xref: /freebsd/sys/net/rtsock.c (revision 7778ab7e0cc22f0824eb1d1047a7ef8b4785267a)
1 /*-
2  * Copyright (c) 1988, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
30  * $FreeBSD$
31  */
32 #include "opt_compat.h"
33 #include "opt_sctp.h"
34 #include "opt_mpath.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/jail.h>
40 #include <sys/kernel.h>
41 #include <sys/domain.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/rwlock.h>
49 #include <sys/signalvar.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/systm.h>
54 
55 #include <net/if.h>
56 #include <net/if_dl.h>
57 #include <net/if_llatbl.h>
58 #include <net/if_types.h>
59 #include <net/netisr.h>
60 #include <net/raw_cb.h>
61 #include <net/route.h>
62 #include <net/vnet.h>
63 
64 #include <netinet/in.h>
65 #include <netinet/if_ether.h>
66 #ifdef INET6
67 #include <netinet6/scope6_var.h>
68 #endif
69 
70 #if defined(INET) || defined(INET6)
71 #ifdef SCTP
72 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
73 #endif /* SCTP */
74 #endif
75 
76 #ifdef COMPAT_FREEBSD32
77 #include <sys/mount.h>
78 #include <compat/freebsd32/freebsd32.h>
79 
80 struct if_data32 {
81 	uint8_t	ifi_type;
82 	uint8_t	ifi_physical;
83 	uint8_t	ifi_addrlen;
84 	uint8_t	ifi_hdrlen;
85 	uint8_t	ifi_link_state;
86 	uint8_t	ifi_spare_char1;
87 	uint8_t	ifi_spare_char2;
88 	uint8_t	ifi_datalen;
89 	uint32_t ifi_mtu;
90 	uint32_t ifi_metric;
91 	uint32_t ifi_baudrate;
92 	uint32_t ifi_ipackets;
93 	uint32_t ifi_ierrors;
94 	uint32_t ifi_opackets;
95 	uint32_t ifi_oerrors;
96 	uint32_t ifi_collisions;
97 	uint32_t ifi_ibytes;
98 	uint32_t ifi_obytes;
99 	uint32_t ifi_imcasts;
100 	uint32_t ifi_omcasts;
101 	uint32_t ifi_iqdrops;
102 	uint32_t ifi_noproto;
103 	uint32_t ifi_hwassist;
104 	int32_t	ifi_epoch;
105 	struct	timeval32 ifi_lastchange;
106 };
107 
108 struct if_msghdr32 {
109 	uint16_t ifm_msglen;
110 	uint8_t	ifm_version;
111 	uint8_t	ifm_type;
112 	int32_t	ifm_addrs;
113 	int32_t	ifm_flags;
114 	uint16_t ifm_index;
115 	struct	if_data32 ifm_data;
116 };
117 #endif
118 
119 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
120 
121 /* NB: these are not modified */
122 static struct	sockaddr route_src = { 2, PF_ROUTE, };
123 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
124 
125 /*
126  * Used by rtsock/raw_input callback code to decide whether to filter the update
127  * notification to a socket bound to a particular FIB.
128  */
129 #define	RTS_FILTER_FIB	M_PROTO8
130 #define	RTS_ALLFIBS	-1
131 
132 static struct {
133 	int	ip_count;	/* attached w/ AF_INET */
134 	int	ip6_count;	/* attached w/ AF_INET6 */
135 	int	ipx_count;	/* attached w/ AF_IPX */
136 	int	any_count;	/* total attached */
137 } route_cb;
138 
139 struct mtx rtsock_mtx;
140 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
141 
142 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
143 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
144 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
145 
146 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
147 
148 struct walkarg {
149 	int	w_tmemsize;
150 	int	w_op, w_arg;
151 	caddr_t	w_tmem;
152 	struct sysctl_req *w_req;
153 };
154 
155 static void	rts_input(struct mbuf *m);
156 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
157 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
158 			caddr_t cp, struct walkarg *w);
159 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
160 			struct rt_addrinfo *rtinfo);
161 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
162 static int	sysctl_iflist(int af, struct walkarg *w);
163 static int	sysctl_ifmalist(int af, struct walkarg *w);
164 static int	route_output(struct mbuf *m, struct socket *so);
165 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
166 			struct rt_metrics_lite *out);
167 static void	rt_getmetrics(const struct rt_metrics_lite *in,
168 			struct rt_metrics *out);
169 static void	rt_dispatch(struct mbuf *, const struct sockaddr *);
170 
171 static struct netisr_handler rtsock_nh = {
172 	.nh_name = "rtsock",
173 	.nh_handler = rts_input,
174 	.nh_proto = NETISR_ROUTE,
175 	.nh_policy = NETISR_POLICY_SOURCE,
176 };
177 
178 static int
179 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
180 {
181 	int error, qlimit;
182 
183 	netisr_getqlimit(&rtsock_nh, &qlimit);
184 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
185         if (error || !req->newptr)
186                 return (error);
187 	if (qlimit < 1)
188 		return (EINVAL);
189 	return (netisr_setqlimit(&rtsock_nh, qlimit));
190 }
191 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
192     0, 0, sysctl_route_netisr_maxqlen, "I",
193     "maximum routing socket dispatch queue length");
194 
195 static void
196 rts_init(void)
197 {
198 	int tmp;
199 
200 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
201 		rtsock_nh.nh_qlimit = tmp;
202 	netisr_register(&rtsock_nh);
203 }
204 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
205 
206 static int
207 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
208     struct rawcb *rp)
209 {
210 	int fibnum;
211 
212 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
213 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
214 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
215 
216 	/* No filtering requested. */
217 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
218 		return (0);
219 
220 	/* Check if it is a rts and the fib matches the one of the socket. */
221 	fibnum = M_GETFIB(m);
222 	if (proto->sp_family != PF_ROUTE ||
223 	    rp->rcb_socket == NULL ||
224 	    rp->rcb_socket->so_fibnum == fibnum)
225 		return (0);
226 
227 	/* Filtering requested and no match, the socket shall be skipped. */
228 	return (1);
229 }
230 
231 static void
232 rts_input(struct mbuf *m)
233 {
234 	struct sockproto route_proto;
235 	unsigned short *family;
236 	struct m_tag *tag;
237 
238 	route_proto.sp_family = PF_ROUTE;
239 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
240 	if (tag != NULL) {
241 		family = (unsigned short *)(tag + 1);
242 		route_proto.sp_protocol = *family;
243 		m_tag_delete(m, tag);
244 	} else
245 		route_proto.sp_protocol = 0;
246 
247 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
248 }
249 
250 /*
251  * It really doesn't make any sense at all for this code to share much
252  * with raw_usrreq.c, since its functionality is so restricted.  XXX
253  */
254 static void
255 rts_abort(struct socket *so)
256 {
257 
258 	raw_usrreqs.pru_abort(so);
259 }
260 
261 static void
262 rts_close(struct socket *so)
263 {
264 
265 	raw_usrreqs.pru_close(so);
266 }
267 
268 /* pru_accept is EOPNOTSUPP */
269 
270 static int
271 rts_attach(struct socket *so, int proto, struct thread *td)
272 {
273 	struct rawcb *rp;
274 	int s, error;
275 
276 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
277 
278 	/* XXX */
279 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
280 	if (rp == NULL)
281 		return ENOBUFS;
282 
283 	/*
284 	 * The splnet() is necessary to block protocols from sending
285 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
286 	 * this PCB is extant but incompletely initialized.
287 	 * Probably we should try to do more of this work beforehand and
288 	 * eliminate the spl.
289 	 */
290 	s = splnet();
291 	so->so_pcb = (caddr_t)rp;
292 	so->so_fibnum = td->td_proc->p_fibnum;
293 	error = raw_attach(so, proto);
294 	rp = sotorawcb(so);
295 	if (error) {
296 		splx(s);
297 		so->so_pcb = NULL;
298 		free(rp, M_PCB);
299 		return error;
300 	}
301 	RTSOCK_LOCK();
302 	switch(rp->rcb_proto.sp_protocol) {
303 	case AF_INET:
304 		route_cb.ip_count++;
305 		break;
306 	case AF_INET6:
307 		route_cb.ip6_count++;
308 		break;
309 	case AF_IPX:
310 		route_cb.ipx_count++;
311 		break;
312 	}
313 	route_cb.any_count++;
314 	RTSOCK_UNLOCK();
315 	soisconnected(so);
316 	so->so_options |= SO_USELOOPBACK;
317 	splx(s);
318 	return 0;
319 }
320 
321 static int
322 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
323 {
324 
325 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
326 }
327 
328 static int
329 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
330 {
331 
332 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
333 }
334 
335 /* pru_connect2 is EOPNOTSUPP */
336 /* pru_control is EOPNOTSUPP */
337 
338 static void
339 rts_detach(struct socket *so)
340 {
341 	struct rawcb *rp = sotorawcb(so);
342 
343 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
344 
345 	RTSOCK_LOCK();
346 	switch(rp->rcb_proto.sp_protocol) {
347 	case AF_INET:
348 		route_cb.ip_count--;
349 		break;
350 	case AF_INET6:
351 		route_cb.ip6_count--;
352 		break;
353 	case AF_IPX:
354 		route_cb.ipx_count--;
355 		break;
356 	}
357 	route_cb.any_count--;
358 	RTSOCK_UNLOCK();
359 	raw_usrreqs.pru_detach(so);
360 }
361 
362 static int
363 rts_disconnect(struct socket *so)
364 {
365 
366 	return (raw_usrreqs.pru_disconnect(so));
367 }
368 
369 /* pru_listen is EOPNOTSUPP */
370 
371 static int
372 rts_peeraddr(struct socket *so, struct sockaddr **nam)
373 {
374 
375 	return (raw_usrreqs.pru_peeraddr(so, nam));
376 }
377 
378 /* pru_rcvd is EOPNOTSUPP */
379 /* pru_rcvoob is EOPNOTSUPP */
380 
381 static int
382 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
383 	 struct mbuf *control, struct thread *td)
384 {
385 
386 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
387 }
388 
389 /* pru_sense is null */
390 
391 static int
392 rts_shutdown(struct socket *so)
393 {
394 
395 	return (raw_usrreqs.pru_shutdown(so));
396 }
397 
398 static int
399 rts_sockaddr(struct socket *so, struct sockaddr **nam)
400 {
401 
402 	return (raw_usrreqs.pru_sockaddr(so, nam));
403 }
404 
405 static struct pr_usrreqs route_usrreqs = {
406 	.pru_abort =		rts_abort,
407 	.pru_attach =		rts_attach,
408 	.pru_bind =		rts_bind,
409 	.pru_connect =		rts_connect,
410 	.pru_detach =		rts_detach,
411 	.pru_disconnect =	rts_disconnect,
412 	.pru_peeraddr =		rts_peeraddr,
413 	.pru_send =		rts_send,
414 	.pru_shutdown =		rts_shutdown,
415 	.pru_sockaddr =		rts_sockaddr,
416 	.pru_close =		rts_close,
417 };
418 
419 #ifndef _SOCKADDR_UNION_DEFINED
420 #define	_SOCKADDR_UNION_DEFINED
421 /*
422  * The union of all possible address formats we handle.
423  */
424 union sockaddr_union {
425 	struct sockaddr		sa;
426 	struct sockaddr_in	sin;
427 	struct sockaddr_in6	sin6;
428 };
429 #endif /* _SOCKADDR_UNION_DEFINED */
430 
431 static int
432 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
433     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
434 {
435 
436 	/* First, see if the returned address is part of the jail. */
437 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
438 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
439 		return (0);
440 	}
441 
442 	switch (info->rti_info[RTAX_DST]->sa_family) {
443 #ifdef INET
444 	case AF_INET:
445 	{
446 		struct in_addr ia;
447 		struct ifaddr *ifa;
448 		int found;
449 
450 		found = 0;
451 		/*
452 		 * Try to find an address on the given outgoing interface
453 		 * that belongs to the jail.
454 		 */
455 		IF_ADDR_LOCK(ifp);
456 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
457 			struct sockaddr *sa;
458 			sa = ifa->ifa_addr;
459 			if (sa->sa_family != AF_INET)
460 				continue;
461 			ia = ((struct sockaddr_in *)sa)->sin_addr;
462 			if (prison_check_ip4(cred, &ia) == 0) {
463 				found = 1;
464 				break;
465 			}
466 		}
467 		IF_ADDR_UNLOCK(ifp);
468 		if (!found) {
469 			/*
470 			 * As a last resort return the 'default' jail address.
471 			 */
472 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
473 			    sin_addr;
474 			if (prison_get_ip4(cred, &ia) != 0)
475 				return (ESRCH);
476 		}
477 		bzero(&saun->sin, sizeof(struct sockaddr_in));
478 		saun->sin.sin_len = sizeof(struct sockaddr_in);
479 		saun->sin.sin_family = AF_INET;
480 		saun->sin.sin_addr.s_addr = ia.s_addr;
481 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
482 		break;
483 	}
484 #endif
485 #ifdef INET6
486 	case AF_INET6:
487 	{
488 		struct in6_addr ia6;
489 		struct ifaddr *ifa;
490 		int found;
491 
492 		found = 0;
493 		/*
494 		 * Try to find an address on the given outgoing interface
495 		 * that belongs to the jail.
496 		 */
497 		IF_ADDR_LOCK(ifp);
498 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
499 			struct sockaddr *sa;
500 			sa = ifa->ifa_addr;
501 			if (sa->sa_family != AF_INET6)
502 				continue;
503 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
504 			    &ia6, sizeof(struct in6_addr));
505 			if (prison_check_ip6(cred, &ia6) == 0) {
506 				found = 1;
507 				break;
508 			}
509 		}
510 		IF_ADDR_UNLOCK(ifp);
511 		if (!found) {
512 			/*
513 			 * As a last resort return the 'default' jail address.
514 			 */
515 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
516 			    sin6_addr;
517 			if (prison_get_ip6(cred, &ia6) != 0)
518 				return (ESRCH);
519 		}
520 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
521 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
522 		saun->sin6.sin6_family = AF_INET6;
523 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
524 		if (sa6_recoverscope(&saun->sin6) != 0)
525 			return (ESRCH);
526 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
527 		break;
528 	}
529 #endif
530 	default:
531 		return (ESRCH);
532 	}
533 	return (0);
534 }
535 
536 /*ARGSUSED*/
537 static int
538 route_output(struct mbuf *m, struct socket *so)
539 {
540 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
541 	struct rt_msghdr *rtm = NULL;
542 	struct rtentry *rt = NULL;
543 	struct radix_node_head *rnh;
544 	struct rt_addrinfo info;
545 	int len, error = 0;
546 	struct ifnet *ifp = NULL;
547 	union sockaddr_union saun;
548 
549 #define senderr(e) { error = e; goto flush;}
550 	if (m == NULL || ((m->m_len < sizeof(long)) &&
551 		       (m = m_pullup(m, sizeof(long))) == NULL))
552 		return (ENOBUFS);
553 	if ((m->m_flags & M_PKTHDR) == 0)
554 		panic("route_output");
555 	len = m->m_pkthdr.len;
556 	if (len < sizeof(*rtm) ||
557 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
558 		info.rti_info[RTAX_DST] = NULL;
559 		senderr(EINVAL);
560 	}
561 	R_Malloc(rtm, struct rt_msghdr *, len);
562 	if (rtm == NULL) {
563 		info.rti_info[RTAX_DST] = NULL;
564 		senderr(ENOBUFS);
565 	}
566 	m_copydata(m, 0, len, (caddr_t)rtm);
567 	if (rtm->rtm_version != RTM_VERSION) {
568 		info.rti_info[RTAX_DST] = NULL;
569 		senderr(EPROTONOSUPPORT);
570 	}
571 	rtm->rtm_pid = curproc->p_pid;
572 	bzero(&info, sizeof(info));
573 	info.rti_addrs = rtm->rtm_addrs;
574 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
575 		info.rti_info[RTAX_DST] = NULL;
576 		senderr(EINVAL);
577 	}
578 	info.rti_flags = rtm->rtm_flags;
579 	if (info.rti_info[RTAX_DST] == NULL ||
580 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
581 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
582 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
583 		senderr(EINVAL);
584 	/*
585 	 * Verify that the caller has the appropriate privilege; RTM_GET
586 	 * is the only operation the non-superuser is allowed.
587 	 */
588 	if (rtm->rtm_type != RTM_GET) {
589 		error = priv_check(curthread, PRIV_NET_ROUTE);
590 		if (error)
591 			senderr(error);
592 	}
593 
594 	/*
595 	 * The given gateway address may be an interface address.
596 	 * For example, issuing a "route change" command on a route
597 	 * entry that was created from a tunnel, and the gateway
598 	 * address given is the local end point. In this case the
599 	 * RTF_GATEWAY flag must be cleared or the destination will
600 	 * not be reachable even though there is no error message.
601 	 */
602 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
603 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
604 		struct route gw_ro;
605 
606 		bzero(&gw_ro, sizeof(gw_ro));
607 		gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
608 		rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
609 		/*
610 		 * A host route through the loopback interface is
611 		 * installed for each interface adddress. In pre 8.0
612 		 * releases the interface address of a PPP link type
613 		 * is not reachable locally. This behavior is fixed as
614 		 * part of the new L2/L3 redesign and rewrite work. The
615 		 * signature of this interface address route is the
616 		 * AF_LINK sa_family type of the rt_gateway, and the
617 		 * rt_ifp has the IFF_LOOPBACK flag set.
618 		 */
619 		if (gw_ro.ro_rt != NULL &&
620 		    gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
621 		    gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)
622 			info.rti_flags &= ~RTF_GATEWAY;
623 		if (gw_ro.ro_rt != NULL)
624 			RTFREE(gw_ro.ro_rt);
625 	}
626 
627 	switch (rtm->rtm_type) {
628 		struct rtentry *saved_nrt;
629 
630 	case RTM_ADD:
631 		if (info.rti_info[RTAX_GATEWAY] == NULL)
632 			senderr(EINVAL);
633 		saved_nrt = NULL;
634 
635 		/* support for new ARP code */
636 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
637 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
638 			error = lla_rt_output(rtm, &info);
639 			break;
640 		}
641 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
642 		    so->so_fibnum);
643 		if (error == 0 && saved_nrt) {
644 			RT_LOCK(saved_nrt);
645 			rt_setmetrics(rtm->rtm_inits,
646 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
647 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
648 			RT_REMREF(saved_nrt);
649 			RT_UNLOCK(saved_nrt);
650 		}
651 		break;
652 
653 	case RTM_DELETE:
654 		saved_nrt = NULL;
655 		/* support for new ARP code */
656 		if (info.rti_info[RTAX_GATEWAY] &&
657 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
658 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
659 			error = lla_rt_output(rtm, &info);
660 			break;
661 		}
662 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
663 		    so->so_fibnum);
664 		if (error == 0) {
665 			RT_LOCK(saved_nrt);
666 			rt = saved_nrt;
667 			goto report;
668 		}
669 		break;
670 
671 	case RTM_GET:
672 	case RTM_CHANGE:
673 	case RTM_LOCK:
674 		rnh = rt_tables_get_rnh(so->so_fibnum,
675 		    info.rti_info[RTAX_DST]->sa_family);
676 		if (rnh == NULL)
677 			senderr(EAFNOSUPPORT);
678 		RADIX_NODE_HEAD_RLOCK(rnh);
679 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
680 			info.rti_info[RTAX_NETMASK], rnh);
681 		if (rt == NULL) {	/* XXX looks bogus */
682 			RADIX_NODE_HEAD_RUNLOCK(rnh);
683 			senderr(ESRCH);
684 		}
685 #ifdef RADIX_MPATH
686 		/*
687 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
688 		 * we require users to specify a matching RTAX_GATEWAY.
689 		 *
690 		 * for RTM_GET, gate is optional even with multipath.
691 		 * if gate == NULL the first match is returned.
692 		 * (no need to call rt_mpath_matchgate if gate == NULL)
693 		 */
694 		if (rn_mpath_capable(rnh) &&
695 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
696 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
697 			if (!rt) {
698 				RADIX_NODE_HEAD_RUNLOCK(rnh);
699 				senderr(ESRCH);
700 			}
701 		}
702 #endif
703 		/*
704 		 * If performing proxied L2 entry insertion, and
705 		 * the actual PPP host entry is found, perform
706 		 * another search to retrieve the prefix route of
707 		 * the local end point of the PPP link.
708 		 */
709 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
710 			struct sockaddr laddr;
711 
712 			if (rt->rt_ifp != NULL &&
713 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
714 				struct ifaddr *ifa;
715 
716 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
717 				if (ifa != NULL)
718 					rt_maskedcopy(ifa->ifa_addr,
719 						      &laddr,
720 						      ifa->ifa_netmask);
721 			} else
722 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
723 					      &laddr,
724 					      rt->rt_ifa->ifa_netmask);
725 			/*
726 			 * refactor rt and no lock operation necessary
727 			 */
728 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
729 			if (rt == NULL) {
730 				RADIX_NODE_HEAD_RUNLOCK(rnh);
731 				senderr(ESRCH);
732 			}
733 		}
734 		RT_LOCK(rt);
735 		RT_ADDREF(rt);
736 		RADIX_NODE_HEAD_RUNLOCK(rnh);
737 
738 		/*
739 		 * Fix for PR: 82974
740 		 *
741 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
742 		 * returns a perfect match in case a netmask is
743 		 * specified.  For host routes only a longest prefix
744 		 * match is returned so it is necessary to compare the
745 		 * existence of the netmask.  If both have a netmask
746 		 * rnh_lookup() did a perfect match and if none of them
747 		 * have a netmask both are host routes which is also a
748 		 * perfect match.
749 		 */
750 
751 		if (rtm->rtm_type != RTM_GET &&
752 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
753 			RT_UNLOCK(rt);
754 			senderr(ESRCH);
755 		}
756 
757 		switch(rtm->rtm_type) {
758 
759 		case RTM_GET:
760 		report:
761 			RT_LOCK_ASSERT(rt);
762 			if ((rt->rt_flags & RTF_HOST) == 0
763 			    ? jailed_without_vnet(curthread->td_ucred)
764 			    : prison_if(curthread->td_ucred,
765 			    rt_key(rt)) != 0) {
766 				RT_UNLOCK(rt);
767 				senderr(ESRCH);
768 			}
769 			info.rti_info[RTAX_DST] = rt_key(rt);
770 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
771 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
772 			info.rti_info[RTAX_GENMASK] = 0;
773 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
774 				ifp = rt->rt_ifp;
775 				if (ifp) {
776 					info.rti_info[RTAX_IFP] =
777 					    ifp->if_addr->ifa_addr;
778 					error = rtm_get_jailed(&info, ifp, rt,
779 					    &saun, curthread->td_ucred);
780 					if (error != 0) {
781 						RT_UNLOCK(rt);
782 						senderr(error);
783 					}
784 					if (ifp->if_flags & IFF_POINTOPOINT)
785 						info.rti_info[RTAX_BRD] =
786 						    rt->rt_ifa->ifa_dstaddr;
787 					rtm->rtm_index = ifp->if_index;
788 				} else {
789 					info.rti_info[RTAX_IFP] = NULL;
790 					info.rti_info[RTAX_IFA] = NULL;
791 				}
792 			} else if ((ifp = rt->rt_ifp) != NULL) {
793 				rtm->rtm_index = ifp->if_index;
794 			}
795 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
796 			if (len > rtm->rtm_msglen) {
797 				struct rt_msghdr *new_rtm;
798 				R_Malloc(new_rtm, struct rt_msghdr *, len);
799 				if (new_rtm == NULL) {
800 					RT_UNLOCK(rt);
801 					senderr(ENOBUFS);
802 				}
803 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
804 				Free(rtm); rtm = new_rtm;
805 			}
806 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
807 			rtm->rtm_flags = rt->rt_flags;
808 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
809 			rtm->rtm_addrs = info.rti_addrs;
810 			break;
811 
812 		case RTM_CHANGE:
813 			/*
814 			 * New gateway could require new ifaddr, ifp;
815 			 * flags may also be different; ifp may be specified
816 			 * by ll sockaddr when protocol address is ambiguous
817 			 */
818 			if (((rt->rt_flags & RTF_GATEWAY) &&
819 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
820 			    info.rti_info[RTAX_IFP] != NULL ||
821 			    (info.rti_info[RTAX_IFA] != NULL &&
822 			     !sa_equal(info.rti_info[RTAX_IFA],
823 				       rt->rt_ifa->ifa_addr))) {
824 				RT_UNLOCK(rt);
825 				RADIX_NODE_HEAD_LOCK(rnh);
826 				error = rt_getifa_fib(&info, rt->rt_fibnum);
827 				/*
828 				 * XXXRW: Really we should release this
829 				 * reference later, but this maintains
830 				 * historical behavior.
831 				 */
832 				if (info.rti_ifa != NULL)
833 					ifa_free(info.rti_ifa);
834 				RADIX_NODE_HEAD_UNLOCK(rnh);
835 				if (error != 0)
836 					senderr(error);
837 				RT_LOCK(rt);
838 			}
839 			if (info.rti_ifa != NULL &&
840 			    info.rti_ifa != rt->rt_ifa &&
841 			    rt->rt_ifa != NULL &&
842 			    rt->rt_ifa->ifa_rtrequest != NULL) {
843 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
844 				    &info);
845 				ifa_free(rt->rt_ifa);
846 			}
847 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
848 				RT_UNLOCK(rt);
849 				RADIX_NODE_HEAD_LOCK(rnh);
850 				RT_LOCK(rt);
851 
852 				error = rt_setgate(rt, rt_key(rt),
853 				    info.rti_info[RTAX_GATEWAY]);
854 				RADIX_NODE_HEAD_UNLOCK(rnh);
855 				if (error != 0) {
856 					RT_UNLOCK(rt);
857 					senderr(error);
858 				}
859 				rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
860 			}
861 			if (info.rti_ifa != NULL &&
862 			    info.rti_ifa != rt->rt_ifa) {
863 				ifa_ref(info.rti_ifa);
864 				rt->rt_ifa = info.rti_ifa;
865 				rt->rt_ifp = info.rti_ifp;
866 			}
867 			/* Allow some flags to be toggled on change. */
868 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
869 				    (rtm->rtm_flags & RTF_FMASK);
870 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
871 					&rt->rt_rmx);
872 			rtm->rtm_index = rt->rt_ifp->if_index;
873 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
874 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
875 			/* FALLTHROUGH */
876 		case RTM_LOCK:
877 			/* We don't support locks anymore */
878 			break;
879 		}
880 		RT_UNLOCK(rt);
881 		break;
882 
883 	default:
884 		senderr(EOPNOTSUPP);
885 	}
886 
887 flush:
888 	if (rtm) {
889 		if (error)
890 			rtm->rtm_errno = error;
891 		else
892 			rtm->rtm_flags |= RTF_DONE;
893 	}
894 	if (rt)		/* XXX can this be true? */
895 		RTFREE(rt);
896     {
897 	struct rawcb *rp = NULL;
898 	/*
899 	 * Check to see if we don't want our own messages.
900 	 */
901 	if ((so->so_options & SO_USELOOPBACK) == 0) {
902 		if (route_cb.any_count <= 1) {
903 			if (rtm)
904 				Free(rtm);
905 			m_freem(m);
906 			return (error);
907 		}
908 		/* There is another listener, so construct message */
909 		rp = sotorawcb(so);
910 	}
911 	if (rtm) {
912 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
913 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
914 			m_freem(m);
915 			m = NULL;
916 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
917 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
918 	}
919 	if (m) {
920 		M_SETFIB(m, so->so_fibnum);
921 		m->m_flags |= RTS_FILTER_FIB;
922 		if (rp) {
923 			/*
924 			 * XXX insure we don't get a copy by
925 			 * invalidating our protocol
926 			 */
927 			unsigned short family = rp->rcb_proto.sp_family;
928 			rp->rcb_proto.sp_family = 0;
929 			rt_dispatch(m, info.rti_info[RTAX_DST]);
930 			rp->rcb_proto.sp_family = family;
931 		} else
932 			rt_dispatch(m, info.rti_info[RTAX_DST]);
933 	}
934 	/* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
935 	if (rtm)
936 		Free(rtm);
937     }
938 	return (error);
939 #undef	sa_equal
940 }
941 
942 static void
943 rt_setmetrics(u_long which, const struct rt_metrics *in,
944 	struct rt_metrics_lite *out)
945 {
946 #define metric(f, e) if (which & (f)) out->e = in->e;
947 	/*
948 	 * Only these are stored in the routing entry since introduction
949 	 * of tcp hostcache. The rest is ignored.
950 	 */
951 	metric(RTV_MTU, rmx_mtu);
952 	metric(RTV_WEIGHT, rmx_weight);
953 	/* Userland -> kernel timebase conversion. */
954 	if (which & RTV_EXPIRE)
955 		out->rmx_expire = in->rmx_expire ?
956 		    in->rmx_expire - time_second + time_uptime : 0;
957 #undef metric
958 }
959 
960 static void
961 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
962 {
963 #define metric(e) out->e = in->e;
964 	bzero(out, sizeof(*out));
965 	metric(rmx_mtu);
966 	metric(rmx_weight);
967 	/* Kernel -> userland timebase conversion. */
968 	out->rmx_expire = in->rmx_expire ?
969 	    in->rmx_expire - time_uptime + time_second : 0;
970 #undef metric
971 }
972 
973 /*
974  * Extract the addresses of the passed sockaddrs.
975  * Do a little sanity checking so as to avoid bad memory references.
976  * This data is derived straight from userland.
977  */
978 static int
979 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
980 {
981 	struct sockaddr *sa;
982 	int i;
983 
984 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
985 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
986 			continue;
987 		sa = (struct sockaddr *)cp;
988 		/*
989 		 * It won't fit.
990 		 */
991 		if (cp + sa->sa_len > cplim)
992 			return (EINVAL);
993 		/*
994 		 * there are no more.. quit now
995 		 * If there are more bits, they are in error.
996 		 * I've seen this. route(1) can evidently generate these.
997 		 * This causes kernel to core dump.
998 		 * for compatibility, If we see this, point to a safe address.
999 		 */
1000 		if (sa->sa_len == 0) {
1001 			rtinfo->rti_info[i] = &sa_zero;
1002 			return (0); /* should be EINVAL but for compat */
1003 		}
1004 		/* accept it */
1005 		rtinfo->rti_info[i] = sa;
1006 		cp += SA_SIZE(sa);
1007 	}
1008 	return (0);
1009 }
1010 
1011 static struct mbuf *
1012 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1013 {
1014 	struct rt_msghdr *rtm;
1015 	struct mbuf *m;
1016 	int i;
1017 	struct sockaddr *sa;
1018 	int len, dlen;
1019 
1020 	switch (type) {
1021 
1022 	case RTM_DELADDR:
1023 	case RTM_NEWADDR:
1024 		len = sizeof(struct ifa_msghdr);
1025 		break;
1026 
1027 	case RTM_DELMADDR:
1028 	case RTM_NEWMADDR:
1029 		len = sizeof(struct ifma_msghdr);
1030 		break;
1031 
1032 	case RTM_IFINFO:
1033 		len = sizeof(struct if_msghdr);
1034 		break;
1035 
1036 	case RTM_IFANNOUNCE:
1037 	case RTM_IEEE80211:
1038 		len = sizeof(struct if_announcemsghdr);
1039 		break;
1040 
1041 	default:
1042 		len = sizeof(struct rt_msghdr);
1043 	}
1044 	if (len > MCLBYTES)
1045 		panic("rt_msg1");
1046 	m = m_gethdr(M_DONTWAIT, MT_DATA);
1047 	if (m && len > MHLEN) {
1048 		MCLGET(m, M_DONTWAIT);
1049 		if ((m->m_flags & M_EXT) == 0) {
1050 			m_free(m);
1051 			m = NULL;
1052 		}
1053 	}
1054 	if (m == NULL)
1055 		return (m);
1056 	m->m_pkthdr.len = m->m_len = len;
1057 	m->m_pkthdr.rcvif = NULL;
1058 	rtm = mtod(m, struct rt_msghdr *);
1059 	bzero((caddr_t)rtm, len);
1060 	for (i = 0; i < RTAX_MAX; i++) {
1061 		if ((sa = rtinfo->rti_info[i]) == NULL)
1062 			continue;
1063 		rtinfo->rti_addrs |= (1 << i);
1064 		dlen = SA_SIZE(sa);
1065 		m_copyback(m, len, dlen, (caddr_t)sa);
1066 		len += dlen;
1067 	}
1068 	if (m->m_pkthdr.len != len) {
1069 		m_freem(m);
1070 		return (NULL);
1071 	}
1072 	rtm->rtm_msglen = len;
1073 	rtm->rtm_version = RTM_VERSION;
1074 	rtm->rtm_type = type;
1075 	return (m);
1076 }
1077 
1078 static int
1079 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1080 {
1081 	int i;
1082 	int len, dlen, second_time = 0;
1083 	caddr_t cp0;
1084 
1085 	rtinfo->rti_addrs = 0;
1086 again:
1087 	switch (type) {
1088 
1089 	case RTM_DELADDR:
1090 	case RTM_NEWADDR:
1091 		len = sizeof(struct ifa_msghdr);
1092 		break;
1093 
1094 	case RTM_IFINFO:
1095 #ifdef COMPAT_FREEBSD32
1096 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1097 			len = sizeof(struct if_msghdr32);
1098 			break;
1099 		}
1100 #endif
1101 		len = sizeof(struct if_msghdr);
1102 		break;
1103 
1104 	case RTM_NEWMADDR:
1105 		len = sizeof(struct ifma_msghdr);
1106 		break;
1107 
1108 	default:
1109 		len = sizeof(struct rt_msghdr);
1110 	}
1111 	cp0 = cp;
1112 	if (cp0)
1113 		cp += len;
1114 	for (i = 0; i < RTAX_MAX; i++) {
1115 		struct sockaddr *sa;
1116 
1117 		if ((sa = rtinfo->rti_info[i]) == NULL)
1118 			continue;
1119 		rtinfo->rti_addrs |= (1 << i);
1120 		dlen = SA_SIZE(sa);
1121 		if (cp) {
1122 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1123 			cp += dlen;
1124 		}
1125 		len += dlen;
1126 	}
1127 	len = ALIGN(len);
1128 	if (cp == NULL && w != NULL && !second_time) {
1129 		struct walkarg *rw = w;
1130 
1131 		if (rw->w_req) {
1132 			if (rw->w_tmemsize < len) {
1133 				if (rw->w_tmem)
1134 					free(rw->w_tmem, M_RTABLE);
1135 				rw->w_tmem = (caddr_t)
1136 					malloc(len, M_RTABLE, M_NOWAIT);
1137 				if (rw->w_tmem)
1138 					rw->w_tmemsize = len;
1139 			}
1140 			if (rw->w_tmem) {
1141 				cp = rw->w_tmem;
1142 				second_time = 1;
1143 				goto again;
1144 			}
1145 		}
1146 	}
1147 	if (cp) {
1148 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1149 
1150 		rtm->rtm_version = RTM_VERSION;
1151 		rtm->rtm_type = type;
1152 		rtm->rtm_msglen = len;
1153 	}
1154 	return (len);
1155 }
1156 
1157 /*
1158  * This routine is called to generate a message from the routing
1159  * socket indicating that a redirect has occured, a routing lookup
1160  * has failed, or that a protocol has detected timeouts to a particular
1161  * destination.
1162  */
1163 void
1164 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1165     int fibnum)
1166 {
1167 	struct rt_msghdr *rtm;
1168 	struct mbuf *m;
1169 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1170 
1171 	if (route_cb.any_count == 0)
1172 		return;
1173 	m = rt_msg1(type, rtinfo);
1174 	if (m == NULL)
1175 		return;
1176 
1177 	if (fibnum != RTS_ALLFIBS) {
1178 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1179 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1180 		M_SETFIB(m, fibnum);
1181 		m->m_flags |= RTS_FILTER_FIB;
1182 	}
1183 
1184 	rtm = mtod(m, struct rt_msghdr *);
1185 	rtm->rtm_flags = RTF_DONE | flags;
1186 	rtm->rtm_errno = error;
1187 	rtm->rtm_addrs = rtinfo->rti_addrs;
1188 	rt_dispatch(m, sa);
1189 }
1190 
1191 void
1192 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1193 {
1194 
1195 	rt_missmsg_fib(type, rtinfo, flags, error, RTS_ALLFIBS);
1196 }
1197 
1198 /*
1199  * This routine is called to generate a message from the routing
1200  * socket indicating that the status of a network interface has changed.
1201  */
1202 void
1203 rt_ifmsg(struct ifnet *ifp)
1204 {
1205 	struct if_msghdr *ifm;
1206 	struct mbuf *m;
1207 	struct rt_addrinfo info;
1208 
1209 	if (route_cb.any_count == 0)
1210 		return;
1211 	bzero((caddr_t)&info, sizeof(info));
1212 	m = rt_msg1(RTM_IFINFO, &info);
1213 	if (m == NULL)
1214 		return;
1215 	ifm = mtod(m, struct if_msghdr *);
1216 	ifm->ifm_index = ifp->if_index;
1217 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1218 	ifm->ifm_data = ifp->if_data;
1219 	ifm->ifm_addrs = 0;
1220 	rt_dispatch(m, NULL);
1221 }
1222 
1223 /*
1224  * This is called to generate messages from the routing socket
1225  * indicating a network interface has had addresses associated with it.
1226  * if we ever reverse the logic and replace messages TO the routing
1227  * socket indicate a request to configure interfaces, then it will
1228  * be unnecessary as the routing socket will automatically generate
1229  * copies of it.
1230  */
1231 void
1232 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
1233     int fibnum)
1234 {
1235 	struct rt_addrinfo info;
1236 	struct sockaddr *sa = NULL;
1237 	int pass;
1238 	struct mbuf *m = NULL;
1239 	struct ifnet *ifp = ifa->ifa_ifp;
1240 
1241 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1242 		("unexpected cmd %u", cmd));
1243 #if defined(INET) || defined(INET6)
1244 #ifdef SCTP
1245 	/*
1246 	 * notify the SCTP stack
1247 	 * this will only get called when an address is added/deleted
1248 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1249 	 */
1250 	sctp_addr_change(ifa, cmd);
1251 #endif /* SCTP */
1252 #endif
1253 	if (route_cb.any_count == 0)
1254 		return;
1255 	for (pass = 1; pass < 3; pass++) {
1256 		bzero((caddr_t)&info, sizeof(info));
1257 		if ((cmd == RTM_ADD && pass == 1) ||
1258 		    (cmd == RTM_DELETE && pass == 2)) {
1259 			struct ifa_msghdr *ifam;
1260 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1261 
1262 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1263 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1264 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1265 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1266 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1267 				continue;
1268 			ifam = mtod(m, struct ifa_msghdr *);
1269 			ifam->ifam_index = ifp->if_index;
1270 			ifam->ifam_metric = ifa->ifa_metric;
1271 			ifam->ifam_flags = ifa->ifa_flags;
1272 			ifam->ifam_addrs = info.rti_addrs;
1273 		}
1274 		if ((cmd == RTM_ADD && pass == 2) ||
1275 		    (cmd == RTM_DELETE && pass == 1)) {
1276 			struct rt_msghdr *rtm;
1277 
1278 			if (rt == NULL)
1279 				continue;
1280 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1281 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1282 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1283 			if ((m = rt_msg1(cmd, &info)) == NULL)
1284 				continue;
1285 			rtm = mtod(m, struct rt_msghdr *);
1286 			rtm->rtm_index = ifp->if_index;
1287 			rtm->rtm_flags |= rt->rt_flags;
1288 			rtm->rtm_errno = error;
1289 			rtm->rtm_addrs = info.rti_addrs;
1290 		}
1291 		if (fibnum != RTS_ALLFIBS) {
1292 			KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: "
1293 			    "fibnum out of range 0 <= %d < %d", __func__,
1294 			     fibnum, rt_numfibs));
1295 			M_SETFIB(m, fibnum);
1296 			m->m_flags |= RTS_FILTER_FIB;
1297 		}
1298 		rt_dispatch(m, sa);
1299 	}
1300 }
1301 
1302 void
1303 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1304 {
1305 
1306 	rt_newaddrmsg_fib(cmd, ifa, error, rt, RTS_ALLFIBS);
1307 }
1308 
1309 /*
1310  * This is the analogue to the rt_newaddrmsg which performs the same
1311  * function but for multicast group memberhips.  This is easier since
1312  * there is no route state to worry about.
1313  */
1314 void
1315 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1316 {
1317 	struct rt_addrinfo info;
1318 	struct mbuf *m = NULL;
1319 	struct ifnet *ifp = ifma->ifma_ifp;
1320 	struct ifma_msghdr *ifmam;
1321 
1322 	if (route_cb.any_count == 0)
1323 		return;
1324 
1325 	bzero((caddr_t)&info, sizeof(info));
1326 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1327 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1328 	/*
1329 	 * If a link-layer address is present, present it as a ``gateway''
1330 	 * (similarly to how ARP entries, e.g., are presented).
1331 	 */
1332 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1333 	m = rt_msg1(cmd, &info);
1334 	if (m == NULL)
1335 		return;
1336 	ifmam = mtod(m, struct ifma_msghdr *);
1337 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1338 	    __func__));
1339 	ifmam->ifmam_index = ifp->if_index;
1340 	ifmam->ifmam_addrs = info.rti_addrs;
1341 	rt_dispatch(m, ifma->ifma_addr);
1342 }
1343 
1344 static struct mbuf *
1345 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1346 	struct rt_addrinfo *info)
1347 {
1348 	struct if_announcemsghdr *ifan;
1349 	struct mbuf *m;
1350 
1351 	if (route_cb.any_count == 0)
1352 		return NULL;
1353 	bzero((caddr_t)info, sizeof(*info));
1354 	m = rt_msg1(type, info);
1355 	if (m != NULL) {
1356 		ifan = mtod(m, struct if_announcemsghdr *);
1357 		ifan->ifan_index = ifp->if_index;
1358 		strlcpy(ifan->ifan_name, ifp->if_xname,
1359 			sizeof(ifan->ifan_name));
1360 		ifan->ifan_what = what;
1361 	}
1362 	return m;
1363 }
1364 
1365 /*
1366  * This is called to generate routing socket messages indicating
1367  * IEEE80211 wireless events.
1368  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1369  */
1370 void
1371 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1372 {
1373 	struct mbuf *m;
1374 	struct rt_addrinfo info;
1375 
1376 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1377 	if (m != NULL) {
1378 		/*
1379 		 * Append the ieee80211 data.  Try to stick it in the
1380 		 * mbuf containing the ifannounce msg; otherwise allocate
1381 		 * a new mbuf and append.
1382 		 *
1383 		 * NB: we assume m is a single mbuf.
1384 		 */
1385 		if (data_len > M_TRAILINGSPACE(m)) {
1386 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1387 			if (n == NULL) {
1388 				m_freem(m);
1389 				return;
1390 			}
1391 			bcopy(data, mtod(n, void *), data_len);
1392 			n->m_len = data_len;
1393 			m->m_next = n;
1394 		} else if (data_len > 0) {
1395 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1396 			m->m_len += data_len;
1397 		}
1398 		if (m->m_flags & M_PKTHDR)
1399 			m->m_pkthdr.len += data_len;
1400 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1401 		rt_dispatch(m, NULL);
1402 	}
1403 }
1404 
1405 /*
1406  * This is called to generate routing socket messages indicating
1407  * network interface arrival and departure.
1408  */
1409 void
1410 rt_ifannouncemsg(struct ifnet *ifp, int what)
1411 {
1412 	struct mbuf *m;
1413 	struct rt_addrinfo info;
1414 
1415 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1416 	if (m != NULL)
1417 		rt_dispatch(m, NULL);
1418 }
1419 
1420 static void
1421 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1422 {
1423 	struct m_tag *tag;
1424 
1425 	/*
1426 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1427 	 * use when injecting the mbuf into the routing socket buffer from
1428 	 * the netisr.
1429 	 */
1430 	if (sa != NULL) {
1431 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1432 		    M_NOWAIT);
1433 		if (tag == NULL) {
1434 			m_freem(m);
1435 			return;
1436 		}
1437 		*(unsigned short *)(tag + 1) = sa->sa_family;
1438 		m_tag_prepend(m, tag);
1439 	}
1440 #ifdef VIMAGE
1441 	if (V_loif)
1442 		m->m_pkthdr.rcvif = V_loif;
1443 	else {
1444 		m_freem(m);
1445 		return;
1446 	}
1447 #endif
1448 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1449 }
1450 
1451 /*
1452  * This is used in dumping the kernel table via sysctl().
1453  */
1454 static int
1455 sysctl_dumpentry(struct radix_node *rn, void *vw)
1456 {
1457 	struct walkarg *w = vw;
1458 	struct rtentry *rt = (struct rtentry *)rn;
1459 	int error = 0, size;
1460 	struct rt_addrinfo info;
1461 
1462 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1463 		return 0;
1464 	if ((rt->rt_flags & RTF_HOST) == 0
1465 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1466 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1467 		return (0);
1468 	bzero((caddr_t)&info, sizeof(info));
1469 	info.rti_info[RTAX_DST] = rt_key(rt);
1470 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1471 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1472 	info.rti_info[RTAX_GENMASK] = 0;
1473 	if (rt->rt_ifp) {
1474 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1475 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1476 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1477 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1478 	}
1479 	size = rt_msg2(RTM_GET, &info, NULL, w);
1480 	if (w->w_req && w->w_tmem) {
1481 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1482 
1483 		rtm->rtm_flags = rt->rt_flags;
1484 		/*
1485 		 * let's be honest about this being a retarded hack
1486 		 */
1487 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1488 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1489 		rtm->rtm_index = rt->rt_ifp->if_index;
1490 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1491 		rtm->rtm_addrs = info.rti_addrs;
1492 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1493 		return (error);
1494 	}
1495 	return (error);
1496 }
1497 
1498 #ifdef COMPAT_FREEBSD32
1499 static void
1500 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1501 {
1502 
1503 	bzero(dst, sizeof(*dst));
1504 	CP(*src, *dst, ifi_type);
1505 	CP(*src, *dst, ifi_physical);
1506 	CP(*src, *dst, ifi_addrlen);
1507 	CP(*src, *dst, ifi_hdrlen);
1508 	CP(*src, *dst, ifi_link_state);
1509 	dst->ifi_datalen = sizeof(struct if_data32);
1510 	CP(*src, *dst, ifi_mtu);
1511 	CP(*src, *dst, ifi_metric);
1512 	CP(*src, *dst, ifi_baudrate);
1513 	CP(*src, *dst, ifi_ipackets);
1514 	CP(*src, *dst, ifi_ierrors);
1515 	CP(*src, *dst, ifi_opackets);
1516 	CP(*src, *dst, ifi_oerrors);
1517 	CP(*src, *dst, ifi_collisions);
1518 	CP(*src, *dst, ifi_ibytes);
1519 	CP(*src, *dst, ifi_obytes);
1520 	CP(*src, *dst, ifi_imcasts);
1521 	CP(*src, *dst, ifi_omcasts);
1522 	CP(*src, *dst, ifi_iqdrops);
1523 	CP(*src, *dst, ifi_noproto);
1524 	CP(*src, *dst, ifi_hwassist);
1525 	CP(*src, *dst, ifi_epoch);
1526 	TV_CP(*src, *dst, ifi_lastchange);
1527 }
1528 #endif
1529 
1530 static int
1531 sysctl_iflist(int af, struct walkarg *w)
1532 {
1533 	struct ifnet *ifp;
1534 	struct ifaddr *ifa;
1535 	struct rt_addrinfo info;
1536 	int len, error = 0;
1537 
1538 	bzero((caddr_t)&info, sizeof(info));
1539 	IFNET_RLOCK();
1540 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1541 		if (w->w_arg && w->w_arg != ifp->if_index)
1542 			continue;
1543 		IF_ADDR_LOCK(ifp);
1544 		ifa = ifp->if_addr;
1545 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1546 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1547 		info.rti_info[RTAX_IFP] = NULL;
1548 		if (w->w_req && w->w_tmem) {
1549 			struct if_msghdr *ifm;
1550 
1551 #ifdef COMPAT_FREEBSD32
1552 			if (w->w_req->flags & SCTL_MASK32) {
1553 				struct if_msghdr32 *ifm32;
1554 
1555 				ifm32 = (struct if_msghdr32 *)w->w_tmem;
1556 				ifm32->ifm_index = ifp->if_index;
1557 				ifm32->ifm_flags = ifp->if_flags |
1558 				    ifp->if_drv_flags;
1559 				copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1560 				ifm32->ifm_addrs = info.rti_addrs;
1561 				error = SYSCTL_OUT(w->w_req, (caddr_t)ifm32,
1562 				    len);
1563 				goto sysctl_out;
1564 			}
1565 #endif
1566 			ifm = (struct if_msghdr *)w->w_tmem;
1567 			ifm->ifm_index = ifp->if_index;
1568 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1569 			ifm->ifm_data = ifp->if_data;
1570 			ifm->ifm_addrs = info.rti_addrs;
1571 			error = SYSCTL_OUT(w->w_req, (caddr_t)ifm, len);
1572 #ifdef COMPAT_FREEBSD32
1573 		sysctl_out:
1574 #endif
1575 			if (error)
1576 				goto done;
1577 		}
1578 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1579 			if (af && af != ifa->ifa_addr->sa_family)
1580 				continue;
1581 			if (prison_if(w->w_req->td->td_ucred,
1582 			    ifa->ifa_addr) != 0)
1583 				continue;
1584 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1585 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1586 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1587 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1588 			if (w->w_req && w->w_tmem) {
1589 				struct ifa_msghdr *ifam;
1590 
1591 				ifam = (struct ifa_msghdr *)w->w_tmem;
1592 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1593 				ifam->ifam_flags = ifa->ifa_flags;
1594 				ifam->ifam_metric = ifa->ifa_metric;
1595 				ifam->ifam_addrs = info.rti_addrs;
1596 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1597 				if (error)
1598 					goto done;
1599 			}
1600 		}
1601 		IF_ADDR_UNLOCK(ifp);
1602 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1603 			info.rti_info[RTAX_BRD] = NULL;
1604 	}
1605 done:
1606 	if (ifp != NULL)
1607 		IF_ADDR_UNLOCK(ifp);
1608 	IFNET_RUNLOCK();
1609 	return (error);
1610 }
1611 
1612 static int
1613 sysctl_ifmalist(int af, struct walkarg *w)
1614 {
1615 	struct ifnet *ifp;
1616 	struct ifmultiaddr *ifma;
1617 	struct	rt_addrinfo info;
1618 	int	len, error = 0;
1619 	struct ifaddr *ifa;
1620 
1621 	bzero((caddr_t)&info, sizeof(info));
1622 	IFNET_RLOCK();
1623 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1624 		if (w->w_arg && w->w_arg != ifp->if_index)
1625 			continue;
1626 		ifa = ifp->if_addr;
1627 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1628 		IF_ADDR_LOCK(ifp);
1629 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1630 			if (af && af != ifma->ifma_addr->sa_family)
1631 				continue;
1632 			if (prison_if(w->w_req->td->td_ucred,
1633 			    ifma->ifma_addr) != 0)
1634 				continue;
1635 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1636 			info.rti_info[RTAX_GATEWAY] =
1637 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1638 			    ifma->ifma_lladdr : NULL;
1639 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1640 			if (w->w_req && w->w_tmem) {
1641 				struct ifma_msghdr *ifmam;
1642 
1643 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1644 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1645 				ifmam->ifmam_flags = 0;
1646 				ifmam->ifmam_addrs = info.rti_addrs;
1647 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1648 				if (error) {
1649 					IF_ADDR_UNLOCK(ifp);
1650 					goto done;
1651 				}
1652 			}
1653 		}
1654 		IF_ADDR_UNLOCK(ifp);
1655 	}
1656 done:
1657 	IFNET_RUNLOCK();
1658 	return (error);
1659 }
1660 
1661 static int
1662 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1663 {
1664 	int	*name = (int *)arg1;
1665 	u_int	namelen = arg2;
1666 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1667 	int	i, lim, error = EINVAL;
1668 	u_char	af;
1669 	struct	walkarg w;
1670 
1671 	name ++;
1672 	namelen--;
1673 	if (req->newptr)
1674 		return (EPERM);
1675 	if (namelen != 3)
1676 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1677 	af = name[0];
1678 	if (af > AF_MAX)
1679 		return (EINVAL);
1680 	bzero(&w, sizeof(w));
1681 	w.w_op = name[1];
1682 	w.w_arg = name[2];
1683 	w.w_req = req;
1684 
1685 	error = sysctl_wire_old_buffer(req, 0);
1686 	if (error)
1687 		return (error);
1688 	switch (w.w_op) {
1689 
1690 	case NET_RT_DUMP:
1691 	case NET_RT_FLAGS:
1692 		if (af == 0) {			/* dump all tables */
1693 			i = 1;
1694 			lim = AF_MAX;
1695 		} else				/* dump only one table */
1696 			i = lim = af;
1697 
1698 		/*
1699 		 * take care of llinfo entries, the caller must
1700 		 * specify an AF
1701 		 */
1702 		if (w.w_op == NET_RT_FLAGS &&
1703 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1704 			if (af != 0)
1705 				error = lltable_sysctl_dumparp(af, w.w_req);
1706 			else
1707 				error = EINVAL;
1708 			break;
1709 		}
1710 		/*
1711 		 * take care of routing entries
1712 		 */
1713 		for (error = 0; error == 0 && i <= lim; i++) {
1714 			rnh = rt_tables_get_rnh(req->td->td_proc->p_fibnum, i);
1715 			if (rnh != NULL) {
1716 				RADIX_NODE_HEAD_LOCK(rnh);
1717 			    	error = rnh->rnh_walktree(rnh,
1718 				    sysctl_dumpentry, &w);
1719 				RADIX_NODE_HEAD_UNLOCK(rnh);
1720 			} else if (af != 0)
1721 				error = EAFNOSUPPORT;
1722 		}
1723 		break;
1724 
1725 	case NET_RT_IFLIST:
1726 		error = sysctl_iflist(af, &w);
1727 		break;
1728 
1729 	case NET_RT_IFMALIST:
1730 		error = sysctl_ifmalist(af, &w);
1731 		break;
1732 	}
1733 	if (w.w_tmem)
1734 		free(w.w_tmem, M_RTABLE);
1735 	return (error);
1736 }
1737 
1738 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1739 
1740 /*
1741  * Definitions of protocols supported in the ROUTE domain.
1742  */
1743 
1744 static struct domain routedomain;		/* or at least forward */
1745 
1746 static struct protosw routesw[] = {
1747 {
1748 	.pr_type =		SOCK_RAW,
1749 	.pr_domain =		&routedomain,
1750 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1751 	.pr_output =		route_output,
1752 	.pr_ctlinput =		raw_ctlinput,
1753 	.pr_init =		raw_init,
1754 	.pr_usrreqs =		&route_usrreqs
1755 }
1756 };
1757 
1758 static struct domain routedomain = {
1759 	.dom_family =		PF_ROUTE,
1760 	.dom_name =		 "route",
1761 	.dom_protosw =		routesw,
1762 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1763 };
1764 
1765 VNET_DOMAIN_SET(route);
1766