xref: /freebsd/sys/net/rtsock.c (revision f7167e0ea0bf5aaabff9490453b3b71b3f1b4d51)
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_mpath.h"
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 
37 #include <sys/param.h>
38 #include <sys/jail.h>
39 #include <sys/kernel.h>
40 #include <sys/domain.h>
41 #include <sys/lock.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/priv.h>
45 #include <sys/proc.h>
46 #include <sys/protosw.h>
47 #include <sys/rwlock.h>
48 #include <sys/signalvar.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/sysctl.h>
52 #include <sys/systm.h>
53 
54 #include <net/if.h>
55 #include <net/if_var.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 #include <netinet/ip_carp.h>
67 #ifdef INET6
68 #include <netinet6/ip6_var.h>
69 #include <netinet6/scope6_var.h>
70 #endif
71 
72 #ifdef COMPAT_FREEBSD32
73 #include <sys/mount.h>
74 #include <compat/freebsd32/freebsd32.h>
75 
76 struct if_data32 {
77 	uint8_t	ifi_type;
78 	uint8_t	ifi_physical;
79 	uint8_t	ifi_addrlen;
80 	uint8_t	ifi_hdrlen;
81 	uint8_t	ifi_link_state;
82 	uint8_t	ifi_vhid;
83 	uint8_t	ifi_baudrate_pf;
84 	uint8_t	ifi_datalen;
85 	uint32_t ifi_mtu;
86 	uint32_t ifi_metric;
87 	uint32_t ifi_baudrate;
88 	uint32_t ifi_ipackets;
89 	uint32_t ifi_ierrors;
90 	uint32_t ifi_opackets;
91 	uint32_t ifi_oerrors;
92 	uint32_t ifi_collisions;
93 	uint32_t ifi_ibytes;
94 	uint32_t ifi_obytes;
95 	uint32_t ifi_imcasts;
96 	uint32_t ifi_omcasts;
97 	uint32_t ifi_iqdrops;
98 	uint32_t ifi_noproto;
99 	uint32_t ifi_hwassist;
100 	int32_t	ifi_epoch;
101 	struct	timeval32 ifi_lastchange;
102 };
103 
104 struct if_msghdr32 {
105 	uint16_t ifm_msglen;
106 	uint8_t	ifm_version;
107 	uint8_t	ifm_type;
108 	int32_t	ifm_addrs;
109 	int32_t	ifm_flags;
110 	uint16_t ifm_index;
111 	struct	if_data32 ifm_data;
112 };
113 
114 struct if_msghdrl32 {
115 	uint16_t ifm_msglen;
116 	uint8_t	ifm_version;
117 	uint8_t	ifm_type;
118 	int32_t	ifm_addrs;
119 	int32_t	ifm_flags;
120 	uint16_t ifm_index;
121 	uint16_t _ifm_spare1;
122 	uint16_t ifm_len;
123 	uint16_t ifm_data_off;
124 	struct	if_data32 ifm_data;
125 };
126 
127 struct ifa_msghdrl32 {
128 	uint16_t ifam_msglen;
129 	uint8_t	ifam_version;
130 	uint8_t	ifam_type;
131 	int32_t	ifam_addrs;
132 	int32_t	ifam_flags;
133 	uint16_t ifam_index;
134 	uint16_t _ifam_spare1;
135 	uint16_t ifam_len;
136 	uint16_t ifam_data_off;
137 	int32_t	ifam_metric;
138 	struct	if_data32 ifam_data;
139 };
140 #endif /* COMPAT_FREEBSD32 */
141 
142 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
143 
144 /* NB: these are not modified */
145 static struct	sockaddr route_src = { 2, PF_ROUTE, };
146 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
147 
148 /* These are external hooks for CARP. */
149 int	(*carp_get_vhid_p)(struct ifaddr *);
150 
151 /*
152  * Used by rtsock/raw_input callback code to decide whether to filter the update
153  * notification to a socket bound to a particular FIB.
154  */
155 #define	RTS_FILTER_FIB	M_PROTO8
156 
157 static struct {
158 	int	ip_count;	/* attached w/ AF_INET */
159 	int	ip6_count;	/* attached w/ AF_INET6 */
160 	int	ipx_count;	/* attached w/ AF_IPX */
161 	int	any_count;	/* total attached */
162 } route_cb;
163 
164 struct mtx rtsock_mtx;
165 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
166 
167 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
168 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
169 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
170 
171 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
172 
173 struct walkarg {
174 	int	w_tmemsize;
175 	int	w_op, w_arg;
176 	caddr_t	w_tmem;
177 	struct sysctl_req *w_req;
178 };
179 
180 static void	rts_input(struct mbuf *m);
181 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
182 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
183 			caddr_t cp, struct walkarg *w);
184 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
185 			struct rt_addrinfo *rtinfo);
186 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
187 static int	sysctl_iflist(int af, struct walkarg *w);
188 static int	sysctl_ifmalist(int af, struct walkarg *w);
189 static int	route_output(struct mbuf *m, struct socket *so);
190 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
191 			struct rt_metrics_lite *out);
192 static void	rt_getmetrics(const struct rt_metrics_lite *in,
193 			struct rt_metrics *out);
194 static void	rt_dispatch(struct mbuf *, sa_family_t);
195 
196 static struct netisr_handler rtsock_nh = {
197 	.nh_name = "rtsock",
198 	.nh_handler = rts_input,
199 	.nh_proto = NETISR_ROUTE,
200 	.nh_policy = NETISR_POLICY_SOURCE,
201 };
202 
203 static int
204 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
205 {
206 	int error, qlimit;
207 
208 	netisr_getqlimit(&rtsock_nh, &qlimit);
209 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
210         if (error || !req->newptr)
211                 return (error);
212 	if (qlimit < 1)
213 		return (EINVAL);
214 	return (netisr_setqlimit(&rtsock_nh, qlimit));
215 }
216 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
217     0, 0, sysctl_route_netisr_maxqlen, "I",
218     "maximum routing socket dispatch queue length");
219 
220 static void
221 rts_init(void)
222 {
223 	int tmp;
224 
225 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
226 		rtsock_nh.nh_qlimit = tmp;
227 	netisr_register(&rtsock_nh);
228 }
229 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
230 
231 static int
232 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
233     struct rawcb *rp)
234 {
235 	int fibnum;
236 
237 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
238 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
239 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
240 
241 	/* No filtering requested. */
242 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
243 		return (0);
244 
245 	/* Check if it is a rts and the fib matches the one of the socket. */
246 	fibnum = M_GETFIB(m);
247 	if (proto->sp_family != PF_ROUTE ||
248 	    rp->rcb_socket == NULL ||
249 	    rp->rcb_socket->so_fibnum == fibnum)
250 		return (0);
251 
252 	/* Filtering requested and no match, the socket shall be skipped. */
253 	return (1);
254 }
255 
256 static void
257 rts_input(struct mbuf *m)
258 {
259 	struct sockproto route_proto;
260 	unsigned short *family;
261 	struct m_tag *tag;
262 
263 	route_proto.sp_family = PF_ROUTE;
264 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
265 	if (tag != NULL) {
266 		family = (unsigned short *)(tag + 1);
267 		route_proto.sp_protocol = *family;
268 		m_tag_delete(m, tag);
269 	} else
270 		route_proto.sp_protocol = 0;
271 
272 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
273 }
274 
275 /*
276  * It really doesn't make any sense at all for this code to share much
277  * with raw_usrreq.c, since its functionality is so restricted.  XXX
278  */
279 static void
280 rts_abort(struct socket *so)
281 {
282 
283 	raw_usrreqs.pru_abort(so);
284 }
285 
286 static void
287 rts_close(struct socket *so)
288 {
289 
290 	raw_usrreqs.pru_close(so);
291 }
292 
293 /* pru_accept is EOPNOTSUPP */
294 
295 static int
296 rts_attach(struct socket *so, int proto, struct thread *td)
297 {
298 	struct rawcb *rp;
299 	int error;
300 
301 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
302 
303 	/* XXX */
304 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
305 	if (rp == NULL)
306 		return ENOBUFS;
307 
308 	so->so_pcb = (caddr_t)rp;
309 	so->so_fibnum = td->td_proc->p_fibnum;
310 	error = raw_attach(so, proto);
311 	rp = sotorawcb(so);
312 	if (error) {
313 		so->so_pcb = NULL;
314 		free(rp, M_PCB);
315 		return error;
316 	}
317 	RTSOCK_LOCK();
318 	switch(rp->rcb_proto.sp_protocol) {
319 	case AF_INET:
320 		route_cb.ip_count++;
321 		break;
322 	case AF_INET6:
323 		route_cb.ip6_count++;
324 		break;
325 	case AF_IPX:
326 		route_cb.ipx_count++;
327 		break;
328 	}
329 	route_cb.any_count++;
330 	RTSOCK_UNLOCK();
331 	soisconnected(so);
332 	so->so_options |= SO_USELOOPBACK;
333 	return 0;
334 }
335 
336 static int
337 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
338 {
339 
340 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
341 }
342 
343 static int
344 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
345 {
346 
347 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
348 }
349 
350 /* pru_connect2 is EOPNOTSUPP */
351 /* pru_control is EOPNOTSUPP */
352 
353 static void
354 rts_detach(struct socket *so)
355 {
356 	struct rawcb *rp = sotorawcb(so);
357 
358 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
359 
360 	RTSOCK_LOCK();
361 	switch(rp->rcb_proto.sp_protocol) {
362 	case AF_INET:
363 		route_cb.ip_count--;
364 		break;
365 	case AF_INET6:
366 		route_cb.ip6_count--;
367 		break;
368 	case AF_IPX:
369 		route_cb.ipx_count--;
370 		break;
371 	}
372 	route_cb.any_count--;
373 	RTSOCK_UNLOCK();
374 	raw_usrreqs.pru_detach(so);
375 }
376 
377 static int
378 rts_disconnect(struct socket *so)
379 {
380 
381 	return (raw_usrreqs.pru_disconnect(so));
382 }
383 
384 /* pru_listen is EOPNOTSUPP */
385 
386 static int
387 rts_peeraddr(struct socket *so, struct sockaddr **nam)
388 {
389 
390 	return (raw_usrreqs.pru_peeraddr(so, nam));
391 }
392 
393 /* pru_rcvd is EOPNOTSUPP */
394 /* pru_rcvoob is EOPNOTSUPP */
395 
396 static int
397 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
398 	 struct mbuf *control, struct thread *td)
399 {
400 
401 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
402 }
403 
404 /* pru_sense is null */
405 
406 static int
407 rts_shutdown(struct socket *so)
408 {
409 
410 	return (raw_usrreqs.pru_shutdown(so));
411 }
412 
413 static int
414 rts_sockaddr(struct socket *so, struct sockaddr **nam)
415 {
416 
417 	return (raw_usrreqs.pru_sockaddr(so, nam));
418 }
419 
420 static struct pr_usrreqs route_usrreqs = {
421 	.pru_abort =		rts_abort,
422 	.pru_attach =		rts_attach,
423 	.pru_bind =		rts_bind,
424 	.pru_connect =		rts_connect,
425 	.pru_detach =		rts_detach,
426 	.pru_disconnect =	rts_disconnect,
427 	.pru_peeraddr =		rts_peeraddr,
428 	.pru_send =		rts_send,
429 	.pru_shutdown =		rts_shutdown,
430 	.pru_sockaddr =		rts_sockaddr,
431 	.pru_close =		rts_close,
432 };
433 
434 #ifndef _SOCKADDR_UNION_DEFINED
435 #define	_SOCKADDR_UNION_DEFINED
436 /*
437  * The union of all possible address formats we handle.
438  */
439 union sockaddr_union {
440 	struct sockaddr		sa;
441 	struct sockaddr_in	sin;
442 	struct sockaddr_in6	sin6;
443 };
444 #endif /* _SOCKADDR_UNION_DEFINED */
445 
446 static int
447 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
448     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
449 {
450 
451 	/* First, see if the returned address is part of the jail. */
452 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
453 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
454 		return (0);
455 	}
456 
457 	switch (info->rti_info[RTAX_DST]->sa_family) {
458 #ifdef INET
459 	case AF_INET:
460 	{
461 		struct in_addr ia;
462 		struct ifaddr *ifa;
463 		int found;
464 
465 		found = 0;
466 		/*
467 		 * Try to find an address on the given outgoing interface
468 		 * that belongs to the jail.
469 		 */
470 		IF_ADDR_RLOCK(ifp);
471 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
472 			struct sockaddr *sa;
473 			sa = ifa->ifa_addr;
474 			if (sa->sa_family != AF_INET)
475 				continue;
476 			ia = ((struct sockaddr_in *)sa)->sin_addr;
477 			if (prison_check_ip4(cred, &ia) == 0) {
478 				found = 1;
479 				break;
480 			}
481 		}
482 		IF_ADDR_RUNLOCK(ifp);
483 		if (!found) {
484 			/*
485 			 * As a last resort return the 'default' jail address.
486 			 */
487 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
488 			    sin_addr;
489 			if (prison_get_ip4(cred, &ia) != 0)
490 				return (ESRCH);
491 		}
492 		bzero(&saun->sin, sizeof(struct sockaddr_in));
493 		saun->sin.sin_len = sizeof(struct sockaddr_in);
494 		saun->sin.sin_family = AF_INET;
495 		saun->sin.sin_addr.s_addr = ia.s_addr;
496 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
497 		break;
498 	}
499 #endif
500 #ifdef INET6
501 	case AF_INET6:
502 	{
503 		struct in6_addr ia6;
504 		struct ifaddr *ifa;
505 		int found;
506 
507 		found = 0;
508 		/*
509 		 * Try to find an address on the given outgoing interface
510 		 * that belongs to the jail.
511 		 */
512 		IF_ADDR_RLOCK(ifp);
513 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
514 			struct sockaddr *sa;
515 			sa = ifa->ifa_addr;
516 			if (sa->sa_family != AF_INET6)
517 				continue;
518 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
519 			    &ia6, sizeof(struct in6_addr));
520 			if (prison_check_ip6(cred, &ia6) == 0) {
521 				found = 1;
522 				break;
523 			}
524 		}
525 		IF_ADDR_RUNLOCK(ifp);
526 		if (!found) {
527 			/*
528 			 * As a last resort return the 'default' jail address.
529 			 */
530 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
531 			    sin6_addr;
532 			if (prison_get_ip6(cred, &ia6) != 0)
533 				return (ESRCH);
534 		}
535 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
536 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
537 		saun->sin6.sin6_family = AF_INET6;
538 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
539 		if (sa6_recoverscope(&saun->sin6) != 0)
540 			return (ESRCH);
541 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
542 		break;
543 	}
544 #endif
545 	default:
546 		return (ESRCH);
547 	}
548 	return (0);
549 }
550 
551 /*ARGSUSED*/
552 static int
553 route_output(struct mbuf *m, struct socket *so)
554 {
555 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
556 	struct rt_msghdr *rtm = NULL;
557 	struct rtentry *rt = NULL;
558 	struct radix_node_head *rnh;
559 	struct rt_addrinfo info;
560 #ifdef INET6
561 	struct sockaddr_storage ss;
562 	struct sockaddr_in6 *sin6;
563 	int i, rti_need_deembed = 0;
564 #endif
565 	int len, error = 0;
566 	struct ifnet *ifp = NULL;
567 	union sockaddr_union saun;
568 	sa_family_t saf = AF_UNSPEC;
569 
570 #define senderr(e) { error = e; goto flush;}
571 	if (m == NULL || ((m->m_len < sizeof(long)) &&
572 		       (m = m_pullup(m, sizeof(long))) == NULL))
573 		return (ENOBUFS);
574 	if ((m->m_flags & M_PKTHDR) == 0)
575 		panic("route_output");
576 	len = m->m_pkthdr.len;
577 	if (len < sizeof(*rtm) ||
578 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
579 		info.rti_info[RTAX_DST] = NULL;
580 		senderr(EINVAL);
581 	}
582 	R_Malloc(rtm, struct rt_msghdr *, len);
583 	if (rtm == NULL) {
584 		info.rti_info[RTAX_DST] = NULL;
585 		senderr(ENOBUFS);
586 	}
587 	m_copydata(m, 0, len, (caddr_t)rtm);
588 	if (rtm->rtm_version != RTM_VERSION) {
589 		info.rti_info[RTAX_DST] = NULL;
590 		senderr(EPROTONOSUPPORT);
591 	}
592 	rtm->rtm_pid = curproc->p_pid;
593 	bzero(&info, sizeof(info));
594 	info.rti_addrs = rtm->rtm_addrs;
595 	/*
596 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
597 	 * link-local address because rtrequest requires addresses with
598 	 * embedded scope id.
599 	 */
600 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
601 		info.rti_info[RTAX_DST] = NULL;
602 		senderr(EINVAL);
603 	}
604 	info.rti_flags = rtm->rtm_flags;
605 	if (info.rti_info[RTAX_DST] == NULL ||
606 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
607 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
608 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
609 		senderr(EINVAL);
610 	saf = info.rti_info[RTAX_DST]->sa_family;
611 	/*
612 	 * Verify that the caller has the appropriate privilege; RTM_GET
613 	 * is the only operation the non-superuser is allowed.
614 	 */
615 	if (rtm->rtm_type != RTM_GET) {
616 		error = priv_check(curthread, PRIV_NET_ROUTE);
617 		if (error)
618 			senderr(error);
619 	}
620 
621 	/*
622 	 * The given gateway address may be an interface address.
623 	 * For example, issuing a "route change" command on a route
624 	 * entry that was created from a tunnel, and the gateway
625 	 * address given is the local end point. In this case the
626 	 * RTF_GATEWAY flag must be cleared or the destination will
627 	 * not be reachable even though there is no error message.
628 	 */
629 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
630 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
631 		struct route gw_ro;
632 
633 		bzero(&gw_ro, sizeof(gw_ro));
634 		gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
635 		rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
636 		/*
637 		 * A host route through the loopback interface is
638 		 * installed for each interface adddress. In pre 8.0
639 		 * releases the interface address of a PPP link type
640 		 * is not reachable locally. This behavior is fixed as
641 		 * part of the new L2/L3 redesign and rewrite work. The
642 		 * signature of this interface address route is the
643 		 * AF_LINK sa_family type of the rt_gateway, and the
644 		 * rt_ifp has the IFF_LOOPBACK flag set.
645 		 */
646 		if (gw_ro.ro_rt != NULL &&
647 		    gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
648 		    gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) {
649 			info.rti_flags &= ~RTF_GATEWAY;
650 			info.rti_flags |= RTF_GWFLAG_COMPAT;
651 		}
652 		if (gw_ro.ro_rt != NULL)
653 			RTFREE(gw_ro.ro_rt);
654 	}
655 
656 	switch (rtm->rtm_type) {
657 		struct rtentry *saved_nrt;
658 
659 	case RTM_ADD:
660 		if (info.rti_info[RTAX_GATEWAY] == NULL)
661 			senderr(EINVAL);
662 		saved_nrt = NULL;
663 
664 		/* support for new ARP code */
665 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
666 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
667 			error = lla_rt_output(rtm, &info);
668 #ifdef INET6
669 			if (error == 0)
670 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
671 #endif
672 			break;
673 		}
674 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
675 		    so->so_fibnum);
676 		if (error == 0 && saved_nrt) {
677 #ifdef INET6
678 			rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
679 #endif
680 			RT_LOCK(saved_nrt);
681 			rt_setmetrics(rtm->rtm_inits,
682 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
683 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
684 			RT_REMREF(saved_nrt);
685 			RT_UNLOCK(saved_nrt);
686 		}
687 		break;
688 
689 	case RTM_DELETE:
690 		saved_nrt = NULL;
691 		/* support for new ARP code */
692 		if (info.rti_info[RTAX_GATEWAY] &&
693 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
694 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
695 			error = lla_rt_output(rtm, &info);
696 #ifdef INET6
697 			if (error == 0)
698 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
699 #endif
700 			break;
701 		}
702 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
703 		    so->so_fibnum);
704 		if (error == 0) {
705 			RT_LOCK(saved_nrt);
706 			rt = saved_nrt;
707 			goto report;
708 		}
709 #ifdef INET6
710 		/* rt_msg2() will not be used when RTM_DELETE fails. */
711 		rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
712 #endif
713 		break;
714 
715 	case RTM_GET:
716 	case RTM_CHANGE:
717 	case RTM_LOCK:
718 		rnh = rt_tables_get_rnh(so->so_fibnum,
719 		    info.rti_info[RTAX_DST]->sa_family);
720 		if (rnh == NULL)
721 			senderr(EAFNOSUPPORT);
722 
723 		RADIX_NODE_HEAD_RLOCK(rnh);
724 
725 		if (info.rti_info[RTAX_NETMASK] == NULL &&
726 		    rtm->rtm_type == RTM_GET) {
727 			/*
728 			 * Provide logest prefix match for
729 			 * address lookup (no mask).
730 			 * 'route -n get addr'
731 			 */
732 			rt = (struct rtentry *) rnh->rnh_matchaddr(
733 			    info.rti_info[RTAX_DST], rnh);
734 		} else
735 			rt = (struct rtentry *) rnh->rnh_lookup(
736 			    info.rti_info[RTAX_DST],
737 			    info.rti_info[RTAX_NETMASK], rnh);
738 
739 		if (rt == NULL) {
740 			RADIX_NODE_HEAD_RUNLOCK(rnh);
741 			senderr(ESRCH);
742 		}
743 #ifdef RADIX_MPATH
744 		/*
745 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
746 		 * we require users to specify a matching RTAX_GATEWAY.
747 		 *
748 		 * for RTM_GET, gate is optional even with multipath.
749 		 * if gate == NULL the first match is returned.
750 		 * (no need to call rt_mpath_matchgate if gate == NULL)
751 		 */
752 		if (rn_mpath_capable(rnh) &&
753 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
754 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
755 			if (!rt) {
756 				RADIX_NODE_HEAD_RUNLOCK(rnh);
757 				senderr(ESRCH);
758 			}
759 		}
760 #endif
761 		/*
762 		 * If performing proxied L2 entry insertion, and
763 		 * the actual PPP host entry is found, perform
764 		 * another search to retrieve the prefix route of
765 		 * the local end point of the PPP link.
766 		 */
767 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
768 			struct sockaddr laddr;
769 
770 			if (rt->rt_ifp != NULL &&
771 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
772 				struct ifaddr *ifa;
773 
774 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
775 				if (ifa != NULL)
776 					rt_maskedcopy(ifa->ifa_addr,
777 						      &laddr,
778 						      ifa->ifa_netmask);
779 			} else
780 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
781 					      &laddr,
782 					      rt->rt_ifa->ifa_netmask);
783 			/*
784 			 * refactor rt and no lock operation necessary
785 			 */
786 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
787 			if (rt == NULL) {
788 				RADIX_NODE_HEAD_RUNLOCK(rnh);
789 				senderr(ESRCH);
790 			}
791 		}
792 		RT_LOCK(rt);
793 		RT_ADDREF(rt);
794 		RADIX_NODE_HEAD_RUNLOCK(rnh);
795 
796 		switch(rtm->rtm_type) {
797 
798 		case RTM_GET:
799 		report:
800 			RT_LOCK_ASSERT(rt);
801 			if ((rt->rt_flags & RTF_HOST) == 0
802 			    ? jailed_without_vnet(curthread->td_ucred)
803 			    : prison_if(curthread->td_ucred,
804 			    rt_key(rt)) != 0) {
805 				RT_UNLOCK(rt);
806 				senderr(ESRCH);
807 			}
808 			info.rti_info[RTAX_DST] = rt_key(rt);
809 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
810 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
811 			info.rti_info[RTAX_GENMASK] = 0;
812 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
813 				ifp = rt->rt_ifp;
814 				if (ifp) {
815 					info.rti_info[RTAX_IFP] =
816 					    ifp->if_addr->ifa_addr;
817 					error = rtm_get_jailed(&info, ifp, rt,
818 					    &saun, curthread->td_ucred);
819 					if (error != 0) {
820 						RT_UNLOCK(rt);
821 						senderr(error);
822 					}
823 					if (ifp->if_flags & IFF_POINTOPOINT)
824 						info.rti_info[RTAX_BRD] =
825 						    rt->rt_ifa->ifa_dstaddr;
826 					rtm->rtm_index = ifp->if_index;
827 				} else {
828 					info.rti_info[RTAX_IFP] = NULL;
829 					info.rti_info[RTAX_IFA] = NULL;
830 				}
831 			} else if ((ifp = rt->rt_ifp) != NULL) {
832 				rtm->rtm_index = ifp->if_index;
833 			}
834 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
835 			if (len > rtm->rtm_msglen) {
836 				struct rt_msghdr *new_rtm;
837 				R_Malloc(new_rtm, struct rt_msghdr *, len);
838 				if (new_rtm == NULL) {
839 					RT_UNLOCK(rt);
840 					senderr(ENOBUFS);
841 				}
842 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
843 				Free(rtm); rtm = new_rtm;
844 			}
845 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
846 			if (rt->rt_flags & RTF_GWFLAG_COMPAT)
847 				rtm->rtm_flags = RTF_GATEWAY |
848 					(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
849 			else
850 				rtm->rtm_flags = rt->rt_flags;
851 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
852 			rtm->rtm_addrs = info.rti_addrs;
853 			break;
854 
855 		case RTM_CHANGE:
856 			/*
857 			 * New gateway could require new ifaddr, ifp;
858 			 * flags may also be different; ifp may be specified
859 			 * by ll sockaddr when protocol address is ambiguous
860 			 */
861 			if (((rt->rt_flags & RTF_GATEWAY) &&
862 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
863 			    info.rti_info[RTAX_IFP] != NULL ||
864 			    (info.rti_info[RTAX_IFA] != NULL &&
865 			     !sa_equal(info.rti_info[RTAX_IFA],
866 				       rt->rt_ifa->ifa_addr))) {
867 				RT_UNLOCK(rt);
868 				RADIX_NODE_HEAD_LOCK(rnh);
869 				error = rt_getifa_fib(&info, rt->rt_fibnum);
870 				/*
871 				 * XXXRW: Really we should release this
872 				 * reference later, but this maintains
873 				 * historical behavior.
874 				 */
875 				if (info.rti_ifa != NULL)
876 					ifa_free(info.rti_ifa);
877 				RADIX_NODE_HEAD_UNLOCK(rnh);
878 				if (error != 0)
879 					senderr(error);
880 				RT_LOCK(rt);
881 			}
882 			if (info.rti_ifa != NULL &&
883 			    info.rti_ifa != rt->rt_ifa &&
884 			    rt->rt_ifa != NULL &&
885 			    rt->rt_ifa->ifa_rtrequest != NULL) {
886 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
887 				    &info);
888 				ifa_free(rt->rt_ifa);
889 			}
890 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
891 				RT_UNLOCK(rt);
892 				RADIX_NODE_HEAD_LOCK(rnh);
893 				RT_LOCK(rt);
894 
895 				error = rt_setgate(rt, rt_key(rt),
896 				    info.rti_info[RTAX_GATEWAY]);
897 				RADIX_NODE_HEAD_UNLOCK(rnh);
898 				if (error != 0) {
899 					RT_UNLOCK(rt);
900 					senderr(error);
901 				}
902 				rt->rt_flags &= ~RTF_GATEWAY;
903 				rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
904 			}
905 			if (info.rti_ifa != NULL &&
906 			    info.rti_ifa != rt->rt_ifa) {
907 				ifa_ref(info.rti_ifa);
908 				rt->rt_ifa = info.rti_ifa;
909 				rt->rt_ifp = info.rti_ifp;
910 			}
911 			/* Allow some flags to be toggled on change. */
912 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
913 				    (rtm->rtm_flags & RTF_FMASK);
914 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
915 					&rt->rt_rmx);
916 			rtm->rtm_index = rt->rt_ifp->if_index;
917 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
918 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
919 			/* FALLTHROUGH */
920 		case RTM_LOCK:
921 			/* We don't support locks anymore */
922 			break;
923 		}
924 		RT_UNLOCK(rt);
925 		break;
926 
927 	default:
928 		senderr(EOPNOTSUPP);
929 	}
930 
931 flush:
932 	if (rtm) {
933 		if (error)
934 			rtm->rtm_errno = error;
935 		else
936 			rtm->rtm_flags |= RTF_DONE;
937 	}
938 	if (rt)		/* XXX can this be true? */
939 		RTFREE(rt);
940     {
941 	struct rawcb *rp = NULL;
942 	/*
943 	 * Check to see if we don't want our own messages.
944 	 */
945 	if ((so->so_options & SO_USELOOPBACK) == 0) {
946 		if (route_cb.any_count <= 1) {
947 			if (rtm)
948 				Free(rtm);
949 			m_freem(m);
950 			return (error);
951 		}
952 		/* There is another listener, so construct message */
953 		rp = sotorawcb(so);
954 	}
955 	if (rtm) {
956 #ifdef INET6
957 		if (rti_need_deembed) {
958 			/* sin6_scope_id is recovered before sending rtm. */
959 			sin6 = (struct sockaddr_in6 *)&ss;
960 			for (i = 0; i < RTAX_MAX; i++) {
961 				if (info.rti_info[i] == NULL)
962 					continue;
963 				if (info.rti_info[i]->sa_family != AF_INET6)
964 					continue;
965 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
966 				if (sa6_recoverscope(sin6) == 0)
967 					bcopy(sin6, info.rti_info[i],
968 						    sizeof(*sin6));
969 			}
970 		}
971 #endif
972 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
973 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
974 			m_freem(m);
975 			m = NULL;
976 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
977 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
978 	}
979 	if (m) {
980 		M_SETFIB(m, so->so_fibnum);
981 		m->m_flags |= RTS_FILTER_FIB;
982 		if (rp) {
983 			/*
984 			 * XXX insure we don't get a copy by
985 			 * invalidating our protocol
986 			 */
987 			unsigned short family = rp->rcb_proto.sp_family;
988 			rp->rcb_proto.sp_family = 0;
989 			rt_dispatch(m, saf);
990 			rp->rcb_proto.sp_family = family;
991 		} else
992 			rt_dispatch(m, saf);
993 	}
994 	/* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
995 	if (rtm)
996 		Free(rtm);
997     }
998 	return (error);
999 #undef	sa_equal
1000 }
1001 
1002 static void
1003 rt_setmetrics(u_long which, const struct rt_metrics *in,
1004 	struct rt_metrics_lite *out)
1005 {
1006 #define metric(f, e) if (which & (f)) out->e = in->e;
1007 	/*
1008 	 * Only these are stored in the routing entry since introduction
1009 	 * of tcp hostcache. The rest is ignored.
1010 	 */
1011 	metric(RTV_MTU, rmx_mtu);
1012 	metric(RTV_WEIGHT, rmx_weight);
1013 	/* Userland -> kernel timebase conversion. */
1014 	if (which & RTV_EXPIRE)
1015 		out->rmx_expire = in->rmx_expire ?
1016 		    in->rmx_expire - time_second + time_uptime : 0;
1017 #undef metric
1018 }
1019 
1020 static void
1021 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
1022 {
1023 #define metric(e) out->e = in->e;
1024 	bzero(out, sizeof(*out));
1025 	metric(rmx_mtu);
1026 	metric(rmx_weight);
1027 	/* Kernel -> userland timebase conversion. */
1028 	out->rmx_expire = in->rmx_expire ?
1029 	    in->rmx_expire - time_uptime + time_second : 0;
1030 #undef metric
1031 }
1032 
1033 /*
1034  * Extract the addresses of the passed sockaddrs.
1035  * Do a little sanity checking so as to avoid bad memory references.
1036  * This data is derived straight from userland.
1037  */
1038 static int
1039 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1040 {
1041 	struct sockaddr *sa;
1042 	int i;
1043 
1044 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1045 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1046 			continue;
1047 		sa = (struct sockaddr *)cp;
1048 		/*
1049 		 * It won't fit.
1050 		 */
1051 		if (cp + sa->sa_len > cplim)
1052 			return (EINVAL);
1053 		/*
1054 		 * there are no more.. quit now
1055 		 * If there are more bits, they are in error.
1056 		 * I've seen this. route(1) can evidently generate these.
1057 		 * This causes kernel to core dump.
1058 		 * for compatibility, If we see this, point to a safe address.
1059 		 */
1060 		if (sa->sa_len == 0) {
1061 			rtinfo->rti_info[i] = &sa_zero;
1062 			return (0); /* should be EINVAL but for compat */
1063 		}
1064 		/* accept it */
1065 #ifdef INET6
1066 		if (sa->sa_family == AF_INET6)
1067 			sa6_embedscope((struct sockaddr_in6 *)sa,
1068 			    V_ip6_use_defzone);
1069 #endif
1070 		rtinfo->rti_info[i] = sa;
1071 		cp += SA_SIZE(sa);
1072 	}
1073 	return (0);
1074 }
1075 
1076 /*
1077  * Used by the routing socket.
1078  */
1079 static struct mbuf *
1080 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1081 {
1082 	struct rt_msghdr *rtm;
1083 	struct mbuf *m;
1084 	int i;
1085 	struct sockaddr *sa;
1086 #ifdef INET6
1087 	struct sockaddr_storage ss;
1088 	struct sockaddr_in6 *sin6;
1089 #endif
1090 	int len, dlen;
1091 
1092 	switch (type) {
1093 
1094 	case RTM_DELADDR:
1095 	case RTM_NEWADDR:
1096 		len = sizeof(struct ifa_msghdr);
1097 		break;
1098 
1099 	case RTM_DELMADDR:
1100 	case RTM_NEWMADDR:
1101 		len = sizeof(struct ifma_msghdr);
1102 		break;
1103 
1104 	case RTM_IFINFO:
1105 		len = sizeof(struct if_msghdr);
1106 		break;
1107 
1108 	case RTM_IFANNOUNCE:
1109 	case RTM_IEEE80211:
1110 		len = sizeof(struct if_announcemsghdr);
1111 		break;
1112 
1113 	default:
1114 		len = sizeof(struct rt_msghdr);
1115 	}
1116 
1117 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1118 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1119 	if (len > MHLEN)
1120 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1121 	else
1122 		m = m_gethdr(M_NOWAIT, MT_DATA);
1123 	if (m == NULL)
1124 		return (m);
1125 
1126 	m->m_pkthdr.len = m->m_len = len;
1127 	rtm = mtod(m, struct rt_msghdr *);
1128 	bzero((caddr_t)rtm, len);
1129 	for (i = 0; i < RTAX_MAX; i++) {
1130 		if ((sa = rtinfo->rti_info[i]) == NULL)
1131 			continue;
1132 		rtinfo->rti_addrs |= (1 << i);
1133 		dlen = SA_SIZE(sa);
1134 #ifdef INET6
1135 		if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1136 			sin6 = (struct sockaddr_in6 *)&ss;
1137 			bcopy(sa, sin6, sizeof(*sin6));
1138 			if (sa6_recoverscope(sin6) == 0)
1139 				sa = (struct sockaddr *)sin6;
1140 		}
1141 #endif
1142 		m_copyback(m, len, dlen, (caddr_t)sa);
1143 		len += dlen;
1144 	}
1145 	if (m->m_pkthdr.len != len) {
1146 		m_freem(m);
1147 		return (NULL);
1148 	}
1149 	rtm->rtm_msglen = len;
1150 	rtm->rtm_version = RTM_VERSION;
1151 	rtm->rtm_type = type;
1152 	return (m);
1153 }
1154 
1155 /*
1156  * Used by the sysctl code and routing socket.
1157  */
1158 static int
1159 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1160 {
1161 	int i;
1162 	int len, dlen, second_time = 0;
1163 	caddr_t cp0;
1164 #ifdef INET6
1165 	struct sockaddr_storage ss;
1166 	struct sockaddr_in6 *sin6;
1167 #endif
1168 
1169 	rtinfo->rti_addrs = 0;
1170 again:
1171 	switch (type) {
1172 
1173 	case RTM_DELADDR:
1174 	case RTM_NEWADDR:
1175 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1176 #ifdef COMPAT_FREEBSD32
1177 			if (w->w_req->flags & SCTL_MASK32)
1178 				len = sizeof(struct ifa_msghdrl32);
1179 			else
1180 #endif
1181 				len = sizeof(struct ifa_msghdrl);
1182 		} else
1183 			len = sizeof(struct ifa_msghdr);
1184 		break;
1185 
1186 	case RTM_IFINFO:
1187 #ifdef COMPAT_FREEBSD32
1188 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1189 			if (w->w_op == NET_RT_IFLISTL)
1190 				len = sizeof(struct if_msghdrl32);
1191 			else
1192 				len = sizeof(struct if_msghdr32);
1193 			break;
1194 		}
1195 #endif
1196 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1197 			len = sizeof(struct if_msghdrl);
1198 		else
1199 			len = sizeof(struct if_msghdr);
1200 		break;
1201 
1202 	case RTM_NEWMADDR:
1203 		len = sizeof(struct ifma_msghdr);
1204 		break;
1205 
1206 	default:
1207 		len = sizeof(struct rt_msghdr);
1208 	}
1209 	cp0 = cp;
1210 	if (cp0)
1211 		cp += len;
1212 	for (i = 0; i < RTAX_MAX; i++) {
1213 		struct sockaddr *sa;
1214 
1215 		if ((sa = rtinfo->rti_info[i]) == NULL)
1216 			continue;
1217 		rtinfo->rti_addrs |= (1 << i);
1218 		dlen = SA_SIZE(sa);
1219 		if (cp) {
1220 #ifdef INET6
1221 			if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1222 				sin6 = (struct sockaddr_in6 *)&ss;
1223 				bcopy(sa, sin6, sizeof(*sin6));
1224 				if (sa6_recoverscope(sin6) == 0)
1225 					sa = (struct sockaddr *)sin6;
1226 			}
1227 #endif
1228 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1229 			cp += dlen;
1230 		}
1231 		len += dlen;
1232 	}
1233 	len = ALIGN(len);
1234 	if (cp == NULL && w != NULL && !second_time) {
1235 		struct walkarg *rw = w;
1236 
1237 		if (rw->w_req) {
1238 			if (rw->w_tmemsize < len) {
1239 				if (rw->w_tmem)
1240 					free(rw->w_tmem, M_RTABLE);
1241 				rw->w_tmem = (caddr_t)
1242 					malloc(len, M_RTABLE, M_NOWAIT);
1243 				if (rw->w_tmem)
1244 					rw->w_tmemsize = len;
1245 			}
1246 			if (rw->w_tmem) {
1247 				cp = rw->w_tmem;
1248 				second_time = 1;
1249 				goto again;
1250 			}
1251 		}
1252 	}
1253 	if (cp) {
1254 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1255 
1256 		rtm->rtm_version = RTM_VERSION;
1257 		rtm->rtm_type = type;
1258 		rtm->rtm_msglen = len;
1259 	}
1260 	return (len);
1261 }
1262 
1263 /*
1264  * This routine is called to generate a message from the routing
1265  * socket indicating that a redirect has occured, a routing lookup
1266  * has failed, or that a protocol has detected timeouts to a particular
1267  * destination.
1268  */
1269 void
1270 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1271     int fibnum)
1272 {
1273 	struct rt_msghdr *rtm;
1274 	struct mbuf *m;
1275 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1276 
1277 	if (route_cb.any_count == 0)
1278 		return;
1279 	m = rt_msg1(type, rtinfo);
1280 	if (m == NULL)
1281 		return;
1282 
1283 	if (fibnum != RT_ALL_FIBS) {
1284 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1285 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1286 		M_SETFIB(m, fibnum);
1287 		m->m_flags |= RTS_FILTER_FIB;
1288 	}
1289 
1290 	rtm = mtod(m, struct rt_msghdr *);
1291 	rtm->rtm_flags = RTF_DONE | flags;
1292 	rtm->rtm_errno = error;
1293 	rtm->rtm_addrs = rtinfo->rti_addrs;
1294 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1295 }
1296 
1297 void
1298 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1299 {
1300 
1301 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1302 }
1303 
1304 /*
1305  * This routine is called to generate a message from the routing
1306  * socket indicating that the status of a network interface has changed.
1307  */
1308 void
1309 rt_ifmsg(struct ifnet *ifp)
1310 {
1311 	struct if_msghdr *ifm;
1312 	struct mbuf *m;
1313 	struct rt_addrinfo info;
1314 
1315 	if (route_cb.any_count == 0)
1316 		return;
1317 	bzero((caddr_t)&info, sizeof(info));
1318 	m = rt_msg1(RTM_IFINFO, &info);
1319 	if (m == NULL)
1320 		return;
1321 	ifm = mtod(m, struct if_msghdr *);
1322 	ifm->ifm_index = ifp->if_index;
1323 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1324 	ifm->ifm_data = ifp->if_data;
1325 	ifm->ifm_addrs = 0;
1326 	rt_dispatch(m, AF_UNSPEC);
1327 }
1328 
1329 /*
1330  * Announce interface address arrival/withdraw.
1331  * Please do not call directly, use rt_addrmsg().
1332  * Assume input data to be valid.
1333  * Returns 0 on success.
1334  */
1335 int
1336 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1337 {
1338 	struct rt_addrinfo info;
1339 	struct sockaddr *sa;
1340 	int ncmd;
1341 	struct mbuf *m;
1342 	struct ifa_msghdr *ifam;
1343 	struct ifnet *ifp = ifa->ifa_ifp;
1344 
1345 	if (route_cb.any_count == 0)
1346 		return (0);
1347 
1348 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1349 
1350 	bzero((caddr_t)&info, sizeof(info));
1351 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1352 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1353 	info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1354 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1355 	if ((m = rt_msg1(ncmd, &info)) == NULL)
1356 		return (ENOBUFS);
1357 	ifam = mtod(m, struct ifa_msghdr *);
1358 	ifam->ifam_index = ifp->if_index;
1359 	ifam->ifam_metric = ifa->ifa_metric;
1360 	ifam->ifam_flags = ifa->ifa_flags;
1361 	ifam->ifam_addrs = info.rti_addrs;
1362 
1363 	if (fibnum != RT_ALL_FIBS) {
1364 		M_SETFIB(m, fibnum);
1365 		m->m_flags |= RTS_FILTER_FIB;
1366 	}
1367 
1368 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1369 
1370 	return (0);
1371 }
1372 
1373 /*
1374  * Announce route addition/removal.
1375  * Please do not call directly, use rt_routemsg().
1376  * Note that @rt data MAY be inconsistent/invalid:
1377  * if some userland app sends us "invalid" route message (invalid mask,
1378  * no dst, wrong address families, etc...) we need to pass it back
1379  * to app (and any other rtsock consumers) with rtm_errno field set to
1380  * non-zero value.
1381  *
1382  * Returns 0 on success.
1383  */
1384 int
1385 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1386     int fibnum)
1387 {
1388 	struct rt_addrinfo info;
1389 	struct sockaddr *sa;
1390 	struct mbuf *m;
1391 	struct rt_msghdr *rtm;
1392 
1393 	if (route_cb.any_count == 0)
1394 		return (0);
1395 
1396 	bzero((caddr_t)&info, sizeof(info));
1397 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1398 	info.rti_info[RTAX_DST] = sa = rt_key(rt);
1399 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1400 	if ((m = rt_msg1(cmd, &info)) == NULL)
1401 		return (ENOBUFS);
1402 	rtm = mtod(m, struct rt_msghdr *);
1403 	rtm->rtm_index = ifp->if_index;
1404 	rtm->rtm_flags |= rt->rt_flags;
1405 	rtm->rtm_errno = error;
1406 	rtm->rtm_addrs = info.rti_addrs;
1407 
1408 	if (fibnum != RT_ALL_FIBS) {
1409 		M_SETFIB(m, fibnum);
1410 		m->m_flags |= RTS_FILTER_FIB;
1411 	}
1412 
1413 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1414 
1415 	return (0);
1416 }
1417 
1418 /*
1419  * This is the analogue to the rt_newaddrmsg which performs the same
1420  * function but for multicast group memberhips.  This is easier since
1421  * there is no route state to worry about.
1422  */
1423 void
1424 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1425 {
1426 	struct rt_addrinfo info;
1427 	struct mbuf *m = NULL;
1428 	struct ifnet *ifp = ifma->ifma_ifp;
1429 	struct ifma_msghdr *ifmam;
1430 
1431 	if (route_cb.any_count == 0)
1432 		return;
1433 
1434 	bzero((caddr_t)&info, sizeof(info));
1435 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1436 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1437 	/*
1438 	 * If a link-layer address is present, present it as a ``gateway''
1439 	 * (similarly to how ARP entries, e.g., are presented).
1440 	 */
1441 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1442 	m = rt_msg1(cmd, &info);
1443 	if (m == NULL)
1444 		return;
1445 	ifmam = mtod(m, struct ifma_msghdr *);
1446 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1447 	    __func__));
1448 	ifmam->ifmam_index = ifp->if_index;
1449 	ifmam->ifmam_addrs = info.rti_addrs;
1450 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1451 }
1452 
1453 static struct mbuf *
1454 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1455 	struct rt_addrinfo *info)
1456 {
1457 	struct if_announcemsghdr *ifan;
1458 	struct mbuf *m;
1459 
1460 	if (route_cb.any_count == 0)
1461 		return NULL;
1462 	bzero((caddr_t)info, sizeof(*info));
1463 	m = rt_msg1(type, info);
1464 	if (m != NULL) {
1465 		ifan = mtod(m, struct if_announcemsghdr *);
1466 		ifan->ifan_index = ifp->if_index;
1467 		strlcpy(ifan->ifan_name, ifp->if_xname,
1468 			sizeof(ifan->ifan_name));
1469 		ifan->ifan_what = what;
1470 	}
1471 	return m;
1472 }
1473 
1474 /*
1475  * This is called to generate routing socket messages indicating
1476  * IEEE80211 wireless events.
1477  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1478  */
1479 void
1480 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1481 {
1482 	struct mbuf *m;
1483 	struct rt_addrinfo info;
1484 
1485 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1486 	if (m != NULL) {
1487 		/*
1488 		 * Append the ieee80211 data.  Try to stick it in the
1489 		 * mbuf containing the ifannounce msg; otherwise allocate
1490 		 * a new mbuf and append.
1491 		 *
1492 		 * NB: we assume m is a single mbuf.
1493 		 */
1494 		if (data_len > M_TRAILINGSPACE(m)) {
1495 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1496 			if (n == NULL) {
1497 				m_freem(m);
1498 				return;
1499 			}
1500 			bcopy(data, mtod(n, void *), data_len);
1501 			n->m_len = data_len;
1502 			m->m_next = n;
1503 		} else if (data_len > 0) {
1504 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1505 			m->m_len += data_len;
1506 		}
1507 		if (m->m_flags & M_PKTHDR)
1508 			m->m_pkthdr.len += data_len;
1509 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1510 		rt_dispatch(m, AF_UNSPEC);
1511 	}
1512 }
1513 
1514 /*
1515  * This is called to generate routing socket messages indicating
1516  * network interface arrival and departure.
1517  */
1518 void
1519 rt_ifannouncemsg(struct ifnet *ifp, int what)
1520 {
1521 	struct mbuf *m;
1522 	struct rt_addrinfo info;
1523 
1524 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1525 	if (m != NULL)
1526 		rt_dispatch(m, AF_UNSPEC);
1527 }
1528 
1529 static void
1530 rt_dispatch(struct mbuf *m, sa_family_t saf)
1531 {
1532 	struct m_tag *tag;
1533 
1534 	/*
1535 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1536 	 * use when injecting the mbuf into the routing socket buffer from
1537 	 * the netisr.
1538 	 */
1539 	if (saf != AF_UNSPEC) {
1540 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1541 		    M_NOWAIT);
1542 		if (tag == NULL) {
1543 			m_freem(m);
1544 			return;
1545 		}
1546 		*(unsigned short *)(tag + 1) = saf;
1547 		m_tag_prepend(m, tag);
1548 	}
1549 #ifdef VIMAGE
1550 	if (V_loif)
1551 		m->m_pkthdr.rcvif = V_loif;
1552 	else {
1553 		m_freem(m);
1554 		return;
1555 	}
1556 #endif
1557 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1558 }
1559 
1560 /*
1561  * This is used in dumping the kernel table via sysctl().
1562  */
1563 static int
1564 sysctl_dumpentry(struct radix_node *rn, void *vw)
1565 {
1566 	struct walkarg *w = vw;
1567 	struct rtentry *rt = (struct rtentry *)rn;
1568 	int error = 0, size;
1569 	struct rt_addrinfo info;
1570 
1571 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1572 		return 0;
1573 	if ((rt->rt_flags & RTF_HOST) == 0
1574 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1575 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1576 		return (0);
1577 	bzero((caddr_t)&info, sizeof(info));
1578 	info.rti_info[RTAX_DST] = rt_key(rt);
1579 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1580 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1581 	info.rti_info[RTAX_GENMASK] = 0;
1582 	if (rt->rt_ifp) {
1583 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1584 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1585 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1586 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1587 	}
1588 	size = rt_msg2(RTM_GET, &info, NULL, w);
1589 	if (w->w_req && w->w_tmem) {
1590 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1591 
1592 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1593 			rtm->rtm_flags = RTF_GATEWAY |
1594 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1595 		else
1596 			rtm->rtm_flags = rt->rt_flags;
1597 		/*
1598 		 * let's be honest about this being a retarded hack
1599 		 */
1600 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1601 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1602 		rtm->rtm_index = rt->rt_ifp->if_index;
1603 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1604 		rtm->rtm_addrs = info.rti_addrs;
1605 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1606 		return (error);
1607 	}
1608 	return (error);
1609 }
1610 
1611 #ifdef COMPAT_FREEBSD32
1612 static void
1613 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1614 {
1615 
1616 	bzero(dst, sizeof(*dst));
1617 	CP(*src, *dst, ifi_type);
1618 	CP(*src, *dst, ifi_physical);
1619 	CP(*src, *dst, ifi_addrlen);
1620 	CP(*src, *dst, ifi_hdrlen);
1621 	CP(*src, *dst, ifi_link_state);
1622 	CP(*src, *dst, ifi_vhid);
1623 	CP(*src, *dst, ifi_baudrate_pf);
1624 	dst->ifi_datalen = sizeof(struct if_data32);
1625 	CP(*src, *dst, ifi_mtu);
1626 	CP(*src, *dst, ifi_metric);
1627 	CP(*src, *dst, ifi_baudrate);
1628 	CP(*src, *dst, ifi_ipackets);
1629 	CP(*src, *dst, ifi_ierrors);
1630 	CP(*src, *dst, ifi_opackets);
1631 	CP(*src, *dst, ifi_oerrors);
1632 	CP(*src, *dst, ifi_collisions);
1633 	CP(*src, *dst, ifi_ibytes);
1634 	CP(*src, *dst, ifi_obytes);
1635 	CP(*src, *dst, ifi_imcasts);
1636 	CP(*src, *dst, ifi_omcasts);
1637 	CP(*src, *dst, ifi_iqdrops);
1638 	CP(*src, *dst, ifi_noproto);
1639 	CP(*src, *dst, ifi_hwassist);
1640 	CP(*src, *dst, ifi_epoch);
1641 	TV_CP(*src, *dst, ifi_lastchange);
1642 }
1643 #endif
1644 
1645 static int
1646 sysctl_iflist_ifml(struct ifnet *ifp, struct rt_addrinfo *info,
1647     struct walkarg *w, int len)
1648 {
1649 	struct if_msghdrl *ifm;
1650 
1651 #ifdef COMPAT_FREEBSD32
1652 	if (w->w_req->flags & SCTL_MASK32) {
1653 		struct if_msghdrl32 *ifm32;
1654 
1655 		ifm32 = (struct if_msghdrl32 *)w->w_tmem;
1656 		ifm32->ifm_addrs = info->rti_addrs;
1657 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1658 		ifm32->ifm_index = ifp->if_index;
1659 		ifm32->_ifm_spare1 = 0;
1660 		ifm32->ifm_len = sizeof(*ifm32);
1661 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1662 
1663 		copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1664 		/* Fixup if_data carp(4) vhid. */
1665 		if (carp_get_vhid_p != NULL)
1666 			ifm32->ifm_data.ifi_vhid =
1667 			    (*carp_get_vhid_p)(ifp->if_addr);
1668 
1669 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1670 	}
1671 #endif
1672 	ifm = (struct if_msghdrl *)w->w_tmem;
1673 	ifm->ifm_addrs = info->rti_addrs;
1674 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1675 	ifm->ifm_index = ifp->if_index;
1676 	ifm->_ifm_spare1 = 0;
1677 	ifm->ifm_len = sizeof(*ifm);
1678 	ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1679 
1680 	ifm->ifm_data = ifp->if_data;
1681 	/* Fixup if_data carp(4) vhid. */
1682 	if (carp_get_vhid_p != NULL)
1683 		ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1684 
1685 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1686 }
1687 
1688 static int
1689 sysctl_iflist_ifm(struct ifnet *ifp, struct rt_addrinfo *info,
1690     struct walkarg *w, int len)
1691 {
1692 	struct if_msghdr *ifm;
1693 
1694 #ifdef COMPAT_FREEBSD32
1695 	if (w->w_req->flags & SCTL_MASK32) {
1696 		struct if_msghdr32 *ifm32;
1697 
1698 		ifm32 = (struct if_msghdr32 *)w->w_tmem;
1699 		ifm32->ifm_addrs = info->rti_addrs;
1700 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1701 		ifm32->ifm_index = ifp->if_index;
1702 
1703 		copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1704 		/* Fixup if_data carp(4) vhid. */
1705 		if (carp_get_vhid_p != NULL)
1706 			ifm32->ifm_data.ifi_vhid =
1707 			    (*carp_get_vhid_p)(ifp->if_addr);
1708 
1709 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1710 	}
1711 #endif
1712 	ifm = (struct if_msghdr *)w->w_tmem;
1713 	ifm->ifm_addrs = info->rti_addrs;
1714 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1715 	ifm->ifm_index = ifp->if_index;
1716 
1717 	ifm->ifm_data = ifp->if_data;
1718 	/* Fixup if_data carp(4) vhid. */
1719 	if (carp_get_vhid_p != NULL)
1720 		ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1721 
1722 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1723 }
1724 
1725 static int
1726 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1727     struct walkarg *w, int len)
1728 {
1729 	struct ifa_msghdrl *ifam;
1730 
1731 #ifdef COMPAT_FREEBSD32
1732 	if (w->w_req->flags & SCTL_MASK32) {
1733 		struct ifa_msghdrl32 *ifam32;
1734 
1735 		ifam32 = (struct ifa_msghdrl32 *)w->w_tmem;
1736 		ifam32->ifam_addrs = info->rti_addrs;
1737 		ifam32->ifam_flags = ifa->ifa_flags;
1738 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
1739 		ifam32->_ifam_spare1 = 0;
1740 		ifam32->ifam_len = sizeof(*ifam32);
1741 		ifam32->ifam_data_off =
1742 		    offsetof(struct ifa_msghdrl32, ifam_data);
1743 		ifam32->ifam_metric = ifa->ifa_metric;
1744 
1745 		bzero(&ifam32->ifam_data, sizeof(ifam32->ifam_data));
1746 		ifam32->ifam_data.ifi_datalen = sizeof(struct if_data32);
1747 		ifam32->ifam_data.ifi_ipackets =
1748 		    counter_u64_fetch(ifa->ifa_ipackets);
1749 		ifam32->ifam_data.ifi_opackets =
1750 		    counter_u64_fetch(ifa->ifa_opackets);
1751 		ifam32->ifam_data.ifi_ibytes =
1752 		    counter_u64_fetch(ifa->ifa_ibytes);
1753 		ifam32->ifam_data.ifi_obytes =
1754 		    counter_u64_fetch(ifa->ifa_obytes);
1755 
1756 		/* Fixup if_data carp(4) vhid. */
1757 		if (carp_get_vhid_p != NULL)
1758 			ifam32->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1759 
1760 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifam32, len));
1761 	}
1762 #endif
1763 
1764 	ifam = (struct ifa_msghdrl *)w->w_tmem;
1765 	ifam->ifam_addrs = info->rti_addrs;
1766 	ifam->ifam_flags = ifa->ifa_flags;
1767 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1768 	ifam->_ifam_spare1 = 0;
1769 	ifam->ifam_len = sizeof(*ifam);
1770 	ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1771 	ifam->ifam_metric = ifa->ifa_metric;
1772 
1773 	bzero(&ifam->ifam_data, sizeof(ifam->ifam_data));
1774 	ifam->ifam_data.ifi_datalen = sizeof(struct if_data);
1775 	ifam->ifam_data.ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1776 	ifam->ifam_data.ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1777 	ifam->ifam_data.ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1778 	ifam->ifam_data.ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1779 
1780 	/* Fixup if_data carp(4) vhid. */
1781 	if (carp_get_vhid_p != NULL)
1782 		ifam->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1783 
1784 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1785 }
1786 
1787 static int
1788 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1789     struct walkarg *w, int len)
1790 {
1791 	struct ifa_msghdr *ifam;
1792 
1793 	ifam = (struct ifa_msghdr *)w->w_tmem;
1794 	ifam->ifam_addrs = info->rti_addrs;
1795 	ifam->ifam_flags = ifa->ifa_flags;
1796 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1797 	ifam->ifam_metric = ifa->ifa_metric;
1798 
1799 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1800 }
1801 
1802 static int
1803 sysctl_iflist(int af, struct walkarg *w)
1804 {
1805 	struct ifnet *ifp;
1806 	struct ifaddr *ifa;
1807 	struct rt_addrinfo info;
1808 	int len, error = 0;
1809 
1810 	bzero((caddr_t)&info, sizeof(info));
1811 	IFNET_RLOCK_NOSLEEP();
1812 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1813 		if (w->w_arg && w->w_arg != ifp->if_index)
1814 			continue;
1815 		IF_ADDR_RLOCK(ifp);
1816 		ifa = ifp->if_addr;
1817 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1818 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1819 		info.rti_info[RTAX_IFP] = NULL;
1820 		if (w->w_req && w->w_tmem) {
1821 			if (w->w_op == NET_RT_IFLISTL)
1822 				error = sysctl_iflist_ifml(ifp, &info, w, len);
1823 			else
1824 				error = sysctl_iflist_ifm(ifp, &info, w, len);
1825 			if (error)
1826 				goto done;
1827 		}
1828 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1829 			if (af && af != ifa->ifa_addr->sa_family)
1830 				continue;
1831 			if (prison_if(w->w_req->td->td_ucred,
1832 			    ifa->ifa_addr) != 0)
1833 				continue;
1834 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1835 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1836 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1837 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1838 			if (w->w_req && w->w_tmem) {
1839 				if (w->w_op == NET_RT_IFLISTL)
1840 					error = sysctl_iflist_ifaml(ifa, &info,
1841 					    w, len);
1842 				else
1843 					error = sysctl_iflist_ifam(ifa, &info,
1844 					    w, len);
1845 				if (error)
1846 					goto done;
1847 			}
1848 		}
1849 		IF_ADDR_RUNLOCK(ifp);
1850 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1851 			info.rti_info[RTAX_BRD] = NULL;
1852 	}
1853 done:
1854 	if (ifp != NULL)
1855 		IF_ADDR_RUNLOCK(ifp);
1856 	IFNET_RUNLOCK_NOSLEEP();
1857 	return (error);
1858 }
1859 
1860 static int
1861 sysctl_ifmalist(int af, struct walkarg *w)
1862 {
1863 	struct ifnet *ifp;
1864 	struct ifmultiaddr *ifma;
1865 	struct	rt_addrinfo info;
1866 	int	len, error = 0;
1867 	struct ifaddr *ifa;
1868 
1869 	bzero((caddr_t)&info, sizeof(info));
1870 	IFNET_RLOCK_NOSLEEP();
1871 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1872 		if (w->w_arg && w->w_arg != ifp->if_index)
1873 			continue;
1874 		ifa = ifp->if_addr;
1875 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1876 		IF_ADDR_RLOCK(ifp);
1877 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1878 			if (af && af != ifma->ifma_addr->sa_family)
1879 				continue;
1880 			if (prison_if(w->w_req->td->td_ucred,
1881 			    ifma->ifma_addr) != 0)
1882 				continue;
1883 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1884 			info.rti_info[RTAX_GATEWAY] =
1885 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1886 			    ifma->ifma_lladdr : NULL;
1887 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1888 			if (w->w_req && w->w_tmem) {
1889 				struct ifma_msghdr *ifmam;
1890 
1891 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1892 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1893 				ifmam->ifmam_flags = 0;
1894 				ifmam->ifmam_addrs = info.rti_addrs;
1895 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1896 				if (error) {
1897 					IF_ADDR_RUNLOCK(ifp);
1898 					goto done;
1899 				}
1900 			}
1901 		}
1902 		IF_ADDR_RUNLOCK(ifp);
1903 	}
1904 done:
1905 	IFNET_RUNLOCK_NOSLEEP();
1906 	return (error);
1907 }
1908 
1909 static int
1910 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1911 {
1912 	int	*name = (int *)arg1;
1913 	u_int	namelen = arg2;
1914 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1915 	int	i, lim, error = EINVAL;
1916 	int	fib = 0;
1917 	u_char	af;
1918 	struct	walkarg w;
1919 
1920 	name ++;
1921 	namelen--;
1922 	if (req->newptr)
1923 		return (EPERM);
1924 	if (name[1] == NET_RT_DUMP) {
1925 		if (namelen == 3)
1926 			fib = req->td->td_proc->p_fibnum;
1927 		else if (namelen == 4)
1928 			fib = (name[3] == RT_ALL_FIBS) ?
1929 			    req->td->td_proc->p_fibnum : name[3];
1930 		else
1931 			return ((namelen < 3) ? EISDIR : ENOTDIR);
1932 		if (fib < 0 || fib >= rt_numfibs)
1933 			return (EINVAL);
1934 	} else if (namelen != 3)
1935 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1936 	af = name[0];
1937 	if (af > AF_MAX)
1938 		return (EINVAL);
1939 	bzero(&w, sizeof(w));
1940 	w.w_op = name[1];
1941 	w.w_arg = name[2];
1942 	w.w_req = req;
1943 
1944 	error = sysctl_wire_old_buffer(req, 0);
1945 	if (error)
1946 		return (error);
1947 	switch (w.w_op) {
1948 
1949 	case NET_RT_DUMP:
1950 	case NET_RT_FLAGS:
1951 		if (af == 0) {			/* dump all tables */
1952 			i = 1;
1953 			lim = AF_MAX;
1954 		} else				/* dump only one table */
1955 			i = lim = af;
1956 
1957 		/*
1958 		 * take care of llinfo entries, the caller must
1959 		 * specify an AF
1960 		 */
1961 		if (w.w_op == NET_RT_FLAGS &&
1962 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1963 			if (af != 0)
1964 				error = lltable_sysctl_dumparp(af, w.w_req);
1965 			else
1966 				error = EINVAL;
1967 			break;
1968 		}
1969 		/*
1970 		 * take care of routing entries
1971 		 */
1972 		for (error = 0; error == 0 && i <= lim; i++) {
1973 			rnh = rt_tables_get_rnh(fib, i);
1974 			if (rnh != NULL) {
1975 				RADIX_NODE_HEAD_RLOCK(rnh);
1976 			    	error = rnh->rnh_walktree(rnh,
1977 				    sysctl_dumpentry, &w);
1978 				RADIX_NODE_HEAD_RUNLOCK(rnh);
1979 			} else if (af != 0)
1980 				error = EAFNOSUPPORT;
1981 		}
1982 		break;
1983 
1984 	case NET_RT_IFLIST:
1985 	case NET_RT_IFLISTL:
1986 		error = sysctl_iflist(af, &w);
1987 		break;
1988 
1989 	case NET_RT_IFMALIST:
1990 		error = sysctl_ifmalist(af, &w);
1991 		break;
1992 	}
1993 	if (w.w_tmem)
1994 		free(w.w_tmem, M_RTABLE);
1995 	return (error);
1996 }
1997 
1998 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1999 
2000 /*
2001  * Definitions of protocols supported in the ROUTE domain.
2002  */
2003 
2004 static struct domain routedomain;		/* or at least forward */
2005 
2006 static struct protosw routesw[] = {
2007 {
2008 	.pr_type =		SOCK_RAW,
2009 	.pr_domain =		&routedomain,
2010 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2011 	.pr_output =		route_output,
2012 	.pr_ctlinput =		raw_ctlinput,
2013 	.pr_init =		raw_init,
2014 	.pr_usrreqs =		&route_usrreqs
2015 }
2016 };
2017 
2018 static struct domain routedomain = {
2019 	.dom_family =		PF_ROUTE,
2020 	.dom_name =		 "route",
2021 	.dom_protosw =		routesw,
2022 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
2023 };
2024 
2025 VNET_DOMAIN_SET(route);
2026