2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1988, 1991, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)rtsock.c 8.7 (Berkeley) 10/12/95
35 #include "opt_route.h"
37 #include "opt_inet6.h"
39 #include <sys/param.h>
41 #include <sys/kernel.h>
42 #include <sys/domain.h>
44 #include <sys/malloc.h>
48 #include <sys/protosw.h>
49 #include <sys/rmlock.h>
50 #include <sys/rwlock.h>
51 #include <sys/signalvar.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/sysctl.h>
55 #include <sys/systm.h>
58 #include <net/if_var.h>
59 #include <net/if_dl.h>
60 #include <net/if_llatbl.h>
61 #include <net/if_types.h>
62 #include <net/netisr.h>
63 #include <net/raw_cb.h>
64 #include <net/route.h>
65 #include <net/route/route_ctl.h>
66 #include <net/route/route_var.h>
69 #include <netinet/in.h>
70 #include <netinet/if_ether.h>
71 #include <netinet/ip_carp.h>
73 #include <netinet6/ip6_var.h>
74 #include <netinet6/scope6_var.h>
76 #include <net/route/nhop.h>
78 #ifdef COMPAT_FREEBSD32
79 #include <sys/mount.h>
80 #include <compat/freebsd32/freebsd32.h>
90 struct if_data ifm_data;
100 uint16_t _ifm_spare1;
102 uint16_t ifm_data_off;
103 uint32_t _ifm_spare2;
104 struct if_data ifm_data;
107 struct ifa_msghdrl32 {
108 uint16_t ifam_msglen;
109 uint8_t ifam_version;
114 uint16_t _ifam_spare1;
116 uint16_t ifam_data_off;
118 struct if_data ifam_data;
121 #define SA_SIZE32(sa) \
122 ( (((struct sockaddr *)(sa))->sa_len == 0) ? \
124 1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
126 #endif /* COMPAT_FREEBSD32 */
128 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
130 /* NB: these are not modified */
131 static struct sockaddr route_src = { 2, PF_ROUTE, };
132 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, };
134 /* These are external hooks for CARP. */
135 int (*carp_get_vhid_p)(struct ifaddr *);
138 * Used by rtsock/raw_input callback code to decide whether to filter the update
139 * notification to a socket bound to a particular FIB.
141 #define RTS_FILTER_FIB M_PROTO8
144 int ip_count; /* attached w/ AF_INET */
145 int ip6_count; /* attached w/ AF_INET6 */
146 int any_count; /* total attached */
148 VNET_DEFINE_STATIC(route_cb_t, route_cb);
149 #define V_route_cb VNET(route_cb)
151 struct mtx rtsock_mtx;
152 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
154 #define RTSOCK_LOCK() mtx_lock(&rtsock_mtx)
155 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
156 #define RTSOCK_LOCK_ASSERT() mtx_assert(&rtsock_mtx, MA_OWNED)
158 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
165 struct sysctl_req *w_req;
166 struct sockaddr *dst;
167 struct sockaddr *mask;
170 static void rts_input(struct mbuf *m);
171 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
172 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
173 struct walkarg *w, int *plen);
174 static int rt_xaddrs(caddr_t cp, caddr_t cplim,
175 struct rt_addrinfo *rtinfo);
176 static int sysctl_dumpentry(struct rtentry *rt, void *vw);
177 static int sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
178 uint32_t weight, struct walkarg *w);
179 static int sysctl_iflist(int af, struct walkarg *w);
180 static int sysctl_ifmalist(int af, struct walkarg *w);
181 static int route_output(struct mbuf *m, struct socket *so, ...);
182 static void rt_getmetrics(const struct rtentry *rt,
183 const struct nhop_object *nh, struct rt_metrics *out);
184 static void rt_dispatch(struct mbuf *, sa_family_t);
185 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
186 struct rt_msghdr *rtm, struct rib_cmd_info *rc);
187 static int update_rtm_from_rc(struct rt_addrinfo *info,
188 struct rt_msghdr **prtm, int alloc_len,
189 struct rib_cmd_info *rc, struct nhop_object *nh);
190 static void send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
191 struct mbuf *m, sa_family_t saf, u_int fibnum,
193 static bool can_export_rte(struct ucred *td_ucred, bool rt_is_host,
194 const struct sockaddr *rt_dst);
196 static struct netisr_handler rtsock_nh = {
198 .nh_handler = rts_input,
199 .nh_proto = NETISR_ROUTE,
200 .nh_policy = NETISR_POLICY_SOURCE,
204 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
208 netisr_getqlimit(&rtsock_nh, &qlimit);
209 error = sysctl_handle_int(oidp, &qlimit, 0, req);
210 if (error || !req->newptr)
214 return (netisr_setqlimit(&rtsock_nh, qlimit));
216 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
217 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
218 0, 0, sysctl_route_netisr_maxqlen, "I",
219 "maximum routing socket dispatch queue length");
226 if (IS_DEFAULT_VNET(curvnet)) {
227 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
228 rtsock_nh.nh_qlimit = tmp;
229 netisr_register(&rtsock_nh);
233 netisr_register_vnet(&rtsock_nh);
236 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
241 vnet_rts_uninit(void)
244 netisr_unregister_vnet(&rtsock_nh);
246 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
251 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
256 KASSERT(m != NULL, ("%s: m is NULL", __func__));
257 KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
258 KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
260 /* No filtering requested. */
261 if ((m->m_flags & RTS_FILTER_FIB) == 0)
264 /* Check if it is a rts and the fib matches the one of the socket. */
265 fibnum = M_GETFIB(m);
266 if (proto->sp_family != PF_ROUTE ||
267 rp->rcb_socket == NULL ||
268 rp->rcb_socket->so_fibnum == fibnum)
271 /* Filtering requested and no match, the socket shall be skipped. */
276 rts_input(struct mbuf *m)
278 struct sockproto route_proto;
279 unsigned short *family;
282 route_proto.sp_family = PF_ROUTE;
283 tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
285 family = (unsigned short *)(tag + 1);
286 route_proto.sp_protocol = *family;
287 m_tag_delete(m, tag);
289 route_proto.sp_protocol = 0;
291 raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
295 * It really doesn't make any sense at all for this code to share much
296 * with raw_usrreq.c, since its functionality is so restricted. XXX
299 rts_abort(struct socket *so)
302 raw_usrreqs.pru_abort(so);
306 rts_close(struct socket *so)
309 raw_usrreqs.pru_close(so);
312 /* pru_accept is EOPNOTSUPP */
315 rts_attach(struct socket *so, int proto, struct thread *td)
320 KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
323 rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
325 so->so_pcb = (caddr_t)rp;
326 so->so_fibnum = td->td_proc->p_fibnum;
327 error = raw_attach(so, proto);
335 switch(rp->rcb_proto.sp_protocol) {
337 V_route_cb.ip_count++;
340 V_route_cb.ip6_count++;
343 V_route_cb.any_count++;
346 so->so_options |= SO_USELOOPBACK;
351 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
354 return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
358 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
361 return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
364 /* pru_connect2 is EOPNOTSUPP */
365 /* pru_control is EOPNOTSUPP */
368 rts_detach(struct socket *so)
370 struct rawcb *rp = sotorawcb(so);
372 KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
375 switch(rp->rcb_proto.sp_protocol) {
377 V_route_cb.ip_count--;
380 V_route_cb.ip6_count--;
383 V_route_cb.any_count--;
385 raw_usrreqs.pru_detach(so);
389 rts_disconnect(struct socket *so)
392 return (raw_usrreqs.pru_disconnect(so));
395 /* pru_listen is EOPNOTSUPP */
398 rts_peeraddr(struct socket *so, struct sockaddr **nam)
401 return (raw_usrreqs.pru_peeraddr(so, nam));
404 /* pru_rcvd is EOPNOTSUPP */
405 /* pru_rcvoob is EOPNOTSUPP */
408 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
409 struct mbuf *control, struct thread *td)
412 return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
415 /* pru_sense is null */
418 rts_shutdown(struct socket *so)
421 return (raw_usrreqs.pru_shutdown(so));
425 rts_sockaddr(struct socket *so, struct sockaddr **nam)
428 return (raw_usrreqs.pru_sockaddr(so, nam));
431 static struct pr_usrreqs route_usrreqs = {
432 .pru_abort = rts_abort,
433 .pru_attach = rts_attach,
434 .pru_bind = rts_bind,
435 .pru_connect = rts_connect,
436 .pru_detach = rts_detach,
437 .pru_disconnect = rts_disconnect,
438 .pru_peeraddr = rts_peeraddr,
439 .pru_send = rts_send,
440 .pru_shutdown = rts_shutdown,
441 .pru_sockaddr = rts_sockaddr,
442 .pru_close = rts_close,
445 #ifndef _SOCKADDR_UNION_DEFINED
446 #define _SOCKADDR_UNION_DEFINED
448 * The union of all possible address formats we handle.
450 union sockaddr_union {
452 struct sockaddr_in sin;
453 struct sockaddr_in6 sin6;
455 #endif /* _SOCKADDR_UNION_DEFINED */
458 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
459 struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
461 #if defined(INET) || defined(INET6)
462 struct epoch_tracker et;
465 /* First, see if the returned address is part of the jail. */
466 if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
467 info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
471 switch (info->rti_info[RTAX_DST]->sa_family) {
481 * Try to find an address on the given outgoing interface
482 * that belongs to the jail.
485 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
488 if (sa->sa_family != AF_INET)
490 ia = ((struct sockaddr_in *)sa)->sin_addr;
491 if (prison_check_ip4(cred, &ia) == 0) {
499 * As a last resort return the 'default' jail address.
501 ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
503 if (prison_get_ip4(cred, &ia) != 0)
506 bzero(&saun->sin, sizeof(struct sockaddr_in));
507 saun->sin.sin_len = sizeof(struct sockaddr_in);
508 saun->sin.sin_family = AF_INET;
509 saun->sin.sin_addr.s_addr = ia.s_addr;
510 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
523 * Try to find an address on the given outgoing interface
524 * that belongs to the jail.
527 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
530 if (sa->sa_family != AF_INET6)
532 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
533 &ia6, sizeof(struct in6_addr));
534 if (prison_check_ip6(cred, &ia6) == 0) {
542 * As a last resort return the 'default' jail address.
544 ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
546 if (prison_get_ip6(cred, &ia6) != 0)
549 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
550 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
551 saun->sin6.sin6_family = AF_INET6;
552 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
553 if (sa6_recoverscope(&saun->sin6) != 0)
555 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
566 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
571 if (V_loif == NULL) {
572 printf("Unable to add blackhole/reject nhop without loopback");
575 info->rti_ifp = V_loif;
577 saf = info->rti_info[RTAX_DST]->sa_family;
579 CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
580 if (ifa->ifa_addr->sa_family == saf) {
585 if (info->rti_ifa == NULL)
588 bzero(saun, sizeof(union sockaddr_union));
592 saun->sin.sin_family = AF_INET;
593 saun->sin.sin_len = sizeof(struct sockaddr_in);
594 saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
599 saun->sin6.sin6_family = AF_INET6;
600 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
601 saun->sin6.sin6_addr = in6addr_loopback;
607 info->rti_info[RTAX_GATEWAY] = &saun->sa;
608 info->rti_flags |= RTF_GATEWAY;
614 * Fills in @info based on userland-provided @rtm message.
616 * Returns 0 on success.
619 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info)
624 rtm->rtm_pid = curproc->p_pid;
625 info->rti_addrs = rtm->rtm_addrs;
627 info->rti_mflags = rtm->rtm_inits;
628 info->rti_rmx = &rtm->rtm_rmx;
631 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
632 * link-local address because rtrequest requires addresses with
635 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
638 info->rti_flags = rtm->rtm_flags;
639 if (info->rti_info[RTAX_DST] == NULL ||
640 info->rti_info[RTAX_DST]->sa_family >= AF_MAX ||
641 (info->rti_info[RTAX_GATEWAY] != NULL &&
642 info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
644 saf = info->rti_info[RTAX_DST]->sa_family;
646 * Verify that the caller has the appropriate privilege; RTM_GET
647 * is the only operation the non-superuser is allowed.
649 if (rtm->rtm_type != RTM_GET) {
650 error = priv_check(curthread, PRIV_NET_ROUTE);
656 * The given gateway address may be an interface address.
657 * For example, issuing a "route change" command on a route
658 * entry that was created from a tunnel, and the gateway
659 * address given is the local end point. In this case the
660 * RTF_GATEWAY flag must be cleared or the destination will
661 * not be reachable even though there is no error message.
663 if (info->rti_info[RTAX_GATEWAY] != NULL &&
664 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
665 struct rt_addrinfo ginfo;
666 struct sockaddr *gdst;
667 struct sockaddr_storage ss;
669 bzero(&ginfo, sizeof(ginfo));
670 bzero(&ss, sizeof(ss));
671 ss.ss_len = sizeof(ss);
673 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
674 gdst = info->rti_info[RTAX_GATEWAY];
677 * A host route through the loopback interface is
678 * installed for each interface adddress. In pre 8.0
679 * releases the interface address of a PPP link type
680 * is not reachable locally. This behavior is fixed as
681 * part of the new L2/L3 redesign and rewrite work. The
682 * signature of this interface address route is the
683 * AF_LINK sa_family type of the gateway, and the
684 * rt_ifp has the IFF_LOOPBACK flag set.
686 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
687 if (ss.ss_family == AF_LINK &&
688 ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
689 info->rti_flags &= ~RTF_GATEWAY;
690 info->rti_flags |= RTF_GWFLAG_COMPAT;
692 rib_free_info(&ginfo);
699 static struct nhop_object *
700 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
702 if (!NH_IS_NHGRP(nh))
705 struct weightened_nhop *wn;
707 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
710 for (int i = 0; i < num_nhops; i++) {
711 if (match_nhop_gw(wn[i].nh, gw))
719 * Handles RTM_GET message from routing socket, returning matching rt.
722 * 0 on success, with locked and referenced matching rt in @rt_nrt
726 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
727 struct rt_msghdr *rtm, struct rib_cmd_info *rc)
730 struct rib_head *rnh;
731 struct nhop_object *nh;
734 saf = info->rti_info[RTAX_DST]->sa_family;
736 rnh = rt_tables_get_rnh(fibnum, saf);
738 return (EAFNOSUPPORT);
742 if (info->rti_info[RTAX_NETMASK] == NULL) {
744 * Provide longest prefix match for
745 * address lookup (no mask).
746 * 'route -n get addr'
748 rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
749 info->rti_info[RTAX_DST], &rnh->head);
751 rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
752 info->rti_info[RTAX_DST],
753 info->rti_info[RTAX_NETMASK], &rnh->head);
755 if (rc->rc_rt == NULL) {
760 nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
766 * If performing proxied L2 entry insertion, and
767 * the actual PPP host entry is found, perform
768 * another search to retrieve the prefix route of
769 * the local end point of the PPP link.
770 * TODO: move this logic to userland.
772 if (rtm->rtm_flags & RTF_ANNOUNCE) {
773 struct sockaddr laddr;
775 if (nh->nh_ifp != NULL &&
776 nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
779 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
782 rt_maskedcopy(ifa->ifa_addr,
786 rt_maskedcopy(nh->nh_ifa->ifa_addr,
788 nh->nh_ifa->ifa_netmask);
790 * refactor rt and no lock operation necessary
792 rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
794 if (rc->rc_rt == NULL) {
798 nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
805 rc->rc_nh_weight = rc->rc_rt->rt_weight;
812 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
815 if (family == AF_INET) {
816 struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
817 struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
819 bzero(dst4, sizeof(struct sockaddr_in));
820 bzero(mask4, sizeof(struct sockaddr_in));
822 dst4->sin_family = AF_INET;
823 dst4->sin_len = sizeof(struct sockaddr_in);
824 mask4->sin_family = AF_INET;
825 mask4->sin_len = sizeof(struct sockaddr_in);
829 if (family == AF_INET6) {
830 struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
831 struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
833 bzero(dst6, sizeof(struct sockaddr_in6));
834 bzero(mask6, sizeof(struct sockaddr_in6));
836 dst6->sin6_family = AF_INET6;
837 dst6->sin6_len = sizeof(struct sockaddr_in6);
838 mask6->sin6_family = AF_INET6;
839 mask6->sin6_len = sizeof(struct sockaddr_in6);
845 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
846 struct sockaddr *mask)
849 if (dst->sa_family == AF_INET) {
850 struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
851 struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
852 uint32_t scopeid = 0;
853 rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
859 if (dst->sa_family == AF_INET6) {
860 struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
861 struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
862 uint32_t scopeid = 0;
863 rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
864 &mask6->sin6_addr, &scopeid);
865 dst6->sin6_scope_id = scopeid;
873 * Update sockaddrs, flags, etc in @prtm based on @rc data.
874 * rtm can be reallocated.
876 * Returns 0 on success, along with pointer to (potentially reallocated)
881 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
882 int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
885 union sockaddr_union saun;
886 struct rt_msghdr *rtm, *orig_rtm = NULL;
891 union sockaddr_union sa_dst, sa_mask;
892 int family = info->rti_info[RTAX_DST]->sa_family;
893 init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
894 export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
896 info->rti_info[RTAX_DST] = &sa_dst.sa;
897 info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
898 info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
899 info->rti_info[RTAX_GENMASK] = 0;
901 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
903 info->rti_info[RTAX_IFP] =
904 ifp->if_addr->ifa_addr;
905 error = rtm_get_jailed(info, ifp, nh,
906 &saun, curthread->td_ucred);
909 if (ifp->if_flags & IFF_POINTOPOINT)
910 info->rti_info[RTAX_BRD] =
911 nh->nh_ifa->ifa_dstaddr;
912 rtm->rtm_index = ifp->if_index;
914 info->rti_info[RTAX_IFP] = NULL;
915 info->rti_info[RTAX_IFA] = NULL;
917 } else if (ifp != NULL)
918 rtm->rtm_index = ifp->if_index;
920 /* Check if we need to realloc storage */
921 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
922 if (len > alloc_len) {
923 struct rt_msghdr *tmp_rtm;
925 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
928 bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
934 * Delay freeing original rtm as info contains
935 * data referencing it.
939 w.w_tmem = (caddr_t)rtm;
940 w.w_tmemsize = alloc_len;
941 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
943 rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
944 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
945 rtm->rtm_flags = RTF_GATEWAY |
946 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
947 rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
948 rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
949 rtm->rtm_addrs = info->rti_addrs;
951 if (orig_rtm != NULL)
952 free(orig_rtm, M_TEMP);
960 save_del_notification(struct rib_cmd_info *rc, void *_cbdata)
962 struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
964 if (rc->rc_cmd == RTM_DELETE)
969 save_add_notification(struct rib_cmd_info *rc, void *_cbdata)
971 struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
973 if (rc->rc_cmd == RTM_ADD)
980 route_output(struct mbuf *m, struct socket *so, ...)
982 struct rt_msghdr *rtm = NULL;
983 struct rtentry *rt = NULL;
984 struct rt_addrinfo info;
985 struct epoch_tracker et;
987 struct sockaddr_storage ss;
988 struct sockaddr_in6 *sin6;
989 int i, rti_need_deembed = 0;
991 int alloc_len = 0, len, error = 0, fibnum;
992 sa_family_t saf = AF_UNSPEC;
993 struct rib_cmd_info rc;
994 struct nhop_object *nh;
996 fibnum = so->so_fibnum;
997 #define senderr(e) { error = e; goto flush;}
998 if (m == NULL || ((m->m_len < sizeof(long)) &&
999 (m = m_pullup(m, sizeof(long))) == NULL))
1001 if ((m->m_flags & M_PKTHDR) == 0)
1002 panic("route_output");
1003 NET_EPOCH_ENTER(et);
1004 len = m->m_pkthdr.len;
1005 if (len < sizeof(*rtm) ||
1006 len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1010 * Most of current messages are in range 200-240 bytes,
1011 * minimize possible re-allocation on reply using larger size
1012 * buffer aligned on 1k boundaty.
1014 alloc_len = roundup2(len, 1024);
1015 if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
1018 m_copydata(m, 0, len, (caddr_t)rtm);
1019 bzero(&info, sizeof(info));
1022 if (rtm->rtm_version != RTM_VERSION) {
1023 /* Do not touch message since format is unknown */
1026 senderr(EPROTONOSUPPORT);
1030 * Starting from here, it is possible
1031 * to alter original message and insert
1032 * caller PID and error value.
1035 if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) {
1039 saf = info.rti_info[RTAX_DST]->sa_family;
1041 /* support for new ARP code */
1042 if (rtm->rtm_flags & RTF_LLDATA) {
1043 error = lla_rt_output(rtm, &info);
1046 rti_need_deembed = 1;
1051 union sockaddr_union gw_saun;
1052 int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1053 if (blackhole_flags != 0) {
1054 if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1055 error = fill_blackholeinfo(&info, &gw_saun);
1060 /* TODO: rebuild rtm from scratch */
1063 switch (rtm->rtm_type) {
1066 if (rtm->rtm_type == RTM_ADD) {
1067 if (info.rti_info[RTAX_GATEWAY] == NULL)
1070 error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1073 rti_need_deembed = 1;
1076 if (NH_IS_NHGRP(rc.rc_nh_new) ||
1077 (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1078 struct rib_cmd_info rc_simple = {};
1079 rib_decompose_notification(&rc,
1080 save_add_notification, (void *)&rc_simple);
1085 rtm->rtm_index = nh->nh_ifp->if_index;
1086 rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1091 error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1094 if (NH_IS_NHGRP(rc.rc_nh_old) ||
1095 (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1096 struct rib_cmd_info rc_simple = {};
1097 rib_decompose_notification(&rc,
1098 save_del_notification, (void *)&rc_simple);
1106 /* rt_msg2() will not be used when RTM_DELETE fails. */
1107 rti_need_deembed = 1;
1112 error = handle_rtm_get(&info, fibnum, rtm, &rc);
1118 if (!can_export_rte(curthread->td_ucred,
1119 info.rti_info[RTAX_NETMASK] == NULL,
1120 info.rti_info[RTAX_DST])) {
1124 error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1126 * Note that some sockaddr pointers may have changed to
1127 * point to memory outsize @rtm. Some may be pointing
1128 * to the on-stack variables.
1129 * Given that, any pointer in @info CANNOT BE USED.
1133 * scopeid deembedding has been performed while
1134 * writing updated rtm in rtsock_msg_buffer().
1135 * With that in mind, skip deembedding procedure below.
1138 rti_need_deembed = 0;
1145 senderr(EOPNOTSUPP);
1154 if (rti_need_deembed) {
1155 /* sin6_scope_id is recovered before sending rtm. */
1156 sin6 = (struct sockaddr_in6 *)&ss;
1157 for (i = 0; i < RTAX_MAX; i++) {
1158 if (info.rti_info[i] == NULL)
1160 if (info.rti_info[i]->sa_family != AF_INET6)
1162 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1163 if (sa6_recoverscope(sin6) == 0)
1164 bcopy(sin6, info.rti_info[i],
1170 send_rtm_reply(so, rtm, m, saf, fibnum, error);
1176 * Sends the prepared reply message in @rtm to all rtsock clients.
1177 * Frees @m and @rtm.
1181 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1182 sa_family_t saf, u_int fibnum, int rtm_errno)
1184 struct rawcb *rp = NULL;
1187 * Check to see if we don't want our own messages.
1189 if ((so->so_options & SO_USELOOPBACK) == 0) {
1190 if (V_route_cb.any_count <= 1) {
1196 /* There is another listener, so construct message */
1202 rtm->rtm_errno = rtm_errno;
1204 rtm->rtm_flags |= RTF_DONE;
1206 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1207 if (m->m_pkthdr.len < rtm->rtm_msglen) {
1210 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
1211 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1216 M_SETFIB(m, fibnum);
1217 m->m_flags |= RTS_FILTER_FIB;
1220 * XXX insure we don't get a copy by
1221 * invalidating our protocol
1223 unsigned short family = rp->rcb_proto.sp_family;
1224 rp->rcb_proto.sp_family = 0;
1225 rt_dispatch(m, saf);
1226 rp->rcb_proto.sp_family = family;
1228 rt_dispatch(m, saf);
1233 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1234 struct rt_metrics *out)
1237 bzero(out, sizeof(*out));
1238 out->rmx_mtu = nh->nh_mtu;
1239 out->rmx_weight = rt->rt_weight;
1240 out->rmx_nhidx = nhop_get_idx(nh);
1241 /* Kernel -> userland timebase conversion. */
1242 out->rmx_expire = rt->rt_expire ?
1243 rt->rt_expire - time_uptime + time_second : 0;
1247 * Extract the addresses of the passed sockaddrs.
1248 * Do a little sanity checking so as to avoid bad memory references.
1249 * This data is derived straight from userland.
1252 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1254 struct sockaddr *sa;
1257 for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1258 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1260 sa = (struct sockaddr *)cp;
1264 if (cp + sa->sa_len > cplim)
1267 * there are no more.. quit now
1268 * If there are more bits, they are in error.
1269 * I've seen this. route(1) can evidently generate these.
1270 * This causes kernel to core dump.
1271 * for compatibility, If we see this, point to a safe address.
1273 if (sa->sa_len == 0) {
1274 rtinfo->rti_info[i] = &sa_zero;
1275 return (0); /* should be EINVAL but for compat */
1279 if (sa->sa_family == AF_INET6)
1280 sa6_embedscope((struct sockaddr_in6 *)sa,
1283 rtinfo->rti_info[i] = sa;
1290 * Fill in @dmask with valid netmask leaving original @smask
1291 * intact. Mostly used with radix netmasks.
1294 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1295 struct sockaddr_storage *dmask)
1297 if (dst == NULL || smask == NULL)
1300 memset(dmask, 0, dst->sa_len);
1301 memcpy(dmask, smask, smask->sa_len);
1302 dmask->ss_len = dst->sa_len;
1303 dmask->ss_family = dst->sa_family;
1305 return ((struct sockaddr *)dmask);
1309 * Writes information related to @rtinfo object to newly-allocated mbuf.
1310 * Assumes MCLBYTES is enough to construct any message.
1311 * Used for OS notifications of vaious events (if/ifa announces,etc)
1313 * Returns allocated mbuf or NULL on failure.
1315 static struct mbuf *
1316 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1318 struct sockaddr_storage ss;
1319 struct rt_msghdr *rtm;
1322 struct sockaddr *sa;
1324 struct sockaddr_in6 *sin6;
1331 len = sizeof(struct ifa_msghdr);
1336 len = sizeof(struct ifma_msghdr);
1340 len = sizeof(struct if_msghdr);
1343 case RTM_IFANNOUNCE:
1345 len = sizeof(struct if_announcemsghdr);
1349 len = sizeof(struct rt_msghdr);
1352 /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1353 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1355 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1357 m = m_gethdr(M_NOWAIT, MT_DATA);
1361 m->m_pkthdr.len = m->m_len = len;
1362 rtm = mtod(m, struct rt_msghdr *);
1363 bzero((caddr_t)rtm, len);
1364 for (i = 0; i < RTAX_MAX; i++) {
1365 if ((sa = rtinfo->rti_info[i]) == NULL)
1367 rtinfo->rti_addrs |= (1 << i);
1370 KASSERT(dlen <= sizeof(ss),
1371 ("%s: sockaddr size overflow", __func__));
1372 bzero(&ss, sizeof(ss));
1373 bcopy(sa, &ss, sa->sa_len);
1374 sa = (struct sockaddr *)&ss;
1376 if (sa->sa_family == AF_INET6) {
1377 sin6 = (struct sockaddr_in6 *)sa;
1378 (void)sa6_recoverscope(sin6);
1381 m_copyback(m, len, dlen, (caddr_t)sa);
1384 if (m->m_pkthdr.len != len) {
1388 rtm->rtm_msglen = len;
1389 rtm->rtm_version = RTM_VERSION;
1390 rtm->rtm_type = type;
1395 * Writes information related to @rtinfo object to preallocated buffer.
1396 * Stores needed size in @plen. If @w is NULL, calculates size without
1398 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1400 * Returns 0 on success.
1404 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1406 struct sockaddr_storage ss;
1407 int len, buflen = 0, dlen, i;
1409 struct rt_msghdr *rtm = NULL;
1411 struct sockaddr_in6 *sin6;
1413 #ifdef COMPAT_FREEBSD32
1414 bool compat32 = false;
1420 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1421 #ifdef COMPAT_FREEBSD32
1422 if (w->w_req->flags & SCTL_MASK32) {
1423 len = sizeof(struct ifa_msghdrl32);
1427 len = sizeof(struct ifa_msghdrl);
1429 len = sizeof(struct ifa_msghdr);
1433 #ifdef COMPAT_FREEBSD32
1434 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1435 if (w->w_op == NET_RT_IFLISTL)
1436 len = sizeof(struct if_msghdrl32);
1438 len = sizeof(struct if_msghdr32);
1443 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1444 len = sizeof(struct if_msghdrl);
1446 len = sizeof(struct if_msghdr);
1450 len = sizeof(struct ifma_msghdr);
1454 len = sizeof(struct rt_msghdr);
1458 rtm = (struct rt_msghdr *)w->w_tmem;
1459 buflen = w->w_tmemsize - len;
1460 cp = (caddr_t)w->w_tmem + len;
1463 rtinfo->rti_addrs = 0;
1464 for (i = 0; i < RTAX_MAX; i++) {
1465 struct sockaddr *sa;
1467 if ((sa = rtinfo->rti_info[i]) == NULL)
1469 rtinfo->rti_addrs |= (1 << i);
1470 #ifdef COMPAT_FREEBSD32
1472 dlen = SA_SIZE32(sa);
1476 if (cp != NULL && buflen >= dlen) {
1477 KASSERT(dlen <= sizeof(ss),
1478 ("%s: sockaddr size overflow", __func__));
1479 bzero(&ss, sizeof(ss));
1480 bcopy(sa, &ss, sa->sa_len);
1481 sa = (struct sockaddr *)&ss;
1483 if (sa->sa_family == AF_INET6) {
1484 sin6 = (struct sockaddr_in6 *)sa;
1485 (void)sa6_recoverscope(sin6);
1488 bcopy((caddr_t)sa, cp, (unsigned)dlen);
1491 } else if (cp != NULL) {
1493 * Buffer too small. Count needed size
1494 * and return with error.
1503 dlen = ALIGN(len) - len;
1515 /* fill header iff buffer is large enough */
1516 rtm->rtm_version = RTM_VERSION;
1517 rtm->rtm_type = type;
1518 rtm->rtm_msglen = len;
1523 if (w != NULL && cp == NULL)
1530 * This routine is called to generate a message from the routing
1531 * socket indicating that a redirect has occurred, a routing lookup
1532 * has failed, or that a protocol has detected timeouts to a particular
1536 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1539 struct rt_msghdr *rtm;
1541 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1543 if (V_route_cb.any_count == 0)
1545 m = rtsock_msg_mbuf(type, rtinfo);
1549 if (fibnum != RT_ALL_FIBS) {
1550 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1551 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1552 M_SETFIB(m, fibnum);
1553 m->m_flags |= RTS_FILTER_FIB;
1556 rtm = mtod(m, struct rt_msghdr *);
1557 rtm->rtm_flags = RTF_DONE | flags;
1558 rtm->rtm_errno = error;
1559 rtm->rtm_addrs = rtinfo->rti_addrs;
1560 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1564 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1567 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1571 * This routine is called to generate a message from the routing
1572 * socket indicating that the status of a network interface has changed.
1575 rt_ifmsg(struct ifnet *ifp)
1577 struct if_msghdr *ifm;
1579 struct rt_addrinfo info;
1581 if (V_route_cb.any_count == 0)
1583 bzero((caddr_t)&info, sizeof(info));
1584 m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1587 ifm = mtod(m, struct if_msghdr *);
1588 ifm->ifm_index = ifp->if_index;
1589 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1590 if_data_copy(ifp, &ifm->ifm_data);
1592 rt_dispatch(m, AF_UNSPEC);
1596 * Announce interface address arrival/withdraw.
1597 * Please do not call directly, use rt_addrmsg().
1598 * Assume input data to be valid.
1599 * Returns 0 on success.
1602 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1604 struct rt_addrinfo info;
1605 struct sockaddr *sa;
1608 struct ifa_msghdr *ifam;
1609 struct ifnet *ifp = ifa->ifa_ifp;
1610 struct sockaddr_storage ss;
1612 if (V_route_cb.any_count == 0)
1615 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1617 bzero((caddr_t)&info, sizeof(info));
1618 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1619 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1620 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1621 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1622 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1623 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1625 ifam = mtod(m, struct ifa_msghdr *);
1626 ifam->ifam_index = ifp->if_index;
1627 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1628 ifam->ifam_flags = ifa->ifa_flags;
1629 ifam->ifam_addrs = info.rti_addrs;
1631 if (fibnum != RT_ALL_FIBS) {
1632 M_SETFIB(m, fibnum);
1633 m->m_flags |= RTS_FILTER_FIB;
1636 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1642 * Announce route addition/removal to rtsock based on @rt data.
1643 * Callers are advives to use rt_routemsg() instead of using this
1644 * function directly.
1645 * Assume @rt data is consistent.
1647 * Returns 0 on success.
1650 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
1653 union sockaddr_union dst, mask;
1654 struct rt_addrinfo info;
1656 if (V_route_cb.any_count == 0)
1659 int family = rt_get_family(rt);
1660 init_sockaddrs_family(family, &dst.sa, &mask.sa);
1661 export_rtaddrs(rt, &dst.sa, &mask.sa);
1663 bzero((caddr_t)&info, sizeof(info));
1664 info.rti_info[RTAX_DST] = &dst.sa;
1665 info.rti_info[RTAX_NETMASK] = &mask.sa;
1666 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1667 info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
1668 info.rti_ifp = nh->nh_ifp;
1670 return (rtsock_routemsg_info(cmd, &info, fibnum));
1674 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1676 struct rt_msghdr *rtm;
1677 struct sockaddr *sa;
1680 if (V_route_cb.any_count == 0)
1683 if (info->rti_flags & RTF_HOST)
1684 info->rti_info[RTAX_NETMASK] = NULL;
1686 m = rtsock_msg_mbuf(cmd, info);
1690 if (fibnum != RT_ALL_FIBS) {
1691 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1692 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1693 M_SETFIB(m, fibnum);
1694 m->m_flags |= RTS_FILTER_FIB;
1697 rtm = mtod(m, struct rt_msghdr *);
1698 rtm->rtm_addrs = info->rti_addrs;
1699 if (info->rti_ifp != NULL)
1700 rtm->rtm_index = info->rti_ifp->if_index;
1701 /* Add RTF_DONE to indicate command 'completion' required by API */
1702 info->rti_flags |= RTF_DONE;
1703 /* Reported routes has to be up */
1704 if (cmd == RTM_ADD || cmd == RTM_CHANGE)
1705 info->rti_flags |= RTF_UP;
1706 rtm->rtm_flags = info->rti_flags;
1708 sa = info->rti_info[RTAX_DST];
1709 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1715 * This is the analogue to the rt_newaddrmsg which performs the same
1716 * function but for multicast group memberhips. This is easier since
1717 * there is no route state to worry about.
1720 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1722 struct rt_addrinfo info;
1723 struct mbuf *m = NULL;
1724 struct ifnet *ifp = ifma->ifma_ifp;
1725 struct ifma_msghdr *ifmam;
1727 if (V_route_cb.any_count == 0)
1730 bzero((caddr_t)&info, sizeof(info));
1731 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1732 if (ifp && ifp->if_addr)
1733 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1735 info.rti_info[RTAX_IFP] = NULL;
1737 * If a link-layer address is present, present it as a ``gateway''
1738 * (similarly to how ARP entries, e.g., are presented).
1740 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1741 m = rtsock_msg_mbuf(cmd, &info);
1744 ifmam = mtod(m, struct ifma_msghdr *);
1745 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1747 ifmam->ifmam_index = ifp->if_index;
1748 ifmam->ifmam_addrs = info.rti_addrs;
1749 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1752 static struct mbuf *
1753 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1754 struct rt_addrinfo *info)
1756 struct if_announcemsghdr *ifan;
1759 if (V_route_cb.any_count == 0)
1761 bzero((caddr_t)info, sizeof(*info));
1762 m = rtsock_msg_mbuf(type, info);
1764 ifan = mtod(m, struct if_announcemsghdr *);
1765 ifan->ifan_index = ifp->if_index;
1766 strlcpy(ifan->ifan_name, ifp->if_xname,
1767 sizeof(ifan->ifan_name));
1768 ifan->ifan_what = what;
1774 * This is called to generate routing socket messages indicating
1775 * IEEE80211 wireless events.
1776 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1779 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1782 struct rt_addrinfo info;
1784 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1787 * Append the ieee80211 data. Try to stick it in the
1788 * mbuf containing the ifannounce msg; otherwise allocate
1789 * a new mbuf and append.
1791 * NB: we assume m is a single mbuf.
1793 if (data_len > M_TRAILINGSPACE(m)) {
1794 struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1799 bcopy(data, mtod(n, void *), data_len);
1800 n->m_len = data_len;
1802 } else if (data_len > 0) {
1803 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1804 m->m_len += data_len;
1806 if (m->m_flags & M_PKTHDR)
1807 m->m_pkthdr.len += data_len;
1808 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1809 rt_dispatch(m, AF_UNSPEC);
1814 * This is called to generate routing socket messages indicating
1815 * network interface arrival and departure.
1818 rt_ifannouncemsg(struct ifnet *ifp, int what)
1821 struct rt_addrinfo info;
1823 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1825 rt_dispatch(m, AF_UNSPEC);
1829 rt_dispatch(struct mbuf *m, sa_family_t saf)
1834 * Preserve the family from the sockaddr, if any, in an m_tag for
1835 * use when injecting the mbuf into the routing socket buffer from
1838 if (saf != AF_UNSPEC) {
1839 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1845 *(unsigned short *)(tag + 1) = saf;
1846 m_tag_prepend(m, tag);
1850 m->m_pkthdr.rcvif = V_loif;
1856 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */
1860 * Checks if rte can be exported v.r.t jails/vnets.
1862 * Returns 1 if it can, 0 otherwise.
1865 can_export_rte(struct ucred *td_ucred, bool rt_is_host,
1866 const struct sockaddr *rt_dst)
1869 if ((!rt_is_host) ? jailed_without_vnet(td_ucred)
1870 : prison_if(td_ucred, rt_dst) != 0)
1877 * This is used in dumping the kernel table via sysctl().
1880 sysctl_dumpentry(struct rtentry *rt, void *vw)
1882 struct walkarg *w = vw;
1883 struct nhop_object *nh;
1888 export_rtaddrs(rt, w->dst, w->mask);
1889 if (!can_export_rte(w->w_req->td->td_ucred, rt_is_host(rt), w->dst))
1891 nh = rt_get_raw_nhop(rt);
1893 if (NH_IS_NHGRP(nh)) {
1894 struct weightened_nhop *wn;
1896 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
1897 for (int i = 0; i < num_nhops; i++) {
1898 error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
1904 error = sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
1911 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
1914 struct rt_addrinfo info;
1915 int error = 0, size;
1918 rtflags = nhop_get_rtflags(nh);
1920 if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
1923 bzero((caddr_t)&info, sizeof(info));
1924 info.rti_info[RTAX_DST] = w->dst;
1925 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1926 info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
1927 info.rti_info[RTAX_GENMASK] = 0;
1928 if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
1929 info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
1930 info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
1931 if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
1932 info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
1934 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1936 if (w->w_req && w->w_tmem) {
1937 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1939 bzero(&rtm->rtm_index,
1940 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1943 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
1944 * and RTF_UP (if entry is linked, which is always true here).
1945 * Given that, use nhop rtflags & add RTF_UP.
1947 rtm->rtm_flags = rtflags | RTF_UP;
1948 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
1949 rtm->rtm_flags = RTF_GATEWAY |
1950 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
1951 rt_getmetrics(rt, nh, &rtm->rtm_rmx);
1952 rtm->rtm_rmx.rmx_weight = weight;
1953 rtm->rtm_index = nh->nh_ifp->if_index;
1954 rtm->rtm_addrs = info.rti_addrs;
1955 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1962 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1963 struct rt_addrinfo *info, struct walkarg *w, int len)
1965 struct if_msghdrl *ifm;
1966 struct if_data *ifd;
1968 ifm = (struct if_msghdrl *)w->w_tmem;
1970 #ifdef COMPAT_FREEBSD32
1971 if (w->w_req->flags & SCTL_MASK32) {
1972 struct if_msghdrl32 *ifm32;
1974 ifm32 = (struct if_msghdrl32 *)ifm;
1975 ifm32->ifm_addrs = info->rti_addrs;
1976 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1977 ifm32->ifm_index = ifp->if_index;
1978 ifm32->_ifm_spare1 = 0;
1979 ifm32->ifm_len = sizeof(*ifm32);
1980 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1981 ifm32->_ifm_spare2 = 0;
1982 ifd = &ifm32->ifm_data;
1986 ifm->ifm_addrs = info->rti_addrs;
1987 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1988 ifm->ifm_index = ifp->if_index;
1989 ifm->_ifm_spare1 = 0;
1990 ifm->ifm_len = sizeof(*ifm);
1991 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1992 ifm->_ifm_spare2 = 0;
1993 ifd = &ifm->ifm_data;
1996 memcpy(ifd, src_ifd, sizeof(*ifd));
1998 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2002 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2003 struct rt_addrinfo *info, struct walkarg *w, int len)
2005 struct if_msghdr *ifm;
2006 struct if_data *ifd;
2008 ifm = (struct if_msghdr *)w->w_tmem;
2010 #ifdef COMPAT_FREEBSD32
2011 if (w->w_req->flags & SCTL_MASK32) {
2012 struct if_msghdr32 *ifm32;
2014 ifm32 = (struct if_msghdr32 *)ifm;
2015 ifm32->ifm_addrs = info->rti_addrs;
2016 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2017 ifm32->ifm_index = ifp->if_index;
2018 ifm32->_ifm_spare1 = 0;
2019 ifd = &ifm32->ifm_data;
2023 ifm->ifm_addrs = info->rti_addrs;
2024 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2025 ifm->ifm_index = ifp->if_index;
2026 ifm->_ifm_spare1 = 0;
2027 ifd = &ifm->ifm_data;
2030 memcpy(ifd, src_ifd, sizeof(*ifd));
2032 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2036 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2037 struct walkarg *w, int len)
2039 struct ifa_msghdrl *ifam;
2040 struct if_data *ifd;
2042 ifam = (struct ifa_msghdrl *)w->w_tmem;
2044 #ifdef COMPAT_FREEBSD32
2045 if (w->w_req->flags & SCTL_MASK32) {
2046 struct ifa_msghdrl32 *ifam32;
2048 ifam32 = (struct ifa_msghdrl32 *)ifam;
2049 ifam32->ifam_addrs = info->rti_addrs;
2050 ifam32->ifam_flags = ifa->ifa_flags;
2051 ifam32->ifam_index = ifa->ifa_ifp->if_index;
2052 ifam32->_ifam_spare1 = 0;
2053 ifam32->ifam_len = sizeof(*ifam32);
2054 ifam32->ifam_data_off =
2055 offsetof(struct ifa_msghdrl32, ifam_data);
2056 ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2057 ifd = &ifam32->ifam_data;
2061 ifam->ifam_addrs = info->rti_addrs;
2062 ifam->ifam_flags = ifa->ifa_flags;
2063 ifam->ifam_index = ifa->ifa_ifp->if_index;
2064 ifam->_ifam_spare1 = 0;
2065 ifam->ifam_len = sizeof(*ifam);
2066 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2067 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2068 ifd = &ifam->ifam_data;
2071 bzero(ifd, sizeof(*ifd));
2072 ifd->ifi_datalen = sizeof(struct if_data);
2073 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2074 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2075 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2076 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2078 /* Fixup if_data carp(4) vhid. */
2079 if (carp_get_vhid_p != NULL)
2080 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2082 return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2086 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2087 struct walkarg *w, int len)
2089 struct ifa_msghdr *ifam;
2091 ifam = (struct ifa_msghdr *)w->w_tmem;
2092 ifam->ifam_addrs = info->rti_addrs;
2093 ifam->ifam_flags = ifa->ifa_flags;
2094 ifam->ifam_index = ifa->ifa_ifp->if_index;
2095 ifam->_ifam_spare1 = 0;
2096 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2098 return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2102 sysctl_iflist(int af, struct walkarg *w)
2107 struct rt_addrinfo info;
2109 struct sockaddr_storage ss;
2111 bzero((caddr_t)&info, sizeof(info));
2112 bzero(&ifd, sizeof(ifd));
2113 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2114 if (w->w_arg && w->w_arg != ifp->if_index)
2116 if_data_copy(ifp, &ifd);
2118 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2119 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2122 info.rti_info[RTAX_IFP] = NULL;
2123 if (w->w_req && w->w_tmem) {
2124 if (w->w_op == NET_RT_IFLISTL)
2125 error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2128 error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2133 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2134 if (af && af != ifa->ifa_addr->sa_family)
2136 if (prison_if(w->w_req->td->td_ucred,
2137 ifa->ifa_addr) != 0)
2139 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2140 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2141 ifa->ifa_addr, ifa->ifa_netmask, &ss);
2142 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2143 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2146 if (w->w_req && w->w_tmem) {
2147 if (w->w_op == NET_RT_IFLISTL)
2148 error = sysctl_iflist_ifaml(ifa, &info,
2151 error = sysctl_iflist_ifam(ifa, &info,
2157 info.rti_info[RTAX_IFA] = NULL;
2158 info.rti_info[RTAX_NETMASK] = NULL;
2159 info.rti_info[RTAX_BRD] = NULL;
2166 sysctl_ifmalist(int af, struct walkarg *w)
2168 struct rt_addrinfo info;
2170 struct ifmultiaddr *ifma;
2177 bzero((caddr_t)&info, sizeof(info));
2179 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2180 if (w->w_arg && w->w_arg != ifp->if_index)
2183 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2184 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2185 if (af && af != ifma->ifma_addr->sa_family)
2187 if (prison_if(w->w_req->td->td_ucred,
2188 ifma->ifma_addr) != 0)
2190 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2191 info.rti_info[RTAX_GATEWAY] =
2192 (ifma->ifma_addr->sa_family != AF_LINK) ?
2193 ifma->ifma_lladdr : NULL;
2194 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2197 if (w->w_req && w->w_tmem) {
2198 struct ifma_msghdr *ifmam;
2200 ifmam = (struct ifma_msghdr *)w->w_tmem;
2201 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2202 ifmam->ifmam_flags = 0;
2203 ifmam->ifmam_addrs = info.rti_addrs;
2204 ifmam->_ifmam_spare1 = 0;
2205 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2217 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2219 union sockaddr_union sa_dst, sa_mask;
2222 w->dst = (struct sockaddr *)&sa_dst;
2223 w->mask = (struct sockaddr *)&sa_mask;
2225 init_sockaddrs_family(family, w->dst, w->mask);
2227 rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2231 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2233 struct epoch_tracker et;
2234 int *name = (int *)arg1;
2235 u_int namelen = arg2;
2236 struct rib_head *rnh = NULL; /* silence compiler. */
2237 int i, lim, error = EINVAL;
2246 if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2248 fib = req->td->td_proc->p_fibnum;
2249 else if (namelen == 4)
2250 fib = (name[3] == RT_ALL_FIBS) ?
2251 req->td->td_proc->p_fibnum : name[3];
2253 return ((namelen < 3) ? EISDIR : ENOTDIR);
2254 if (fib < 0 || fib >= rt_numfibs)
2256 } else if (namelen != 3)
2257 return ((namelen < 3) ? EISDIR : ENOTDIR);
2261 bzero(&w, sizeof(w));
2266 error = sysctl_wire_old_buffer(req, 0);
2271 * Allocate reply buffer in advance.
2272 * All rtsock messages has maximum length of u_short.
2274 w.w_tmemsize = 65536;
2275 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2277 NET_EPOCH_ENTER(et);
2281 if (af == 0) { /* dump all tables */
2284 } else /* dump only one table */
2288 * take care of llinfo entries, the caller must
2291 if (w.w_op == NET_RT_FLAGS &&
2292 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2294 error = lltable_sysctl_dumparp(af, w.w_req);
2300 * take care of routing entries
2302 for (error = 0; error == 0 && i <= lim; i++) {
2303 rnh = rt_tables_get_rnh(fib, i);
2305 rtable_sysctl_dump(fib, i, &w);
2307 error = EAFNOSUPPORT;
2312 /* Allow dumping one specific af/fib at a time */
2318 if (fib < 0 || fib > rt_numfibs) {
2322 rnh = rt_tables_get_rnh(fib, af);
2324 error = EAFNOSUPPORT;
2327 if (w.w_op == NET_RT_NHOP)
2328 error = nhops_dump_sysctl(rnh, w.w_req);
2331 error = nhgrp_dump_sysctl(rnh, w.w_req);
2337 case NET_RT_IFLISTL:
2338 error = sysctl_iflist(af, &w);
2341 case NET_RT_IFMALIST:
2342 error = sysctl_ifmalist(af, &w);
2347 free(w.w_tmem, M_TEMP);
2351 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2352 sysctl_rtsock, "Return route tables and interface/address lists");
2355 * Definitions of protocols supported in the ROUTE domain.
2358 static struct domain routedomain; /* or at least forward */
2360 static struct protosw routesw[] = {
2362 .pr_type = SOCK_RAW,
2363 .pr_domain = &routedomain,
2364 .pr_flags = PR_ATOMIC|PR_ADDR,
2365 .pr_output = route_output,
2366 .pr_ctlinput = raw_ctlinput,
2367 .pr_init = raw_init,
2368 .pr_usrreqs = &route_usrreqs
2372 static struct domain routedomain = {
2373 .dom_family = PF_ROUTE,
2374 .dom_name = "route",
2375 .dom_protosw = routesw,
2376 .dom_protoswNPROTOSW = &routesw[nitems(routesw)]
2379 VNET_DOMAIN_SET(route);