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>
59 #include <ddb/db_lex.h>
63 #include <net/if_var.h>
64 #include <net/if_dl.h>
65 #include <net/if_llatbl.h>
66 #include <net/if_types.h>
67 #include <net/netisr.h>
68 #include <net/raw_cb.h>
69 #include <net/route.h>
70 #include <net/route_var.h>
73 #include <netinet/in.h>
74 #include <netinet/if_ether.h>
75 #include <netinet/ip_carp.h>
77 #include <netinet6/ip6_var.h>
78 #include <netinet6/scope6_var.h>
80 #include <net/route/nhop.h>
81 #include <net/route/shared.h>
83 #ifdef COMPAT_FREEBSD32
84 #include <sys/mount.h>
85 #include <compat/freebsd32/freebsd32.h>
95 struct if_data ifm_data;
105 uint16_t _ifm_spare1;
107 uint16_t ifm_data_off;
108 uint32_t _ifm_spare2;
109 struct if_data ifm_data;
112 struct ifa_msghdrl32 {
113 uint16_t ifam_msglen;
114 uint8_t ifam_version;
119 uint16_t _ifam_spare1;
121 uint16_t ifam_data_off;
123 struct if_data ifam_data;
126 #define SA_SIZE32(sa) \
127 ( (((struct sockaddr *)(sa))->sa_len == 0) ? \
129 1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
131 #endif /* COMPAT_FREEBSD32 */
133 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
135 /* NB: these are not modified */
136 static struct sockaddr route_src = { 2, PF_ROUTE, };
137 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, };
139 /* These are external hooks for CARP. */
140 int (*carp_get_vhid_p)(struct ifaddr *);
143 * Used by rtsock/raw_input callback code to decide whether to filter the update
144 * notification to a socket bound to a particular FIB.
146 #define RTS_FILTER_FIB M_PROTO8
149 int ip_count; /* attached w/ AF_INET */
150 int ip6_count; /* attached w/ AF_INET6 */
151 int any_count; /* total attached */
153 VNET_DEFINE_STATIC(route_cb_t, route_cb);
154 #define V_route_cb VNET(route_cb)
156 struct mtx rtsock_mtx;
157 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
159 #define RTSOCK_LOCK() mtx_lock(&rtsock_mtx)
160 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
161 #define RTSOCK_LOCK_ASSERT() mtx_assert(&rtsock_mtx, MA_OWNED)
163 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
170 struct sysctl_req *w_req;
173 static void rts_input(struct mbuf *m);
174 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
175 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
176 struct walkarg *w, int *plen);
177 static int rt_xaddrs(caddr_t cp, caddr_t cplim,
178 struct rt_addrinfo *rtinfo);
179 static int sysctl_dumpentry(struct radix_node *rn, void *vw);
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, struct rt_metrics *out);
184 static void rt_dispatch(struct mbuf *, sa_family_t);
185 static struct sockaddr *rtsock_fix_netmask(struct sockaddr *dst,
186 struct sockaddr *smask, struct sockaddr_storage *dmask);
187 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
188 struct rt_msghdr *rtm, struct rtentry **ret_nrt);
189 static int update_rtm_from_rte(struct rt_addrinfo *info,
190 struct rt_msghdr **prtm, int alloc_len,
192 static void send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
193 struct mbuf *m, sa_family_t saf, u_int fibnum,
195 static int can_export_rte(struct ucred *td_ucred, const struct rtentry *rt);
197 static struct netisr_handler rtsock_nh = {
199 .nh_handler = rts_input,
200 .nh_proto = NETISR_ROUTE,
201 .nh_policy = NETISR_POLICY_SOURCE,
205 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
209 netisr_getqlimit(&rtsock_nh, &qlimit);
210 error = sysctl_handle_int(oidp, &qlimit, 0, req);
211 if (error || !req->newptr)
215 return (netisr_setqlimit(&rtsock_nh, qlimit));
217 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
218 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
219 0, 0, sysctl_route_netisr_maxqlen, "I",
220 "maximum routing socket dispatch queue length");
227 if (IS_DEFAULT_VNET(curvnet)) {
228 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
229 rtsock_nh.nh_qlimit = tmp;
230 netisr_register(&rtsock_nh);
234 netisr_register_vnet(&rtsock_nh);
237 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
242 vnet_rts_uninit(void)
245 netisr_unregister_vnet(&rtsock_nh);
247 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
252 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
257 KASSERT(m != NULL, ("%s: m is NULL", __func__));
258 KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
259 KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
261 /* No filtering requested. */
262 if ((m->m_flags & RTS_FILTER_FIB) == 0)
265 /* Check if it is a rts and the fib matches the one of the socket. */
266 fibnum = M_GETFIB(m);
267 if (proto->sp_family != PF_ROUTE ||
268 rp->rcb_socket == NULL ||
269 rp->rcb_socket->so_fibnum == fibnum)
272 /* Filtering requested and no match, the socket shall be skipped. */
277 rts_input(struct mbuf *m)
279 struct sockproto route_proto;
280 unsigned short *family;
283 route_proto.sp_family = PF_ROUTE;
284 tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
286 family = (unsigned short *)(tag + 1);
287 route_proto.sp_protocol = *family;
288 m_tag_delete(m, tag);
290 route_proto.sp_protocol = 0;
292 raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
296 * It really doesn't make any sense at all for this code to share much
297 * with raw_usrreq.c, since its functionality is so restricted. XXX
300 rts_abort(struct socket *so)
303 raw_usrreqs.pru_abort(so);
307 rts_close(struct socket *so)
310 raw_usrreqs.pru_close(so);
313 /* pru_accept is EOPNOTSUPP */
316 rts_attach(struct socket *so, int proto, struct thread *td)
321 KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
324 rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
326 so->so_pcb = (caddr_t)rp;
327 so->so_fibnum = td->td_proc->p_fibnum;
328 error = raw_attach(so, proto);
336 switch(rp->rcb_proto.sp_protocol) {
338 V_route_cb.ip_count++;
341 V_route_cb.ip6_count++;
344 V_route_cb.any_count++;
347 so->so_options |= SO_USELOOPBACK;
352 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
355 return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
359 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
362 return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
365 /* pru_connect2 is EOPNOTSUPP */
366 /* pru_control is EOPNOTSUPP */
369 rts_detach(struct socket *so)
371 struct rawcb *rp = sotorawcb(so);
373 KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
376 switch(rp->rcb_proto.sp_protocol) {
378 V_route_cb.ip_count--;
381 V_route_cb.ip6_count--;
384 V_route_cb.any_count--;
386 raw_usrreqs.pru_detach(so);
390 rts_disconnect(struct socket *so)
393 return (raw_usrreqs.pru_disconnect(so));
396 /* pru_listen is EOPNOTSUPP */
399 rts_peeraddr(struct socket *so, struct sockaddr **nam)
402 return (raw_usrreqs.pru_peeraddr(so, nam));
405 /* pru_rcvd is EOPNOTSUPP */
406 /* pru_rcvoob is EOPNOTSUPP */
409 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
410 struct mbuf *control, struct thread *td)
413 return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
416 /* pru_sense is null */
419 rts_shutdown(struct socket *so)
422 return (raw_usrreqs.pru_shutdown(so));
426 rts_sockaddr(struct socket *so, struct sockaddr **nam)
429 return (raw_usrreqs.pru_sockaddr(so, nam));
432 static struct pr_usrreqs route_usrreqs = {
433 .pru_abort = rts_abort,
434 .pru_attach = rts_attach,
435 .pru_bind = rts_bind,
436 .pru_connect = rts_connect,
437 .pru_detach = rts_detach,
438 .pru_disconnect = rts_disconnect,
439 .pru_peeraddr = rts_peeraddr,
440 .pru_send = rts_send,
441 .pru_shutdown = rts_shutdown,
442 .pru_sockaddr = rts_sockaddr,
443 .pru_close = rts_close,
446 #ifndef _SOCKADDR_UNION_DEFINED
447 #define _SOCKADDR_UNION_DEFINED
449 * The union of all possible address formats we handle.
451 union sockaddr_union {
453 struct sockaddr_in sin;
454 struct sockaddr_in6 sin6;
456 #endif /* _SOCKADDR_UNION_DEFINED */
459 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
460 struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
462 #if defined(INET) || defined(INET6)
463 struct epoch_tracker et;
466 /* First, see if the returned address is part of the jail. */
467 if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
468 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
472 switch (info->rti_info[RTAX_DST]->sa_family) {
482 * Try to find an address on the given outgoing interface
483 * that belongs to the jail.
486 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
489 if (sa->sa_family != AF_INET)
491 ia = ((struct sockaddr_in *)sa)->sin_addr;
492 if (prison_check_ip4(cred, &ia) == 0) {
500 * As a last resort return the 'default' jail address.
502 ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
504 if (prison_get_ip4(cred, &ia) != 0)
507 bzero(&saun->sin, sizeof(struct sockaddr_in));
508 saun->sin.sin_len = sizeof(struct sockaddr_in);
509 saun->sin.sin_family = AF_INET;
510 saun->sin.sin_addr.s_addr = ia.s_addr;
511 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
524 * Try to find an address on the given outgoing interface
525 * that belongs to the jail.
528 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
531 if (sa->sa_family != AF_INET6)
533 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
534 &ia6, sizeof(struct in6_addr));
535 if (prison_check_ip6(cred, &ia6) == 0) {
543 * As a last resort return the 'default' jail address.
545 ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
547 if (prison_get_ip6(cred, &ia6) != 0)
550 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
551 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
552 saun->sin6.sin6_family = AF_INET6;
553 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
554 if (sa6_recoverscope(&saun->sin6) != 0)
556 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
567 * Fills in @info based on userland-provided @rtm message.
569 * Returns 0 on success.
572 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info)
577 rtm->rtm_pid = curproc->p_pid;
578 info->rti_addrs = rtm->rtm_addrs;
580 info->rti_mflags = rtm->rtm_inits;
581 info->rti_rmx = &rtm->rtm_rmx;
584 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
585 * link-local address because rtrequest requires addresses with
588 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
591 if (rtm->rtm_flags & RTF_RNH_LOCKED)
593 info->rti_flags = rtm->rtm_flags;
594 if (info->rti_info[RTAX_DST] == NULL ||
595 info->rti_info[RTAX_DST]->sa_family >= AF_MAX ||
596 (info->rti_info[RTAX_GATEWAY] != NULL &&
597 info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
599 saf = info->rti_info[RTAX_DST]->sa_family;
601 * Verify that the caller has the appropriate privilege; RTM_GET
602 * is the only operation the non-superuser is allowed.
604 if (rtm->rtm_type != RTM_GET) {
605 error = priv_check(curthread, PRIV_NET_ROUTE);
611 * The given gateway address may be an interface address.
612 * For example, issuing a "route change" command on a route
613 * entry that was created from a tunnel, and the gateway
614 * address given is the local end point. In this case the
615 * RTF_GATEWAY flag must be cleared or the destination will
616 * not be reachable even though there is no error message.
618 if (info->rti_info[RTAX_GATEWAY] != NULL &&
619 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
620 struct rt_addrinfo ginfo;
621 struct sockaddr *gdst;
622 struct sockaddr_storage ss;
624 bzero(&ginfo, sizeof(ginfo));
625 bzero(&ss, sizeof(ss));
626 ss.ss_len = sizeof(ss);
628 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
629 gdst = info->rti_info[RTAX_GATEWAY];
632 * A host route through the loopback interface is
633 * installed for each interface adddress. In pre 8.0
634 * releases the interface address of a PPP link type
635 * is not reachable locally. This behavior is fixed as
636 * part of the new L2/L3 redesign and rewrite work. The
637 * signature of this interface address route is the
638 * AF_LINK sa_family type of the rt_gateway, and the
639 * rt_ifp has the IFF_LOOPBACK flag set.
641 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
642 if (ss.ss_family == AF_LINK &&
643 ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
644 info->rti_flags &= ~RTF_GATEWAY;
645 info->rti_flags |= RTF_GWFLAG_COMPAT;
647 rib_free_info(&ginfo);
655 * Handles RTM_GET message from routing socket, returning matching rt.
658 * 0 on success, with locked and referenced matching rt in @rt_nrt
662 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
663 struct rt_msghdr *rtm, struct rtentry **ret_nrt)
667 struct rib_head *rnh;
670 saf = info->rti_info[RTAX_DST]->sa_family;
672 rnh = rt_tables_get_rnh(fibnum, saf);
674 return (EAFNOSUPPORT);
678 if (info->rti_info[RTAX_NETMASK] == NULL) {
680 * Provide longest prefix match for
681 * address lookup (no mask).
682 * 'route -n get addr'
684 rt = (struct rtentry *) rnh->rnh_matchaddr(
685 info->rti_info[RTAX_DST], &rnh->head);
687 rt = (struct rtentry *) rnh->rnh_lookup(
688 info->rti_info[RTAX_DST],
689 info->rti_info[RTAX_NETMASK], &rnh->head);
697 * for RTM_GET, gate is optional even with multipath.
698 * if gate == NULL the first match is returned.
699 * (no need to call rt_mpath_matchgate if gate == NULL)
701 if (rt_mpath_capable(rnh) && info->rti_info[RTAX_GATEWAY]) {
702 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
710 * If performing proxied L2 entry insertion, and
711 * the actual PPP host entry is found, perform
712 * another search to retrieve the prefix route of
713 * the local end point of the PPP link.
715 if (rtm->rtm_flags & RTF_ANNOUNCE) {
716 struct sockaddr laddr;
718 if (rt->rt_ifp != NULL &&
719 rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
720 struct epoch_tracker et;
724 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
728 rt_maskedcopy(ifa->ifa_addr,
732 rt_maskedcopy(rt->rt_ifa->ifa_addr,
734 rt->rt_ifa->ifa_netmask);
736 * refactor rt and no lock operation necessary
738 rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
755 * Update sockaddrs, flags, etc in @prtm based on @rt data.
756 * Assumes @rt is locked.
757 * rtm can be reallocated.
759 * Returns 0 on success, along with pointer to (potentially reallocated)
764 update_rtm_from_rte(struct rt_addrinfo *info, struct rt_msghdr **prtm,
765 int alloc_len, struct rtentry *rt)
767 struct sockaddr_storage netmask_ss;
769 union sockaddr_union saun;
770 struct rt_msghdr *rtm, *orig_rtm = NULL;
778 info->rti_info[RTAX_DST] = rt_key(rt);
779 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
780 info->rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
781 rt_mask(rt), &netmask_ss);
782 info->rti_info[RTAX_GENMASK] = 0;
784 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
786 info->rti_info[RTAX_IFP] =
787 ifp->if_addr->ifa_addr;
788 error = rtm_get_jailed(info, ifp, rt,
789 &saun, curthread->td_ucred);
792 if (ifp->if_flags & IFF_POINTOPOINT)
793 info->rti_info[RTAX_BRD] =
794 rt->rt_ifa->ifa_dstaddr;
795 rtm->rtm_index = ifp->if_index;
797 info->rti_info[RTAX_IFP] = NULL;
798 info->rti_info[RTAX_IFA] = NULL;
800 } else if (ifp != NULL)
801 rtm->rtm_index = ifp->if_index;
803 /* Check if we need to realloc storage */
804 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
805 if (len > alloc_len) {
806 struct rt_msghdr *tmp_rtm;
808 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
811 bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
817 * Delay freeing original rtm as info contains
818 * data referencing it.
822 w.w_tmem = (caddr_t)rtm;
823 w.w_tmemsize = alloc_len;
824 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
826 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
827 rtm->rtm_flags = RTF_GATEWAY |
828 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
830 rtm->rtm_flags = rt->rt_flags;
831 rt_getmetrics(rt, &rtm->rtm_rmx);
832 rtm->rtm_addrs = info->rti_addrs;
834 if (orig_rtm != NULL)
835 free(orig_rtm, M_TEMP);
843 route_output(struct mbuf *m, struct socket *so, ...)
845 struct rt_msghdr *rtm = NULL;
846 struct rtentry *rt = NULL;
847 struct rt_addrinfo info;
848 struct epoch_tracker et;
850 struct sockaddr_storage ss;
851 struct sockaddr_in6 *sin6;
852 int i, rti_need_deembed = 0;
854 int alloc_len = 0, len, error = 0, fibnum;
855 sa_family_t saf = AF_UNSPEC;
858 fibnum = so->so_fibnum;
860 #define senderr(e) { error = e; goto flush;}
861 if (m == NULL || ((m->m_len < sizeof(long)) &&
862 (m = m_pullup(m, sizeof(long))) == NULL))
864 if ((m->m_flags & M_PKTHDR) == 0)
865 panic("route_output");
867 len = m->m_pkthdr.len;
868 if (len < sizeof(*rtm) ||
869 len != mtod(m, struct rt_msghdr *)->rtm_msglen)
873 * Most of current messages are in range 200-240 bytes,
874 * minimize possible re-allocation on reply using larger size
875 * buffer aligned on 1k boundaty.
877 alloc_len = roundup2(len, 1024);
878 if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
881 m_copydata(m, 0, len, (caddr_t)rtm);
882 bzero(&info, sizeof(info));
883 bzero(&w, sizeof(w));
885 if (rtm->rtm_version != RTM_VERSION) {
886 /* Do not touch message since format is unknown */
889 senderr(EPROTONOSUPPORT);
893 * Starting from here, it is possible
894 * to alter original message and insert
895 * caller PID and error value.
898 if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) {
902 saf = info.rti_info[RTAX_DST]->sa_family;
904 /* support for new ARP code */
905 if (rtm->rtm_flags & RTF_LLDATA) {
906 error = lla_rt_output(rtm, &info);
909 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
914 switch (rtm->rtm_type) {
915 struct rtentry *saved_nrt;
919 if (rtm->rtm_type == RTM_ADD) {
920 if (info.rti_info[RTAX_GATEWAY] == NULL)
924 error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
926 if (error == 0 && saved_nrt != NULL) {
928 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
931 rtm->rtm_index = saved_nrt->rt_ifp->if_index;
932 RT_REMREF(saved_nrt);
933 RT_UNLOCK(saved_nrt);
939 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
946 /* rt_msg2() will not be used when RTM_DELETE fails. */
947 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
952 error = handle_rtm_get(&info, fibnum, rtm, &rt);
958 if (!can_export_rte(curthread->td_ucred, rt)) {
962 error = update_rtm_from_rte(&info, &rtm, alloc_len, rt);
964 * Note that some sockaddr pointers may have changed to
965 * point to memory outsize @rtm. Some may be pointing
966 * to the on-stack variables.
967 * Given that, any pointer in @info CANNOT BE USED.
971 * scopeid deembedding has been performed while
972 * writing updated rtm in rtsock_msg_buffer().
973 * With that in mind, skip deembedding procedure below.
976 rti_need_deembed = 0;
994 if (rti_need_deembed) {
995 /* sin6_scope_id is recovered before sending rtm. */
996 sin6 = (struct sockaddr_in6 *)&ss;
997 for (i = 0; i < RTAX_MAX; i++) {
998 if (info.rti_info[i] == NULL)
1000 if (info.rti_info[i]->sa_family != AF_INET6)
1002 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1003 if (sa6_recoverscope(sin6) == 0)
1004 bcopy(sin6, info.rti_info[i],
1010 send_rtm_reply(so, rtm, m, saf, fibnum, error);
1016 * Sends the prepared reply message in @rtm to all rtsock clients.
1017 * Frees @m and @rtm.
1021 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1022 sa_family_t saf, u_int fibnum, int rtm_errno)
1024 struct rawcb *rp = NULL;
1027 * Check to see if we don't want our own messages.
1029 if ((so->so_options & SO_USELOOPBACK) == 0) {
1030 if (V_route_cb.any_count <= 1) {
1036 /* There is another listener, so construct message */
1042 rtm->rtm_errno = rtm_errno;
1044 rtm->rtm_flags |= RTF_DONE;
1046 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1047 if (m->m_pkthdr.len < rtm->rtm_msglen) {
1050 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
1051 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1056 M_SETFIB(m, fibnum);
1057 m->m_flags |= RTS_FILTER_FIB;
1060 * XXX insure we don't get a copy by
1061 * invalidating our protocol
1063 unsigned short family = rp->rcb_proto.sp_family;
1064 rp->rcb_proto.sp_family = 0;
1065 rt_dispatch(m, saf);
1066 rp->rcb_proto.sp_family = family;
1068 rt_dispatch(m, saf);
1074 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
1077 bzero(out, sizeof(*out));
1078 out->rmx_mtu = rt->rt_mtu;
1079 out->rmx_weight = rt->rt_weight;
1080 out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
1081 out->rmx_nhidx = nhop_get_idx(rt->rt_nhop);
1082 /* Kernel -> userland timebase conversion. */
1083 out->rmx_expire = rt->rt_expire ?
1084 rt->rt_expire - time_uptime + time_second : 0;
1088 * Extract the addresses of the passed sockaddrs.
1089 * Do a little sanity checking so as to avoid bad memory references.
1090 * This data is derived straight from userland.
1093 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1095 struct sockaddr *sa;
1098 for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1099 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1101 sa = (struct sockaddr *)cp;
1105 if (cp + sa->sa_len > cplim)
1108 * there are no more.. quit now
1109 * If there are more bits, they are in error.
1110 * I've seen this. route(1) can evidently generate these.
1111 * This causes kernel to core dump.
1112 * for compatibility, If we see this, point to a safe address.
1114 if (sa->sa_len == 0) {
1115 rtinfo->rti_info[i] = &sa_zero;
1116 return (0); /* should be EINVAL but for compat */
1120 if (sa->sa_family == AF_INET6)
1121 sa6_embedscope((struct sockaddr_in6 *)sa,
1124 rtinfo->rti_info[i] = sa;
1131 * Fill in @dmask with valid netmask leaving original @smask
1132 * intact. Mostly used with radix netmasks.
1134 static struct sockaddr *
1135 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask,
1136 struct sockaddr_storage *dmask)
1138 if (dst == NULL || smask == NULL)
1141 memset(dmask, 0, dst->sa_len);
1142 memcpy(dmask, smask, smask->sa_len);
1143 dmask->ss_len = dst->sa_len;
1144 dmask->ss_family = dst->sa_family;
1146 return ((struct sockaddr *)dmask);
1150 * Writes information related to @rtinfo object to newly-allocated mbuf.
1151 * Assumes MCLBYTES is enough to construct any message.
1152 * Used for OS notifications of vaious events (if/ifa announces,etc)
1154 * Returns allocated mbuf or NULL on failure.
1156 static struct mbuf *
1157 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1159 struct rt_msghdr *rtm;
1162 struct sockaddr *sa;
1164 struct sockaddr_storage ss;
1165 struct sockaddr_in6 *sin6;
1173 len = sizeof(struct ifa_msghdr);
1178 len = sizeof(struct ifma_msghdr);
1182 len = sizeof(struct if_msghdr);
1185 case RTM_IFANNOUNCE:
1187 len = sizeof(struct if_announcemsghdr);
1191 len = sizeof(struct rt_msghdr);
1194 /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1195 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1197 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1199 m = m_gethdr(M_NOWAIT, MT_DATA);
1203 m->m_pkthdr.len = m->m_len = len;
1204 rtm = mtod(m, struct rt_msghdr *);
1205 bzero((caddr_t)rtm, len);
1206 for (i = 0; i < RTAX_MAX; i++) {
1207 if ((sa = rtinfo->rti_info[i]) == NULL)
1209 rtinfo->rti_addrs |= (1 << i);
1212 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1213 sin6 = (struct sockaddr_in6 *)&ss;
1214 bcopy(sa, sin6, sizeof(*sin6));
1215 if (sa6_recoverscope(sin6) == 0)
1216 sa = (struct sockaddr *)sin6;
1219 m_copyback(m, len, dlen, (caddr_t)sa);
1222 if (m->m_pkthdr.len != len) {
1226 rtm->rtm_msglen = len;
1227 rtm->rtm_version = RTM_VERSION;
1228 rtm->rtm_type = type;
1233 * Writes information related to @rtinfo object to preallocated buffer.
1234 * Stores needed size in @plen. If @w is NULL, calculates size without
1236 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1238 * Returns 0 on success.
1242 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1245 int len, buflen = 0, dlen;
1247 struct rt_msghdr *rtm = NULL;
1249 struct sockaddr_storage ss;
1250 struct sockaddr_in6 *sin6;
1252 #ifdef COMPAT_FREEBSD32
1253 bool compat32 = false;
1260 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1261 #ifdef COMPAT_FREEBSD32
1262 if (w->w_req->flags & SCTL_MASK32) {
1263 len = sizeof(struct ifa_msghdrl32);
1267 len = sizeof(struct ifa_msghdrl);
1269 len = sizeof(struct ifa_msghdr);
1273 #ifdef COMPAT_FREEBSD32
1274 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1275 if (w->w_op == NET_RT_IFLISTL)
1276 len = sizeof(struct if_msghdrl32);
1278 len = sizeof(struct if_msghdr32);
1283 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1284 len = sizeof(struct if_msghdrl);
1286 len = sizeof(struct if_msghdr);
1290 len = sizeof(struct ifma_msghdr);
1294 len = sizeof(struct rt_msghdr);
1298 rtm = (struct rt_msghdr *)w->w_tmem;
1299 buflen = w->w_tmemsize - len;
1300 cp = (caddr_t)w->w_tmem + len;
1303 rtinfo->rti_addrs = 0;
1304 for (i = 0; i < RTAX_MAX; i++) {
1305 struct sockaddr *sa;
1307 if ((sa = rtinfo->rti_info[i]) == NULL)
1309 rtinfo->rti_addrs |= (1 << i);
1310 #ifdef COMPAT_FREEBSD32
1312 dlen = SA_SIZE32(sa);
1316 if (cp != NULL && buflen >= dlen) {
1318 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1319 sin6 = (struct sockaddr_in6 *)&ss;
1320 bcopy(sa, sin6, sizeof(*sin6));
1321 if (sa6_recoverscope(sin6) == 0)
1322 sa = (struct sockaddr *)sin6;
1325 bcopy((caddr_t)sa, cp, (unsigned)dlen);
1328 } else if (cp != NULL) {
1330 * Buffer too small. Count needed size
1331 * and return with error.
1340 dlen = ALIGN(len) - len;
1352 /* fill header iff buffer is large enough */
1353 rtm->rtm_version = RTM_VERSION;
1354 rtm->rtm_type = type;
1355 rtm->rtm_msglen = len;
1360 if (w != NULL && cp == NULL)
1367 * This routine is called to generate a message from the routing
1368 * socket indicating that a redirect has occurred, a routing lookup
1369 * has failed, or that a protocol has detected timeouts to a particular
1373 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1376 struct rt_msghdr *rtm;
1378 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1380 if (V_route_cb.any_count == 0)
1382 m = rtsock_msg_mbuf(type, rtinfo);
1386 if (fibnum != RT_ALL_FIBS) {
1387 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1388 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1389 M_SETFIB(m, fibnum);
1390 m->m_flags |= RTS_FILTER_FIB;
1393 rtm = mtod(m, struct rt_msghdr *);
1394 rtm->rtm_flags = RTF_DONE | flags;
1395 rtm->rtm_errno = error;
1396 rtm->rtm_addrs = rtinfo->rti_addrs;
1397 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1401 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1404 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1408 * This routine is called to generate a message from the routing
1409 * socket indicating that the status of a network interface has changed.
1412 rt_ifmsg(struct ifnet *ifp)
1414 struct if_msghdr *ifm;
1416 struct rt_addrinfo info;
1418 if (V_route_cb.any_count == 0)
1420 bzero((caddr_t)&info, sizeof(info));
1421 m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1424 ifm = mtod(m, struct if_msghdr *);
1425 ifm->ifm_index = ifp->if_index;
1426 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1427 if_data_copy(ifp, &ifm->ifm_data);
1429 rt_dispatch(m, AF_UNSPEC);
1433 * Announce interface address arrival/withdraw.
1434 * Please do not call directly, use rt_addrmsg().
1435 * Assume input data to be valid.
1436 * Returns 0 on success.
1439 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1441 struct rt_addrinfo info;
1442 struct sockaddr *sa;
1445 struct ifa_msghdr *ifam;
1446 struct ifnet *ifp = ifa->ifa_ifp;
1447 struct sockaddr_storage ss;
1449 if (V_route_cb.any_count == 0)
1452 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1454 bzero((caddr_t)&info, sizeof(info));
1455 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1456 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1457 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1458 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1459 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1460 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1462 ifam = mtod(m, struct ifa_msghdr *);
1463 ifam->ifam_index = ifp->if_index;
1464 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1465 ifam->ifam_flags = ifa->ifa_flags;
1466 ifam->ifam_addrs = info.rti_addrs;
1468 if (fibnum != RT_ALL_FIBS) {
1469 M_SETFIB(m, fibnum);
1470 m->m_flags |= RTS_FILTER_FIB;
1473 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1479 * Announce route addition/removal to rtsock based on @rt data.
1480 * Callers are advives to use rt_routemsg() instead of using this
1481 * function directly.
1482 * Assume @rt data is consistent.
1484 * Returns 0 on success.
1487 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs,
1490 struct sockaddr_storage ss;
1491 struct rt_addrinfo info;
1493 if (V_route_cb.any_count == 0)
1496 bzero((caddr_t)&info, sizeof(info));
1497 info.rti_info[RTAX_DST] = rt_key(rt);
1498 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss);
1499 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1500 info.rti_flags = rt->rt_flags;
1503 return (rtsock_routemsg_info(cmd, &info, fibnum));
1507 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1509 struct rt_msghdr *rtm;
1510 struct sockaddr *sa;
1513 if (V_route_cb.any_count == 0)
1516 if (info->rti_flags & RTF_HOST)
1517 info->rti_info[RTAX_NETMASK] = NULL;
1519 m = rtsock_msg_mbuf(cmd, info);
1523 if (fibnum != RT_ALL_FIBS) {
1524 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1525 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1526 M_SETFIB(m, fibnum);
1527 m->m_flags |= RTS_FILTER_FIB;
1530 rtm = mtod(m, struct rt_msghdr *);
1531 rtm->rtm_addrs = info->rti_addrs;
1532 if (info->rti_ifp != NULL)
1533 rtm->rtm_index = info->rti_ifp->if_index;
1534 /* Add RTF_DONE to indicate command 'completion' required by API */
1535 info->rti_flags |= RTF_DONE;
1536 /* Reported routes has to be up */
1537 if (cmd == RTM_ADD || cmd == RTM_CHANGE)
1538 info->rti_flags |= RTF_UP;
1539 rtm->rtm_flags = info->rti_flags;
1541 sa = info->rti_info[RTAX_DST];
1542 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1548 * This is the analogue to the rt_newaddrmsg which performs the same
1549 * function but for multicast group memberhips. This is easier since
1550 * there is no route state to worry about.
1553 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1555 struct rt_addrinfo info;
1556 struct mbuf *m = NULL;
1557 struct ifnet *ifp = ifma->ifma_ifp;
1558 struct ifma_msghdr *ifmam;
1560 if (V_route_cb.any_count == 0)
1563 bzero((caddr_t)&info, sizeof(info));
1564 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1565 if (ifp && ifp->if_addr)
1566 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1568 info.rti_info[RTAX_IFP] = NULL;
1570 * If a link-layer address is present, present it as a ``gateway''
1571 * (similarly to how ARP entries, e.g., are presented).
1573 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1574 m = rtsock_msg_mbuf(cmd, &info);
1577 ifmam = mtod(m, struct ifma_msghdr *);
1578 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1580 ifmam->ifmam_index = ifp->if_index;
1581 ifmam->ifmam_addrs = info.rti_addrs;
1582 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1585 static struct mbuf *
1586 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1587 struct rt_addrinfo *info)
1589 struct if_announcemsghdr *ifan;
1592 if (V_route_cb.any_count == 0)
1594 bzero((caddr_t)info, sizeof(*info));
1595 m = rtsock_msg_mbuf(type, info);
1597 ifan = mtod(m, struct if_announcemsghdr *);
1598 ifan->ifan_index = ifp->if_index;
1599 strlcpy(ifan->ifan_name, ifp->if_xname,
1600 sizeof(ifan->ifan_name));
1601 ifan->ifan_what = what;
1607 * This is called to generate routing socket messages indicating
1608 * IEEE80211 wireless events.
1609 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1612 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1615 struct rt_addrinfo info;
1617 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1620 * Append the ieee80211 data. Try to stick it in the
1621 * mbuf containing the ifannounce msg; otherwise allocate
1622 * a new mbuf and append.
1624 * NB: we assume m is a single mbuf.
1626 if (data_len > M_TRAILINGSPACE(m)) {
1627 struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1632 bcopy(data, mtod(n, void *), data_len);
1633 n->m_len = data_len;
1635 } else if (data_len > 0) {
1636 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1637 m->m_len += data_len;
1639 if (m->m_flags & M_PKTHDR)
1640 m->m_pkthdr.len += data_len;
1641 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1642 rt_dispatch(m, AF_UNSPEC);
1647 * This is called to generate routing socket messages indicating
1648 * network interface arrival and departure.
1651 rt_ifannouncemsg(struct ifnet *ifp, int what)
1654 struct rt_addrinfo info;
1656 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1658 rt_dispatch(m, AF_UNSPEC);
1662 rt_dispatch(struct mbuf *m, sa_family_t saf)
1667 * Preserve the family from the sockaddr, if any, in an m_tag for
1668 * use when injecting the mbuf into the routing socket buffer from
1671 if (saf != AF_UNSPEC) {
1672 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1678 *(unsigned short *)(tag + 1) = saf;
1679 m_tag_prepend(m, tag);
1683 m->m_pkthdr.rcvif = V_loif;
1689 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */
1693 * Checks if rte can be exported v.r.t jails/vnets.
1695 * Returns 1 if it can, 0 otherwise.
1698 can_export_rte(struct ucred *td_ucred, const struct rtentry *rt)
1701 if ((rt->rt_flags & RTF_HOST) == 0
1702 ? jailed_without_vnet(td_ucred)
1703 : prison_if(td_ucred, rt_key_const(rt)) != 0)
1709 * This is used in dumping the kernel table via sysctl().
1712 sysctl_dumpentry(struct radix_node *rn, void *vw)
1714 struct walkarg *w = vw;
1715 struct rtentry *rt = (struct rtentry *)rn;
1716 int error = 0, size;
1717 struct rt_addrinfo info;
1718 struct sockaddr_storage ss;
1722 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1724 if (!can_export_rte(w->w_req->td->td_ucred, rt))
1726 bzero((caddr_t)&info, sizeof(info));
1727 info.rti_info[RTAX_DST] = rt_key(rt);
1728 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1729 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1731 info.rti_info[RTAX_GENMASK] = 0;
1732 if (rt->rt_ifp && !(rt->rt_ifp->if_flags & IFF_DYING)) {
1733 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1734 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1735 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1736 info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1738 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1740 if (w->w_req && w->w_tmem) {
1741 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1743 bzero(&rtm->rtm_index,
1744 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1745 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1746 rtm->rtm_flags = RTF_GATEWAY |
1747 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1749 rtm->rtm_flags = rt->rt_flags;
1750 rt_getmetrics(rt, &rtm->rtm_rmx);
1751 rtm->rtm_index = rt->rt_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);
2164 * Unfortunately, RTF_ values are expressed as raw masks rather than powers of
2165 * 2, so we cannot use them as nice C99 initializer indices below.
2167 static const char * const rtf_flag_strings[] = {
2194 [30] = "RNH_LOCKED",
2195 [31] = "GWFLAG_COMPAT",
2198 static const char * __pure
2199 rt_flag_name(unsigned idx)
2201 if (idx >= nitems(rtf_flag_strings))
2202 return ("INVALID_FLAG");
2203 if (rtf_flag_strings[idx] == NULL)
2205 return (rtf_flag_strings[idx]);
2209 rt_dumpaddr_ddb(const char *name, const struct sockaddr *sa)
2211 char buf[INET6_ADDRSTRLEN], *res;
2216 else if (sa->sa_family == AF_INET) {
2217 res = inet_ntop(AF_INET,
2218 &((const struct sockaddr_in *)sa)->sin_addr,
2220 } else if (sa->sa_family == AF_INET6) {
2221 res = inet_ntop(AF_INET6,
2222 &((const struct sockaddr_in6 *)sa)->sin6_addr,
2224 } else if (sa->sa_family == AF_LINK) {
2229 db_printf("%s <%s> ", name, res);
2233 db_printf("%s <af:%d> ", name, sa->sa_family);
2237 rt_dumpentry_ddb(struct radix_node *rn, void *arg __unused)
2239 struct sockaddr_storage ss;
2243 /* If RNTORT is important, put it in a header. */
2246 rt_dumpaddr_ddb("dst", rt_key(rt));
2247 rt_dumpaddr_ddb("gateway", rt->rt_gateway);
2248 rt_dumpaddr_ddb("netmask", rtsock_fix_netmask(rt_key(rt), rt_mask(rt),
2250 if (rt->rt_ifp != NULL && (rt->rt_ifp->if_flags & IFF_DYING) == 0) {
2251 rt_dumpaddr_ddb("ifp", rt->rt_ifp->if_addr->ifa_addr);
2252 rt_dumpaddr_ddb("ifa", rt->rt_ifa->ifa_addr);
2255 db_printf("flags ");
2256 flags = rt->rt_flags;
2260 while ((idx = ffs(flags)) > 0) {
2263 if (flags != rt->rt_flags)
2265 db_printf("%s", rt_flag_name(idx));
2267 flags &= ~(1ul << idx);
2274 DB_SHOW_COMMAND(routetable, db_show_routetable_cmd)
2276 struct rib_head *rnh;
2286 for (; i <= lim; i++) {
2287 rnh = rt_tables_get_rnh(0, i);
2290 db_printf("%s: AF %d not supported?\n",
2297 if (!have_addr && i > 1)
2300 db_printf("Route table for AF %d%s%s%s:\n", i,
2301 (i == AF_INET || i == AF_INET6) ? " (" : "",
2302 (i == AF_INET) ? "INET" : (i == AF_INET6) ? "INET6" : "",
2303 (i == AF_INET || i == AF_INET6) ? ")" : "");
2305 error = rnh->rnh_walktree(&rnh->head, rt_dumpentry_ddb, NULL);
2307 db_printf("%s: walktree(%d): %d\n", __func__, i,
2312 _DB_FUNC(_show, route, db_show_route_cmd, db_show_table, CS_OWN, NULL)
2314 char buf[INET6_ADDRSTRLEN], *bp;
2315 const void *dst_addrp;
2316 struct sockaddr *dstp;
2319 struct sockaddr_in dest_sin;
2320 struct sockaddr_in6 dest_sin6;
2322 uint16_t hextets[8];
2324 int t, af, exp, tokflags;
2327 * Undecoded address family. No double-colon expansion seen yet.
2331 /* Assume INET6 to start; we can work back if guess was wrong. */
2332 tokflags = DRT_WSPACE | DRT_HEX | DRT_HEXADECIMAL;
2335 * db_command has lexed 'show route' for us.
2337 t = db_read_token_flags(tokflags);
2339 t = db_read_token_flags(tokflags);
2342 * tEOL: Just 'show route' isn't a valid mode.
2343 * tMINUS: It's either '-h' or some invalid option. Regardless, usage.
2345 if (t == tEOL || t == tMINUS)
2350 tets = nitems(hextets);
2353 * Each loop iteration, we expect to read one octet (v4) or hextet
2354 * (v6), followed by an appropriate field separator ('.' or ':' or
2357 * At the start of each loop, we're looking for a number (octet or
2360 * INET6 addresses have a special case where they may begin with '::'.
2362 for (i = 0; i < tets; i++) {
2363 t = db_read_token_flags(tokflags);
2365 if (t == tCOLONCOLON) {
2366 /* INET6 with leading '::' or invalid. */
2368 db_printf("Parse error: unexpected extra "
2377 } else if (t == tNUMBER) {
2379 * Lexer separates out '-' as tMINUS, but make the
2380 * assumption explicit here.
2382 MPASS(db_tok_number >= 0);
2384 if (af == AF_INET && db_tok_number > UINT8_MAX) {
2385 db_printf("Not a valid v4 octet: %ld\n",
2386 (long)db_tok_number);
2389 hextets[i] = db_tok_number;
2390 } else if (t == tEOL) {
2392 * We can only detect the end of an IPv6 address in
2393 * compact representation with EOL.
2395 if (af != AF_INET6 || exp < 0) {
2396 db_printf("Parse failed. Got unexpected EOF "
2397 "when the address is not a compact-"
2398 "representation IPv6 address.\n");
2403 db_printf("Parse failed. Unexpected token %d.\n", t);
2407 /* Next, look for a separator, if appropriate. */
2411 t = db_read_token_flags(tokflags);
2417 if (t == tCOLONCOLON) {
2428 /* Need to fixup the first parsed number. */
2429 if (hextets[0] > 0x255 ||
2430 (hextets[0] & 0xf0) > 0x90 ||
2431 (hextets[0] & 0xf) > 9) {
2432 db_printf("Not a valid v4 octet: %x\n",
2438 dn = (hn >> 8) * 100 +
2439 ((hn >> 4) & 0xf) * 10 +
2444 /* Switch to decimal for remaining octets. */
2445 tokflags &= ~DRT_RADIX_MASK;
2446 tokflags |= DRT_DECIMAL;
2452 db_printf("Parse error. Unexpected token %d.\n", t);
2454 } else if (af == AF_INET) {
2457 db_printf("Expected '.' (%d) between octets but got "
2458 "(%d).\n", tDOT, t);
2461 } else if (af == AF_INET6) {
2464 if (t == tCOLONCOLON) {
2471 db_printf("Got bogus second '::' in v6 "
2477 * Handle in the earlier part of the loop
2478 * because we need to handle trailing :: too.
2484 db_printf("Expected ':' (%d) or '::' (%d) between "
2485 "hextets but got (%d).\n", tCOLON, tCOLONCOLON, t);
2490 /* Check for trailing garbage. */
2492 t = db_read_token_flags(tokflags);
2494 db_printf("Got unexpected garbage after address "
2501 * Need to expand compact INET6 addresses.
2503 * Technically '::' for a single ':0:' is MUST NOT but just in case,
2504 * don't bother expanding that form (exp >= 0 && i == tets case).
2506 if (af == AF_INET6 && exp >= 0 && i < tets) {
2508 memmove(&hextets[exp + 1 + (nitems(hextets) - i)],
2510 (i - (exp + 1)) * sizeof(hextets[0]));
2512 memset(&hextets[exp + 1], 0, (nitems(hextets) - i) *
2513 sizeof(hextets[0]));
2516 memset(&u, 0, sizeof(u));
2517 if (af == AF_INET) {
2518 u.dest_sin.sin_family = AF_INET;
2519 u.dest_sin.sin_len = sizeof(u.dest_sin);
2520 u.dest_sin.sin_addr.s_addr = htonl(
2521 ((uint32_t)hextets[0] << 24) |
2522 ((uint32_t)hextets[1] << 16) |
2523 ((uint32_t)hextets[2] << 8) |
2524 (uint32_t)hextets[3]);
2525 dstp = (void *)&u.dest_sin;
2526 dst_addrp = &u.dest_sin.sin_addr;
2527 } else if (af == AF_INET6) {
2528 u.dest_sin6.sin6_family = AF_INET6;
2529 u.dest_sin6.sin6_len = sizeof(u.dest_sin6);
2530 for (i = 0; i < nitems(hextets); i++)
2531 u.dest_sin6.sin6_addr.s6_addr16[i] = htons(hextets[i]);
2532 dstp = (void *)&u.dest_sin6;
2533 dst_addrp = &u.dest_sin6.sin6_addr;
2537 /* Appease Clang false positive: */
2541 bp = inet_ntop(af, dst_addrp, buf, sizeof(buf));
2543 db_printf("Looking up route to destination '%s'\n", bp);
2546 rt = rtalloc1(dstp, 0, RTF_RNH_LOCKED);
2550 db_printf("Could not get route for that server.\n");
2554 rt_dumpentry_ddb((void *)rt, NULL);
2559 db_printf("Usage: 'show route <address>'\n"
2560 " Currently accepts only dotted-decimal INET or colon-separated\n"
2561 " hextet INET6 addresses.\n");