
266 struct rtattr
*tb_encap
[MPLS_IPTUNNEL_MAX
+ 1] = {0};
267 mpls_lse_t
*lses
= NULL
;
272 mpls_label_t label
= 0;
274 netlink_parse_rtattr_nested(tb_encap
, MPLS_IPTUNNEL_MAX
, tb
);
275 lses
= (mpls_lse_t
*)RTA_DATA(tb_encap
[MPLS_IPTUNNEL_DST
]);
276 while (!bos
&& num_labels
< MPLS_MAX_LABELS
) {
277 mpls_lse_decode(lses
[num_labels
], &label
, &ttl
, &exp
, &bos
);
278 labels
[num_labels
++] = label
;
284 /* Looking up routing table by netlink interface. */
285 static int netlink_route_change_read_unicast(struct nlmsghdr
*h
, ns_id_t ns_id
,
290 struct rtattr
*tb
[RTA_MAX
+ 1];
293 struct prefix_ipv6 src_p
= {};
296 char anyaddr
[16] = {0};
298 int proto
= ZEBRA_ROUTE_KERNEL
;
303 uint8_t distance
= 0;
308 void *prefsrc
= NULL
; /* IPv4 preferred source host address */
309 void *src
= NULL
; /* IPv6 srcdest source prefix */
310 enum blackhole_type bh_type
= BLACKHOLE_UNSPEC
;
313 mpls_label_t labels
[MPLS_MAX_LABELS
] = {0};
318 if (startup
&& h
->nlmsg_type
!= RTM_NEWROUTE
)
320 switch (rtm
->rtm_type
) {
324 bh_type
= BLACKHOLE_NULL
;
326 case RTN_UNREACHABLE
:
327 bh_type
= BLACKHOLE_REJECT
;
330 bh_type
= BLACKHOLE_ADMINPROHIB
;
336 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
338 zlog_err("%s: Message received from netlink is of a broken size %d %zu",
339 __PRETTY_FUNCTION__
, h
->nlmsg_len
,
340 (size_t)NLMSG_LENGTH(sizeof(struct rtmsg
)));
344 memset(tb
, 0, sizeof tb
);
345 netlink_parse_rtattr(tb
, RTA_MAX
, RTM_RTA(rtm
), len
);
347 if (rtm
->rtm_flags
& RTM_F_CLONED
)
349 if (rtm
->rtm_protocol
== RTPROT_REDIRECT
)
351 if (rtm
->rtm_protocol
== RTPROT_KERNEL
)
354 if (!startup
&& is_selfroute(rtm
->rtm_protocol
)
355 && h
->nlmsg_type
== RTM_NEWROUTE
) {
356 if (IS_ZEBRA_DEBUG_KERNEL
)
357 zlog_debug("Route type: %d Received that we think we have originated, ignoring",
362 /* We don't care about change notifications for the MPLS table. */
363 /* TODO: Revisit this. */
364 if (rtm
->rtm_family
== AF_MPLS
)
367 /* Table corresponding to route. */
369 table
= *(int *)RTA_DATA(tb
[RTA_TABLE
]);
371 table
= rtm
->rtm_table
;
374 vrf_id
= vrf_lookup_by_table(table
, ns_id
);
375 if (vrf_id
== VRF_DEFAULT
) {
376 if (!is_zebra_valid_kernel_table(table
)
377 && !is_zebra_main_routing_table(table
))
381 /* Route which inserted by Zebra. */
382 if (is_selfroute(rtm
->rtm_protocol
)) {
383 flags
|= ZEBRA_FLAG_SELFROUTE
;
384 proto
= proto2zebra(rtm
->rtm_protocol
, rtm
->rtm_family
);
387 index
= *(int *)RTA_DATA(tb
[RTA_OIF
]);
390 dest
= RTA_DATA(tb
[RTA_DST
]);
395 src
= RTA_DATA(tb
[RTA_SRC
]);
400 prefsrc
= RTA_DATA(tb
[RTA_PREFSRC
]);
403 gate
= RTA_DATA(tb
[RTA_GATEWAY
]);
405 if (tb
[RTA_PRIORITY
])
406 metric
= *(int *)RTA_DATA(tb
[RTA_PRIORITY
]);
408 #if defined(SUPPORT_REALMS)
410 tag
= *(uint32_t *)RTA_DATA(tb
[RTA_FLOW
]);
413 if (tb
[RTA_METRICS
]) {
414 struct rtattr
*mxrta
[RTAX_MAX
+ 1];
416 memset(mxrta
, 0, sizeof mxrta
);
417 netlink_parse_rtattr(mxrta
, RTAX_MAX
, RTA_DATA(tb
[RTA_METRICS
]),
418 RTA_PAYLOAD(tb
[RTA_METRICS
]));
421 mtu
= *(uint32_t *)RTA_DATA(mxrta
[RTAX_MTU
]);
424 if (rtm
->rtm_family
== AF_INET
) {
426 if (rtm
->rtm_dst_len
> IPV4_MAX_BITLEN
) {
428 "Invalid destination prefix length: %u received from kernel route change",
432 memcpy(&p
.u
.prefix4
, dest
, 4);
433 p
.prefixlen
= rtm
->rtm_dst_len
;
435 if (rtm
->rtm_src_len
!= 0) {
436 char buf
[PREFIX_STRLEN
];
438 EC_ZEBRA_UNSUPPORTED_V4_SRCDEST
,
439 "unsupported IPv4 sourcedest route (dest %s vrf %u)",
440 prefix2str(&p
, buf
, sizeof(buf
)), vrf_id
);
444 /* Force debug below to not display anything for source */
446 } else if (rtm
->rtm_family
== AF_INET6
) {
448 if (rtm
->rtm_dst_len
> IPV6_MAX_BITLEN
) {
450 "Invalid destination prefix length: %u received from kernel route change",
454 memcpy(&p
.u
.prefix6
, dest
, 16);
455 p
.prefixlen
= rtm
->rtm_dst_len
;
457 src_p
.family
= AF_INET6
;
458 if (rtm
->rtm_src_len
> IPV6_MAX_BITLEN
) {
460 "Invalid source prefix length: %u received from kernel route change",
464 memcpy(&src_p
.prefix
, src
, 16);
465 src_p
.prefixlen
= rtm
->rtm_src_len
;
469 * For ZEBRA_ROUTE_KERNEL types:
471 * The metric/priority of the route received from the kernel
472 * is a 32 bit number. We are going to interpret the high
473 * order byte as the Admin Distance and the low order 3 bytes
476 * This will allow us to do two things:
477 * 1) Allow the creation of kernel routes that can be
478 * overridden by zebra.
479 * 2) Allow the old behavior for 'most' kernel route types
480 * if a user enters 'ip route ...' v4 routes get a metric
481 * of 0 and v6 routes get a metric of 1024. Both of these
482 * values will end up with a admin distance of 0, which
483 * will cause them to win for the purposes of zebra.
485 if (proto
== ZEBRA_ROUTE_KERNEL
) {
486 distance
= (metric
>> 24) & 0xFF;
487 metric
= (metric
& 0x00FFFFFF);
490 if (IS_ZEBRA_DEBUG_KERNEL
) {
491 char buf
[PREFIX_STRLEN
];
492 char buf2
[PREFIX_STRLEN
];
493 zlog_debug("%s %s%s%s vrf %u(%u) metric: %d Admin Distance: %d",
494 nl_msg_type_to_str(h
->nlmsg_type
),
495 prefix2str(&p
, buf
, sizeof(buf
)),
496 src_p
.prefixlen
? " from " : "",
498 ? prefix2str(&src_p
, buf2
, sizeof(buf2
))
500 vrf_id
, table
, metric
, distance
);
504 if (rtm
->rtm_family
== AF_INET6
)
507 if (h
->nlmsg_type
== RTM_NEWROUTE
) {
508 struct interface
*ifp
;
509 vrf_id_t nh_vrf_id
= vrf_id
;
511 if (!tb
[RTA_MULTIPATH
]) {
513 size_t sz
= (afi
== AFI_IP
) ? 4 : 16;
515 memset(&nh
, 0, sizeof(nh
));
517 if (bh_type
== BLACKHOLE_UNSPEC
) {
519 nh
.type
= NEXTHOP_TYPE_IFINDEX
;
520 else if (index
&& gate
)
523 ? NEXTHOP_TYPE_IPV4_IFINDEX
524 : NEXTHOP_TYPE_IPV6_IFINDEX
;
525 else if (!index
&& gate
)
526 nh
.type
= (afi
== AFI_IP
)
530 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
531 nh
.bh_type
= bh_type
;
534 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
535 nh
.bh_type
= bh_type
;
539 memcpy(&nh
.src
, prefsrc
, sz
);
541 memcpy(&nh
.gate
, gate
, sz
);
544 ifp
= if_lookup_by_index_per_ns(
545 zebra_ns_lookup(ns_id
),
548 nh_vrf_id
= ifp
->vrf_id
;
550 nh
.vrf_id
= nh_vrf_id
;
552 if (tb
[RTA_ENCAP
] && tb
[RTA_ENCAP_TYPE
]
553 && *(uint16_t *)RTA_DATA(tb
[RTA_ENCAP_TYPE
])
554 == LWTUNNEL_ENCAP_MPLS
) {
556 parse_encap_mpls(tb
[RTA_ENCAP
], labels
);
560 nexthop_add_labels(&nh
, ZEBRA_LSP_STATIC
,
563 rib_add(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
, &p
,
564 &src_p
, &nh
, table
, metric
, mtu
, distance
, tag
);
566 /* This is a multipath route */
568 struct route_entry
*re
;
569 struct rtnexthop
*rtnh
=
570 (struct rtnexthop
*)RTA_DATA(tb
[RTA_MULTIPATH
]);
572 len
= RTA_PAYLOAD(tb
[RTA_MULTIPATH
]);
574 re
= XCALLOC(MTYPE_RE
, sizeof(struct route_entry
));
576 re
->distance
= distance
;
583 re
->uptime
= time(NULL
);
587 struct nexthop
*nh
= NULL
;
589 if (len
< (int)sizeof(*rtnh
)
590 || rtnh
->rtnh_len
> len
)
593 index
= rtnh
->rtnh_ifindex
;
596 * Yes we are looking this up
597 * for every nexthop and just
598 * using the last one looked
601 ifp
= if_lookup_by_index_per_ns(
602 zebra_ns_lookup(ns_id
),
605 nh_vrf_id
= ifp
->vrf_id
;
608 EC_ZEBRA_UNKNOWN_INTERFACE
,
609 "%s: Unknown interface %u specified, defaulting to VRF_DEFAULT",
612 nh_vrf_id
= VRF_DEFAULT
;
618 if (rtnh
->rtnh_len
> sizeof(*rtnh
)) {
619 memset(tb
, 0, sizeof(tb
));
620 netlink_parse_rtattr(
621 tb
, RTA_MAX
, RTNH_DATA(rtnh
),
622 rtnh
->rtnh_len
- sizeof(*rtnh
));
626 if (tb
[RTA_ENCAP
] && tb
[RTA_ENCAP_TYPE
]
627 && *(uint16_t *)RTA_DATA(
629 == LWTUNNEL_ENCAP_MPLS
) {
630 num_labels
= parse_encap_mpls(
631 tb
[RTA_ENCAP
], labels
);
636 if (rtm
->rtm_family
== AF_INET
) {
638 nh
= route_entry_nexthop_ipv4_ifindex_add(
643 nh
= route_entry_nexthop_ipv4_add(
647 } else if (rtm
->rtm_family
650 nh
= route_entry_nexthop_ipv6_ifindex_add(
654 nh
= route_entry_nexthop_ipv6_add(
659 nh
= route_entry_nexthop_ifindex_add(
660 re
, index
, nh_vrf_id
);
662 if (nh
&& num_labels
)
663 nexthop_add_labels(nh
, ZEBRA_LSP_STATIC
,
666 if (rtnh
->rtnh_len
== 0)
669 len
-= NLMSG_ALIGN(rtnh
->rtnh_len
);
670 rtnh
= RTNH_NEXT(rtnh
);
673 zserv_nexthop_num_warn(__func__
,
674 (const struct prefix
*)&p
,
676 if (re
->nexthop_num
== 0)
679 rib_add_multipath(afi
, SAFI_UNICAST
, &p
,
683 if (!tb
[RTA_MULTIPATH
]) {
685 size_t sz
= (afi
== AFI_IP
) ? 4 : 16;
687 memset(&nh
, 0, sizeof(nh
));
688 if (bh_type
== BLACKHOLE_UNSPEC
) {
690 nh
.type
= NEXTHOP_TYPE_IFINDEX
;
691 else if (index
&& gate
)
694 ? NEXTHOP_TYPE_IPV4_IFINDEX
695 : NEXTHOP_TYPE_IPV6_IFINDEX
;
696 else if (!index
&& gate
)
697 nh
.type
= (afi
== AFI_IP
)
701 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
702 nh
.bh_type
= BLACKHOLE_UNSPEC
;
705 nh
.type
= NEXTHOP_TYPE_BLACKHOLE
;
706 nh
.bh_type
= bh_type
;
710 memcpy(&nh
.gate
, gate
, sz
);
711 rib_delete(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
,
712 &p
, &src_p
, &nh
, table
, metric
, distance
,
715 /* XXX: need to compare the entire list of nexthops
716 * here for NLM_F_APPEND stupidity */
717 rib_delete(afi
, SAFI_UNICAST
, vrf_id
, proto
, 0, flags
,
718 &p
, &src_p
, NULL
, table
, metric
, distance
,
726 static struct mcast_route_data
*mroute
= NULL
;
728 static int netlink_route_change_read_multicast(struct nlmsghdr
*h
,
729 ns_id_t ns_id
, int startup
)
733 struct rtattr
*tb
[RTA_MAX
+ 1];
734 struct mcast_route_data
*m
;
735 struct mcast_route_data mr
;
742 char oif_list
[256] = "\0";
749 memset(&mr
, 0, sizeof(mr
));
755 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
757 memset(tb
, 0, sizeof tb
);
758 netlink_parse_rtattr(tb
, RTA_MAX
, RTM_RTA(rtm
), len
);
761 table
= *(int *)RTA_DATA(tb
[RTA_TABLE
]);
763 table
= rtm
->rtm_table
;
765 vrf
= vrf_lookup_by_table(table
, ns_id
);
768 iif
= *(int *)RTA_DATA(tb
[RTA_IIF
]);
771 m
->sg
.src
= *(struct in_addr
*)RTA_DATA(tb
[RTA_SRC
]);
774 m
->sg
.grp
= *(struct in_addr
*)RTA_DATA(tb
[RTA_DST
]);
776 if ((RTA_EXPIRES
<= RTA_MAX
) && tb
[RTA_EXPIRES
])
777 m
->lastused
= *(unsigned long long *)RTA_DATA(tb
[RTA_EXPIRES
]);
779 if (tb
[RTA_MULTIPATH
]) {
780 struct rtnexthop
*rtnh
=
781 (struct rtnexthop
*)RTA_DATA(tb
[RTA_MULTIPATH
]);
783 len
= RTA_PAYLOAD(tb
[RTA_MULTIPATH
]);
785 if (len
< (int)sizeof(*rtnh
) || rtnh
->rtnh_len
> len
)
788 oif
[oif_count
] = rtnh
->rtnh_ifindex
;
791 if (rtnh
->rtnh_len
== 0)
794 len
-= NLMSG_ALIGN(rtnh
->rtnh_len
);
795 rtnh
= RTNH_NEXT(rtnh
);
799 if (IS_ZEBRA_DEBUG_KERNEL
) {
800 struct interface
*ifp
;
801 strlcpy(sbuf
, inet_ntoa(m
->sg
.src
), sizeof(sbuf
));
802 strlcpy(gbuf
, inet_ntoa(m
->sg
.grp
), sizeof(gbuf
));
803 for (count
= 0; count
< oif_count
; count
++) {
804 ifp
= if_lookup_by_index(oif
[count
], vrf
);
807 sprintf(temp
, "%s(%d) ", ifp
? ifp
->name
: "Unknown",
809 strcat(oif_list
, temp
);
811 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(vrf
);
812 ifp
= if_lookup_by_index(iif
, vrf
);
813 zlog_debug("MCAST VRF: %s(%d) %s (%s,%s) IIF: %s(%d) OIF: %s jiffies: %lld",
814 zvrf
->vrf
->name
, vrf
,
815 nl_msg_type_to_str(h
->nlmsg_type
),
816 sbuf
, gbuf
, ifp
? ifp
->name
: "Unknown", iif
,
817 oif_list
, m
->lastused
);
822 int netlink_route_change(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
829 if (!(h
->nlmsg_type
== RTM_NEWROUTE
|| h
->nlmsg_type
== RTM_DELROUTE
)) {
830 /* If this is not route add/delete message print warning. */
831 zlog_debug("Kernel message: %s NS %u\n",
832 nl_msg_type_to_str(h
->nlmsg_type
), ns_id
);
836 if (!(rtm
->rtm_family
== AF_INET
||
837 rtm
->rtm_family
== AF_INET6
||
838 rtm
->rtm_family
== RTNL_FAMILY_IPMR
)) {
840 EC_ZEBRA_UNKNOWN_FAMILY
,
841 "Invalid address family: %u received from kernel route change: %s",
842 rtm
->rtm_family
, nl_msg_type_to_str(h
->nlmsg_type
));
846 /* Connected route. */
847 if (IS_ZEBRA_DEBUG_KERNEL
)
848 zlog_debug("%s %s %s proto %s NS %u",
849 nl_msg_type_to_str(h
->nlmsg_type
),
850 nl_family_to_str(rtm
->rtm_family
),
851 nl_rttype_to_str(rtm
->rtm_type
),
852 nl_rtproto_to_str(rtm
->rtm_protocol
), ns_id
);
855 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct rtmsg
));
857 zlog_err("%s: Message received from netlink is of a broken size: %d %zu",
860 (size_t)NLMSG_LENGTH(sizeof(struct rtmsg
)));
864 if (rtm
->rtm_type
== RTN_MULTICAST
)
865 netlink_route_change_read_multicast(h
, ns_id
, startup
);
867 netlink_route_change_read_unicast(h
, ns_id
, startup
);
871 /* Request for specific route information from the kernel */
872 static int netlink_request_route(struct zebra_ns
*zns
, int family
, int type
)
879 /* Form the request, specifying filter (rtattr) if needed. */
880 memset(&req
, 0, sizeof(req
));
881 req
.n
.nlmsg_type
= type
;
882 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
883 req
.rtm
.rtm_family
= family
;
885 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
888 /* Routing table read function using netlink interface. Only called
890 int netlink_route_read(struct zebra_ns
*zns
)
894 /* Get IPv4 routing table. */
895 ret
= netlink_request_route(zns
, AF_INET
, RTM_GETROUTE
);
898 ret
= netlink_parse_info(netlink_route_change_read_unicast
,
899 &zns
->netlink_cmd
, zns
, 0, 1);
903 /* Get IPv6 routing table. */
904 ret
= netlink_request_route(zns
, AF_INET6
, RTM_GETROUTE
);
907 ret
= netlink_parse_info(netlink_route_change_read_unicast
,
908 &zns
->netlink_cmd
, zns
, 0, 1);
915 static void _netlink_route_nl_add_gateway_info(uint8_t route_family
,
917 struct nlmsghdr
*nlmsg
,
918 size_t req_size
, int bytelen
,
919 struct nexthop
*nexthop
)
921 if (route_family
== AF_MPLS
) {
922 struct gw_family_t gw_fam
;
924 gw_fam
.family
= gw_family
;
925 if (gw_family
== AF_INET
)
926 memcpy(&gw_fam
.gate
.ipv4
, &nexthop
->gate
.ipv4
, bytelen
);
928 memcpy(&gw_fam
.gate
.ipv6
, &nexthop
->gate
.ipv6
, bytelen
);
929 addattr_l(nlmsg
, req_size
, RTA_VIA
, &gw_fam
.family
,
932 if (gw_family
== AF_INET
)
933 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
,
934 &nexthop
->gate
.ipv4
, bytelen
);
936 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
,
937 &nexthop
->gate
.ipv6
, bytelen
);
941 static void _netlink_route_rta_add_gateway_info(uint8_t route_family
,
944 struct rtnexthop
*rtnh
,
945 size_t req_size
, int bytelen
,
946 struct nexthop
*nexthop
)
948 if (route_family
== AF_MPLS
) {
949 struct gw_family_t gw_fam
;
951 gw_fam
.family
= gw_family
;
952 if (gw_family
== AF_INET
)
953 memcpy(&gw_fam
.gate
.ipv4
, &nexthop
->gate
.ipv4
, bytelen
);
955 memcpy(&gw_fam
.gate
.ipv6
, &nexthop
->gate
.ipv6
, bytelen
);
956 rta_addattr_l(rta
, req_size
, RTA_VIA
, &gw_fam
.family
,
958 rtnh
->rtnh_len
+= RTA_LENGTH(bytelen
+ 2);
960 if (gw_family
== AF_INET
)
961 rta_addattr_l(rta
, req_size
, RTA_GATEWAY
,
962 &nexthop
->gate
.ipv4
, bytelen
);
964 rta_addattr_l(rta
, req_size
, RTA_GATEWAY
,
965 &nexthop
->gate
.ipv6
, bytelen
);
966 rtnh
->rtnh_len
+= sizeof(struct rtattr
) + bytelen
;
970 /* This function takes a nexthop as argument and adds
971 * the appropriate netlink attributes to an existing
974 * @param routedesc: Human readable description of route type
975 * (direct/recursive, single-/multipath)
976 * @param bytelen: Length of addresses in bytes.
977 * @param nexthop: Nexthop information
978 * @param nlmsg: nlmsghdr structure to fill in.
979 * @param req_size: The size allocated for the message.
981 static void _netlink_route_build_singlepath(const char *routedesc
, int bytelen
,
982 struct nexthop
*nexthop
,
983 struct nlmsghdr
*nlmsg
,
985 size_t req_size
, int cmd
)
987 struct mpls_label_stack
*nh_label
;
988 mpls_lse_t out_lse
[MPLS_MAX_LABELS
];
993 * label_buf is *only* currently used within debugging.
994 * As such when we assign it we are guarding it inside
995 * a debug test. If you want to change this make sure
996 * you fix this assumption
1001 for (struct nexthop
*nh
= nexthop
; nh
; nh
= nh
->rparent
) {
1002 char label_buf1
[20];
1004 nh_label
= nh
->nh_label
;
1005 if (!nh_label
|| !nh_label
->num_labels
)
1008 for (int i
= 0; i
< nh_label
->num_labels
; i
++) {
1009 if (nh_label
->label
[i
] == MPLS_LABEL_IMPLICIT_NULL
)
1012 if (IS_ZEBRA_DEBUG_KERNEL
) {
1014 sprintf(label_buf
, "label %u",
1015 nh_label
->label
[i
]);
1017 sprintf(label_buf1
, "/%u",
1018 nh_label
->label
[i
]);
1019 strlcat(label_buf
, label_buf1
,
1024 out_lse
[num_labels
] =
1025 mpls_lse_encode(nh_label
->label
[i
], 0, 0, 0);
1031 /* Set the BoS bit */
1032 out_lse
[num_labels
- 1] |= htonl(1 << MPLS_LS_S_SHIFT
);
1034 if (rtmsg
->rtm_family
== AF_MPLS
)
1035 addattr_l(nlmsg
, req_size
, RTA_NEWDST
, &out_lse
,
1036 num_labels
* sizeof(mpls_lse_t
));
1038 struct rtattr
*nest
;
1039 uint16_t encap
= LWTUNNEL_ENCAP_MPLS
;
1041 addattr_l(nlmsg
, req_size
, RTA_ENCAP_TYPE
, &encap
,
1043 nest
= addattr_nest(nlmsg
, req_size
, RTA_ENCAP
);
1044 addattr_l(nlmsg
, req_size
, MPLS_IPTUNNEL_DST
, &out_lse
,
1045 num_labels
* sizeof(mpls_lse_t
));
1046 addattr_nest_end(nlmsg
, nest
);
1050 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
))
1051 rtmsg
->rtm_flags
|= RTNH_F_ONLINK
;
1053 if (rtmsg
->rtm_family
== AF_INET
1054 && (nexthop
->type
== NEXTHOP_TYPE_IPV6
1055 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
)) {
1056 rtmsg
->rtm_flags
|= RTNH_F_ONLINK
;
1057 addattr_l(nlmsg
, req_size
, RTA_GATEWAY
, &ipv4_ll
, 4);
1058 addattr32(nlmsg
, req_size
, RTA_OIF
, nexthop
->ifindex
);
1060 if (nexthop
->rmap_src
.ipv4
.s_addr
&& (cmd
== RTM_NEWROUTE
))
1061 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1062 &nexthop
->rmap_src
.ipv4
, bytelen
);
1063 else if (nexthop
->src
.ipv4
.s_addr
&& (cmd
== RTM_NEWROUTE
))
1064 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1065 &nexthop
->src
.ipv4
, bytelen
);
1067 if (IS_ZEBRA_DEBUG_KERNEL
)
1069 " 5549: _netlink_route_build_singlepath() (%s): "
1070 "nexthop via %s %s if %u(%u)",
1071 routedesc
, ipv4_ll_buf
, label_buf
,
1072 nexthop
->ifindex
, nexthop
->vrf_id
);
1076 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
1077 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
1078 /* Send deletes to the kernel without specifying the next-hop */
1079 if (cmd
!= RTM_DELROUTE
)
1080 _netlink_route_nl_add_gateway_info(
1081 rtmsg
->rtm_family
, AF_INET
, nlmsg
, req_size
,
1084 if (cmd
== RTM_NEWROUTE
) {
1085 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1086 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1087 &nexthop
->rmap_src
.ipv4
, bytelen
);
1088 else if (nexthop
->src
.ipv4
.s_addr
)
1089 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1090 &nexthop
->src
.ipv4
, bytelen
);
1093 if (IS_ZEBRA_DEBUG_KERNEL
)
1095 "netlink_route_multipath() (%s): "
1096 "nexthop via %s %s if %u(%u)",
1097 routedesc
, inet_ntoa(nexthop
->gate
.ipv4
),
1098 label_buf
, nexthop
->ifindex
, nexthop
->vrf_id
);
1101 if (nexthop
->type
== NEXTHOP_TYPE_IPV6
1102 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1103 _netlink_route_nl_add_gateway_info(rtmsg
->rtm_family
, AF_INET6
,
1104 nlmsg
, req_size
, bytelen
,
1107 if (cmd
== RTM_NEWROUTE
) {
1108 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1109 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1110 &nexthop
->rmap_src
.ipv6
, bytelen
);
1111 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1112 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1113 &nexthop
->src
.ipv6
, bytelen
);
1116 if (IS_ZEBRA_DEBUG_KERNEL
)
1118 "netlink_route_multipath() (%s): "
1119 "nexthop via %s %s if %u(%u)",
1120 routedesc
, inet6_ntoa(nexthop
->gate
.ipv6
),
1121 label_buf
, nexthop
->ifindex
, nexthop
->vrf_id
);
1125 * We have the ifindex so we should always send it
1126 * This is especially useful if we are doing route
1129 if (nexthop
->type
!= NEXTHOP_TYPE_BLACKHOLE
)
1130 addattr32(nlmsg
, req_size
, RTA_OIF
, nexthop
->ifindex
);
1132 if (nexthop
->type
== NEXTHOP_TYPE_IFINDEX
1133 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
1134 if (cmd
== RTM_NEWROUTE
) {
1135 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1136 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1137 &nexthop
->rmap_src
.ipv4
, bytelen
);
1138 else if (nexthop
->src
.ipv4
.s_addr
)
1139 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1140 &nexthop
->src
.ipv4
, bytelen
);
1143 if (IS_ZEBRA_DEBUG_KERNEL
)
1145 "netlink_route_multipath() (%s): "
1146 "nexthop via if %u(%u)",
1147 routedesc
, nexthop
->ifindex
, nexthop
->vrf_id
);
1150 if (nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1151 if (cmd
== RTM_NEWROUTE
) {
1152 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1153 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1154 &nexthop
->rmap_src
.ipv6
, bytelen
);
1155 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1156 addattr_l(nlmsg
, req_size
, RTA_PREFSRC
,
1157 &nexthop
->src
.ipv6
, bytelen
);
1160 if (IS_ZEBRA_DEBUG_KERNEL
)
1162 "netlink_route_multipath() (%s): "
1163 "nexthop via if %u(%u)",
1164 routedesc
, nexthop
->ifindex
, nexthop
->vrf_id
);
1168 /* This function takes a nexthop as argument and
1169 * appends to the given rtattr/rtnexthop pair the
1170 * representation of the nexthop. If the nexthop
1171 * defines a preferred source, the src parameter
1172 * will be modified to point to that src, otherwise
1173 * it will be kept unmodified.
1175 * @param routedesc: Human readable description of route type
1176 * (direct/recursive, single-/multipath)
1177 * @param bytelen: Length of addresses in bytes.
1178 * @param nexthop: Nexthop information
1179 * @param rta: rtnetlink attribute structure
1180 * @param rtnh: pointer to an rtnetlink nexthop structure
1181 * @param src: pointer pointing to a location where
1182 * the prefsrc should be stored.
1184 static void _netlink_route_build_multipath(const char *routedesc
, int bytelen
,
1185 struct nexthop
*nexthop
,
1187 struct rtnexthop
*rtnh
,
1188 struct rtmsg
*rtmsg
,
1191 struct mpls_label_stack
*nh_label
;
1192 mpls_lse_t out_lse
[MPLS_MAX_LABELS
];
1194 char label_buf
[256];
1196 rtnh
->rtnh_len
= sizeof(*rtnh
);
1197 rtnh
->rtnh_flags
= 0;
1198 rtnh
->rtnh_hops
= 0;
1199 rta
->rta_len
+= rtnh
->rtnh_len
;
1202 * label_buf is *only* currently used within debugging.
1203 * As such when we assign it we are guarding it inside
1204 * a debug test. If you want to change this make sure
1205 * you fix this assumption
1207 label_buf
[0] = '\0';
1210 for (struct nexthop
*nh
= nexthop
; nh
; nh
= nh
->rparent
) {
1211 char label_buf1
[20];
1213 nh_label
= nh
->nh_label
;
1214 if (!nh_label
|| !nh_label
->num_labels
)
1217 for (int i
= 0; i
< nh_label
->num_labels
; i
++) {
1218 if (nh_label
->label
[i
] == MPLS_LABEL_IMPLICIT_NULL
)
1221 if (IS_ZEBRA_DEBUG_KERNEL
) {
1223 sprintf(label_buf
, "label %u",
1224 nh_label
->label
[i
]);
1226 sprintf(label_buf1
, "/%u",
1227 nh_label
->label
[i
]);
1228 strlcat(label_buf
, label_buf1
,
1233 out_lse
[num_labels
] =
1234 mpls_lse_encode(nh_label
->label
[i
], 0, 0, 0);
1240 /* Set the BoS bit */
1241 out_lse
[num_labels
- 1] |= htonl(1 << MPLS_LS_S_SHIFT
);
1243 if (rtmsg
->rtm_family
== AF_MPLS
) {
1244 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_NEWDST
,
1246 num_labels
* sizeof(mpls_lse_t
));
1248 RTA_LENGTH(num_labels
* sizeof(mpls_lse_t
));
1250 struct rtattr
*nest
;
1251 uint16_t encap
= LWTUNNEL_ENCAP_MPLS
;
1252 int len
= rta
->rta_len
;
1254 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_ENCAP_TYPE
,
1255 &encap
, sizeof(uint16_t));
1256 nest
= rta_nest(rta
, NL_PKT_BUF_SIZE
, RTA_ENCAP
);
1257 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, MPLS_IPTUNNEL_DST
,
1259 num_labels
* sizeof(mpls_lse_t
));
1260 rta_nest_end(rta
, nest
);
1261 rtnh
->rtnh_len
+= rta
->rta_len
- len
;
1265 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ONLINK
))
1266 rtnh
->rtnh_flags
|= RTNH_F_ONLINK
;
1268 if (rtmsg
->rtm_family
== AF_INET
1269 && (nexthop
->type
== NEXTHOP_TYPE_IPV6
1270 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
)) {
1272 rtnh
->rtnh_flags
|= RTNH_F_ONLINK
;
1273 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTA_GATEWAY
, &ipv4_ll
,
1275 rtnh
->rtnh_len
+= sizeof(struct rtattr
) + bytelen
;
1276 rtnh
->rtnh_ifindex
= nexthop
->ifindex
;
1278 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1279 *src
= &nexthop
->rmap_src
;
1280 else if (nexthop
->src
.ipv4
.s_addr
)
1281 *src
= &nexthop
->src
;
1283 if (IS_ZEBRA_DEBUG_KERNEL
)
1285 " 5549: netlink_route_build_multipath() (%s): "
1286 "nexthop via %s %s if %u",
1287 routedesc
, ipv4_ll_buf
, label_buf
,
1292 if (nexthop
->type
== NEXTHOP_TYPE_IPV4
1293 || nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
) {
1294 _netlink_route_rta_add_gateway_info(rtmsg
->rtm_family
, AF_INET
,
1295 rta
, rtnh
, NL_PKT_BUF_SIZE
,
1297 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1298 *src
= &nexthop
->rmap_src
;
1299 else if (nexthop
->src
.ipv4
.s_addr
)
1300 *src
= &nexthop
->src
;
1302 if (IS_ZEBRA_DEBUG_KERNEL
)
1304 "netlink_route_multipath() (%s): "
1305 "nexthop via %s %s if %u",
1306 routedesc
, inet_ntoa(nexthop
->gate
.ipv4
),
1307 label_buf
, nexthop
->ifindex
);
1309 if (nexthop
->type
== NEXTHOP_TYPE_IPV6
1310 || nexthop
->type
== NEXTHOP_TYPE_IPV6_IFINDEX
) {
1311 _netlink_route_rta_add_gateway_info(rtmsg
->rtm_family
, AF_INET6
,
1312 rta
, rtnh
, NL_PKT_BUF_SIZE
,
1315 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->rmap_src
.ipv6
))
1316 *src
= &nexthop
->rmap_src
;
1317 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop
->src
.ipv6
))
1318 *src
= &nexthop
->src
;
1320 if (IS_ZEBRA_DEBUG_KERNEL
)
1322 "netlink_route_multipath() (%s): "
1323 "nexthop via %s %s if %u",
1324 routedesc
, inet6_ntoa(nexthop
->gate
.ipv6
),
1325 label_buf
, nexthop
->ifindex
);
1329 * We have figured out the ifindex so we should always send it
1330 * This is especially useful if we are doing route
1333 if (nexthop
->type
!= NEXTHOP_TYPE_BLACKHOLE
)
1334 rtnh
->rtnh_ifindex
= nexthop
->ifindex
;
1337 if (nexthop
->type
== NEXTHOP_TYPE_IPV4_IFINDEX
1338 || nexthop
->type
== NEXTHOP_TYPE_IFINDEX
) {
1339 if (nexthop
->rmap_src
.ipv4
.s_addr
)
1340 *src
= &nexthop
->rmap_src
;
1341 else if (nexthop
->src
.ipv4
.s_addr
)
1342 *src
= &nexthop
->src
;
1344 if (IS_ZEBRA_DEBUG_KERNEL
)
1346 "netlink_route_multipath() (%s): "
1347 "nexthop via if %u",
1348 routedesc
, nexthop
->ifindex
);
1352 static inline void _netlink_mpls_build_singlepath(const char *routedesc
,
1353 zebra_nhlfe_t
*nhlfe
,
1354 struct nlmsghdr
*nlmsg
,
1355 struct rtmsg
*rtmsg
,
1356 size_t req_size
, int cmd
)
1361 family
= NHLFE_FAMILY(nhlfe
);
1362 bytelen
= (family
== AF_INET
? 4 : 16);
1363 _netlink_route_build_singlepath(routedesc
, bytelen
, nhlfe
->nexthop
,
1364 nlmsg
, rtmsg
, req_size
, cmd
);
1369 _netlink_mpls_build_multipath(const char *routedesc
, zebra_nhlfe_t
*nhlfe
,
1370 struct rtattr
*rta
, struct rtnexthop
*rtnh
,
1371 struct rtmsg
*rtmsg
, union g_addr
**src
)
1376 family
= NHLFE_FAMILY(nhlfe
);
1377 bytelen
= (family
== AF_INET
? 4 : 16);
1378 _netlink_route_build_multipath(routedesc
, bytelen
, nhlfe
->nexthop
, rta
,
1383 /* Log debug information for netlink_route_multipath
1384 * if debug logging is enabled.
1386 * @param cmd: Netlink command which is to be processed
1387 * @param p: Prefix for which the change is due
1388 * @param family: Address family which the change concerns
1389 * @param zvrf: The vrf we are in
1390 * @param tableid: The table we are working on
1392 static void _netlink_route_debug(int cmd
, const struct prefix
*p
,
1393 int family
, vrf_id_t vrfid
,
1396 if (IS_ZEBRA_DEBUG_KERNEL
) {
1397 char buf
[PREFIX_STRLEN
];
1399 "netlink_route_multipath(): %s %s vrf %u(%u)",
1400 nl_msg_type_to_str(cmd
),
1401 prefix2str(p
, buf
, sizeof(buf
)),
1406 static void _netlink_mpls_debug(int cmd
, uint32_t label
, const char *routedesc
)
1408 if (IS_ZEBRA_DEBUG_KERNEL
)
1409 zlog_debug("netlink_mpls_multipath() (%s): %s %u/20", routedesc
,
1410 nl_msg_type_to_str(cmd
), label
);
1413 static int netlink_neigh_update(int cmd
, int ifindex
, uint32_t addr
, char *lla
,
1414 int llalen
, ns_id_t ns_id
)
1422 struct zebra_ns
*zns
= zebra_ns_lookup(ns_id
);
1424 memset(&req
, 0, sizeof(req
));
1426 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1427 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
1428 req
.n
.nlmsg_type
= cmd
; // RTM_NEWNEIGH or RTM_DELNEIGH
1429 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1431 req
.ndm
.ndm_family
= AF_INET
;
1432 req
.ndm
.ndm_state
= NUD_PERMANENT
;
1433 req
.ndm
.ndm_ifindex
= ifindex
;
1434 req
.ndm
.ndm_type
= RTN_UNICAST
;
1436 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &addr
, 4);
1437 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, lla
, llalen
);
1439 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1443 /* Routing table change via netlink interface. */
1444 /* Update flag indicates whether this is a "replace" or not. */
1445 static int netlink_route_multipath(int cmd
, const struct prefix
*p
,
1446 const struct prefix
*src_p
,
1447 struct route_entry
*re
,
1451 struct sockaddr_nl snl
;
1452 struct nexthop
*nexthop
= NULL
;
1453 unsigned int nexthop_num
;
1454 int family
= PREFIX_FAMILY(p
);
1455 const char *routedesc
;
1462 char buf
[NL_PKT_BUF_SIZE
];
1465 struct zebra_ns
*zns
;
1466 struct zebra_vrf
*zvrf
= vrf_info_lookup(re
->vrf_id
);
1469 memset(&req
, 0, sizeof req
- NL_PKT_BUF_SIZE
);
1471 bytelen
= (family
== AF_INET
? 4 : 16);
1473 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
1474 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
1475 if ((cmd
== RTM_NEWROUTE
) && update
)
1476 req
.n
.nlmsg_flags
|= NLM_F_REPLACE
;
1477 req
.n
.nlmsg_type
= cmd
;
1478 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1480 req
.r
.rtm_family
= family
;
1481 req
.r
.rtm_dst_len
= p
->prefixlen
;
1482 req
.r
.rtm_src_len
= src_p
? src_p
->prefixlen
: 0;
1483 req
.r
.rtm_protocol
= zebra2proto(re
->type
);
1484 req
.r
.rtm_scope
= RT_SCOPE_UNIVERSE
;
1487 * blackhole routes are not RTN_UNICAST, they are
1488 * RTN_ BLACKHOLE|UNREACHABLE|PROHIBIT
1489 * so setting this value as a RTN_UNICAST would
1490 * cause the route lookup of just the prefix
1491 * to fail. So no need to specify this for
1492 * the RTM_DELROUTE case
1494 if (cmd
!= RTM_DELROUTE
)
1495 req
.r
.rtm_type
= RTN_UNICAST
;
1497 addattr_l(&req
.n
, sizeof req
, RTA_DST
, &p
->u
.prefix
, bytelen
);
1499 addattr_l(&req
.n
, sizeof req
, RTA_SRC
, &src_p
->u
.prefix
,
1503 /* Hardcode the metric for all routes coming from zebra. Metric isn't
1505 * either by the kernel or by zebra. Its purely for calculating best
1507 * by the routing protocol and for communicating with protocol peers.
1509 addattr32(&req
.n
, sizeof req
, RTA_PRIORITY
, NL_DEFAULT_ROUTE_METRIC
);
1510 #if defined(SUPPORT_REALMS)
1511 if (re
->tag
> 0 && re
->tag
<= 255)
1512 addattr32(&req
.n
, sizeof req
, RTA_FLOW
, re
->tag
);
1514 /* Table corresponding to this route. */
1515 if (re
->table
< 256)
1516 req
.r
.rtm_table
= re
->table
;
1518 req
.r
.rtm_table
= RT_TABLE_UNSPEC
;
1519 addattr32(&req
.n
, sizeof req
, RTA_TABLE
, re
->table
);
1522 _netlink_route_debug(cmd
, p
, family
, zvrf_id(zvrf
), re
->table
);
1525 * If we are not updating the route and we have received
1526 * a route delete, then all we need to fill in is the
1527 * prefix information to tell the kernel to schwack
1530 if (!update
&& cmd
== RTM_DELROUTE
)
1533 if (re
->mtu
|| re
->nexthop_mtu
) {
1534 char buf
[NL_PKT_BUF_SIZE
];
1535 struct rtattr
*rta
= (void *)buf
;
1536 uint32_t mtu
= re
->mtu
;
1537 if (!mtu
|| (re
->nexthop_mtu
&& re
->nexthop_mtu
< mtu
))
1538 mtu
= re
->nexthop_mtu
;
1539 rta
->rta_type
= RTA_METRICS
;
1540 rta
->rta_len
= RTA_LENGTH(0);
1541 rta_addattr_l(rta
, NL_PKT_BUF_SIZE
, RTAX_MTU
, &mtu
, sizeof mtu
);
1542 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_METRICS
, RTA_DATA(rta
),
1546 /* Count overall nexthops so we can decide whether to use singlepath
1547 * or multipath case. */
1549 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1550 if (CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_RECURSIVE
))
1552 if (cmd
== RTM_NEWROUTE
&& !NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1554 if (cmd
== RTM_DELROUTE
1555 && !CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
1561 /* Singlepath case. */
1562 if (nexthop_num
== 1 || multipath_num
== 1) {
1564 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1566 * So we want to cover 2 types of blackhole
1568 * 1) A normal blackhole route( ala from a static
1570 * 2) A recursively resolved blackhole route
1572 if (nexthop
->type
== NEXTHOP_TYPE_BLACKHOLE
) {
1573 switch (nexthop
->bh_type
) {
1574 case BLACKHOLE_ADMINPROHIB
:
1575 req
.r
.rtm_type
= RTN_PROHIBIT
;
1577 case BLACKHOLE_REJECT
:
1578 req
.r
.rtm_type
= RTN_UNREACHABLE
;
1581 req
.r
.rtm_type
= RTN_BLACKHOLE
;
1586 if (CHECK_FLAG(nexthop
->flags
,
1587 NEXTHOP_FLAG_RECURSIVE
)) {
1589 if (family
== AF_INET
) {
1590 if (nexthop
->rmap_src
.ipv4
1597 } else if (nexthop
->src
.ipv4
1605 } else if (family
== AF_INET6
) {
1606 if (!IN6_IS_ADDR_UNSPECIFIED(
1614 !IN6_IS_ADDR_UNSPECIFIED(
1627 if ((cmd
== RTM_NEWROUTE
1628 && NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1629 || (cmd
== RTM_DELROUTE
1630 && CHECK_FLAG(nexthop
->flags
,
1631 NEXTHOP_FLAG_FIB
))) {
1632 routedesc
= nexthop
->rparent
1633 ? "recursive, single-path"
1636 _netlink_route_build_singlepath(
1637 routedesc
, bytelen
, nexthop
, &req
.n
,
1638 &req
.r
, sizeof req
, cmd
);
1643 if (setsrc
&& (cmd
== RTM_NEWROUTE
)) {
1644 if (family
== AF_INET
)
1645 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1646 &src
.ipv4
, bytelen
);
1647 else if (family
== AF_INET6
)
1648 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1649 &src
.ipv6
, bytelen
);
1652 char buf
[NL_PKT_BUF_SIZE
];
1653 struct rtattr
*rta
= (void *)buf
;
1654 struct rtnexthop
*rtnh
;
1655 union g_addr
*src1
= NULL
;
1657 rta
->rta_type
= RTA_MULTIPATH
;
1658 rta
->rta_len
= RTA_LENGTH(0);
1659 rtnh
= RTA_DATA(rta
);
1662 for (ALL_NEXTHOPS(re
->ng
, nexthop
)) {
1663 if (nexthop_num
>= multipath_num
)
1666 if (CHECK_FLAG(nexthop
->flags
,
1667 NEXTHOP_FLAG_RECURSIVE
)) {
1668 /* This only works for IPv4 now */
1670 if (family
== AF_INET
) {
1671 if (nexthop
->rmap_src
.ipv4
1678 } else if (nexthop
->src
.ipv4
1686 } else if (family
== AF_INET6
) {
1687 if (!IN6_IS_ADDR_UNSPECIFIED(
1695 !IN6_IS_ADDR_UNSPECIFIED(
1708 if ((cmd
== RTM_NEWROUTE
1709 && NEXTHOP_IS_ACTIVE(nexthop
->flags
))
1710 || (cmd
== RTM_DELROUTE
1711 && CHECK_FLAG(nexthop
->flags
,
1712 NEXTHOP_FLAG_FIB
))) {
1713 routedesc
= nexthop
->rparent
1714 ? "recursive, multipath"
1718 _netlink_route_build_multipath(
1719 routedesc
, bytelen
, nexthop
, rta
, rtnh
,
1721 rtnh
= RTNH_NEXT(rtnh
);
1723 if (!setsrc
&& src1
) {
1724 if (family
== AF_INET
)
1725 src
.ipv4
= src1
->ipv4
;
1726 else if (family
== AF_INET6
)
1727 src
.ipv6
= src1
->ipv6
;
1733 if (setsrc
&& (cmd
== RTM_NEWROUTE
)) {
1734 if (family
== AF_INET
)
1735 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1736 &src
.ipv4
, bytelen
);
1737 else if (family
== AF_INET6
)
1738 addattr_l(&req
.n
, sizeof req
, RTA_PREFSRC
,
1739 &src
.ipv6
, bytelen
);
1740 if (IS_ZEBRA_DEBUG_KERNEL
)
1741 zlog_debug("Setting source");
1744 if (rta
->rta_len
> RTA_LENGTH(0))
1745 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_MULTIPATH
,
1746 RTA_DATA(rta
), RTA_PAYLOAD(rta
));
1749 /* If there is no useful nexthop then return. */
1750 if (nexthop_num
== 0) {
1751 if (IS_ZEBRA_DEBUG_KERNEL
)
1753 "netlink_route_multipath(): No useful nexthop.");
1759 /* Destination netlink address. */
1760 memset(&snl
, 0, sizeof snl
);
1761 snl
.nl_family
= AF_NETLINK
;
1763 /* Talk to netlink socket. */
1764 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1768 int kernel_get_ipmr_sg_stats(struct zebra_vrf
*zvrf
, void *in
)
1771 struct mcast_route_data
*mr
= (struct mcast_route_data
*)in
;
1779 struct zebra_ns
*zns
;
1782 memset(&req
, 0, sizeof(req
));
1784 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1785 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
1786 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
1788 req
.ndm
.ndm_family
= RTNL_FAMILY_IPMR
;
1789 req
.n
.nlmsg_type
= RTM_GETROUTE
;
1791 addattr_l(&req
.n
, sizeof(req
), RTA_IIF
, &mroute
->ifindex
, 4);
1792 addattr_l(&req
.n
, sizeof(req
), RTA_OIF
, &mroute
->ifindex
, 4);
1793 addattr_l(&req
.n
, sizeof(req
), RTA_SRC
, &mroute
->sg
.src
.s_addr
, 4);
1794 addattr_l(&req
.n
, sizeof(req
), RTA_DST
, &mroute
->sg
.grp
.s_addr
, 4);
1795 addattr_l(&req
.n
, sizeof(req
), RTA_TABLE
, &zvrf
->table_id
, 4);
1797 suc
= netlink_talk(netlink_route_change_read_multicast
, &req
.n
,
1798 &zns
->netlink_cmd
, zns
, 0);
1804 enum dp_req_result
kernel_route_rib(struct route_node
*rn
,
1805 const struct prefix
*p
,
1806 const struct prefix
*src_p
,
1807 struct route_entry
*old
,
1808 struct route_entry
*new)
1815 if (p
->family
== AF_INET
|| v6_rr_semantics
)
1816 ret
= netlink_route_multipath(RTM_NEWROUTE
, p
, src_p
,
1817 new, (old
) ? 1 : 0);
1820 * So v6 route replace semantics are not in
1821 * the kernel at this point as I understand it.
1822 * So let's do a delete than an add.
1823 * In the future once v6 route replace semantics
1824 * are in we can figure out what to do here to
1825 * allow working with old and new kernels.
1827 * I'm also intentionally ignoring the failure case
1828 * of the route delete. If that happens yeah we're
1832 netlink_route_multipath(RTM_DELROUTE
, p
, src_p
,
1834 ret
= netlink_route_multipath(RTM_NEWROUTE
, p
, src_p
,
1837 kernel_route_rib_pass_fail(rn
, p
, new,
1838 (!ret
) ? DP_INSTALL_SUCCESS
1839 : DP_INSTALL_FAILURE
);
1840 return DP_REQUEST_SUCCESS
;
1844 ret
= netlink_route_multipath(RTM_DELROUTE
, p
, src_p
, old
, 0);
1846 kernel_route_rib_pass_fail(rn
, p
, old
,
1847 (!ret
) ? DP_DELETE_SUCCESS
1848 : DP_DELETE_FAILURE
);
1851 return DP_REQUEST_SUCCESS
;
1854 int kernel_neigh_update(int add
, int ifindex
, uint32_t addr
, char *lla
,
1855 int llalen
, ns_id_t ns_id
)
1857 return netlink_neigh_update(add
? RTM_NEWNEIGH
: RTM_DELNEIGH
, ifindex
,
1858 addr
, lla
, llalen
, ns_id
);
1862 * Add remote VTEP to the flood list for this VxLAN interface (VNI). This
1863 * is done by adding an FDB entry with a MAC of 00:00:00:00:00:00.
1865 static int netlink_vxlan_flood_list_update(struct interface
*ifp
,
1866 struct in_addr
*vtep_ip
, int cmd
)
1868 struct zebra_ns
*zns
;
1874 uint8_t dst_mac
[6] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0};
1875 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
1878 memset(&req
, 0, sizeof(req
));
1880 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
1881 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
1882 if (cmd
== RTM_NEWNEIGH
)
1883 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_APPEND
);
1884 req
.n
.nlmsg_type
= cmd
;
1885 req
.ndm
.ndm_family
= PF_BRIDGE
;
1886 req
.ndm
.ndm_state
= NUD_NOARP
| NUD_PERMANENT
;
1887 req
.ndm
.ndm_flags
|= NTF_SELF
; // Handle by "self", not "master"
1890 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, &dst_mac
, 6);
1891 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
1892 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &vtep_ip
->s_addr
, 4);
1894 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
1899 * Add remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1901 * a "flood" MAC FDB entry.
1903 int kernel_add_vtep(vni_t vni
, struct interface
*ifp
, struct in_addr
*vtep_ip
)
1905 if (IS_ZEBRA_DEBUG_VXLAN
)
1906 zlog_debug("Install %s into flood list for VNI %u intf %s(%u)",
1907 inet_ntoa(*vtep_ip
), vni
, ifp
->name
, ifp
->ifindex
);
1909 return netlink_vxlan_flood_list_update(ifp
, vtep_ip
, RTM_NEWNEIGH
);
1913 * Remove remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1914 * deleting the "flood" MAC FDB entry.
1916 int kernel_del_vtep(vni_t vni
, struct interface
*ifp
, struct in_addr
*vtep_ip
)
1918 if (IS_ZEBRA_DEBUG_VXLAN
)
1920 "Uninstall %s from flood list for VNI %u intf %s(%u)",
1921 inet_ntoa(*vtep_ip
), vni
, ifp
->name
, ifp
->ifindex
);
1923 return netlink_vxlan_flood_list_update(ifp
, vtep_ip
, RTM_DELNEIGH
);
1927 #define NDA_RTA(r) \
1928 ((struct rtattr *)(((char *)(r)) + NLMSG_ALIGN(sizeof(struct ndmsg))))
1931 static int netlink_macfdb_change(struct nlmsghdr
*h
, int len
, ns_id_t ns_id
)
1934 struct interface
*ifp
;
1935 struct zebra_if
*zif
;
1936 struct rtattr
*tb
[NDA_MAX
+ 1];
1937 struct interface
*br_if
;
1940 struct prefix vtep_ip
;
1941 int vid_present
= 0, dst_present
= 0;
1942 char buf
[ETHER_ADDR_STRLEN
];
1947 ndm
= NLMSG_DATA(h
);
1949 /* We only process macfdb notifications if EVPN is enabled */
1950 if (!is_evpn_enabled())
1953 /* The interface should exist. */
1954 ifp
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
1956 if (!ifp
|| !ifp
->info
)
1959 /* The interface should be something we're interested in. */
1960 if (!IS_ZEBRA_IF_BRIDGE_SLAVE(ifp
))
1963 /* Drop "permanent" entries. */
1964 if (ndm
->ndm_state
& NUD_PERMANENT
)
1967 zif
= (struct zebra_if
*)ifp
->info
;
1968 if ((br_if
= zif
->brslave_info
.br_if
) == NULL
) {
1969 zlog_debug("%s family %s IF %s(%u) brIF %u - no bridge master",
1970 nl_msg_type_to_str(h
->nlmsg_type
),
1971 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1972 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
);
1976 /* Parse attributes and extract fields of interest. */
1977 memset(tb
, 0, sizeof tb
);
1978 netlink_parse_rtattr(tb
, NDA_MAX
, NDA_RTA(ndm
), len
);
1980 if (!tb
[NDA_LLADDR
]) {
1981 zlog_debug("%s family %s IF %s(%u) brIF %u - no LLADDR",
1982 nl_msg_type_to_str(h
->nlmsg_type
),
1983 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1984 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
);
1988 if (RTA_PAYLOAD(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
1990 "%s family %s IF %s(%u) brIF %u - LLADDR is not MAC, len %lu",
1991 nl_msg_type_to_str(h
->nlmsg_type
),
1992 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
1993 ndm
->ndm_ifindex
, zif
->brslave_info
.bridge_ifindex
,
1994 (unsigned long)RTA_PAYLOAD(tb
[NDA_LLADDR
]));
1998 memcpy(&mac
, RTA_DATA(tb
[NDA_LLADDR
]), ETH_ALEN
);
2000 if ((NDA_VLAN
<= NDA_MAX
) && tb
[NDA_VLAN
]) {
2002 vid
= *(uint16_t *)RTA_DATA(tb
[NDA_VLAN
]);
2003 sprintf(vid_buf
, " VLAN %u", vid
);
2007 /* TODO: Only IPv4 supported now. */
2009 vtep_ip
.family
= AF_INET
;
2010 vtep_ip
.prefixlen
= IPV4_MAX_BITLEN
;
2011 memcpy(&(vtep_ip
.u
.prefix4
.s_addr
), RTA_DATA(tb
[NDA_DST
]),
2013 sprintf(dst_buf
, " dst %s", inet_ntoa(vtep_ip
.u
.prefix4
));
2016 sticky
= (ndm
->ndm_state
& NUD_NOARP
) ? 1 : 0;
2018 if (IS_ZEBRA_DEBUG_KERNEL
)
2019 zlog_debug("Rx %s family %s IF %s(%u)%s %sMAC %s%s",
2020 nl_msg_type_to_str(h
->nlmsg_type
),
2021 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2022 ndm
->ndm_ifindex
, vid_present
? vid_buf
: "",
2023 sticky
? "sticky " : "",
2024 prefix_mac2str(&mac
, buf
, sizeof(buf
)),
2025 dst_present
? dst_buf
: "");
2027 if (filter_vlan
&& vid
!= filter_vlan
)
2030 /* If add or update, do accordingly if learnt on a "local" interface; if
2031 * the notification is over VxLAN, this has to be related to
2033 * so perform an implicit delete of any local entry (if it exists).
2035 if (h
->nlmsg_type
== RTM_NEWNEIGH
) {
2036 /* Drop "permanent" entries. */
2037 if (ndm
->ndm_state
& NUD_PERMANENT
)
2040 if (IS_ZEBRA_IF_VXLAN(ifp
))
2041 return zebra_vxlan_check_del_local_mac(ifp
, br_if
, &mac
,
2044 return zebra_vxlan_local_mac_add_update(ifp
, br_if
, &mac
, vid
,
2048 /* This is a delete notification.
2049 * 1. For a MAC over VxLan, check if it needs to be refreshed(readded)
2050 * 2. For a MAC over "local" interface, delete the mac
2051 * Note: We will get notifications from both bridge driver and VxLAN
2053 * Ignore the notification from VxLan driver as it is also generated
2054 * when mac moves from remote to local.
2059 if (IS_ZEBRA_IF_VXLAN(ifp
))
2060 return zebra_vxlan_check_readd_remote_mac(ifp
, br_if
, &mac
,
2063 return zebra_vxlan_local_mac_del(ifp
, br_if
, &mac
, vid
);
2066 static int netlink_macfdb_table(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
2071 if (h
->nlmsg_type
!= RTM_NEWNEIGH
)
2074 /* Length validity. */
2075 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2079 /* We are interested only in AF_BRIDGE notifications. */
2080 ndm
= NLMSG_DATA(h
);
2081 if (ndm
->ndm_family
!= AF_BRIDGE
)
2084 return netlink_macfdb_change(h
, len
, ns_id
);
2087 /* Request for MAC FDB information from the kernel */
2088 static int netlink_request_macs(struct zebra_ns
*zns
, int family
, int type
,
2089 ifindex_t master_ifindex
)
2093 struct ifinfomsg ifm
;
2097 /* Form the request, specifying filter (rtattr) if needed. */
2098 memset(&req
, 0, sizeof(req
));
2099 req
.n
.nlmsg_type
= type
;
2100 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ifinfomsg
));
2101 req
.ifm
.ifi_family
= family
;
2103 addattr32(&req
.n
, sizeof(req
), IFLA_MASTER
, master_ifindex
);
2105 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
2109 * MAC forwarding database read using netlink interface. This is invoked
2112 int netlink_macfdb_read(struct zebra_ns
*zns
)
2116 /* Get bridge FDB table. */
2117 ret
= netlink_request_macs(zns
, AF_BRIDGE
, RTM_GETNEIGH
, 0);
2120 /* We are reading entire table. */
2122 ret
= netlink_parse_info(netlink_macfdb_table
, &zns
->netlink_cmd
, zns
,
2129 * MAC forwarding database read using netlink interface. This is for a
2130 * specific bridge and matching specific access VLAN (if VLAN-aware bridge).
2132 int netlink_macfdb_read_for_bridge(struct zebra_ns
*zns
, struct interface
*ifp
,
2133 struct interface
*br_if
)
2135 struct zebra_if
*br_zif
;
2136 struct zebra_if
*zif
;
2137 struct zebra_l2info_vxlan
*vxl
;
2141 /* Save VLAN we're filtering on, if needed. */
2142 br_zif
= (struct zebra_if
*)br_if
->info
;
2143 zif
= (struct zebra_if
*)ifp
->info
;
2144 vxl
= &zif
->l2info
.vxl
;
2145 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif
))
2146 filter_vlan
= vxl
->access_vlan
;
2148 /* Get bridge FDB table for specific bridge - we do the VLAN filtering.
2150 ret
= netlink_request_macs(zns
, AF_BRIDGE
, RTM_GETNEIGH
,
2154 ret
= netlink_parse_info(netlink_macfdb_table
, &zns
->netlink_cmd
, zns
,
2157 /* Reset VLAN filter. */
2162 static int netlink_macfdb_update(struct interface
*ifp
, vlanid_t vid
,
2163 struct ethaddr
*mac
, struct in_addr vtep_ip
,
2164 int local
, int cmd
, uint8_t sticky
)
2166 struct zebra_ns
*zns
;
2173 struct zebra_if
*zif
;
2174 struct interface
*br_if
;
2175 struct zebra_if
*br_zif
;
2176 char buf
[ETHER_ADDR_STRLEN
];
2177 int vid_present
= 0, dst_present
= 0;
2180 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
2184 if ((br_if
= zif
->brslave_info
.br_if
) == NULL
) {
2185 zlog_debug("MAC %s on IF %s(%u) - no mapping to bridge",
2186 (cmd
== RTM_NEWNEIGH
) ? "add" : "del", ifp
->name
,
2191 memset(&req
, 0, sizeof(req
));
2193 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2194 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
2195 if (cmd
== RTM_NEWNEIGH
)
2196 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_REPLACE
);
2197 req
.n
.nlmsg_type
= cmd
;
2198 req
.ndm
.ndm_family
= AF_BRIDGE
;
2199 req
.ndm
.ndm_flags
|= NTF_SELF
| NTF_MASTER
;
2200 req
.ndm
.ndm_state
= NUD_REACHABLE
;
2203 req
.ndm
.ndm_state
|= NUD_NOARP
;
2205 req
.ndm
.ndm_flags
|= NTF_EXT_LEARNED
;
2207 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, mac
, 6);
2208 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
2210 dst_alen
= 4; // TODO: hardcoded
2211 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &vtep_ip
, dst_alen
);
2213 sprintf(dst_buf
, " dst %s", inet_ntoa(vtep_ip
));
2215 br_zif
= (struct zebra_if
*)br_if
->info
;
2216 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif
) && vid
> 0) {
2217 addattr16(&req
.n
, sizeof(req
), NDA_VLAN
, vid
);
2219 sprintf(vid_buf
, " VLAN %u", vid
);
2221 addattr32(&req
.n
, sizeof(req
), NDA_MASTER
, br_if
->ifindex
);
2223 if (IS_ZEBRA_DEBUG_KERNEL
)
2224 zlog_debug("Tx %s family %s IF %s(%u)%s %sMAC %s%s",
2225 nl_msg_type_to_str(cmd
),
2226 nl_family_to_str(req
.ndm
.ndm_family
), ifp
->name
,
2227 ifp
->ifindex
, vid_present
? vid_buf
: "",
2228 sticky
? "sticky " : "",
2229 prefix_mac2str(mac
, buf
, sizeof(buf
)),
2230 dst_present
? dst_buf
: "");
2232 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2237 (NUD_PERMANENT | NUD_NOARP | NUD_REACHABLE | NUD_PROBE | NUD_STALE \
2240 static int netlink_ipneigh_change(struct nlmsghdr
*h
, int len
, ns_id_t ns_id
)
2243 struct interface
*ifp
;
2244 struct zebra_if
*zif
;
2245 struct rtattr
*tb
[NDA_MAX
+ 1];
2246 struct interface
*link_if
;
2249 char buf
[ETHER_ADDR_STRLEN
];
2250 char buf2
[INET6_ADDRSTRLEN
];
2251 int mac_present
= 0;
2252 uint8_t ext_learned
;
2253 uint8_t router_flag
;
2255 ndm
= NLMSG_DATA(h
);
2257 /* The interface should exist. */
2258 ifp
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
2260 if (!ifp
|| !ifp
->info
)
2263 zif
= (struct zebra_if
*)ifp
->info
;
2265 /* Parse attributes and extract fields of interest. */
2266 memset(tb
, 0, sizeof tb
);
2267 netlink_parse_rtattr(tb
, NDA_MAX
, NDA_RTA(ndm
), len
);
2270 zlog_debug("%s family %s IF %s(%u) - no DST",
2271 nl_msg_type_to_str(h
->nlmsg_type
),
2272 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2277 memset(&ip
, 0, sizeof(struct ipaddr
));
2278 ip
.ipa_type
= (ndm
->ndm_family
== AF_INET
) ? IPADDR_V4
: IPADDR_V6
;
2279 memcpy(&ip
.ip
.addr
, RTA_DATA(tb
[NDA_DST
]), RTA_PAYLOAD(tb
[NDA_DST
]));
2281 /* Drop some "permanent" entries. */
2282 if (ndm
->ndm_state
& NUD_PERMANENT
) {
2283 char b
[16] = "169.254.0.1";
2284 struct in_addr ipv4_ll
;
2286 if (ndm
->ndm_family
!= AF_INET
)
2289 if (!zif
->v6_2_v4_ll_neigh_entry
)
2292 if (h
->nlmsg_type
!= RTM_DELNEIGH
)
2295 inet_pton(AF_INET
, b
, &ipv4_ll
);
2296 if (ipv4_ll
.s_addr
!= ip
.ip
._v4_addr
.s_addr
)
2299 if_nbr_ipv6ll_to_ipv4ll_neigh_update(
2300 ifp
, &zif
->v6_2_v4_ll_addr6
, true);
2304 /* The neighbor is present on an SVI. From this, we locate the
2306 * bridge because we're only interested in neighbors on a VxLAN bridge.
2307 * The bridge is located based on the nature of the SVI:
2308 * (a) In the case of a VLAN-aware bridge, the SVI is a L3 VLAN
2310 * and is linked to the bridge
2311 * (b) In the case of a VLAN-unaware bridge, the SVI is the bridge
2315 if (IS_ZEBRA_IF_VLAN(ifp
)) {
2316 link_if
= if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id
),
2320 } else if (IS_ZEBRA_IF_BRIDGE(ifp
))
2325 memset(&mac
, 0, sizeof(struct ethaddr
));
2326 if (h
->nlmsg_type
== RTM_NEWNEIGH
) {
2327 if (tb
[NDA_LLADDR
]) {
2328 if (RTA_PAYLOAD(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
2330 "%s family %s IF %s(%u) - LLADDR is not MAC, len %lu",
2331 nl_msg_type_to_str(h
->nlmsg_type
),
2332 nl_family_to_str(ndm
->ndm_family
),
2333 ifp
->name
, ndm
->ndm_ifindex
,
2334 (unsigned long)RTA_PAYLOAD(
2340 memcpy(&mac
, RTA_DATA(tb
[NDA_LLADDR
]), ETH_ALEN
);
2343 ext_learned
= (ndm
->ndm_flags
& NTF_EXT_LEARNED
) ? 1 : 0;
2344 router_flag
= (ndm
->ndm_flags
& NTF_ROUTER
) ? 1 : 0;
2346 if (IS_ZEBRA_DEBUG_KERNEL
)
2348 "Rx %s family %s IF %s(%u) IP %s MAC %s state 0x%x flags 0x%x",
2349 nl_msg_type_to_str(h
->nlmsg_type
),
2350 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2352 ipaddr2str(&ip
, buf2
, sizeof(buf2
)),
2354 ? prefix_mac2str(&mac
, buf
, sizeof(buf
))
2356 ndm
->ndm_state
, ndm
->ndm_flags
);
2358 /* If the neighbor state is valid for use, process as an add or
2360 * else process as a delete. Note that the delete handling may
2362 * in re-adding the neighbor if it is a valid "remote" neighbor.
2364 if (ndm
->ndm_state
& NUD_VALID
)
2365 return zebra_vxlan_handle_kernel_neigh_update(
2366 ifp
, link_if
, &ip
, &mac
, ndm
->ndm_state
,
2367 ext_learned
, router_flag
);
2369 return zebra_vxlan_handle_kernel_neigh_del(ifp
, link_if
, &ip
);
2372 if (IS_ZEBRA_DEBUG_KERNEL
)
2373 zlog_debug("Rx %s family %s IF %s(%u) IP %s",
2374 nl_msg_type_to_str(h
->nlmsg_type
),
2375 nl_family_to_str(ndm
->ndm_family
), ifp
->name
,
2377 ipaddr2str(&ip
, buf2
, sizeof(buf2
)));
2379 /* Process the delete - it may result in re-adding the neighbor if it is
2380 * a valid "remote" neighbor.
2382 return zebra_vxlan_handle_kernel_neigh_del(ifp
, link_if
, &ip
);
2385 static int netlink_neigh_table(struct nlmsghdr
*h
, ns_id_t ns_id
, int startup
)
2390 if (h
->nlmsg_type
!= RTM_NEWNEIGH
)
2393 /* Length validity. */
2394 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2398 /* We are interested only in AF_INET or AF_INET6 notifications. */
2399 ndm
= NLMSG_DATA(h
);
2400 if (ndm
->ndm_family
!= AF_INET
&& ndm
->ndm_family
!= AF_INET6
)
2403 return netlink_neigh_change(h
, len
);
2406 /* Request for IP neighbor information from the kernel */
2407 static int netlink_request_neigh(struct zebra_ns
*zns
, int family
, int type
,
2416 /* Form the request, specifying filter (rtattr) if needed. */
2417 memset(&req
, 0, sizeof(req
));
2418 req
.n
.nlmsg_type
= type
;
2419 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2420 req
.ndm
.ndm_family
= family
;
2422 addattr32(&req
.n
, sizeof(req
), NDA_IFINDEX
, ifindex
);
2424 return netlink_request(&zns
->netlink_cmd
, &req
.n
);
2428 * IP Neighbor table read using netlink interface. This is invoked
2431 int netlink_neigh_read(struct zebra_ns
*zns
)
2435 /* Get IP neighbor table. */
2436 ret
= netlink_request_neigh(zns
, AF_UNSPEC
, RTM_GETNEIGH
, 0);
2439 ret
= netlink_parse_info(netlink_neigh_table
, &zns
->netlink_cmd
, zns
, 0,
2446 * IP Neighbor table read using netlink interface. This is for a specific
2449 int netlink_neigh_read_for_vlan(struct zebra_ns
*zns
, struct interface
*vlan_if
)
2453 ret
= netlink_request_neigh(zns
, AF_UNSPEC
, RTM_GETNEIGH
,
2457 ret
= netlink_parse_info(netlink_neigh_table
, &zns
->netlink_cmd
, zns
, 0,
2463 int netlink_neigh_change(struct nlmsghdr
*h
, ns_id_t ns_id
)
2468 if (!(h
->nlmsg_type
== RTM_NEWNEIGH
|| h
->nlmsg_type
== RTM_DELNEIGH
))
2471 /* Length validity. */
2472 len
= h
->nlmsg_len
- NLMSG_LENGTH(sizeof(struct ndmsg
));
2474 zlog_err("%s: Message received from netlink is of a broken size %d %zu",
2475 __PRETTY_FUNCTION__
, h
->nlmsg_len
,
2476 (size_t)NLMSG_LENGTH(sizeof(struct ndmsg
)));
2480 /* Is this a notification for the MAC FDB or IP neighbor table? */
2481 ndm
= NLMSG_DATA(h
);
2482 if (ndm
->ndm_family
== AF_BRIDGE
)
2483 return netlink_macfdb_change(h
, len
, ns_id
);
2485 if (ndm
->ndm_type
!= RTN_UNICAST
)
2488 if (ndm
->ndm_family
== AF_INET
|| ndm
->ndm_family
== AF_INET6
)
2489 return netlink_ipneigh_change(h
, len
, ns_id
);
2492 EC_ZEBRA_UNKNOWN_FAMILY
,
2493 "Invalid address family: %u received from kernel neighbor change: %s",
2494 ndm
->ndm_family
, nl_msg_type_to_str(h
->nlmsg_type
));
2501 static int netlink_neigh_update2(struct interface
*ifp
, struct ipaddr
*ip
,
2502 struct ethaddr
*mac
, uint8_t flags
,
2503 uint16_t state
, int cmd
)
2512 struct zebra_ns
*zns
;
2513 char buf
[INET6_ADDRSTRLEN
];
2514 char buf2
[ETHER_ADDR_STRLEN
];
2515 struct zebra_vrf
*zvrf
= zebra_vrf_lookup_by_id(ifp
->vrf_id
);
2518 memset(&req
, 0, sizeof(req
));
2520 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct ndmsg
));
2521 req
.n
.nlmsg_flags
= NLM_F_REQUEST
;
2522 if (cmd
== RTM_NEWNEIGH
)
2523 req
.n
.nlmsg_flags
|= (NLM_F_CREATE
| NLM_F_REPLACE
);
2524 req
.n
.nlmsg_type
= cmd
; // RTM_NEWNEIGH or RTM_DELNEIGH
2525 req
.ndm
.ndm_family
= IS_IPADDR_V4(ip
) ? AF_INET
: AF_INET6
;
2526 req
.ndm
.ndm_state
= state
;
2527 req
.ndm
.ndm_ifindex
= ifp
->ifindex
;
2528 req
.ndm
.ndm_type
= RTN_UNICAST
;
2529 req
.ndm
.ndm_flags
= flags
;
2531 ipa_len
= IS_IPADDR_V4(ip
) ? IPV4_MAX_BYTELEN
: IPV6_MAX_BYTELEN
;
2532 addattr_l(&req
.n
, sizeof(req
), NDA_DST
, &ip
->ip
.addr
, ipa_len
);
2534 addattr_l(&req
.n
, sizeof(req
), NDA_LLADDR
, mac
, 6);
2536 if (IS_ZEBRA_DEBUG_KERNEL
)
2537 zlog_debug("Tx %s family %s IF %s(%u) Neigh %s MAC %s flags 0x%x",
2538 nl_msg_type_to_str(cmd
),
2539 nl_family_to_str(req
.ndm
.ndm_family
), ifp
->name
,
2540 ifp
->ifindex
, ipaddr2str(ip
, buf
, sizeof(buf
)),
2541 mac
? prefix_mac2str(mac
, buf2
, sizeof(buf2
))
2544 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2548 int kernel_add_mac(struct interface
*ifp
, vlanid_t vid
, struct ethaddr
*mac
,
2549 struct in_addr vtep_ip
, uint8_t sticky
)
2551 return netlink_macfdb_update(ifp
, vid
, mac
, vtep_ip
, 0, RTM_NEWNEIGH
,
2555 int kernel_del_mac(struct interface
*ifp
, vlanid_t vid
, struct ethaddr
*mac
,
2556 struct in_addr vtep_ip
, int local
)
2558 return netlink_macfdb_update(ifp
, vid
, mac
, vtep_ip
, local
,
2562 int kernel_add_neigh(struct interface
*ifp
, struct ipaddr
*ip
,
2563 struct ethaddr
*mac
, uint8_t flags
)
2565 return netlink_neigh_update2(ifp
, ip
, mac
, flags
,
2566 NUD_NOARP
, RTM_NEWNEIGH
);
2569 int kernel_del_neigh(struct interface
*ifp
, struct ipaddr
*ip
)
2571 return netlink_neigh_update2(ifp
, ip
, NULL
, 0, 0, RTM_DELNEIGH
);
2575 * MPLS label forwarding table change via netlink interface.
2577 int netlink_mpls_multipath(int cmd
, zebra_lsp_t
*lsp
)
2580 zebra_nhlfe_t
*nhlfe
;
2581 struct nexthop
*nexthop
= NULL
;
2582 unsigned int nexthop_num
;
2583 const char *routedesc
;
2584 struct zebra_ns
*zns
= zebra_ns_lookup(NS_DEFAULT
);
2590 char buf
[NL_PKT_BUF_SIZE
];
2593 memset(&req
, 0, sizeof req
- NL_PKT_BUF_SIZE
);
2596 * Count # nexthops so we can decide whether to use singlepath
2597 * or multipath case.
2600 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2601 nexthop
= nhlfe
->nexthop
;
2604 if (cmd
== RTM_NEWROUTE
) {
2605 /* Count all selected NHLFEs */
2606 if (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2607 && CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_ACTIVE
))
2611 /* Count all installed NHLFEs */
2612 if (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_INSTALLED
)
2613 && CHECK_FLAG(nexthop
->flags
, NEXTHOP_FLAG_FIB
))
2618 if ((nexthop_num
== 0) || (!lsp
->best_nhlfe
&& (cmd
!= RTM_DELROUTE
)))
2621 req
.n
.nlmsg_len
= NLMSG_LENGTH(sizeof(struct rtmsg
));
2622 req
.n
.nlmsg_flags
= NLM_F_CREATE
| NLM_F_REQUEST
;
2623 req
.n
.nlmsg_type
= cmd
;
2624 req
.n
.nlmsg_pid
= zns
->netlink_cmd
.snl
.nl_pid
;
2626 req
.r
.rtm_family
= AF_MPLS
;
2627 req
.r
.rtm_table
= RT_TABLE_MAIN
;
2628 req
.r
.rtm_dst_len
= MPLS_LABEL_LEN_BITS
;
2629 req
.r
.rtm_scope
= RT_SCOPE_UNIVERSE
;
2630 req
.r
.rtm_type
= RTN_UNICAST
;
2632 if (cmd
== RTM_NEWROUTE
) {
2633 /* We do a replace to handle update. */
2634 req
.n
.nlmsg_flags
|= NLM_F_REPLACE
;
2636 /* set the protocol value if installing */
2637 route_type
= re_type_from_lsp_type(lsp
->best_nhlfe
->type
);
2638 req
.r
.rtm_protocol
= zebra2proto(route_type
);
2641 /* Fill destination */
2642 lse
= mpls_lse_encode(lsp
->ile
.in_label
, 0, 0, 1);
2643 addattr_l(&req
.n
, sizeof req
, RTA_DST
, &lse
, sizeof(mpls_lse_t
));
2645 /* Fill nexthops (paths) based on single-path or multipath. The paths
2646 * chosen depend on the operation.
2648 if (nexthop_num
== 1 || multipath_num
== 1) {
2649 routedesc
= "single-path";
2650 _netlink_mpls_debug(cmd
, lsp
->ile
.in_label
, routedesc
);
2653 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2654 nexthop
= nhlfe
->nexthop
;
2658 if ((cmd
== RTM_NEWROUTE
2659 && (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2660 && CHECK_FLAG(nexthop
->flags
,
2661 NEXTHOP_FLAG_ACTIVE
)))
2662 || (cmd
== RTM_DELROUTE
2663 && (CHECK_FLAG(nhlfe
->flags
,
2664 NHLFE_FLAG_INSTALLED
)
2665 && CHECK_FLAG(nexthop
->flags
,
2666 NEXTHOP_FLAG_FIB
)))) {
2667 /* Add the gateway */
2668 _netlink_mpls_build_singlepath(routedesc
, nhlfe
,
2675 } else /* Multipath case */
2677 char buf
[NL_PKT_BUF_SIZE
];
2678 struct rtattr
*rta
= (void *)buf
;
2679 struct rtnexthop
*rtnh
;
2680 union g_addr
*src1
= NULL
;
2682 rta
->rta_type
= RTA_MULTIPATH
;
2683 rta
->rta_len
= RTA_LENGTH(0);
2684 rtnh
= RTA_DATA(rta
);
2686 routedesc
= "multipath";
2687 _netlink_mpls_debug(cmd
, lsp
->ile
.in_label
, routedesc
);
2690 for (nhlfe
= lsp
->nhlfe_list
; nhlfe
; nhlfe
= nhlfe
->next
) {
2691 nexthop
= nhlfe
->nexthop
;
2695 if (nexthop_num
>= multipath_num
)
2698 if ((cmd
== RTM_NEWROUTE
2699 && (CHECK_FLAG(nhlfe
->flags
, NHLFE_FLAG_SELECTED
)
2700 && CHECK_FLAG(nexthop
->flags
,
2701 NEXTHOP_FLAG_ACTIVE
)))
2702 || (cmd
== RTM_DELROUTE
2703 && (CHECK_FLAG(nhlfe
->flags
,
2704 NHLFE_FLAG_INSTALLED
)
2705 && CHECK_FLAG(nexthop
->flags
,
2706 NEXTHOP_FLAG_FIB
)))) {
2709 /* Build the multipath */
2710 _netlink_mpls_build_multipath(routedesc
, nhlfe
,
2713 rtnh
= RTNH_NEXT(rtnh
);
2717 /* Add the multipath */
2718 if (rta
->rta_len
> RTA_LENGTH(0))
2719 addattr_l(&req
.n
, NL_PKT_BUF_SIZE
, RTA_MULTIPATH
,
2720 RTA_DATA(rta
), RTA_PAYLOAD(rta
));
2723 /* Talk to netlink socket. */
2724 return netlink_talk(netlink_talk_filter
, &req
.n
, &zns
->netlink_cmd
, zns
,
2727 #endif /* HAVE_NETLINK */