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>
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, 0, "");
167 struct sysctl_req *w_req;
170 static void rts_input(struct mbuf *m);
171 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
172 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
173 struct walkarg *w, int *plen);
174 static int rt_xaddrs(caddr_t cp, caddr_t cplim,
175 struct rt_addrinfo *rtinfo);
176 static int sysctl_dumpentry(struct radix_node *rn, void *vw);
177 static int sysctl_iflist(int af, struct walkarg *w);
178 static int sysctl_ifmalist(int af, struct walkarg *w);
179 static int route_output(struct mbuf *m, struct socket *so, ...);
180 static void rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
181 static void rt_dispatch(struct mbuf *, sa_family_t);
182 static struct sockaddr *rtsock_fix_netmask(struct sockaddr *dst,
183 struct sockaddr *smask, struct sockaddr_storage *dmask);
184 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
185 struct rt_msghdr *rtm, struct rtentry **ret_nrt);
186 static int update_rtm_from_rte(struct rt_addrinfo *info,
187 struct rt_msghdr **prtm, int alloc_len,
189 static void send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
190 struct mbuf *m, sa_family_t saf, u_int fibnum,
192 static int can_export_rte(struct ucred *td_ucred, const struct rtentry *rt);
194 static struct netisr_handler rtsock_nh = {
196 .nh_handler = rts_input,
197 .nh_proto = NETISR_ROUTE,
198 .nh_policy = NETISR_POLICY_SOURCE,
202 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
206 netisr_getqlimit(&rtsock_nh, &qlimit);
207 error = sysctl_handle_int(oidp, &qlimit, 0, req);
208 if (error || !req->newptr)
212 return (netisr_setqlimit(&rtsock_nh, qlimit));
214 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
215 0, 0, sysctl_route_netisr_maxqlen, "I",
216 "maximum routing socket dispatch queue length");
223 if (IS_DEFAULT_VNET(curvnet)) {
224 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
225 rtsock_nh.nh_qlimit = tmp;
226 netisr_register(&rtsock_nh);
230 netisr_register_vnet(&rtsock_nh);
233 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
238 vnet_rts_uninit(void)
241 netisr_unregister_vnet(&rtsock_nh);
243 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
248 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
253 KASSERT(m != NULL, ("%s: m is NULL", __func__));
254 KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
255 KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
257 /* No filtering requested. */
258 if ((m->m_flags & RTS_FILTER_FIB) == 0)
261 /* Check if it is a rts and the fib matches the one of the socket. */
262 fibnum = M_GETFIB(m);
263 if (proto->sp_family != PF_ROUTE ||
264 rp->rcb_socket == NULL ||
265 rp->rcb_socket->so_fibnum == fibnum)
268 /* Filtering requested and no match, the socket shall be skipped. */
273 rts_input(struct mbuf *m)
275 struct sockproto route_proto;
276 unsigned short *family;
279 route_proto.sp_family = PF_ROUTE;
280 tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
282 family = (unsigned short *)(tag + 1);
283 route_proto.sp_protocol = *family;
284 m_tag_delete(m, tag);
286 route_proto.sp_protocol = 0;
288 raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
292 * It really doesn't make any sense at all for this code to share much
293 * with raw_usrreq.c, since its functionality is so restricted. XXX
296 rts_abort(struct socket *so)
299 raw_usrreqs.pru_abort(so);
303 rts_close(struct socket *so)
306 raw_usrreqs.pru_close(so);
309 /* pru_accept is EOPNOTSUPP */
312 rts_attach(struct socket *so, int proto, struct thread *td)
317 KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
320 rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
322 so->so_pcb = (caddr_t)rp;
323 so->so_fibnum = td->td_proc->p_fibnum;
324 error = raw_attach(so, proto);
332 switch(rp->rcb_proto.sp_protocol) {
334 V_route_cb.ip_count++;
337 V_route_cb.ip6_count++;
340 V_route_cb.any_count++;
343 so->so_options |= SO_USELOOPBACK;
348 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
351 return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
355 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
358 return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
361 /* pru_connect2 is EOPNOTSUPP */
362 /* pru_control is EOPNOTSUPP */
365 rts_detach(struct socket *so)
367 struct rawcb *rp = sotorawcb(so);
369 KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
372 switch(rp->rcb_proto.sp_protocol) {
374 V_route_cb.ip_count--;
377 V_route_cb.ip6_count--;
380 V_route_cb.any_count--;
382 raw_usrreqs.pru_detach(so);
386 rts_disconnect(struct socket *so)
389 return (raw_usrreqs.pru_disconnect(so));
392 /* pru_listen is EOPNOTSUPP */
395 rts_peeraddr(struct socket *so, struct sockaddr **nam)
398 return (raw_usrreqs.pru_peeraddr(so, nam));
401 /* pru_rcvd is EOPNOTSUPP */
402 /* pru_rcvoob is EOPNOTSUPP */
405 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
406 struct mbuf *control, struct thread *td)
409 return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
412 /* pru_sense is null */
415 rts_shutdown(struct socket *so)
418 return (raw_usrreqs.pru_shutdown(so));
422 rts_sockaddr(struct socket *so, struct sockaddr **nam)
425 return (raw_usrreqs.pru_sockaddr(so, nam));
428 static struct pr_usrreqs route_usrreqs = {
429 .pru_abort = rts_abort,
430 .pru_attach = rts_attach,
431 .pru_bind = rts_bind,
432 .pru_connect = rts_connect,
433 .pru_detach = rts_detach,
434 .pru_disconnect = rts_disconnect,
435 .pru_peeraddr = rts_peeraddr,
436 .pru_send = rts_send,
437 .pru_shutdown = rts_shutdown,
438 .pru_sockaddr = rts_sockaddr,
439 .pru_close = rts_close,
442 #ifndef _SOCKADDR_UNION_DEFINED
443 #define _SOCKADDR_UNION_DEFINED
445 * The union of all possible address formats we handle.
447 union sockaddr_union {
449 struct sockaddr_in sin;
450 struct sockaddr_in6 sin6;
452 #endif /* _SOCKADDR_UNION_DEFINED */
455 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
456 struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
458 #if defined(INET) || defined(INET6)
459 struct epoch_tracker et;
462 /* First, see if the returned address is part of the jail. */
463 if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
464 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
468 switch (info->rti_info[RTAX_DST]->sa_family) {
478 * Try to find an address on the given outgoing interface
479 * that belongs to the jail.
482 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
485 if (sa->sa_family != AF_INET)
487 ia = ((struct sockaddr_in *)sa)->sin_addr;
488 if (prison_check_ip4(cred, &ia) == 0) {
496 * As a last resort return the 'default' jail address.
498 ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
500 if (prison_get_ip4(cred, &ia) != 0)
503 bzero(&saun->sin, sizeof(struct sockaddr_in));
504 saun->sin.sin_len = sizeof(struct sockaddr_in);
505 saun->sin.sin_family = AF_INET;
506 saun->sin.sin_addr.s_addr = ia.s_addr;
507 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
520 * Try to find an address on the given outgoing interface
521 * that belongs to the jail.
524 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
527 if (sa->sa_family != AF_INET6)
529 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
530 &ia6, sizeof(struct in6_addr));
531 if (prison_check_ip6(cred, &ia6) == 0) {
539 * As a last resort return the 'default' jail address.
541 ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
543 if (prison_get_ip6(cred, &ia6) != 0)
546 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
547 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
548 saun->sin6.sin6_family = AF_INET6;
549 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
550 if (sa6_recoverscope(&saun->sin6) != 0)
552 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
563 * Fills in @info based on userland-provided @rtm message.
565 * Returns 0 on success.
568 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info)
573 rtm->rtm_pid = curproc->p_pid;
574 info->rti_addrs = rtm->rtm_addrs;
576 info->rti_mflags = rtm->rtm_inits;
577 info->rti_rmx = &rtm->rtm_rmx;
580 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
581 * link-local address because rtrequest requires addresses with
584 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
587 if (rtm->rtm_flags & RTF_RNH_LOCKED)
589 info->rti_flags = rtm->rtm_flags;
590 if (info->rti_info[RTAX_DST] == NULL ||
591 info->rti_info[RTAX_DST]->sa_family >= AF_MAX ||
592 (info->rti_info[RTAX_GATEWAY] != NULL &&
593 info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
595 saf = info->rti_info[RTAX_DST]->sa_family;
597 * Verify that the caller has the appropriate privilege; RTM_GET
598 * is the only operation the non-superuser is allowed.
600 if (rtm->rtm_type != RTM_GET) {
601 error = priv_check(curthread, PRIV_NET_ROUTE);
607 * The given gateway address may be an interface address.
608 * For example, issuing a "route change" command on a route
609 * entry that was created from a tunnel, and the gateway
610 * address given is the local end point. In this case the
611 * RTF_GATEWAY flag must be cleared or the destination will
612 * not be reachable even though there is no error message.
614 if (info->rti_info[RTAX_GATEWAY] != NULL &&
615 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
616 struct rt_addrinfo ginfo;
617 struct sockaddr *gdst;
618 struct sockaddr_storage ss;
620 bzero(&ginfo, sizeof(ginfo));
621 bzero(&ss, sizeof(ss));
622 ss.ss_len = sizeof(ss);
624 ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
625 gdst = info->rti_info[RTAX_GATEWAY];
628 * A host route through the loopback interface is
629 * installed for each interface adddress. In pre 8.0
630 * releases the interface address of a PPP link type
631 * is not reachable locally. This behavior is fixed as
632 * part of the new L2/L3 redesign and rewrite work. The
633 * signature of this interface address route is the
634 * AF_LINK sa_family type of the rt_gateway, and the
635 * rt_ifp has the IFF_LOOPBACK flag set.
637 if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
638 if (ss.ss_family == AF_LINK &&
639 ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
640 info->rti_flags &= ~RTF_GATEWAY;
641 info->rti_flags |= RTF_GWFLAG_COMPAT;
643 rib_free_info(&ginfo);
651 * Handles RTM_GET message from routing socket, returning matching rt.
654 * 0 on success, with locked and referenced matching rt in @rt_nrt
658 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
659 struct rt_msghdr *rtm, struct rtentry **ret_nrt)
663 struct rib_head *rnh;
666 saf = info->rti_info[RTAX_DST]->sa_family;
668 rnh = rt_tables_get_rnh(fibnum, saf);
670 return (EAFNOSUPPORT);
674 if (info->rti_info[RTAX_NETMASK] == NULL) {
676 * Provide longest prefix match for
677 * address lookup (no mask).
678 * 'route -n get addr'
680 rt = (struct rtentry *) rnh->rnh_matchaddr(
681 info->rti_info[RTAX_DST], &rnh->head);
683 rt = (struct rtentry *) rnh->rnh_lookup(
684 info->rti_info[RTAX_DST],
685 info->rti_info[RTAX_NETMASK], &rnh->head);
693 * for RTM_GET, gate is optional even with multipath.
694 * if gate == NULL the first match is returned.
695 * (no need to call rt_mpath_matchgate if gate == NULL)
697 if (rt_mpath_capable(rnh) && info->rti_info[RTAX_GATEWAY]) {
698 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
706 * If performing proxied L2 entry insertion, and
707 * the actual PPP host entry is found, perform
708 * another search to retrieve the prefix route of
709 * the local end point of the PPP link.
711 if (rtm->rtm_flags & RTF_ANNOUNCE) {
712 struct sockaddr laddr;
714 if (rt->rt_ifp != NULL &&
715 rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
716 struct epoch_tracker et;
720 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
724 rt_maskedcopy(ifa->ifa_addr,
728 rt_maskedcopy(rt->rt_ifa->ifa_addr,
730 rt->rt_ifa->ifa_netmask);
732 * refactor rt and no lock operation necessary
734 rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
751 * Update sockaddrs, flags, etc in @prtm based on @rt data.
752 * Assumes @rt is locked.
753 * rtm can be reallocated.
755 * Returns 0 on success, along with pointer to (potentially reallocated)
760 update_rtm_from_rte(struct rt_addrinfo *info, struct rt_msghdr **prtm,
761 int alloc_len, struct rtentry *rt)
763 struct sockaddr_storage netmask_ss;
765 union sockaddr_union saun;
766 struct rt_msghdr *rtm, *orig_rtm = NULL;
774 info->rti_info[RTAX_DST] = rt_key(rt);
775 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
776 info->rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
777 rt_mask(rt), &netmask_ss);
778 info->rti_info[RTAX_GENMASK] = 0;
780 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
782 info->rti_info[RTAX_IFP] =
783 ifp->if_addr->ifa_addr;
784 error = rtm_get_jailed(info, ifp, rt,
785 &saun, curthread->td_ucred);
788 if (ifp->if_flags & IFF_POINTOPOINT)
789 info->rti_info[RTAX_BRD] =
790 rt->rt_ifa->ifa_dstaddr;
791 rtm->rtm_index = ifp->if_index;
793 info->rti_info[RTAX_IFP] = NULL;
794 info->rti_info[RTAX_IFA] = NULL;
796 } else if (ifp != NULL)
797 rtm->rtm_index = ifp->if_index;
799 /* Check if we need to realloc storage */
800 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
801 if (len > alloc_len) {
802 struct rt_msghdr *tmp_rtm;
804 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
807 bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
813 * Delay freeing original rtm as info contains
814 * data referencing it.
818 w.w_tmem = (caddr_t)rtm;
819 w.w_tmemsize = alloc_len;
820 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
822 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
823 rtm->rtm_flags = RTF_GATEWAY |
824 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
826 rtm->rtm_flags = rt->rt_flags;
827 rt_getmetrics(rt, &rtm->rtm_rmx);
828 rtm->rtm_addrs = info->rti_addrs;
830 if (orig_rtm != NULL)
831 free(orig_rtm, M_TEMP);
839 route_output(struct mbuf *m, struct socket *so, ...)
841 struct rt_msghdr *rtm = NULL;
842 struct rtentry *rt = NULL;
843 struct rt_addrinfo info;
844 struct epoch_tracker et;
846 struct sockaddr_storage ss;
847 struct sockaddr_in6 *sin6;
848 int i, rti_need_deembed = 0;
850 int alloc_len = 0, len, error = 0, fibnum;
851 sa_family_t saf = AF_UNSPEC;
854 fibnum = so->so_fibnum;
856 #define senderr(e) { error = e; goto flush;}
857 if (m == NULL || ((m->m_len < sizeof(long)) &&
858 (m = m_pullup(m, sizeof(long))) == NULL))
860 if ((m->m_flags & M_PKTHDR) == 0)
861 panic("route_output");
863 len = m->m_pkthdr.len;
864 if (len < sizeof(*rtm) ||
865 len != mtod(m, struct rt_msghdr *)->rtm_msglen)
869 * Most of current messages are in range 200-240 bytes,
870 * minimize possible re-allocation on reply using larger size
871 * buffer aligned on 1k boundaty.
873 alloc_len = roundup2(len, 1024);
874 if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
877 m_copydata(m, 0, len, (caddr_t)rtm);
878 bzero(&info, sizeof(info));
879 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 = (V_deembed_scopeid) ? 1 : 0;
910 switch (rtm->rtm_type) {
911 struct rtentry *saved_nrt;
915 if (rtm->rtm_type == RTM_ADD) {
916 if (info.rti_info[RTAX_GATEWAY] == NULL)
920 error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
922 if (error == 0 && saved_nrt != NULL) {
924 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
927 rtm->rtm_index = saved_nrt->rt_ifp->if_index;
928 RT_REMREF(saved_nrt);
929 RT_UNLOCK(saved_nrt);
935 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
942 /* rt_msg2() will not be used when RTM_DELETE fails. */
943 rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
948 error = handle_rtm_get(&info, fibnum, rtm, &rt);
954 if (!can_export_rte(curthread->td_ucred, rt)) {
958 error = update_rtm_from_rte(&info, &rtm, alloc_len, rt);
960 * Note that some sockaddr pointers may have changed to
961 * point to memory outsize @rtm. Some may be pointing
962 * to the on-stack variables.
963 * Given that, any pointer in @info CANNOT BE USED.
967 * scopeid deembedding has been performed while
968 * writing updated rtm in rtsock_msg_buffer().
969 * With that in mind, skip deembedding procedure below.
972 rti_need_deembed = 0;
990 if (rti_need_deembed) {
991 /* sin6_scope_id is recovered before sending rtm. */
992 sin6 = (struct sockaddr_in6 *)&ss;
993 for (i = 0; i < RTAX_MAX; i++) {
994 if (info.rti_info[i] == NULL)
996 if (info.rti_info[i]->sa_family != AF_INET6)
998 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
999 if (sa6_recoverscope(sin6) == 0)
1000 bcopy(sin6, info.rti_info[i],
1006 send_rtm_reply(so, rtm, m, saf, fibnum, error);
1012 * Sends the prepared reply message in @rtm to all rtsock clients.
1013 * Frees @m and @rtm.
1017 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1018 sa_family_t saf, u_int fibnum, int rtm_errno)
1020 struct rawcb *rp = NULL;
1023 * Check to see if we don't want our own messages.
1025 if ((so->so_options & SO_USELOOPBACK) == 0) {
1026 if (V_route_cb.any_count <= 1) {
1032 /* There is another listener, so construct message */
1038 rtm->rtm_errno = rtm_errno;
1040 rtm->rtm_flags |= RTF_DONE;
1042 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1043 if (m->m_pkthdr.len < rtm->rtm_msglen) {
1046 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
1047 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1052 M_SETFIB(m, fibnum);
1053 m->m_flags |= RTS_FILTER_FIB;
1056 * XXX insure we don't get a copy by
1057 * invalidating our protocol
1059 unsigned short family = rp->rcb_proto.sp_family;
1060 rp->rcb_proto.sp_family = 0;
1061 rt_dispatch(m, saf);
1062 rp->rcb_proto.sp_family = family;
1064 rt_dispatch(m, saf);
1070 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
1073 bzero(out, sizeof(*out));
1074 out->rmx_mtu = rt->rt_mtu;
1075 out->rmx_weight = rt->rt_weight;
1076 out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
1077 /* Kernel -> userland timebase conversion. */
1078 out->rmx_expire = rt->rt_expire ?
1079 rt->rt_expire - time_uptime + time_second : 0;
1083 * Extract the addresses of the passed sockaddrs.
1084 * Do a little sanity checking so as to avoid bad memory references.
1085 * This data is derived straight from userland.
1088 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1090 struct sockaddr *sa;
1093 for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1094 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1096 sa = (struct sockaddr *)cp;
1100 if (cp + sa->sa_len > cplim)
1103 * there are no more.. quit now
1104 * If there are more bits, they are in error.
1105 * I've seen this. route(1) can evidently generate these.
1106 * This causes kernel to core dump.
1107 * for compatibility, If we see this, point to a safe address.
1109 if (sa->sa_len == 0) {
1110 rtinfo->rti_info[i] = &sa_zero;
1111 return (0); /* should be EINVAL but for compat */
1115 if (sa->sa_family == AF_INET6)
1116 sa6_embedscope((struct sockaddr_in6 *)sa,
1119 rtinfo->rti_info[i] = sa;
1126 * Fill in @dmask with valid netmask leaving original @smask
1127 * intact. Mostly used with radix netmasks.
1129 static struct sockaddr *
1130 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask,
1131 struct sockaddr_storage *dmask)
1133 if (dst == NULL || smask == NULL)
1136 memset(dmask, 0, dst->sa_len);
1137 memcpy(dmask, smask, smask->sa_len);
1138 dmask->ss_len = dst->sa_len;
1139 dmask->ss_family = dst->sa_family;
1141 return ((struct sockaddr *)dmask);
1145 * Writes information related to @rtinfo object to newly-allocated mbuf.
1146 * Assumes MCLBYTES is enough to construct any message.
1147 * Used for OS notifications of vaious events (if/ifa announces,etc)
1149 * Returns allocated mbuf or NULL on failure.
1151 static struct mbuf *
1152 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1154 struct rt_msghdr *rtm;
1157 struct sockaddr *sa;
1159 struct sockaddr_storage ss;
1160 struct sockaddr_in6 *sin6;
1168 len = sizeof(struct ifa_msghdr);
1173 len = sizeof(struct ifma_msghdr);
1177 len = sizeof(struct if_msghdr);
1180 case RTM_IFANNOUNCE:
1182 len = sizeof(struct if_announcemsghdr);
1186 len = sizeof(struct rt_msghdr);
1189 /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1190 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1192 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1194 m = m_gethdr(M_NOWAIT, MT_DATA);
1198 m->m_pkthdr.len = m->m_len = len;
1199 rtm = mtod(m, struct rt_msghdr *);
1200 bzero((caddr_t)rtm, len);
1201 for (i = 0; i < RTAX_MAX; i++) {
1202 if ((sa = rtinfo->rti_info[i]) == NULL)
1204 rtinfo->rti_addrs |= (1 << i);
1207 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1208 sin6 = (struct sockaddr_in6 *)&ss;
1209 bcopy(sa, sin6, sizeof(*sin6));
1210 if (sa6_recoverscope(sin6) == 0)
1211 sa = (struct sockaddr *)sin6;
1214 m_copyback(m, len, dlen, (caddr_t)sa);
1217 if (m->m_pkthdr.len != len) {
1221 rtm->rtm_msglen = len;
1222 rtm->rtm_version = RTM_VERSION;
1223 rtm->rtm_type = type;
1228 * Writes information related to @rtinfo object to preallocated buffer.
1229 * Stores needed size in @plen. If @w is NULL, calculates size without
1231 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1233 * Returns 0 on success.
1237 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1240 int len, buflen = 0, dlen;
1242 struct rt_msghdr *rtm = NULL;
1244 struct sockaddr_storage ss;
1245 struct sockaddr_in6 *sin6;
1247 #ifdef COMPAT_FREEBSD32
1248 bool compat32 = false;
1255 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1256 #ifdef COMPAT_FREEBSD32
1257 if (w->w_req->flags & SCTL_MASK32) {
1258 len = sizeof(struct ifa_msghdrl32);
1262 len = sizeof(struct ifa_msghdrl);
1264 len = sizeof(struct ifa_msghdr);
1268 #ifdef COMPAT_FREEBSD32
1269 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1270 if (w->w_op == NET_RT_IFLISTL)
1271 len = sizeof(struct if_msghdrl32);
1273 len = sizeof(struct if_msghdr32);
1278 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1279 len = sizeof(struct if_msghdrl);
1281 len = sizeof(struct if_msghdr);
1285 len = sizeof(struct ifma_msghdr);
1289 len = sizeof(struct rt_msghdr);
1293 rtm = (struct rt_msghdr *)w->w_tmem;
1294 buflen = w->w_tmemsize - len;
1295 cp = (caddr_t)w->w_tmem + len;
1298 rtinfo->rti_addrs = 0;
1299 for (i = 0; i < RTAX_MAX; i++) {
1300 struct sockaddr *sa;
1302 if ((sa = rtinfo->rti_info[i]) == NULL)
1304 rtinfo->rti_addrs |= (1 << i);
1305 #ifdef COMPAT_FREEBSD32
1307 dlen = SA_SIZE32(sa);
1311 if (cp != NULL && buflen >= dlen) {
1313 if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1314 sin6 = (struct sockaddr_in6 *)&ss;
1315 bcopy(sa, sin6, sizeof(*sin6));
1316 if (sa6_recoverscope(sin6) == 0)
1317 sa = (struct sockaddr *)sin6;
1320 bcopy((caddr_t)sa, cp, (unsigned)dlen);
1323 } else if (cp != NULL) {
1325 * Buffer too small. Count needed size
1326 * and return with error.
1335 dlen = ALIGN(len) - len;
1347 /* fill header iff buffer is large enough */
1348 rtm->rtm_version = RTM_VERSION;
1349 rtm->rtm_type = type;
1350 rtm->rtm_msglen = len;
1355 if (w != NULL && cp == NULL)
1362 * This routine is called to generate a message from the routing
1363 * socket indicating that a redirect has occurred, a routing lookup
1364 * has failed, or that a protocol has detected timeouts to a particular
1368 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1371 struct rt_msghdr *rtm;
1373 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1375 if (V_route_cb.any_count == 0)
1377 m = rtsock_msg_mbuf(type, rtinfo);
1381 if (fibnum != RT_ALL_FIBS) {
1382 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1383 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1384 M_SETFIB(m, fibnum);
1385 m->m_flags |= RTS_FILTER_FIB;
1388 rtm = mtod(m, struct rt_msghdr *);
1389 rtm->rtm_flags = RTF_DONE | flags;
1390 rtm->rtm_errno = error;
1391 rtm->rtm_addrs = rtinfo->rti_addrs;
1392 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1396 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1399 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1403 * This routine is called to generate a message from the routing
1404 * socket indicating that the status of a network interface has changed.
1407 rt_ifmsg(struct ifnet *ifp)
1409 struct if_msghdr *ifm;
1411 struct rt_addrinfo info;
1413 if (V_route_cb.any_count == 0)
1415 bzero((caddr_t)&info, sizeof(info));
1416 m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1419 ifm = mtod(m, struct if_msghdr *);
1420 ifm->ifm_index = ifp->if_index;
1421 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1422 if_data_copy(ifp, &ifm->ifm_data);
1424 rt_dispatch(m, AF_UNSPEC);
1428 * Announce interface address arrival/withdraw.
1429 * Please do not call directly, use rt_addrmsg().
1430 * Assume input data to be valid.
1431 * Returns 0 on success.
1434 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1436 struct rt_addrinfo info;
1437 struct sockaddr *sa;
1440 struct ifa_msghdr *ifam;
1441 struct ifnet *ifp = ifa->ifa_ifp;
1442 struct sockaddr_storage ss;
1444 if (V_route_cb.any_count == 0)
1447 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1449 bzero((caddr_t)&info, sizeof(info));
1450 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1451 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1452 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1453 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1454 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1455 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1457 ifam = mtod(m, struct ifa_msghdr *);
1458 ifam->ifam_index = ifp->if_index;
1459 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1460 ifam->ifam_flags = ifa->ifa_flags;
1461 ifam->ifam_addrs = info.rti_addrs;
1463 if (fibnum != RT_ALL_FIBS) {
1464 M_SETFIB(m, fibnum);
1465 m->m_flags |= RTS_FILTER_FIB;
1468 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1474 * Announce route addition/removal to rtsock based on @rt data.
1475 * Callers are advives to use rt_routemsg() instead of using this
1476 * function directly.
1477 * Assume @rt data is consistent.
1479 * Returns 0 on success.
1482 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs,
1485 struct sockaddr_storage ss;
1486 struct rt_addrinfo info;
1488 if (V_route_cb.any_count == 0)
1491 bzero((caddr_t)&info, sizeof(info));
1492 info.rti_info[RTAX_DST] = rt_key(rt);
1493 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss);
1494 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1495 info.rti_flags = rt->rt_flags;
1498 return (rtsock_routemsg_info(cmd, &info, fibnum));
1502 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1504 struct rt_msghdr *rtm;
1505 struct sockaddr *sa;
1508 if (V_route_cb.any_count == 0)
1511 if (info->rti_flags & RTF_HOST)
1512 info->rti_info[RTAX_NETMASK] = NULL;
1514 m = rtsock_msg_mbuf(cmd, info);
1518 if (fibnum != RT_ALL_FIBS) {
1519 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1520 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1521 M_SETFIB(m, fibnum);
1522 m->m_flags |= RTS_FILTER_FIB;
1525 rtm = mtod(m, struct rt_msghdr *);
1526 rtm->rtm_addrs = info->rti_addrs;
1527 if (info->rti_ifp != NULL)
1528 rtm->rtm_index = info->rti_ifp->if_index;
1529 /* Add RTF_DONE to indicate command 'completion' required by API */
1530 info->rti_flags |= RTF_DONE;
1531 /* Reported routes has to be up */
1532 if (cmd == RTM_ADD || cmd == RTM_CHANGE)
1533 info->rti_flags |= RTF_UP;
1534 rtm->rtm_flags = info->rti_flags;
1536 sa = info->rti_info[RTAX_DST];
1537 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1543 * This is the analogue to the rt_newaddrmsg which performs the same
1544 * function but for multicast group memberhips. This is easier since
1545 * there is no route state to worry about.
1548 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1550 struct rt_addrinfo info;
1551 struct mbuf *m = NULL;
1552 struct ifnet *ifp = ifma->ifma_ifp;
1553 struct ifma_msghdr *ifmam;
1555 if (V_route_cb.any_count == 0)
1558 bzero((caddr_t)&info, sizeof(info));
1559 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1560 if (ifp && ifp->if_addr)
1561 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1563 info.rti_info[RTAX_IFP] = NULL;
1565 * If a link-layer address is present, present it as a ``gateway''
1566 * (similarly to how ARP entries, e.g., are presented).
1568 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1569 m = rtsock_msg_mbuf(cmd, &info);
1572 ifmam = mtod(m, struct ifma_msghdr *);
1573 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1575 ifmam->ifmam_index = ifp->if_index;
1576 ifmam->ifmam_addrs = info.rti_addrs;
1577 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1580 static struct mbuf *
1581 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1582 struct rt_addrinfo *info)
1584 struct if_announcemsghdr *ifan;
1587 if (V_route_cb.any_count == 0)
1589 bzero((caddr_t)info, sizeof(*info));
1590 m = rtsock_msg_mbuf(type, info);
1592 ifan = mtod(m, struct if_announcemsghdr *);
1593 ifan->ifan_index = ifp->if_index;
1594 strlcpy(ifan->ifan_name, ifp->if_xname,
1595 sizeof(ifan->ifan_name));
1596 ifan->ifan_what = what;
1602 * This is called to generate routing socket messages indicating
1603 * IEEE80211 wireless events.
1604 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1607 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1610 struct rt_addrinfo info;
1612 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1615 * Append the ieee80211 data. Try to stick it in the
1616 * mbuf containing the ifannounce msg; otherwise allocate
1617 * a new mbuf and append.
1619 * NB: we assume m is a single mbuf.
1621 if (data_len > M_TRAILINGSPACE(m)) {
1622 struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1627 bcopy(data, mtod(n, void *), data_len);
1628 n->m_len = data_len;
1630 } else if (data_len > 0) {
1631 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1632 m->m_len += data_len;
1634 if (m->m_flags & M_PKTHDR)
1635 m->m_pkthdr.len += data_len;
1636 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1637 rt_dispatch(m, AF_UNSPEC);
1642 * This is called to generate routing socket messages indicating
1643 * network interface arrival and departure.
1646 rt_ifannouncemsg(struct ifnet *ifp, int what)
1649 struct rt_addrinfo info;
1651 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1653 rt_dispatch(m, AF_UNSPEC);
1657 rt_dispatch(struct mbuf *m, sa_family_t saf)
1662 * Preserve the family from the sockaddr, if any, in an m_tag for
1663 * use when injecting the mbuf into the routing socket buffer from
1666 if (saf != AF_UNSPEC) {
1667 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1673 *(unsigned short *)(tag + 1) = saf;
1674 m_tag_prepend(m, tag);
1678 m->m_pkthdr.rcvif = V_loif;
1684 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */
1688 * Checks if rte can be exported v.r.t jails/vnets.
1690 * Returns 1 if it can, 0 otherwise.
1693 can_export_rte(struct ucred *td_ucred, const struct rtentry *rt)
1696 if ((rt->rt_flags & RTF_HOST) == 0
1697 ? jailed_without_vnet(td_ucred)
1698 : prison_if(td_ucred, rt_key_const(rt)) != 0)
1704 * This is used in dumping the kernel table via sysctl().
1707 sysctl_dumpentry(struct radix_node *rn, void *vw)
1709 struct walkarg *w = vw;
1710 struct rtentry *rt = (struct rtentry *)rn;
1711 int error = 0, size;
1712 struct rt_addrinfo info;
1713 struct sockaddr_storage ss;
1717 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1719 if (!can_export_rte(w->w_req->td->td_ucred, rt))
1721 bzero((caddr_t)&info, sizeof(info));
1722 info.rti_info[RTAX_DST] = rt_key(rt);
1723 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1724 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1726 info.rti_info[RTAX_GENMASK] = 0;
1727 if (rt->rt_ifp && !(rt->rt_ifp->if_flags & IFF_DYING)) {
1728 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1729 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1730 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1731 info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1733 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1735 if (w->w_req && w->w_tmem) {
1736 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1738 bzero(&rtm->rtm_index,
1739 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1740 if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1741 rtm->rtm_flags = RTF_GATEWAY |
1742 (rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1744 rtm->rtm_flags = rt->rt_flags;
1745 rt_getmetrics(rt, &rtm->rtm_rmx);
1746 rtm->rtm_index = rt->rt_ifp->if_index;
1747 rtm->rtm_addrs = info.rti_addrs;
1748 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1755 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1756 struct rt_addrinfo *info, struct walkarg *w, int len)
1758 struct if_msghdrl *ifm;
1759 struct if_data *ifd;
1761 ifm = (struct if_msghdrl *)w->w_tmem;
1763 #ifdef COMPAT_FREEBSD32
1764 if (w->w_req->flags & SCTL_MASK32) {
1765 struct if_msghdrl32 *ifm32;
1767 ifm32 = (struct if_msghdrl32 *)ifm;
1768 ifm32->ifm_addrs = info->rti_addrs;
1769 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1770 ifm32->ifm_index = ifp->if_index;
1771 ifm32->_ifm_spare1 = 0;
1772 ifm32->ifm_len = sizeof(*ifm32);
1773 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1774 ifm32->_ifm_spare2 = 0;
1775 ifd = &ifm32->ifm_data;
1779 ifm->ifm_addrs = info->rti_addrs;
1780 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1781 ifm->ifm_index = ifp->if_index;
1782 ifm->_ifm_spare1 = 0;
1783 ifm->ifm_len = sizeof(*ifm);
1784 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1785 ifm->_ifm_spare2 = 0;
1786 ifd = &ifm->ifm_data;
1789 memcpy(ifd, src_ifd, sizeof(*ifd));
1791 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1795 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1796 struct rt_addrinfo *info, struct walkarg *w, int len)
1798 struct if_msghdr *ifm;
1799 struct if_data *ifd;
1801 ifm = (struct if_msghdr *)w->w_tmem;
1803 #ifdef COMPAT_FREEBSD32
1804 if (w->w_req->flags & SCTL_MASK32) {
1805 struct if_msghdr32 *ifm32;
1807 ifm32 = (struct if_msghdr32 *)ifm;
1808 ifm32->ifm_addrs = info->rti_addrs;
1809 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1810 ifm32->ifm_index = ifp->if_index;
1811 ifm32->_ifm_spare1 = 0;
1812 ifd = &ifm32->ifm_data;
1816 ifm->ifm_addrs = info->rti_addrs;
1817 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1818 ifm->ifm_index = ifp->if_index;
1819 ifm->_ifm_spare1 = 0;
1820 ifd = &ifm->ifm_data;
1823 memcpy(ifd, src_ifd, sizeof(*ifd));
1825 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1829 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1830 struct walkarg *w, int len)
1832 struct ifa_msghdrl *ifam;
1833 struct if_data *ifd;
1835 ifam = (struct ifa_msghdrl *)w->w_tmem;
1837 #ifdef COMPAT_FREEBSD32
1838 if (w->w_req->flags & SCTL_MASK32) {
1839 struct ifa_msghdrl32 *ifam32;
1841 ifam32 = (struct ifa_msghdrl32 *)ifam;
1842 ifam32->ifam_addrs = info->rti_addrs;
1843 ifam32->ifam_flags = ifa->ifa_flags;
1844 ifam32->ifam_index = ifa->ifa_ifp->if_index;
1845 ifam32->_ifam_spare1 = 0;
1846 ifam32->ifam_len = sizeof(*ifam32);
1847 ifam32->ifam_data_off =
1848 offsetof(struct ifa_msghdrl32, ifam_data);
1849 ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1850 ifd = &ifam32->ifam_data;
1854 ifam->ifam_addrs = info->rti_addrs;
1855 ifam->ifam_flags = ifa->ifa_flags;
1856 ifam->ifam_index = ifa->ifa_ifp->if_index;
1857 ifam->_ifam_spare1 = 0;
1858 ifam->ifam_len = sizeof(*ifam);
1859 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1860 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1861 ifd = &ifam->ifam_data;
1864 bzero(ifd, sizeof(*ifd));
1865 ifd->ifi_datalen = sizeof(struct if_data);
1866 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1867 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1868 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1869 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1871 /* Fixup if_data carp(4) vhid. */
1872 if (carp_get_vhid_p != NULL)
1873 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1875 return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1879 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1880 struct walkarg *w, int len)
1882 struct ifa_msghdr *ifam;
1884 ifam = (struct ifa_msghdr *)w->w_tmem;
1885 ifam->ifam_addrs = info->rti_addrs;
1886 ifam->ifam_flags = ifa->ifa_flags;
1887 ifam->ifam_index = ifa->ifa_ifp->if_index;
1888 ifam->_ifam_spare1 = 0;
1889 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1891 return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1895 sysctl_iflist(int af, struct walkarg *w)
1900 struct rt_addrinfo info;
1902 struct sockaddr_storage ss;
1904 bzero((caddr_t)&info, sizeof(info));
1905 bzero(&ifd, sizeof(ifd));
1906 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1907 if (w->w_arg && w->w_arg != ifp->if_index)
1909 if_data_copy(ifp, &ifd);
1911 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1912 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1915 info.rti_info[RTAX_IFP] = NULL;
1916 if (w->w_req && w->w_tmem) {
1917 if (w->w_op == NET_RT_IFLISTL)
1918 error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1921 error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1926 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1927 if (af && af != ifa->ifa_addr->sa_family)
1929 if (prison_if(w->w_req->td->td_ucred,
1930 ifa->ifa_addr) != 0)
1932 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1933 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1934 ifa->ifa_addr, ifa->ifa_netmask, &ss);
1935 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1936 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1939 if (w->w_req && w->w_tmem) {
1940 if (w->w_op == NET_RT_IFLISTL)
1941 error = sysctl_iflist_ifaml(ifa, &info,
1944 error = sysctl_iflist_ifam(ifa, &info,
1950 info.rti_info[RTAX_IFA] = NULL;
1951 info.rti_info[RTAX_NETMASK] = NULL;
1952 info.rti_info[RTAX_BRD] = NULL;
1959 sysctl_ifmalist(int af, struct walkarg *w)
1961 struct rt_addrinfo info;
1963 struct ifmultiaddr *ifma;
1970 bzero((caddr_t)&info, sizeof(info));
1972 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1973 if (w->w_arg && w->w_arg != ifp->if_index)
1976 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1977 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1978 if (af && af != ifma->ifma_addr->sa_family)
1980 if (prison_if(w->w_req->td->td_ucred,
1981 ifma->ifma_addr) != 0)
1983 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1984 info.rti_info[RTAX_GATEWAY] =
1985 (ifma->ifma_addr->sa_family != AF_LINK) ?
1986 ifma->ifma_lladdr : NULL;
1987 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1990 if (w->w_req && w->w_tmem) {
1991 struct ifma_msghdr *ifmam;
1993 ifmam = (struct ifma_msghdr *)w->w_tmem;
1994 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1995 ifmam->ifmam_flags = 0;
1996 ifmam->ifmam_addrs = info.rti_addrs;
1997 ifmam->_ifmam_spare1 = 0;
1998 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2010 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2013 struct epoch_tracker et;
2014 int *name = (int *)arg1;
2015 u_int namelen = arg2;
2016 struct rib_head *rnh = NULL; /* silence compiler. */
2017 int i, lim, error = EINVAL;
2026 if (name[1] == NET_RT_DUMP) {
2028 fib = req->td->td_proc->p_fibnum;
2029 else if (namelen == 4)
2030 fib = (name[3] == RT_ALL_FIBS) ?
2031 req->td->td_proc->p_fibnum : name[3];
2033 return ((namelen < 3) ? EISDIR : ENOTDIR);
2034 if (fib < 0 || fib >= rt_numfibs)
2036 } else if (namelen != 3)
2037 return ((namelen < 3) ? EISDIR : ENOTDIR);
2041 bzero(&w, sizeof(w));
2046 error = sysctl_wire_old_buffer(req, 0);
2051 * Allocate reply buffer in advance.
2052 * All rtsock messages has maximum length of u_short.
2054 w.w_tmemsize = 65536;
2055 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2057 NET_EPOCH_ENTER(et);
2061 if (af == 0) { /* dump all tables */
2064 } else /* dump only one table */
2068 * take care of llinfo entries, the caller must
2071 if (w.w_op == NET_RT_FLAGS &&
2072 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2074 error = lltable_sysctl_dumparp(af, w.w_req);
2080 * take care of routing entries
2082 for (error = 0; error == 0 && i <= lim; i++) {
2083 rnh = rt_tables_get_rnh(fib, i);
2086 error = rnh->rnh_walktree(&rnh->head,
2087 sysctl_dumpentry, &w);
2090 error = EAFNOSUPPORT;
2095 case NET_RT_IFLISTL:
2096 error = sysctl_iflist(af, &w);
2099 case NET_RT_IFMALIST:
2100 error = sysctl_ifmalist(af, &w);
2105 free(w.w_tmem, M_TEMP);
2109 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2110 sysctl_rtsock, "Return route tables and interface/address lists");
2113 * Definitions of protocols supported in the ROUTE domain.
2116 static struct domain routedomain; /* or at least forward */
2118 static struct protosw routesw[] = {
2120 .pr_type = SOCK_RAW,
2121 .pr_domain = &routedomain,
2122 .pr_flags = PR_ATOMIC|PR_ADDR,
2123 .pr_output = route_output,
2124 .pr_ctlinput = raw_ctlinput,
2125 .pr_init = raw_init,
2126 .pr_usrreqs = &route_usrreqs
2130 static struct domain routedomain = {
2131 .dom_family = PF_ROUTE,
2132 .dom_name = "route",
2133 .dom_protosw = routesw,
2134 .dom_protoswNPROTOSW = &routesw[nitems(routesw)]
2137 VNET_DOMAIN_SET(route);
2141 * Unfortunately, RTF_ values are expressed as raw masks rather than powers of
2142 * 2, so we cannot use them as nice C99 initializer indices below.
2144 static const char * const rtf_flag_strings[] = {
2171 [30] = "RNH_LOCKED",
2172 [31] = "GWFLAG_COMPAT",
2175 static const char * __pure
2176 rt_flag_name(unsigned idx)
2178 if (idx >= nitems(rtf_flag_strings))
2179 return ("INVALID_FLAG");
2180 if (rtf_flag_strings[idx] == NULL)
2182 return (rtf_flag_strings[idx]);
2186 rt_dumpaddr_ddb(const char *name, const struct sockaddr *sa)
2188 char buf[INET6_ADDRSTRLEN], *res;
2193 else if (sa->sa_family == AF_INET) {
2194 res = inet_ntop(AF_INET,
2195 &((const struct sockaddr_in *)sa)->sin_addr,
2197 } else if (sa->sa_family == AF_INET6) {
2198 res = inet_ntop(AF_INET6,
2199 &((const struct sockaddr_in6 *)sa)->sin6_addr,
2201 } else if (sa->sa_family == AF_LINK) {
2206 db_printf("%s <%s> ", name, res);
2210 db_printf("%s <af:%d> ", name, sa->sa_family);
2214 rt_dumpentry_ddb(struct radix_node *rn, void *arg __unused)
2216 struct sockaddr_storage ss;
2220 /* If RNTORT is important, put it in a header. */
2223 rt_dumpaddr_ddb("dst", rt_key(rt));
2224 rt_dumpaddr_ddb("gateway", rt->rt_gateway);
2225 rt_dumpaddr_ddb("netmask", rtsock_fix_netmask(rt_key(rt), rt_mask(rt),
2227 if (rt->rt_ifp != NULL && (rt->rt_ifp->if_flags & IFF_DYING) == 0) {
2228 rt_dumpaddr_ddb("ifp", rt->rt_ifp->if_addr->ifa_addr);
2229 rt_dumpaddr_ddb("ifa", rt->rt_ifa->ifa_addr);
2232 db_printf("flags ");
2233 flags = rt->rt_flags;
2237 while ((idx = ffs(flags)) > 0) {
2240 if (flags != rt->rt_flags)
2242 db_printf("%s", rt_flag_name(idx));
2244 flags &= ~(1ul << idx);
2251 DB_SHOW_COMMAND(routetable, db_show_routetable_cmd)
2253 struct rib_head *rnh;
2263 for (; i <= lim; i++) {
2264 rnh = rt_tables_get_rnh(0, i);
2267 db_printf("%s: AF %d not supported?\n",
2274 if (!have_addr && i > 1)
2277 db_printf("Route table for AF %d%s%s%s:\n", i,
2278 (i == AF_INET || i == AF_INET6) ? " (" : "",
2279 (i == AF_INET) ? "INET" : (i == AF_INET6) ? "INET6" : "",
2280 (i == AF_INET || i == AF_INET6) ? ")" : "");
2282 error = rnh->rnh_walktree(&rnh->head, rt_dumpentry_ddb, NULL);
2284 db_printf("%s: walktree(%d): %d\n", __func__, i,
2289 _DB_FUNC(_show, route, db_show_route_cmd, db_show_table, CS_OWN, NULL)
2291 char buf[INET6_ADDRSTRLEN], *bp;
2292 const void *dst_addrp;
2293 struct sockaddr *dstp;
2296 struct sockaddr_in dest_sin;
2297 struct sockaddr_in6 dest_sin6;
2299 uint16_t hextets[8];
2301 int t, af, exp, tokflags;
2304 * Undecoded address family. No double-colon expansion seen yet.
2308 /* Assume INET6 to start; we can work back if guess was wrong. */
2309 tokflags = DRT_WSPACE | DRT_HEX | DRT_HEXADECIMAL;
2312 * db_command has lexed 'show route' for us.
2314 t = db_read_token_flags(tokflags);
2316 t = db_read_token_flags(tokflags);
2319 * tEOL: Just 'show route' isn't a valid mode.
2320 * tMINUS: It's either '-h' or some invalid option. Regardless, usage.
2322 if (t == tEOL || t == tMINUS)
2327 tets = nitems(hextets);
2330 * Each loop iteration, we expect to read one octet (v4) or hextet
2331 * (v6), followed by an appropriate field separator ('.' or ':' or
2334 * At the start of each loop, we're looking for a number (octet or
2337 * INET6 addresses have a special case where they may begin with '::'.
2339 for (i = 0; i < tets; i++) {
2340 t = db_read_token_flags(tokflags);
2342 if (t == tCOLONCOLON) {
2343 /* INET6 with leading '::' or invalid. */
2345 db_printf("Parse error: unexpected extra "
2354 } else if (t == tNUMBER) {
2356 * Lexer separates out '-' as tMINUS, but make the
2357 * assumption explicit here.
2359 MPASS(db_tok_number >= 0);
2361 if (af == AF_INET && db_tok_number > UINT8_MAX) {
2362 db_printf("Not a valid v4 octet: %ld\n",
2363 (long)db_tok_number);
2366 hextets[i] = db_tok_number;
2367 } else if (t == tEOL) {
2369 * We can only detect the end of an IPv6 address in
2370 * compact representation with EOL.
2372 if (af != AF_INET6 || exp < 0) {
2373 db_printf("Parse failed. Got unexpected EOF "
2374 "when the address is not a compact-"
2375 "representation IPv6 address.\n");
2380 db_printf("Parse failed. Unexpected token %d.\n", t);
2384 /* Next, look for a separator, if appropriate. */
2388 t = db_read_token_flags(tokflags);
2394 if (t == tCOLONCOLON) {
2405 /* Need to fixup the first parsed number. */
2406 if (hextets[0] > 0x255 ||
2407 (hextets[0] & 0xf0) > 0x90 ||
2408 (hextets[0] & 0xf) > 9) {
2409 db_printf("Not a valid v4 octet: %x\n",
2415 dn = (hn >> 8) * 100 +
2416 ((hn >> 4) & 0xf) * 10 +
2421 /* Switch to decimal for remaining octets. */
2422 tokflags &= ~DRT_RADIX_MASK;
2423 tokflags |= DRT_DECIMAL;
2429 db_printf("Parse error. Unexpected token %d.\n", t);
2431 } else if (af == AF_INET) {
2434 db_printf("Expected '.' (%d) between octets but got "
2435 "(%d).\n", tDOT, t);
2438 } else if (af == AF_INET6) {
2441 if (t == tCOLONCOLON) {
2448 db_printf("Got bogus second '::' in v6 "
2454 * Handle in the earlier part of the loop
2455 * because we need to handle trailing :: too.
2461 db_printf("Expected ':' (%d) or '::' (%d) between "
2462 "hextets but got (%d).\n", tCOLON, tCOLONCOLON, t);
2467 /* Check for trailing garbage. */
2469 t = db_read_token_flags(tokflags);
2471 db_printf("Got unexpected garbage after address "
2478 * Need to expand compact INET6 addresses.
2480 * Technically '::' for a single ':0:' is MUST NOT but just in case,
2481 * don't bother expanding that form (exp >= 0 && i == tets case).
2483 if (af == AF_INET6 && exp >= 0 && i < tets) {
2485 memmove(&hextets[exp + 1 + (nitems(hextets) - i)],
2487 (i - (exp + 1)) * sizeof(hextets[0]));
2489 memset(&hextets[exp + 1], 0, (nitems(hextets) - i) *
2490 sizeof(hextets[0]));
2493 memset(&u, 0, sizeof(u));
2494 if (af == AF_INET) {
2495 u.dest_sin.sin_family = AF_INET;
2496 u.dest_sin.sin_len = sizeof(u.dest_sin);
2497 u.dest_sin.sin_addr.s_addr = htonl(
2498 ((uint32_t)hextets[0] << 24) |
2499 ((uint32_t)hextets[1] << 16) |
2500 ((uint32_t)hextets[2] << 8) |
2501 (uint32_t)hextets[3]);
2502 dstp = (void *)&u.dest_sin;
2503 dst_addrp = &u.dest_sin.sin_addr;
2504 } else if (af == AF_INET6) {
2505 u.dest_sin6.sin6_family = AF_INET6;
2506 u.dest_sin6.sin6_len = sizeof(u.dest_sin6);
2507 for (i = 0; i < nitems(hextets); i++)
2508 u.dest_sin6.sin6_addr.s6_addr16[i] = htons(hextets[i]);
2509 dstp = (void *)&u.dest_sin6;
2510 dst_addrp = &u.dest_sin6.sin6_addr;
2514 /* Appease Clang false positive: */
2518 bp = inet_ntop(af, dst_addrp, buf, sizeof(buf));
2520 db_printf("Looking up route to destination '%s'\n", bp);
2523 rt = rtalloc1(dstp, 0, RTF_RNH_LOCKED);
2527 db_printf("Could not get route for that server.\n");
2531 rt_dumpentry_ddb((void *)rt, NULL);
2536 db_printf("Usage: 'show route <address>'\n"
2537 " Currently accepts only dotted-decimal INET or colon-separated\n"
2538 " hextet INET6 addresses.\n");