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_mpath.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>
68 #include <net/radix_mpath.h>
72 #include <netinet/in.h>
73 #include <netinet/if_ether.h>
74 #include <netinet/ip_carp.h>
76 #include <netinet6/ip6_var.h>
77 #include <netinet6/scope6_var.h>
79 #include <net/route/nhop.h>
81 #ifdef COMPAT_FREEBSD32
82 #include <sys/mount.h>
83 #include <compat/freebsd32/freebsd32.h>
93 struct if_data ifm_data;
103 uint16_t _ifm_spare1;
105 uint16_t ifm_data_off;
106 uint32_t _ifm_spare2;
107 struct if_data ifm_data;
110 struct ifa_msghdrl32 {
111 uint16_t ifam_msglen;
112 uint8_t ifam_version;
117 uint16_t _ifam_spare1;
119 uint16_t ifam_data_off;
121 struct if_data ifam_data;
124 #define SA_SIZE32(sa) \
125 ( (((struct sockaddr *)(sa))->sa_len == 0) ? \
127 1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
129 #endif /* COMPAT_FREEBSD32 */
131 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
133 /* NB: these are not modified */
134 static struct sockaddr route_src = { 2, PF_ROUTE, };
135 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, };
137 /* These are external hooks for CARP. */
138 int (*carp_get_vhid_p)(struct ifaddr *);
141 * Used by rtsock/raw_input callback code to decide whether to filter the update
142 * notification to a socket bound to a particular FIB.
144 #define RTS_FILTER_FIB M_PROTO8
147 int ip_count; /* attached w/ AF_INET */
148 int ip6_count; /* attached w/ AF_INET6 */
149 int any_count; /* total attached */
151 VNET_DEFINE_STATIC(route_cb_t, route_cb);
152 #define V_route_cb VNET(route_cb)
154 struct mtx rtsock_mtx;
155 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
157 #define RTSOCK_LOCK() mtx_lock(&rtsock_mtx)
158 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
159 #define RTSOCK_LOCK_ASSERT() mtx_assert(&rtsock_mtx, MA_OWNED)
161 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
168 struct sysctl_req *w_req;
171 static void rts_input(struct mbuf *m);
172 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
173 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
174 struct walkarg *w, int *plen);
175 static int rt_xaddrs(caddr_t cp, caddr_t cplim,
176 struct rt_addrinfo *rtinfo);
177 static int sysctl_dumpentry(struct radix_node *rn, void *vw);
178 static int sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
179 uint32_t weight, struct walkarg *w);
180 static int sysctl_iflist(int af, struct walkarg *w);
181 static int sysctl_ifmalist(int af, struct walkarg *w);
182 static int route_output(struct mbuf *m, struct socket *so, ...);
183 static void rt_getmetrics(const struct rtentry *rt,
184 const struct nhop_object *nh, struct rt_metrics *out);
185 static void rt_dispatch(struct mbuf *, sa_family_t);
186 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
187 struct rt_msghdr *rtm, struct rib_cmd_info *rc);
188 static int update_rtm_from_rc(struct rt_addrinfo *info,
189 struct rt_msghdr **prtm, int alloc_len,
190 struct rib_cmd_info *rc, struct nhop_object *nh);
191 static void send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
192 struct mbuf *m, sa_family_t saf, u_int fibnum,
194 static int can_export_rte(struct ucred *td_ucred, const struct rtentry *rt);
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 * Fills in @info based on userland-provided @rtm message.
568 * Returns 0 on success.
571 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info)
576 rtm->rtm_pid = curproc->p_pid;
577 info->rti_addrs = rtm->rtm_addrs;
579 info->rti_mflags = rtm->rtm_inits;
580 info->rti_rmx = &rtm->rtm_rmx;
583 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
584 * link-local address because rtrequest requires addresses with
587 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
590 if (rtm->rtm_flags & RTF_RNH_LOCKED)
592 info->rti_flags = rtm->rtm_flags;
593 if (info->rti_info[RTAX_DST] == NULL ||
594 info->rti_info[RTAX_DST]->sa_family >= AF_MAX ||
595 (info->rti_info[RTAX_GATEWAY] != NULL &&
596 info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
598 saf = info->rti_info[RTAX_DST]->sa_family;
600 * Verify that the caller has the appropriate privilege; RTM_GET
601 * is the only operation the non-superuser is allowed.
603 if (rtm->rtm_type != RTM_GET) {
604 error = priv_check(curthread, PRIV_NET_ROUTE);
610 * The given gateway address may be an interface address.
611 * For example, issuing a "route change" command on a route
612 * entry that was created from a tunnel, and the gateway
613 * address given is the local end point. In this case the
614 * RTF_GATEWAY flag must be cleared or the destination will
615 * not be reachable even though there is no error message.
617 if (info->rti_info[RTAX_GATEWAY] != NULL &&
618 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
619 struct rt_addrinfo ginfo;
620 struct sockaddr *gdst;
621 struct sockaddr_storage ss;
623 bzero(&ginfo, sizeof(ginfo));
624 bzero(&ss, sizeof(ss));
625 ss.ss_len = sizeof(ss);
627 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
628 gdst = info->rti_info[RTAX_GATEWAY];
631 * A host route through the loopback interface is
632 * installed for each interface adddress. In pre 8.0
633 * releases the interface address of a PPP link type
634 * is not reachable locally. This behavior is fixed as
635 * part of the new L2/L3 redesign and rewrite work. The
636 * signature of this interface address route is the
637 * AF_LINK sa_family type of the gateway, and the
638 * rt_ifp has the IFF_LOOPBACK flag set.
640 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
641 if (ss.ss_family == AF_LINK &&
642 ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
643 info->rti_flags &= ~RTF_GATEWAY;
644 info->rti_flags |= RTF_GWFLAG_COMPAT;
646 rib_free_info(&ginfo);
654 * Handles RTM_GET message from routing socket, returning matching rt.
657 * 0 on success, with locked and referenced matching rt in @rt_nrt
661 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
662 struct rt_msghdr *rtm, struct rib_cmd_info *rc)
665 struct rib_head *rnh;
668 saf = info->rti_info[RTAX_DST]->sa_family;
670 rnh = rt_tables_get_rnh(fibnum, saf);
672 return (EAFNOSUPPORT);
676 if (info->rti_info[RTAX_NETMASK] == NULL) {
678 * Provide longest prefix match for
679 * address lookup (no mask).
680 * 'route -n get addr'
682 rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
683 info->rti_info[RTAX_DST], &rnh->head);
685 rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
686 info->rti_info[RTAX_DST],
687 info->rti_info[RTAX_NETMASK], &rnh->head);
689 if (rc->rc_rt == NULL) {
695 * for RTM_GET, gate is optional even with multipath.
696 * if gate == NULL the first match is returned.
697 * (no need to call rt_mpath_matchgate if gate == NULL)
699 if (rt_mpath_capable(rnh) && info->rti_info[RTAX_GATEWAY]) {
700 rc->rc_rt = rt_mpath_matchgate(rc->rc_rt,
701 info->rti_info[RTAX_GATEWAY]);
702 if (rc->rc_rt == NULL) {
709 * If performing proxied L2 entry insertion, and
710 * the actual PPP host entry is found, perform
711 * another search to retrieve the prefix route of
712 * the local end point of the PPP link.
713 * TODO: move this logic to userland.
715 if (rtm->rtm_flags & RTF_ANNOUNCE) {
716 struct sockaddr laddr;
717 struct nhop_object *nh;
719 nh = rc->rc_rt->rt_nhop;
720 if (nh->nh_ifp != NULL &&
721 nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
724 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
727 rt_maskedcopy(ifa->ifa_addr,
731 rt_maskedcopy(nh->nh_ifa->ifa_addr,
733 nh->nh_ifa->ifa_netmask);
735 * refactor rt and no lock operation necessary
737 rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
739 if (rc->rc_rt == NULL) {
744 rc->rc_nh_new = rc->rc_rt->rt_nhop;
745 rc->rc_nh_weight = rc->rc_rt->rt_weight;
752 * Update sockaddrs, flags, etc in @prtm based on @rc data.
753 * rtm can be reallocated.
755 * Returns 0 on success, along with pointer to (potentially reallocated)
760 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
761 int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
763 struct sockaddr_storage netmask_ss;
765 union sockaddr_union saun;
766 struct rt_msghdr *rtm, *orig_rtm = NULL;
772 info->rti_info[RTAX_DST] = rt_key(rc->rc_rt);
773 info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
774 info->rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rc->rc_rt),
775 rt_mask(rc->rc_rt), &netmask_ss);
776 info->rti_info[RTAX_GENMASK] = 0;
778 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
780 info->rti_info[RTAX_IFP] =
781 ifp->if_addr->ifa_addr;
782 error = rtm_get_jailed(info, ifp, nh,
783 &saun, curthread->td_ucred);
786 if (ifp->if_flags & IFF_POINTOPOINT)
787 info->rti_info[RTAX_BRD] =
788 nh->nh_ifa->ifa_dstaddr;
789 rtm->rtm_index = ifp->if_index;
791 info->rti_info[RTAX_IFP] = NULL;
792 info->rti_info[RTAX_IFA] = NULL;
794 } else if (ifp != NULL)
795 rtm->rtm_index = ifp->if_index;
797 /* Check if we need to realloc storage */
798 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
799 if (len > alloc_len) {
800 struct rt_msghdr *tmp_rtm;
802 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
805 bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
811 * Delay freeing original rtm as info contains
812 * data referencing it.
816 w.w_tmem = (caddr_t)rtm;
817 w.w_tmemsize = alloc_len;
818 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
820 rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
821 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
822 rtm->rtm_flags = RTF_GATEWAY |
823 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
824 rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
825 rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
826 rtm->rtm_addrs = info->rti_addrs;
828 if (orig_rtm != NULL)
829 free(orig_rtm, M_TEMP);
837 route_output(struct mbuf *m, struct socket *so, ...)
839 struct rt_msghdr *rtm = NULL;
840 struct rtentry *rt = NULL;
841 struct rt_addrinfo info;
842 struct epoch_tracker et;
844 struct sockaddr_storage ss;
845 struct sockaddr_in6 *sin6;
846 int i, rti_need_deembed = 0;
848 int alloc_len = 0, len, error = 0, fibnum;
849 sa_family_t saf = AF_UNSPEC;
851 struct rib_cmd_info rc;
852 struct nhop_object *nh;
854 fibnum = so->so_fibnum;
855 #define senderr(e) { error = e; goto flush;}
856 if (m == NULL || ((m->m_len < sizeof(long)) &&
857 (m = m_pullup(m, sizeof(long))) == NULL))
859 if ((m->m_flags & M_PKTHDR) == 0)
860 panic("route_output");
862 len = m->m_pkthdr.len;
863 if (len < sizeof(*rtm) ||
864 len != mtod(m, struct rt_msghdr *)->rtm_msglen)
868 * Most of current messages are in range 200-240 bytes,
869 * minimize possible re-allocation on reply using larger size
870 * buffer aligned on 1k boundaty.
872 alloc_len = roundup2(len, 1024);
873 if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
876 m_copydata(m, 0, len, (caddr_t)rtm);
877 bzero(&info, sizeof(info));
878 bzero(&w, sizeof(w));
881 if (rtm->rtm_version != RTM_VERSION) {
882 /* Do not touch message since format is unknown */
885 senderr(EPROTONOSUPPORT);
889 * Starting from here, it is possible
890 * to alter original message and insert
891 * caller PID and error value.
894 if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) {
898 saf = info.rti_info[RTAX_DST]->sa_family;
900 /* support for new ARP code */
901 if (rtm->rtm_flags & RTF_LLDATA) {
902 error = lla_rt_output(rtm, &info);
905 rti_need_deembed = 1;
910 switch (rtm->rtm_type) {
913 if (rtm->rtm_type == RTM_ADD) {
914 if (info.rti_info[RTAX_GATEWAY] == NULL)
917 error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
920 rti_need_deembed = 1;
923 rtm->rtm_index = nh->nh_ifp->if_index;
928 error = rib_action(fibnum, RTM_DELETE, &info, &rc);
934 /* rt_msg2() will not be used when RTM_DELETE fails. */
935 rti_need_deembed = 1;
940 error = handle_rtm_get(&info, fibnum, rtm, &rc);
946 if (!can_export_rte(curthread->td_ucred, rc.rc_rt)) {
950 error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
952 * Note that some sockaddr pointers may have changed to
953 * point to memory outsize @rtm. Some may be pointing
954 * to the on-stack variables.
955 * Given that, any pointer in @info CANNOT BE USED.
959 * scopeid deembedding has been performed while
960 * writing updated rtm in rtsock_msg_buffer().
961 * With that in mind, skip deembedding procedure below.
964 rti_need_deembed = 0;
980 if (rti_need_deembed) {
981 /* sin6_scope_id is recovered before sending rtm. */
982 sin6 = (struct sockaddr_in6 *)&ss;
983 for (i = 0; i < RTAX_MAX; i++) {
984 if (info.rti_info[i] == NULL)
986 if (info.rti_info[i]->sa_family != AF_INET6)
988 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
989 if (sa6_recoverscope(sin6) == 0)
990 bcopy(sin6, info.rti_info[i],
996 send_rtm_reply(so, rtm, m, saf, fibnum, error);
1002 * Sends the prepared reply message in @rtm to all rtsock clients.
1003 * Frees @m and @rtm.
1007 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1008 sa_family_t saf, u_int fibnum, int rtm_errno)
1010 struct rawcb *rp = NULL;
1013 * Check to see if we don't want our own messages.
1015 if ((so->so_options & SO_USELOOPBACK) == 0) {
1016 if (V_route_cb.any_count <= 1) {
1022 /* There is another listener, so construct message */
1028 rtm->rtm_errno = rtm_errno;
1030 rtm->rtm_flags |= RTF_DONE;
1032 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1033 if (m->m_pkthdr.len < rtm->rtm_msglen) {
1036 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
1037 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1042 M_SETFIB(m, fibnum);
1043 m->m_flags |= RTS_FILTER_FIB;
1046 * XXX insure we don't get a copy by
1047 * invalidating our protocol
1049 unsigned short family = rp->rcb_proto.sp_family;
1050 rp->rcb_proto.sp_family = 0;
1051 rt_dispatch(m, saf);
1052 rp->rcb_proto.sp_family = family;
1054 rt_dispatch(m, saf);
1059 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1060 struct rt_metrics *out)
1063 bzero(out, sizeof(*out));
1064 out->rmx_mtu = nh->nh_mtu;
1065 out->rmx_weight = rt->rt_weight;
1066 out->rmx_nhidx = nhop_get_idx(nh);
1067 /* Kernel -> userland timebase conversion. */
1068 out->rmx_expire = rt->rt_expire ?
1069 rt->rt_expire - time_uptime + time_second : 0;
1073 * Extract the addresses of the passed sockaddrs.
1074 * Do a little sanity checking so as to avoid bad memory references.
1075 * This data is derived straight from userland.
1078 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1080 struct sockaddr *sa;
1083 for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1084 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1086 sa = (struct sockaddr *)cp;
1090 if (cp + sa->sa_len > cplim)
1093 * there are no more.. quit now
1094 * If there are more bits, they are in error.
1095 * I've seen this. route(1) can evidently generate these.
1096 * This causes kernel to core dump.
1097 * for compatibility, If we see this, point to a safe address.
1099 if (sa->sa_len == 0) {
1100 rtinfo->rti_info[i] = &sa_zero;
1101 return (0); /* should be EINVAL but for compat */
1105 if (sa->sa_family == AF_INET6)
1106 sa6_embedscope((struct sockaddr_in6 *)sa,
1109 rtinfo->rti_info[i] = sa;
1116 * Fill in @dmask with valid netmask leaving original @smask
1117 * intact. Mostly used with radix netmasks.
1120 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1121 struct sockaddr_storage *dmask)
1123 if (dst == NULL || smask == NULL)
1126 memset(dmask, 0, dst->sa_len);
1127 memcpy(dmask, smask, smask->sa_len);
1128 dmask->ss_len = dst->sa_len;
1129 dmask->ss_family = dst->sa_family;
1131 return ((struct sockaddr *)dmask);
1135 * Writes information related to @rtinfo object to newly-allocated mbuf.
1136 * Assumes MCLBYTES is enough to construct any message.
1137 * Used for OS notifications of vaious events (if/ifa announces,etc)
1139 * Returns allocated mbuf or NULL on failure.
1141 static struct mbuf *
1142 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1144 struct rt_msghdr *rtm;
1147 struct sockaddr *sa;
1149 struct sockaddr_storage ss;
1150 struct sockaddr_in6 *sin6;
1157 len = sizeof(struct ifa_msghdr);
1162 len = sizeof(struct ifma_msghdr);
1166 len = sizeof(struct if_msghdr);
1169 case RTM_IFANNOUNCE:
1171 len = sizeof(struct if_announcemsghdr);
1175 len = sizeof(struct rt_msghdr);
1178 /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1179 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1181 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1183 m = m_gethdr(M_NOWAIT, MT_DATA);
1187 m->m_pkthdr.len = m->m_len = len;
1188 rtm = mtod(m, struct rt_msghdr *);
1189 bzero((caddr_t)rtm, len);
1190 for (i = 0; i < RTAX_MAX; i++) {
1191 if ((sa = rtinfo->rti_info[i]) == NULL)
1193 rtinfo->rti_addrs |= (1 << i);
1196 if (sa->sa_family == AF_INET6) {
1197 sin6 = (struct sockaddr_in6 *)&ss;
1198 bcopy(sa, sin6, sizeof(*sin6));
1199 if (sa6_recoverscope(sin6) == 0)
1200 sa = (struct sockaddr *)sin6;
1203 m_copyback(m, len, dlen, (caddr_t)sa);
1206 if (m->m_pkthdr.len != len) {
1210 rtm->rtm_msglen = len;
1211 rtm->rtm_version = RTM_VERSION;
1212 rtm->rtm_type = type;
1217 * Writes information related to @rtinfo object to preallocated buffer.
1218 * Stores needed size in @plen. If @w is NULL, calculates size without
1220 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1222 * Returns 0 on success.
1226 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1229 int len, buflen = 0, dlen;
1231 struct rt_msghdr *rtm = NULL;
1233 struct sockaddr_storage ss;
1234 struct sockaddr_in6 *sin6;
1236 #ifdef COMPAT_FREEBSD32
1237 bool compat32 = false;
1243 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1244 #ifdef COMPAT_FREEBSD32
1245 if (w->w_req->flags & SCTL_MASK32) {
1246 len = sizeof(struct ifa_msghdrl32);
1250 len = sizeof(struct ifa_msghdrl);
1252 len = sizeof(struct ifa_msghdr);
1256 #ifdef COMPAT_FREEBSD32
1257 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1258 if (w->w_op == NET_RT_IFLISTL)
1259 len = sizeof(struct if_msghdrl32);
1261 len = sizeof(struct if_msghdr32);
1266 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1267 len = sizeof(struct if_msghdrl);
1269 len = sizeof(struct if_msghdr);
1273 len = sizeof(struct ifma_msghdr);
1277 len = sizeof(struct rt_msghdr);
1281 rtm = (struct rt_msghdr *)w->w_tmem;
1282 buflen = w->w_tmemsize - len;
1283 cp = (caddr_t)w->w_tmem + len;
1286 rtinfo->rti_addrs = 0;
1287 for (i = 0; i < RTAX_MAX; i++) {
1288 struct sockaddr *sa;
1290 if ((sa = rtinfo->rti_info[i]) == NULL)
1292 rtinfo->rti_addrs |= (1 << i);
1293 #ifdef COMPAT_FREEBSD32
1295 dlen = SA_SIZE32(sa);
1299 if (cp != NULL && buflen >= dlen) {
1301 if (sa->sa_family == AF_INET6) {
1302 sin6 = (struct sockaddr_in6 *)&ss;
1303 bcopy(sa, sin6, sizeof(*sin6));
1304 if (sa6_recoverscope(sin6) == 0)
1305 sa = (struct sockaddr *)sin6;
1308 bcopy((caddr_t)sa, cp, (unsigned)dlen);
1311 } else if (cp != NULL) {
1313 * Buffer too small. Count needed size
1314 * and return with error.
1323 dlen = ALIGN(len) - len;
1335 /* fill header iff buffer is large enough */
1336 rtm->rtm_version = RTM_VERSION;
1337 rtm->rtm_type = type;
1338 rtm->rtm_msglen = len;
1343 if (w != NULL && cp == NULL)
1350 * This routine is called to generate a message from the routing
1351 * socket indicating that a redirect has occurred, a routing lookup
1352 * has failed, or that a protocol has detected timeouts to a particular
1356 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1359 struct rt_msghdr *rtm;
1361 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1363 if (V_route_cb.any_count == 0)
1365 m = rtsock_msg_mbuf(type, rtinfo);
1369 if (fibnum != RT_ALL_FIBS) {
1370 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1371 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1372 M_SETFIB(m, fibnum);
1373 m->m_flags |= RTS_FILTER_FIB;
1376 rtm = mtod(m, struct rt_msghdr *);
1377 rtm->rtm_flags = RTF_DONE | flags;
1378 rtm->rtm_errno = error;
1379 rtm->rtm_addrs = rtinfo->rti_addrs;
1380 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1384 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1387 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1391 * This routine is called to generate a message from the routing
1392 * socket indicating that the status of a network interface has changed.
1395 rt_ifmsg(struct ifnet *ifp)
1397 struct if_msghdr *ifm;
1399 struct rt_addrinfo info;
1401 if (V_route_cb.any_count == 0)
1403 bzero((caddr_t)&info, sizeof(info));
1404 m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1407 ifm = mtod(m, struct if_msghdr *);
1408 ifm->ifm_index = ifp->if_index;
1409 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1410 if_data_copy(ifp, &ifm->ifm_data);
1412 rt_dispatch(m, AF_UNSPEC);
1416 * Announce interface address arrival/withdraw.
1417 * Please do not call directly, use rt_addrmsg().
1418 * Assume input data to be valid.
1419 * Returns 0 on success.
1422 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1424 struct rt_addrinfo info;
1425 struct sockaddr *sa;
1428 struct ifa_msghdr *ifam;
1429 struct ifnet *ifp = ifa->ifa_ifp;
1430 struct sockaddr_storage ss;
1432 if (V_route_cb.any_count == 0)
1435 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1437 bzero((caddr_t)&info, sizeof(info));
1438 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1439 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1440 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1441 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1442 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1443 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1445 ifam = mtod(m, struct ifa_msghdr *);
1446 ifam->ifam_index = ifp->if_index;
1447 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1448 ifam->ifam_flags = ifa->ifa_flags;
1449 ifam->ifam_addrs = info.rti_addrs;
1451 if (fibnum != RT_ALL_FIBS) {
1452 M_SETFIB(m, fibnum);
1453 m->m_flags |= RTS_FILTER_FIB;
1456 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1462 * Announce route addition/removal to rtsock based on @rt data.
1463 * Callers are advives to use rt_routemsg() instead of using this
1464 * function directly.
1465 * Assume @rt data is consistent.
1467 * Returns 0 on success.
1470 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs,
1473 struct sockaddr_storage ss;
1474 struct rt_addrinfo info;
1475 struct nhop_object *nh;
1477 if (V_route_cb.any_count == 0)
1481 bzero((caddr_t)&info, sizeof(info));
1482 info.rti_info[RTAX_DST] = rt_key(rt);
1483 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss);
1484 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1485 info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
1488 return (rtsock_routemsg_info(cmd, &info, fibnum));
1492 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1494 struct rt_msghdr *rtm;
1495 struct sockaddr *sa;
1498 if (V_route_cb.any_count == 0)
1501 if (info->rti_flags & RTF_HOST)
1502 info->rti_info[RTAX_NETMASK] = NULL;
1504 m = rtsock_msg_mbuf(cmd, info);
1508 if (fibnum != RT_ALL_FIBS) {
1509 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1510 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1511 M_SETFIB(m, fibnum);
1512 m->m_flags |= RTS_FILTER_FIB;
1515 rtm = mtod(m, struct rt_msghdr *);
1516 rtm->rtm_addrs = info->rti_addrs;
1517 if (info->rti_ifp != NULL)
1518 rtm->rtm_index = info->rti_ifp->if_index;
1519 /* Add RTF_DONE to indicate command 'completion' required by API */
1520 info->rti_flags |= RTF_DONE;
1521 /* Reported routes has to be up */
1522 if (cmd == RTM_ADD || cmd == RTM_CHANGE)
1523 info->rti_flags |= RTF_UP;
1524 rtm->rtm_flags = info->rti_flags;
1526 sa = info->rti_info[RTAX_DST];
1527 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1533 * This is the analogue to the rt_newaddrmsg which performs the same
1534 * function but for multicast group memberhips. This is easier since
1535 * there is no route state to worry about.
1538 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1540 struct rt_addrinfo info;
1541 struct mbuf *m = NULL;
1542 struct ifnet *ifp = ifma->ifma_ifp;
1543 struct ifma_msghdr *ifmam;
1545 if (V_route_cb.any_count == 0)
1548 bzero((caddr_t)&info, sizeof(info));
1549 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1550 if (ifp && ifp->if_addr)
1551 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1553 info.rti_info[RTAX_IFP] = NULL;
1555 * If a link-layer address is present, present it as a ``gateway''
1556 * (similarly to how ARP entries, e.g., are presented).
1558 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1559 m = rtsock_msg_mbuf(cmd, &info);
1562 ifmam = mtod(m, struct ifma_msghdr *);
1563 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1565 ifmam->ifmam_index = ifp->if_index;
1566 ifmam->ifmam_addrs = info.rti_addrs;
1567 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1570 static struct mbuf *
1571 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1572 struct rt_addrinfo *info)
1574 struct if_announcemsghdr *ifan;
1577 if (V_route_cb.any_count == 0)
1579 bzero((caddr_t)info, sizeof(*info));
1580 m = rtsock_msg_mbuf(type, info);
1582 ifan = mtod(m, struct if_announcemsghdr *);
1583 ifan->ifan_index = ifp->if_index;
1584 strlcpy(ifan->ifan_name, ifp->if_xname,
1585 sizeof(ifan->ifan_name));
1586 ifan->ifan_what = what;
1592 * This is called to generate routing socket messages indicating
1593 * IEEE80211 wireless events.
1594 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1597 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1600 struct rt_addrinfo info;
1602 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1605 * Append the ieee80211 data. Try to stick it in the
1606 * mbuf containing the ifannounce msg; otherwise allocate
1607 * a new mbuf and append.
1609 * NB: we assume m is a single mbuf.
1611 if (data_len > M_TRAILINGSPACE(m)) {
1612 struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1617 bcopy(data, mtod(n, void *), data_len);
1618 n->m_len = data_len;
1620 } else if (data_len > 0) {
1621 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1622 m->m_len += data_len;
1624 if (m->m_flags & M_PKTHDR)
1625 m->m_pkthdr.len += data_len;
1626 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1627 rt_dispatch(m, AF_UNSPEC);
1632 * This is called to generate routing socket messages indicating
1633 * network interface arrival and departure.
1636 rt_ifannouncemsg(struct ifnet *ifp, int what)
1639 struct rt_addrinfo info;
1641 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1643 rt_dispatch(m, AF_UNSPEC);
1647 rt_dispatch(struct mbuf *m, sa_family_t saf)
1652 * Preserve the family from the sockaddr, if any, in an m_tag for
1653 * use when injecting the mbuf into the routing socket buffer from
1656 if (saf != AF_UNSPEC) {
1657 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1663 *(unsigned short *)(tag + 1) = saf;
1664 m_tag_prepend(m, tag);
1668 m->m_pkthdr.rcvif = V_loif;
1674 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */
1678 * Checks if rte can be exported v.r.t jails/vnets.
1680 * Returns 1 if it can, 0 otherwise.
1683 can_export_rte(struct ucred *td_ucred, const struct rtentry *rt)
1686 if ((rt->rte_flags & RTF_HOST) == 0
1687 ? jailed_without_vnet(td_ucred)
1688 : prison_if(td_ucred, rt_key_const(rt)) != 0)
1694 * This is used in dumping the kernel table via sysctl().
1697 sysctl_dumpentry(struct radix_node *rn, void *vw)
1699 struct walkarg *w = vw;
1700 struct rtentry *rt = (struct rtentry *)rn;
1701 struct nhop_object *nh;
1706 if (w->w_op == NET_RT_FLAGS && !(rt->rte_flags & w->w_arg))
1708 if (!can_export_rte(w->w_req->td->td_ucred, rt))
1711 error = sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
1718 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
1721 struct rt_addrinfo info;
1722 int error = 0, size;
1723 struct sockaddr_storage ss;
1725 bzero((caddr_t)&info, sizeof(info));
1726 info.rti_info[RTAX_DST] = rt_key(rt);
1727 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1728 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1730 info.rti_info[RTAX_GENMASK] = 0;
1731 if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
1732 info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
1733 info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
1734 if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
1735 info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
1737 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1739 if (w->w_req && w->w_tmem) {
1740 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1742 bzero(&rtm->rtm_index,
1743 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1744 if (rt->rte_flags & RTF_GWFLAG_COMPAT)
1745 rtm->rtm_flags = RTF_GATEWAY |
1746 (rt->rte_flags & ~RTF_GWFLAG_COMPAT);
1748 rtm->rtm_flags = rt->rte_flags;
1749 rtm->rtm_flags |= nhop_get_rtflags(nh);
1750 rt_getmetrics(rt, nh, &rtm->rtm_rmx);
1751 rtm->rtm_index = nh->nh_ifp->if_index;
1752 rtm->rtm_addrs = info.rti_addrs;
1753 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1760 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1761 struct rt_addrinfo *info, struct walkarg *w, int len)
1763 struct if_msghdrl *ifm;
1764 struct if_data *ifd;
1766 ifm = (struct if_msghdrl *)w->w_tmem;
1768 #ifdef COMPAT_FREEBSD32
1769 if (w->w_req->flags & SCTL_MASK32) {
1770 struct if_msghdrl32 *ifm32;
1772 ifm32 = (struct if_msghdrl32 *)ifm;
1773 ifm32->ifm_addrs = info->rti_addrs;
1774 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1775 ifm32->ifm_index = ifp->if_index;
1776 ifm32->_ifm_spare1 = 0;
1777 ifm32->ifm_len = sizeof(*ifm32);
1778 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1779 ifm32->_ifm_spare2 = 0;
1780 ifd = &ifm32->ifm_data;
1784 ifm->ifm_addrs = info->rti_addrs;
1785 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1786 ifm->ifm_index = ifp->if_index;
1787 ifm->_ifm_spare1 = 0;
1788 ifm->ifm_len = sizeof(*ifm);
1789 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1790 ifm->_ifm_spare2 = 0;
1791 ifd = &ifm->ifm_data;
1794 memcpy(ifd, src_ifd, sizeof(*ifd));
1796 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1800 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1801 struct rt_addrinfo *info, struct walkarg *w, int len)
1803 struct if_msghdr *ifm;
1804 struct if_data *ifd;
1806 ifm = (struct if_msghdr *)w->w_tmem;
1808 #ifdef COMPAT_FREEBSD32
1809 if (w->w_req->flags & SCTL_MASK32) {
1810 struct if_msghdr32 *ifm32;
1812 ifm32 = (struct if_msghdr32 *)ifm;
1813 ifm32->ifm_addrs = info->rti_addrs;
1814 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1815 ifm32->ifm_index = ifp->if_index;
1816 ifm32->_ifm_spare1 = 0;
1817 ifd = &ifm32->ifm_data;
1821 ifm->ifm_addrs = info->rti_addrs;
1822 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1823 ifm->ifm_index = ifp->if_index;
1824 ifm->_ifm_spare1 = 0;
1825 ifd = &ifm->ifm_data;
1828 memcpy(ifd, src_ifd, sizeof(*ifd));
1830 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1834 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1835 struct walkarg *w, int len)
1837 struct ifa_msghdrl *ifam;
1838 struct if_data *ifd;
1840 ifam = (struct ifa_msghdrl *)w->w_tmem;
1842 #ifdef COMPAT_FREEBSD32
1843 if (w->w_req->flags & SCTL_MASK32) {
1844 struct ifa_msghdrl32 *ifam32;
1846 ifam32 = (struct ifa_msghdrl32 *)ifam;
1847 ifam32->ifam_addrs = info->rti_addrs;
1848 ifam32->ifam_flags = ifa->ifa_flags;
1849 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1850 ifam32->_ifam_spare1 = 0;
1851 ifam32->ifam_len = sizeof(*ifam32);
1852 ifam32->ifam_data_off =
1853 offsetof(struct ifa_msghdrl32, ifam_data);
1854 ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1855 ifd = &ifam32->ifam_data;
1859 ifam->ifam_addrs = info->rti_addrs;
1860 ifam->ifam_flags = ifa->ifa_flags;
1861 ifam->ifam_index = ifa->ifa_ifp->if_index;
1862 ifam->_ifam_spare1 = 0;
1863 ifam->ifam_len = sizeof(*ifam);
1864 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1865 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1866 ifd = &ifam->ifam_data;
1869 bzero(ifd, sizeof(*ifd));
1870 ifd->ifi_datalen = sizeof(struct if_data);
1871 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1872 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1873 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1874 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1876 /* Fixup if_data carp(4) vhid. */
1877 if (carp_get_vhid_p != NULL)
1878 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1880 return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1884 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1885 struct walkarg *w, int len)
1887 struct ifa_msghdr *ifam;
1889 ifam = (struct ifa_msghdr *)w->w_tmem;
1890 ifam->ifam_addrs = info->rti_addrs;
1891 ifam->ifam_flags = ifa->ifa_flags;
1892 ifam->ifam_index = ifa->ifa_ifp->if_index;
1893 ifam->_ifam_spare1 = 0;
1894 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1896 return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1900 sysctl_iflist(int af, struct walkarg *w)
1905 struct rt_addrinfo info;
1907 struct sockaddr_storage ss;
1909 bzero((caddr_t)&info, sizeof(info));
1910 bzero(&ifd, sizeof(ifd));
1911 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1912 if (w->w_arg && w->w_arg != ifp->if_index)
1914 if_data_copy(ifp, &ifd);
1916 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1917 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1920 info.rti_info[RTAX_IFP] = NULL;
1921 if (w->w_req && w->w_tmem) {
1922 if (w->w_op == NET_RT_IFLISTL)
1923 error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1926 error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1931 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1932 if (af && af != ifa->ifa_addr->sa_family)
1934 if (prison_if(w->w_req->td->td_ucred,
1935 ifa->ifa_addr) != 0)
1937 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1938 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1939 ifa->ifa_addr, ifa->ifa_netmask, &ss);
1940 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1941 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1944 if (w->w_req && w->w_tmem) {
1945 if (w->w_op == NET_RT_IFLISTL)
1946 error = sysctl_iflist_ifaml(ifa, &info,
1949 error = sysctl_iflist_ifam(ifa, &info,
1955 info.rti_info[RTAX_IFA] = NULL;
1956 info.rti_info[RTAX_NETMASK] = NULL;
1957 info.rti_info[RTAX_BRD] = NULL;
1964 sysctl_ifmalist(int af, struct walkarg *w)
1966 struct rt_addrinfo info;
1968 struct ifmultiaddr *ifma;
1975 bzero((caddr_t)&info, sizeof(info));
1977 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1978 if (w->w_arg && w->w_arg != ifp->if_index)
1981 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1982 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1983 if (af && af != ifma->ifma_addr->sa_family)
1985 if (prison_if(w->w_req->td->td_ucred,
1986 ifma->ifma_addr) != 0)
1988 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1989 info.rti_info[RTAX_GATEWAY] =
1990 (ifma->ifma_addr->sa_family != AF_LINK) ?
1991 ifma->ifma_lladdr : NULL;
1992 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1995 if (w->w_req && w->w_tmem) {
1996 struct ifma_msghdr *ifmam;
1998 ifmam = (struct ifma_msghdr *)w->w_tmem;
1999 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2000 ifmam->ifmam_flags = 0;
2001 ifmam->ifmam_addrs = info.rti_addrs;
2002 ifmam->_ifmam_spare1 = 0;
2003 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2015 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2018 struct epoch_tracker et;
2019 int *name = (int *)arg1;
2020 u_int namelen = arg2;
2021 struct rib_head *rnh = NULL; /* silence compiler. */
2022 int i, lim, error = EINVAL;
2031 if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP) {
2033 fib = req->td->td_proc->p_fibnum;
2034 else if (namelen == 4)
2035 fib = (name[3] == RT_ALL_FIBS) ?
2036 req->td->td_proc->p_fibnum : name[3];
2038 return ((namelen < 3) ? EISDIR : ENOTDIR);
2039 if (fib < 0 || fib >= rt_numfibs)
2041 } else if (namelen != 3)
2042 return ((namelen < 3) ? EISDIR : ENOTDIR);
2046 bzero(&w, sizeof(w));
2051 error = sysctl_wire_old_buffer(req, 0);
2056 * Allocate reply buffer in advance.
2057 * All rtsock messages has maximum length of u_short.
2059 w.w_tmemsize = 65536;
2060 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2062 NET_EPOCH_ENTER(et);
2066 if (af == 0) { /* dump all tables */
2069 } else /* dump only one table */
2073 * take care of llinfo entries, the caller must
2076 if (w.w_op == NET_RT_FLAGS &&
2077 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2079 error = lltable_sysctl_dumparp(af, w.w_req);
2085 * take care of routing entries
2087 for (error = 0; error == 0 && i <= lim; i++) {
2088 rnh = rt_tables_get_rnh(fib, i);
2091 error = rnh->rnh_walktree(&rnh->head,
2092 sysctl_dumpentry, &w);
2095 error = EAFNOSUPPORT;
2099 /* Allow dumping one specific af/fib at a time */
2105 if (fib < 0 || fib > rt_numfibs) {
2109 rnh = rt_tables_get_rnh(fib, af);
2111 error = EAFNOSUPPORT;
2114 if (w.w_op == NET_RT_NHOP)
2115 error = nhops_dump_sysctl(rnh, w.w_req);
2118 case NET_RT_IFLISTL:
2119 error = sysctl_iflist(af, &w);
2122 case NET_RT_IFMALIST:
2123 error = sysctl_ifmalist(af, &w);
2128 free(w.w_tmem, M_TEMP);
2132 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2133 sysctl_rtsock, "Return route tables and interface/address lists");
2136 * Definitions of protocols supported in the ROUTE domain.
2139 static struct domain routedomain; /* or at least forward */
2141 static struct protosw routesw[] = {
2143 .pr_type = SOCK_RAW,
2144 .pr_domain = &routedomain,
2145 .pr_flags = PR_ATOMIC|PR_ADDR,
2146 .pr_output = route_output,
2147 .pr_ctlinput = raw_ctlinput,
2148 .pr_init = raw_init,
2149 .pr_usrreqs = &route_usrreqs
2153 static struct domain routedomain = {
2154 .dom_family = PF_ROUTE,
2155 .dom_name = "route",
2156 .dom_protosw = routesw,
2157 .dom_protoswNPROTOSW = &routesw[nitems(routesw)]
2160 VNET_DOMAIN_SET(route);