xref: /freebsd/sys/net/rtsock.c (revision a0e793cbf1951d07fc47a0d9ea389d7dacba5213)
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_msghdr32 {
77 	uint16_t ifm_msglen;
78 	uint8_t	ifm_version;
79 	uint8_t	ifm_type;
80 	int32_t	ifm_addrs;
81 	int32_t	ifm_flags;
82 	uint16_t ifm_index;
83 	struct	if_data ifm_data;
84 };
85 
86 struct if_msghdrl32 {
87 	uint16_t ifm_msglen;
88 	uint8_t	ifm_version;
89 	uint8_t	ifm_type;
90 	int32_t	ifm_addrs;
91 	int32_t	ifm_flags;
92 	uint16_t ifm_index;
93 	uint16_t _ifm_spare1;
94 	uint16_t ifm_len;
95 	uint16_t ifm_data_off;
96 	struct	if_data ifm_data;
97 };
98 
99 struct ifa_msghdrl32 {
100 	uint16_t ifam_msglen;
101 	uint8_t	ifam_version;
102 	uint8_t	ifam_type;
103 	int32_t	ifam_addrs;
104 	int32_t	ifam_flags;
105 	uint16_t ifam_index;
106 	uint16_t _ifam_spare1;
107 	uint16_t ifam_len;
108 	uint16_t ifam_data_off;
109 	int32_t	ifam_metric;
110 	struct	if_data ifam_data;
111 };
112 #endif /* COMPAT_FREEBSD32 */
113 
114 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
115 
116 /* NB: these are not modified */
117 static struct	sockaddr route_src = { 2, PF_ROUTE, };
118 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
119 
120 /* These are external hooks for CARP. */
121 int	(*carp_get_vhid_p)(struct ifaddr *);
122 
123 /*
124  * Used by rtsock/raw_input callback code to decide whether to filter the update
125  * notification to a socket bound to a particular FIB.
126  */
127 #define	RTS_FILTER_FIB	M_PROTO8
128 
129 typedef struct {
130 	int	ip_count;	/* attached w/ AF_INET */
131 	int	ip6_count;	/* attached w/ AF_INET6 */
132 	int	any_count;	/* total attached */
133 } route_cb_t;
134 static VNET_DEFINE(route_cb_t, route_cb);
135 #define	V_route_cb VNET(route_cb)
136 
137 struct mtx rtsock_mtx;
138 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
139 
140 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
141 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
142 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
143 
144 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
145 
146 struct walkarg {
147 	int	w_tmemsize;
148 	int	w_op, w_arg;
149 	caddr_t	w_tmem;
150 	struct sysctl_req *w_req;
151 };
152 
153 static void	rts_input(struct mbuf *m);
154 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
155 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
156 			struct walkarg *w, int *plen);
157 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
158 			struct rt_addrinfo *rtinfo);
159 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
160 static int	sysctl_iflist(int af, struct walkarg *w);
161 static int	sysctl_ifmalist(int af, struct walkarg *w);
162 static int	route_output(struct mbuf *m, struct socket *so);
163 static void	rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
164 static void	rt_dispatch(struct mbuf *, sa_family_t);
165 static struct sockaddr	*rtsock_fix_netmask(struct sockaddr *dst,
166 			struct sockaddr *smask, struct sockaddr_storage *dmask);
167 
168 static struct netisr_handler rtsock_nh = {
169 	.nh_name = "rtsock",
170 	.nh_handler = rts_input,
171 	.nh_proto = NETISR_ROUTE,
172 	.nh_policy = NETISR_POLICY_SOURCE,
173 };
174 
175 static int
176 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
177 {
178 	int error, qlimit;
179 
180 	netisr_getqlimit(&rtsock_nh, &qlimit);
181 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
182         if (error || !req->newptr)
183                 return (error);
184 	if (qlimit < 1)
185 		return (EINVAL);
186 	return (netisr_setqlimit(&rtsock_nh, qlimit));
187 }
188 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
189     0, 0, sysctl_route_netisr_maxqlen, "I",
190     "maximum routing socket dispatch queue length");
191 
192 static void
193 rts_init(void)
194 {
195 	int tmp;
196 
197 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
198 		rtsock_nh.nh_qlimit = tmp;
199 	netisr_register(&rtsock_nh);
200 }
201 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
202 
203 static int
204 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
205     struct rawcb *rp)
206 {
207 	int fibnum;
208 
209 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
210 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
211 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
212 
213 	/* No filtering requested. */
214 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
215 		return (0);
216 
217 	/* Check if it is a rts and the fib matches the one of the socket. */
218 	fibnum = M_GETFIB(m);
219 	if (proto->sp_family != PF_ROUTE ||
220 	    rp->rcb_socket == NULL ||
221 	    rp->rcb_socket->so_fibnum == fibnum)
222 		return (0);
223 
224 	/* Filtering requested and no match, the socket shall be skipped. */
225 	return (1);
226 }
227 
228 static void
229 rts_input(struct mbuf *m)
230 {
231 	struct sockproto route_proto;
232 	unsigned short *family;
233 	struct m_tag *tag;
234 
235 	route_proto.sp_family = PF_ROUTE;
236 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
237 	if (tag != NULL) {
238 		family = (unsigned short *)(tag + 1);
239 		route_proto.sp_protocol = *family;
240 		m_tag_delete(m, tag);
241 	} else
242 		route_proto.sp_protocol = 0;
243 
244 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
245 }
246 
247 /*
248  * It really doesn't make any sense at all for this code to share much
249  * with raw_usrreq.c, since its functionality is so restricted.  XXX
250  */
251 static void
252 rts_abort(struct socket *so)
253 {
254 
255 	raw_usrreqs.pru_abort(so);
256 }
257 
258 static void
259 rts_close(struct socket *so)
260 {
261 
262 	raw_usrreqs.pru_close(so);
263 }
264 
265 /* pru_accept is EOPNOTSUPP */
266 
267 static int
268 rts_attach(struct socket *so, int proto, struct thread *td)
269 {
270 	struct rawcb *rp;
271 	int error;
272 
273 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
274 
275 	/* XXX */
276 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
277 	if (rp == NULL)
278 		return ENOBUFS;
279 
280 	so->so_pcb = (caddr_t)rp;
281 	so->so_fibnum = td->td_proc->p_fibnum;
282 	error = raw_attach(so, proto);
283 	rp = sotorawcb(so);
284 	if (error) {
285 		so->so_pcb = NULL;
286 		free(rp, M_PCB);
287 		return error;
288 	}
289 	RTSOCK_LOCK();
290 	switch(rp->rcb_proto.sp_protocol) {
291 	case AF_INET:
292 		V_route_cb.ip_count++;
293 		break;
294 	case AF_INET6:
295 		V_route_cb.ip6_count++;
296 		break;
297 	}
298 	V_route_cb.any_count++;
299 	RTSOCK_UNLOCK();
300 	soisconnected(so);
301 	so->so_options |= SO_USELOOPBACK;
302 	return 0;
303 }
304 
305 static int
306 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
307 {
308 
309 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
310 }
311 
312 static int
313 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
314 {
315 
316 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
317 }
318 
319 /* pru_connect2 is EOPNOTSUPP */
320 /* pru_control is EOPNOTSUPP */
321 
322 static void
323 rts_detach(struct socket *so)
324 {
325 	struct rawcb *rp = sotorawcb(so);
326 
327 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
328 
329 	RTSOCK_LOCK();
330 	switch(rp->rcb_proto.sp_protocol) {
331 	case AF_INET:
332 		V_route_cb.ip_count--;
333 		break;
334 	case AF_INET6:
335 		V_route_cb.ip6_count--;
336 		break;
337 	}
338 	V_route_cb.any_count--;
339 	RTSOCK_UNLOCK();
340 	raw_usrreqs.pru_detach(so);
341 }
342 
343 static int
344 rts_disconnect(struct socket *so)
345 {
346 
347 	return (raw_usrreqs.pru_disconnect(so));
348 }
349 
350 /* pru_listen is EOPNOTSUPP */
351 
352 static int
353 rts_peeraddr(struct socket *so, struct sockaddr **nam)
354 {
355 
356 	return (raw_usrreqs.pru_peeraddr(so, nam));
357 }
358 
359 /* pru_rcvd is EOPNOTSUPP */
360 /* pru_rcvoob is EOPNOTSUPP */
361 
362 static int
363 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
364 	 struct mbuf *control, struct thread *td)
365 {
366 
367 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
368 }
369 
370 /* pru_sense is null */
371 
372 static int
373 rts_shutdown(struct socket *so)
374 {
375 
376 	return (raw_usrreqs.pru_shutdown(so));
377 }
378 
379 static int
380 rts_sockaddr(struct socket *so, struct sockaddr **nam)
381 {
382 
383 	return (raw_usrreqs.pru_sockaddr(so, nam));
384 }
385 
386 static struct pr_usrreqs route_usrreqs = {
387 	.pru_abort =		rts_abort,
388 	.pru_attach =		rts_attach,
389 	.pru_bind =		rts_bind,
390 	.pru_connect =		rts_connect,
391 	.pru_detach =		rts_detach,
392 	.pru_disconnect =	rts_disconnect,
393 	.pru_peeraddr =		rts_peeraddr,
394 	.pru_send =		rts_send,
395 	.pru_shutdown =		rts_shutdown,
396 	.pru_sockaddr =		rts_sockaddr,
397 	.pru_close =		rts_close,
398 };
399 
400 #ifndef _SOCKADDR_UNION_DEFINED
401 #define	_SOCKADDR_UNION_DEFINED
402 /*
403  * The union of all possible address formats we handle.
404  */
405 union sockaddr_union {
406 	struct sockaddr		sa;
407 	struct sockaddr_in	sin;
408 	struct sockaddr_in6	sin6;
409 };
410 #endif /* _SOCKADDR_UNION_DEFINED */
411 
412 static int
413 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
414     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
415 {
416 
417 	/* First, see if the returned address is part of the jail. */
418 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
419 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
420 		return (0);
421 	}
422 
423 	switch (info->rti_info[RTAX_DST]->sa_family) {
424 #ifdef INET
425 	case AF_INET:
426 	{
427 		struct in_addr ia;
428 		struct ifaddr *ifa;
429 		int found;
430 
431 		found = 0;
432 		/*
433 		 * Try to find an address on the given outgoing interface
434 		 * that belongs to the jail.
435 		 */
436 		IF_ADDR_RLOCK(ifp);
437 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
438 			struct sockaddr *sa;
439 			sa = ifa->ifa_addr;
440 			if (sa->sa_family != AF_INET)
441 				continue;
442 			ia = ((struct sockaddr_in *)sa)->sin_addr;
443 			if (prison_check_ip4(cred, &ia) == 0) {
444 				found = 1;
445 				break;
446 			}
447 		}
448 		IF_ADDR_RUNLOCK(ifp);
449 		if (!found) {
450 			/*
451 			 * As a last resort return the 'default' jail address.
452 			 */
453 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
454 			    sin_addr;
455 			if (prison_get_ip4(cred, &ia) != 0)
456 				return (ESRCH);
457 		}
458 		bzero(&saun->sin, sizeof(struct sockaddr_in));
459 		saun->sin.sin_len = sizeof(struct sockaddr_in);
460 		saun->sin.sin_family = AF_INET;
461 		saun->sin.sin_addr.s_addr = ia.s_addr;
462 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
463 		break;
464 	}
465 #endif
466 #ifdef INET6
467 	case AF_INET6:
468 	{
469 		struct in6_addr ia6;
470 		struct ifaddr *ifa;
471 		int found;
472 
473 		found = 0;
474 		/*
475 		 * Try to find an address on the given outgoing interface
476 		 * that belongs to the jail.
477 		 */
478 		IF_ADDR_RLOCK(ifp);
479 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
480 			struct sockaddr *sa;
481 			sa = ifa->ifa_addr;
482 			if (sa->sa_family != AF_INET6)
483 				continue;
484 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
485 			    &ia6, sizeof(struct in6_addr));
486 			if (prison_check_ip6(cred, &ia6) == 0) {
487 				found = 1;
488 				break;
489 			}
490 		}
491 		IF_ADDR_RUNLOCK(ifp);
492 		if (!found) {
493 			/*
494 			 * As a last resort return the 'default' jail address.
495 			 */
496 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
497 			    sin6_addr;
498 			if (prison_get_ip6(cred, &ia6) != 0)
499 				return (ESRCH);
500 		}
501 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
502 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
503 		saun->sin6.sin6_family = AF_INET6;
504 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
505 		if (sa6_recoverscope(&saun->sin6) != 0)
506 			return (ESRCH);
507 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
508 		break;
509 	}
510 #endif
511 	default:
512 		return (ESRCH);
513 	}
514 	return (0);
515 }
516 
517 /*ARGSUSED*/
518 static int
519 route_output(struct mbuf *m, struct socket *so)
520 {
521 	struct rt_msghdr *rtm = NULL;
522 	struct rtentry *rt = NULL;
523 	struct radix_node_head *rnh;
524 	struct rt_addrinfo info;
525 	struct sockaddr_storage ss;
526 #ifdef INET6
527 	struct sockaddr_in6 *sin6;
528 	int i, rti_need_deembed = 0;
529 #endif
530 	int alloc_len = 0, len, error = 0, fibnum;
531 	struct ifnet *ifp = NULL;
532 	union sockaddr_union saun;
533 	sa_family_t saf = AF_UNSPEC;
534 	struct rawcb *rp = NULL;
535 	struct walkarg w;
536 
537 	fibnum = so->so_fibnum;
538 
539 #define senderr(e) { error = e; goto flush;}
540 	if (m == NULL || ((m->m_len < sizeof(long)) &&
541 		       (m = m_pullup(m, sizeof(long))) == NULL))
542 		return (ENOBUFS);
543 	if ((m->m_flags & M_PKTHDR) == 0)
544 		panic("route_output");
545 	len = m->m_pkthdr.len;
546 	if (len < sizeof(*rtm) ||
547 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
548 		senderr(EINVAL);
549 
550 	/*
551 	 * Most of current messages are in range 200-240 bytes,
552 	 * minimize possible re-allocation on reply using larger size
553 	 * buffer aligned on 1k boundaty.
554 	 */
555 	alloc_len = roundup2(len, 1024);
556 	if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
557 		senderr(ENOBUFS);
558 
559 	m_copydata(m, 0, len, (caddr_t)rtm);
560 	bzero(&info, sizeof(info));
561 	bzero(&w, sizeof(w));
562 
563 	if (rtm->rtm_version != RTM_VERSION) {
564 		/* Do not touch message since format is unknown */
565 		free(rtm, M_TEMP);
566 		rtm = NULL;
567 		senderr(EPROTONOSUPPORT);
568 	}
569 
570 	/*
571 	 * Starting from here, it is possible
572 	 * to alter original message and insert
573 	 * caller PID and error value.
574 	 */
575 
576 	rtm->rtm_pid = curproc->p_pid;
577 	info.rti_addrs = rtm->rtm_addrs;
578 
579 	info.rti_mflags = rtm->rtm_inits;
580 	info.rti_rmx = &rtm->rtm_rmx;
581 
582 	/*
583 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
584 	 * link-local address because rtrequest requires addresses with
585 	 * embedded scope id.
586 	 */
587 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info))
588 		senderr(EINVAL);
589 
590 	info.rti_flags = rtm->rtm_flags;
591 	if (info.rti_info[RTAX_DST] == NULL ||
592 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
593 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
594 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
595 		senderr(EINVAL);
596 	saf = info.rti_info[RTAX_DST]->sa_family;
597 	/*
598 	 * Verify that the caller has the appropriate privilege; RTM_GET
599 	 * is the only operation the non-superuser is allowed.
600 	 */
601 	if (rtm->rtm_type != RTM_GET) {
602 		error = priv_check(curthread, PRIV_NET_ROUTE);
603 		if (error)
604 			senderr(error);
605 	}
606 
607 	/*
608 	 * The given gateway address may be an interface address.
609 	 * For example, issuing a "route change" command on a route
610 	 * entry that was created from a tunnel, and the gateway
611 	 * address given is the local end point. In this case the
612 	 * RTF_GATEWAY flag must be cleared or the destination will
613 	 * not be reachable even though there is no error message.
614 	 */
615 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
616 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
617 		struct route gw_ro;
618 
619 		bzero(&gw_ro, sizeof(gw_ro));
620 		gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
621 		rtalloc_ign_fib(&gw_ro, 0, fibnum);
622 		/*
623 		 * A host route through the loopback interface is
624 		 * installed for each interface adddress. In pre 8.0
625 		 * releases the interface address of a PPP link type
626 		 * is not reachable locally. This behavior is fixed as
627 		 * part of the new L2/L3 redesign and rewrite work. The
628 		 * signature of this interface address route is the
629 		 * AF_LINK sa_family type of the rt_gateway, and the
630 		 * rt_ifp has the IFF_LOOPBACK flag set.
631 		 */
632 		if (gw_ro.ro_rt != NULL &&
633 		    gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
634 		    gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) {
635 			info.rti_flags &= ~RTF_GATEWAY;
636 			info.rti_flags |= RTF_GWFLAG_COMPAT;
637 		}
638 		if (gw_ro.ro_rt != NULL)
639 			RTFREE(gw_ro.ro_rt);
640 	}
641 
642 	switch (rtm->rtm_type) {
643 		struct rtentry *saved_nrt;
644 
645 	case RTM_ADD:
646 	case RTM_CHANGE:
647 		if (info.rti_info[RTAX_GATEWAY] == NULL)
648 			senderr(EINVAL);
649 		saved_nrt = NULL;
650 
651 		/* support for new ARP code */
652 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
653 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
654 			error = lla_rt_output(rtm, &info);
655 #ifdef INET6
656 			if (error == 0)
657 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
658 #endif
659 			break;
660 		}
661 		error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
662 		    fibnum);
663 		if (error == 0 && saved_nrt != NULL) {
664 #ifdef INET6
665 			rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
666 #endif
667 			RT_LOCK(saved_nrt);
668 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
669 			RT_REMREF(saved_nrt);
670 			RT_UNLOCK(saved_nrt);
671 		}
672 		break;
673 
674 	case RTM_DELETE:
675 		saved_nrt = NULL;
676 		/* support for new ARP code */
677 		if (info.rti_info[RTAX_GATEWAY] &&
678 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
679 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
680 			error = lla_rt_output(rtm, &info);
681 #ifdef INET6
682 			if (error == 0)
683 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
684 #endif
685 			break;
686 		}
687 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
688 		if (error == 0) {
689 			RT_LOCK(saved_nrt);
690 			rt = saved_nrt;
691 			goto report;
692 		}
693 #ifdef INET6
694 		/* rt_msg2() will not be used when RTM_DELETE fails. */
695 		rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
696 #endif
697 		break;
698 
699 	case RTM_GET:
700 		rnh = rt_tables_get_rnh(fibnum, saf);
701 		if (rnh == NULL)
702 			senderr(EAFNOSUPPORT);
703 
704 		RADIX_NODE_HEAD_RLOCK(rnh);
705 
706 		if (info.rti_info[RTAX_NETMASK] == NULL &&
707 		    rtm->rtm_type == RTM_GET) {
708 			/*
709 			 * Provide logest prefix match for
710 			 * address lookup (no mask).
711 			 * 'route -n get addr'
712 			 */
713 			rt = (struct rtentry *) rnh->rnh_matchaddr(
714 			    info.rti_info[RTAX_DST], rnh);
715 		} else
716 			rt = (struct rtentry *) rnh->rnh_lookup(
717 			    info.rti_info[RTAX_DST],
718 			    info.rti_info[RTAX_NETMASK], rnh);
719 
720 		if (rt == NULL) {
721 			RADIX_NODE_HEAD_RUNLOCK(rnh);
722 			senderr(ESRCH);
723 		}
724 #ifdef RADIX_MPATH
725 		/*
726 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
727 		 * we require users to specify a matching RTAX_GATEWAY.
728 		 *
729 		 * for RTM_GET, gate is optional even with multipath.
730 		 * if gate == NULL the first match is returned.
731 		 * (no need to call rt_mpath_matchgate if gate == NULL)
732 		 */
733 		if (rn_mpath_capable(rnh) &&
734 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
735 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
736 			if (!rt) {
737 				RADIX_NODE_HEAD_RUNLOCK(rnh);
738 				senderr(ESRCH);
739 			}
740 		}
741 #endif
742 		/*
743 		 * If performing proxied L2 entry insertion, and
744 		 * the actual PPP host entry is found, perform
745 		 * another search to retrieve the prefix route of
746 		 * the local end point of the PPP link.
747 		 */
748 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
749 			struct sockaddr laddr;
750 
751 			if (rt->rt_ifp != NULL &&
752 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
753 				struct ifaddr *ifa;
754 
755 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
756 				if (ifa != NULL)
757 					rt_maskedcopy(ifa->ifa_addr,
758 						      &laddr,
759 						      ifa->ifa_netmask);
760 			} else
761 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
762 					      &laddr,
763 					      rt->rt_ifa->ifa_netmask);
764 			/*
765 			 * refactor rt and no lock operation necessary
766 			 */
767 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
768 			if (rt == NULL) {
769 				RADIX_NODE_HEAD_RUNLOCK(rnh);
770 				senderr(ESRCH);
771 			}
772 		}
773 		RT_LOCK(rt);
774 		RT_ADDREF(rt);
775 		RADIX_NODE_HEAD_RUNLOCK(rnh);
776 
777 report:
778 		RT_LOCK_ASSERT(rt);
779 		if ((rt->rt_flags & RTF_HOST) == 0
780 		    ? jailed_without_vnet(curthread->td_ucred)
781 		    : prison_if(curthread->td_ucred,
782 		    rt_key(rt)) != 0) {
783 			RT_UNLOCK(rt);
784 			senderr(ESRCH);
785 		}
786 		info.rti_info[RTAX_DST] = rt_key(rt);
787 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
788 		info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
789 		    rt_mask(rt), &ss);
790 		info.rti_info[RTAX_GENMASK] = 0;
791 		if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
792 			ifp = rt->rt_ifp;
793 			if (ifp) {
794 				info.rti_info[RTAX_IFP] =
795 				    ifp->if_addr->ifa_addr;
796 				error = rtm_get_jailed(&info, ifp, rt,
797 				    &saun, curthread->td_ucred);
798 				if (error != 0) {
799 					RT_UNLOCK(rt);
800 					senderr(error);
801 				}
802 				if (ifp->if_flags & IFF_POINTOPOINT)
803 					info.rti_info[RTAX_BRD] =
804 					    rt->rt_ifa->ifa_dstaddr;
805 				rtm->rtm_index = ifp->if_index;
806 			} else {
807 				info.rti_info[RTAX_IFP] = NULL;
808 				info.rti_info[RTAX_IFA] = NULL;
809 			}
810 		} else if ((ifp = rt->rt_ifp) != NULL) {
811 			rtm->rtm_index = ifp->if_index;
812 		}
813 
814 		/* Check if we need to realloc storage */
815 		rtsock_msg_buffer(rtm->rtm_type, &info, NULL, &len);
816 		if (len > alloc_len) {
817 			struct rt_msghdr *new_rtm;
818 			new_rtm = malloc(len, M_TEMP, M_NOWAIT);
819 			if (new_rtm == NULL) {
820 				RT_UNLOCK(rt);
821 				senderr(ENOBUFS);
822 			}
823 			bcopy(rtm, new_rtm, rtm->rtm_msglen);
824 			free(rtm, M_TEMP);
825 			rtm = new_rtm;
826 			alloc_len = len;
827 		}
828 
829 		w.w_tmem = (caddr_t)rtm;
830 		w.w_tmemsize = alloc_len;
831 		rtsock_msg_buffer(rtm->rtm_type, &info, &w, &len);
832 
833 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
834 			rtm->rtm_flags = RTF_GATEWAY |
835 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
836 		else
837 			rtm->rtm_flags = rt->rt_flags;
838 		rt_getmetrics(rt, &rtm->rtm_rmx);
839 		rtm->rtm_addrs = info.rti_addrs;
840 
841 		RT_UNLOCK(rt);
842 		break;
843 
844 	default:
845 		senderr(EOPNOTSUPP);
846 	}
847 
848 flush:
849 	if (rt != NULL)
850 		RTFREE(rt);
851 	/*
852 	 * Check to see if we don't want our own messages.
853 	 */
854 	if ((so->so_options & SO_USELOOPBACK) == 0) {
855 		if (V_route_cb.any_count <= 1) {
856 			if (rtm != NULL)
857 				free(rtm, M_TEMP);
858 			m_freem(m);
859 			return (error);
860 		}
861 		/* There is another listener, so construct message */
862 		rp = sotorawcb(so);
863 	}
864 
865 	if (rtm != NULL) {
866 #ifdef INET6
867 		if (rti_need_deembed) {
868 			/* sin6_scope_id is recovered before sending rtm. */
869 			sin6 = (struct sockaddr_in6 *)&ss;
870 			for (i = 0; i < RTAX_MAX; i++) {
871 				if (info.rti_info[i] == NULL)
872 					continue;
873 				if (info.rti_info[i]->sa_family != AF_INET6)
874 					continue;
875 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
876 				if (sa6_recoverscope(sin6) == 0)
877 					bcopy(sin6, info.rti_info[i],
878 						    sizeof(*sin6));
879 			}
880 		}
881 #endif
882 		if (error != 0)
883 			rtm->rtm_errno = error;
884 		else
885 			rtm->rtm_flags |= RTF_DONE;
886 
887 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
888 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
889 			m_freem(m);
890 			m = NULL;
891 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
892 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
893 
894 		free(rtm, M_TEMP);
895 	}
896 	if (m != NULL) {
897 		M_SETFIB(m, fibnum);
898 		m->m_flags |= RTS_FILTER_FIB;
899 		if (rp) {
900 			/*
901 			 * XXX insure we don't get a copy by
902 			 * invalidating our protocol
903 			 */
904 			unsigned short family = rp->rcb_proto.sp_family;
905 			rp->rcb_proto.sp_family = 0;
906 			rt_dispatch(m, saf);
907 			rp->rcb_proto.sp_family = family;
908 		} else
909 			rt_dispatch(m, saf);
910 	}
911 
912 	return (error);
913 }
914 
915 static void
916 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
917 {
918 
919 	bzero(out, sizeof(*out));
920 	out->rmx_mtu = rt->rt_mtu;
921 	out->rmx_weight = rt->rt_weight;
922 	out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
923 	/* Kernel -> userland timebase conversion. */
924 	out->rmx_expire = rt->rt_expire ?
925 	    rt->rt_expire - time_uptime + time_second : 0;
926 }
927 
928 /*
929  * Extract the addresses of the passed sockaddrs.
930  * Do a little sanity checking so as to avoid bad memory references.
931  * This data is derived straight from userland.
932  */
933 static int
934 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
935 {
936 	struct sockaddr *sa;
937 	int i;
938 
939 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
940 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
941 			continue;
942 		sa = (struct sockaddr *)cp;
943 		/*
944 		 * It won't fit.
945 		 */
946 		if (cp + sa->sa_len > cplim)
947 			return (EINVAL);
948 		/*
949 		 * there are no more.. quit now
950 		 * If there are more bits, they are in error.
951 		 * I've seen this. route(1) can evidently generate these.
952 		 * This causes kernel to core dump.
953 		 * for compatibility, If we see this, point to a safe address.
954 		 */
955 		if (sa->sa_len == 0) {
956 			rtinfo->rti_info[i] = &sa_zero;
957 			return (0); /* should be EINVAL but for compat */
958 		}
959 		/* accept it */
960 #ifdef INET6
961 		if (sa->sa_family == AF_INET6)
962 			sa6_embedscope((struct sockaddr_in6 *)sa,
963 			    V_ip6_use_defzone);
964 #endif
965 		rtinfo->rti_info[i] = sa;
966 		cp += SA_SIZE(sa);
967 	}
968 	return (0);
969 }
970 
971 /*
972  * Fill in @dmask with valid netmask leaving original @smask
973  * intact. Mostly used with radix netmasks.
974  */
975 static struct sockaddr *
976 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask,
977     struct sockaddr_storage *dmask)
978 {
979 	if (dst == NULL || smask == NULL)
980 		return (NULL);
981 
982 	memset(dmask, 0, dst->sa_len);
983 	memcpy(dmask, smask, smask->sa_len);
984 	dmask->ss_len = dst->sa_len;
985 	dmask->ss_family = dst->sa_family;
986 
987 	return ((struct sockaddr *)dmask);
988 }
989 
990 /*
991  * Writes information related to @rtinfo object to newly-allocated mbuf.
992  * Assumes MCLBYTES is enough to construct any message.
993  * Used for OS notifications of vaious events (if/ifa announces,etc)
994  *
995  * Returns allocated mbuf or NULL on failure.
996  */
997 static struct mbuf *
998 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
999 {
1000 	struct rt_msghdr *rtm;
1001 	struct mbuf *m;
1002 	int i;
1003 	struct sockaddr *sa;
1004 #ifdef INET6
1005 	struct sockaddr_storage ss;
1006 	struct sockaddr_in6 *sin6;
1007 #endif
1008 	int len, dlen;
1009 
1010 	switch (type) {
1011 
1012 	case RTM_DELADDR:
1013 	case RTM_NEWADDR:
1014 		len = sizeof(struct ifa_msghdr);
1015 		break;
1016 
1017 	case RTM_DELMADDR:
1018 	case RTM_NEWMADDR:
1019 		len = sizeof(struct ifma_msghdr);
1020 		break;
1021 
1022 	case RTM_IFINFO:
1023 		len = sizeof(struct if_msghdr);
1024 		break;
1025 
1026 	case RTM_IFANNOUNCE:
1027 	case RTM_IEEE80211:
1028 		len = sizeof(struct if_announcemsghdr);
1029 		break;
1030 
1031 	default:
1032 		len = sizeof(struct rt_msghdr);
1033 	}
1034 
1035 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1036 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1037 	if (len > MHLEN)
1038 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1039 	else
1040 		m = m_gethdr(M_NOWAIT, MT_DATA);
1041 	if (m == NULL)
1042 		return (m);
1043 
1044 	m->m_pkthdr.len = m->m_len = len;
1045 	rtm = mtod(m, struct rt_msghdr *);
1046 	bzero((caddr_t)rtm, len);
1047 	for (i = 0; i < RTAX_MAX; i++) {
1048 		if ((sa = rtinfo->rti_info[i]) == NULL)
1049 			continue;
1050 		rtinfo->rti_addrs |= (1 << i);
1051 		dlen = SA_SIZE(sa);
1052 #ifdef INET6
1053 		if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1054 			sin6 = (struct sockaddr_in6 *)&ss;
1055 			bcopy(sa, sin6, sizeof(*sin6));
1056 			if (sa6_recoverscope(sin6) == 0)
1057 				sa = (struct sockaddr *)sin6;
1058 		}
1059 #endif
1060 		m_copyback(m, len, dlen, (caddr_t)sa);
1061 		len += dlen;
1062 	}
1063 	if (m->m_pkthdr.len != len) {
1064 		m_freem(m);
1065 		return (NULL);
1066 	}
1067 	rtm->rtm_msglen = len;
1068 	rtm->rtm_version = RTM_VERSION;
1069 	rtm->rtm_type = type;
1070 	return (m);
1071 }
1072 
1073 /*
1074  * Writes information related to @rtinfo object to preallocated buffer.
1075  * Stores needed size in @plen. If @w is NULL, calculates size without
1076  * writing.
1077  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1078  *
1079  * Returns 0 on success.
1080  *
1081  */
1082 static int
1083 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1084 {
1085 	int i;
1086 	int len, buflen = 0, dlen;
1087 	caddr_t cp = NULL;
1088 	struct rt_msghdr *rtm = NULL;
1089 #ifdef INET6
1090 	struct sockaddr_storage ss;
1091 	struct sockaddr_in6 *sin6;
1092 #endif
1093 
1094 	switch (type) {
1095 
1096 	case RTM_DELADDR:
1097 	case RTM_NEWADDR:
1098 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1099 #ifdef COMPAT_FREEBSD32
1100 			if (w->w_req->flags & SCTL_MASK32)
1101 				len = sizeof(struct ifa_msghdrl32);
1102 			else
1103 #endif
1104 				len = sizeof(struct ifa_msghdrl);
1105 		} else
1106 			len = sizeof(struct ifa_msghdr);
1107 		break;
1108 
1109 	case RTM_IFINFO:
1110 #ifdef COMPAT_FREEBSD32
1111 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1112 			if (w->w_op == NET_RT_IFLISTL)
1113 				len = sizeof(struct if_msghdrl32);
1114 			else
1115 				len = sizeof(struct if_msghdr32);
1116 			break;
1117 		}
1118 #endif
1119 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1120 			len = sizeof(struct if_msghdrl);
1121 		else
1122 			len = sizeof(struct if_msghdr);
1123 		break;
1124 
1125 	case RTM_NEWMADDR:
1126 		len = sizeof(struct ifma_msghdr);
1127 		break;
1128 
1129 	default:
1130 		len = sizeof(struct rt_msghdr);
1131 	}
1132 
1133 	if (w != NULL) {
1134 		rtm = (struct rt_msghdr *)w->w_tmem;
1135 		buflen = w->w_tmemsize - len;
1136 		cp = (caddr_t)w->w_tmem + len;
1137 	}
1138 
1139 	rtinfo->rti_addrs = 0;
1140 	for (i = 0; i < RTAX_MAX; i++) {
1141 		struct sockaddr *sa;
1142 
1143 		if ((sa = rtinfo->rti_info[i]) == NULL)
1144 			continue;
1145 		rtinfo->rti_addrs |= (1 << i);
1146 		dlen = SA_SIZE(sa);
1147 		if (cp != NULL && buflen >= dlen) {
1148 #ifdef INET6
1149 			if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1150 				sin6 = (struct sockaddr_in6 *)&ss;
1151 				bcopy(sa, sin6, sizeof(*sin6));
1152 				if (sa6_recoverscope(sin6) == 0)
1153 					sa = (struct sockaddr *)sin6;
1154 			}
1155 #endif
1156 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1157 			cp += dlen;
1158 			buflen -= dlen;
1159 		} else if (cp != NULL) {
1160 			/*
1161 			 * Buffer too small. Count needed size
1162 			 * and return with error.
1163 			 */
1164 			cp = NULL;
1165 		}
1166 
1167 		len += dlen;
1168 	}
1169 
1170 	if (cp != NULL) {
1171 		dlen = ALIGN(len) - len;
1172 		if (buflen < dlen)
1173 			cp = NULL;
1174 		else
1175 			buflen -= dlen;
1176 	}
1177 	len = ALIGN(len);
1178 
1179 	if (cp != NULL) {
1180 		/* fill header iff buffer is large enough */
1181 		rtm->rtm_version = RTM_VERSION;
1182 		rtm->rtm_type = type;
1183 		rtm->rtm_msglen = len;
1184 	}
1185 
1186 	*plen = len;
1187 
1188 	if (w != NULL && cp == NULL)
1189 		return (ENOBUFS);
1190 
1191 	return (0);
1192 }
1193 
1194 /*
1195  * This routine is called to generate a message from the routing
1196  * socket indicating that a redirect has occured, a routing lookup
1197  * has failed, or that a protocol has detected timeouts to a particular
1198  * destination.
1199  */
1200 void
1201 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1202     int fibnum)
1203 {
1204 	struct rt_msghdr *rtm;
1205 	struct mbuf *m;
1206 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1207 
1208 	if (V_route_cb.any_count == 0)
1209 		return;
1210 	m = rtsock_msg_mbuf(type, rtinfo);
1211 	if (m == NULL)
1212 		return;
1213 
1214 	if (fibnum != RT_ALL_FIBS) {
1215 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1216 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1217 		M_SETFIB(m, fibnum);
1218 		m->m_flags |= RTS_FILTER_FIB;
1219 	}
1220 
1221 	rtm = mtod(m, struct rt_msghdr *);
1222 	rtm->rtm_flags = RTF_DONE | flags;
1223 	rtm->rtm_errno = error;
1224 	rtm->rtm_addrs = rtinfo->rti_addrs;
1225 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1226 }
1227 
1228 void
1229 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1230 {
1231 
1232 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1233 }
1234 
1235 /*
1236  * This routine is called to generate a message from the routing
1237  * socket indicating that the status of a network interface has changed.
1238  */
1239 void
1240 rt_ifmsg(struct ifnet *ifp)
1241 {
1242 	struct if_msghdr *ifm;
1243 	struct mbuf *m;
1244 	struct rt_addrinfo info;
1245 
1246 	if (V_route_cb.any_count == 0)
1247 		return;
1248 	bzero((caddr_t)&info, sizeof(info));
1249 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1250 	if (m == NULL)
1251 		return;
1252 	ifm = mtod(m, struct if_msghdr *);
1253 	ifm->ifm_index = ifp->if_index;
1254 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1255 	ifm->ifm_data = ifp->if_data;
1256 	ifm->ifm_addrs = 0;
1257 	rt_dispatch(m, AF_UNSPEC);
1258 }
1259 
1260 /*
1261  * Announce interface address arrival/withdraw.
1262  * Please do not call directly, use rt_addrmsg().
1263  * Assume input data to be valid.
1264  * Returns 0 on success.
1265  */
1266 int
1267 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1268 {
1269 	struct rt_addrinfo info;
1270 	struct sockaddr *sa;
1271 	int ncmd;
1272 	struct mbuf *m;
1273 	struct ifa_msghdr *ifam;
1274 	struct ifnet *ifp = ifa->ifa_ifp;
1275 	struct sockaddr_storage ss;
1276 
1277 	if (V_route_cb.any_count == 0)
1278 		return (0);
1279 
1280 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1281 
1282 	bzero((caddr_t)&info, sizeof(info));
1283 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1284 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1285 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1286 	    info.rti_info[RTAX_IFP], ifa->ifa_netmask, &ss);
1287 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1288 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1289 		return (ENOBUFS);
1290 	ifam = mtod(m, struct ifa_msghdr *);
1291 	ifam->ifam_index = ifp->if_index;
1292 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1293 	ifam->ifam_flags = ifa->ifa_flags;
1294 	ifam->ifam_addrs = info.rti_addrs;
1295 
1296 	if (fibnum != RT_ALL_FIBS) {
1297 		M_SETFIB(m, fibnum);
1298 		m->m_flags |= RTS_FILTER_FIB;
1299 	}
1300 
1301 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1302 
1303 	return (0);
1304 }
1305 
1306 /*
1307  * Announce route addition/removal.
1308  * Please do not call directly, use rt_routemsg().
1309  * Note that @rt data MAY be inconsistent/invalid:
1310  * if some userland app sends us "invalid" route message (invalid mask,
1311  * no dst, wrong address families, etc...) we need to pass it back
1312  * to app (and any other rtsock consumers) with rtm_errno field set to
1313  * non-zero value.
1314  *
1315  * Returns 0 on success.
1316  */
1317 int
1318 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1319     int fibnum)
1320 {
1321 	struct rt_addrinfo info;
1322 	struct sockaddr *sa;
1323 	struct mbuf *m;
1324 	struct rt_msghdr *rtm;
1325 	struct sockaddr_storage ss;
1326 
1327 	if (V_route_cb.any_count == 0)
1328 		return (0);
1329 
1330 	bzero((caddr_t)&info, sizeof(info));
1331 	info.rti_info[RTAX_DST] = sa = rt_key(rt);
1332 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(sa, rt_mask(rt), &ss);
1333 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1334 	if ((m = rtsock_msg_mbuf(cmd, &info)) == NULL)
1335 		return (ENOBUFS);
1336 	rtm = mtod(m, struct rt_msghdr *);
1337 	rtm->rtm_index = ifp->if_index;
1338 	rtm->rtm_flags |= rt->rt_flags;
1339 	rtm->rtm_errno = error;
1340 	rtm->rtm_addrs = info.rti_addrs;
1341 
1342 	if (fibnum != RT_ALL_FIBS) {
1343 		M_SETFIB(m, fibnum);
1344 		m->m_flags |= RTS_FILTER_FIB;
1345 	}
1346 
1347 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1348 
1349 	return (0);
1350 }
1351 
1352 /*
1353  * This is the analogue to the rt_newaddrmsg which performs the same
1354  * function but for multicast group memberhips.  This is easier since
1355  * there is no route state to worry about.
1356  */
1357 void
1358 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1359 {
1360 	struct rt_addrinfo info;
1361 	struct mbuf *m = NULL;
1362 	struct ifnet *ifp = ifma->ifma_ifp;
1363 	struct ifma_msghdr *ifmam;
1364 
1365 	if (V_route_cb.any_count == 0)
1366 		return;
1367 
1368 	bzero((caddr_t)&info, sizeof(info));
1369 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1370 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1371 	/*
1372 	 * If a link-layer address is present, present it as a ``gateway''
1373 	 * (similarly to how ARP entries, e.g., are presented).
1374 	 */
1375 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1376 	m = rtsock_msg_mbuf(cmd, &info);
1377 	if (m == NULL)
1378 		return;
1379 	ifmam = mtod(m, struct ifma_msghdr *);
1380 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1381 	    __func__));
1382 	ifmam->ifmam_index = ifp->if_index;
1383 	ifmam->ifmam_addrs = info.rti_addrs;
1384 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1385 }
1386 
1387 static struct mbuf *
1388 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1389 	struct rt_addrinfo *info)
1390 {
1391 	struct if_announcemsghdr *ifan;
1392 	struct mbuf *m;
1393 
1394 	if (V_route_cb.any_count == 0)
1395 		return NULL;
1396 	bzero((caddr_t)info, sizeof(*info));
1397 	m = rtsock_msg_mbuf(type, info);
1398 	if (m != NULL) {
1399 		ifan = mtod(m, struct if_announcemsghdr *);
1400 		ifan->ifan_index = ifp->if_index;
1401 		strlcpy(ifan->ifan_name, ifp->if_xname,
1402 			sizeof(ifan->ifan_name));
1403 		ifan->ifan_what = what;
1404 	}
1405 	return m;
1406 }
1407 
1408 /*
1409  * This is called to generate routing socket messages indicating
1410  * IEEE80211 wireless events.
1411  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1412  */
1413 void
1414 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1415 {
1416 	struct mbuf *m;
1417 	struct rt_addrinfo info;
1418 
1419 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1420 	if (m != NULL) {
1421 		/*
1422 		 * Append the ieee80211 data.  Try to stick it in the
1423 		 * mbuf containing the ifannounce msg; otherwise allocate
1424 		 * a new mbuf and append.
1425 		 *
1426 		 * NB: we assume m is a single mbuf.
1427 		 */
1428 		if (data_len > M_TRAILINGSPACE(m)) {
1429 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1430 			if (n == NULL) {
1431 				m_freem(m);
1432 				return;
1433 			}
1434 			bcopy(data, mtod(n, void *), data_len);
1435 			n->m_len = data_len;
1436 			m->m_next = n;
1437 		} else if (data_len > 0) {
1438 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1439 			m->m_len += data_len;
1440 		}
1441 		if (m->m_flags & M_PKTHDR)
1442 			m->m_pkthdr.len += data_len;
1443 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1444 		rt_dispatch(m, AF_UNSPEC);
1445 	}
1446 }
1447 
1448 /*
1449  * This is called to generate routing socket messages indicating
1450  * network interface arrival and departure.
1451  */
1452 void
1453 rt_ifannouncemsg(struct ifnet *ifp, int what)
1454 {
1455 	struct mbuf *m;
1456 	struct rt_addrinfo info;
1457 
1458 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1459 	if (m != NULL)
1460 		rt_dispatch(m, AF_UNSPEC);
1461 }
1462 
1463 static void
1464 rt_dispatch(struct mbuf *m, sa_family_t saf)
1465 {
1466 	struct m_tag *tag;
1467 
1468 	/*
1469 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1470 	 * use when injecting the mbuf into the routing socket buffer from
1471 	 * the netisr.
1472 	 */
1473 	if (saf != AF_UNSPEC) {
1474 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1475 		    M_NOWAIT);
1476 		if (tag == NULL) {
1477 			m_freem(m);
1478 			return;
1479 		}
1480 		*(unsigned short *)(tag + 1) = saf;
1481 		m_tag_prepend(m, tag);
1482 	}
1483 #ifdef VIMAGE
1484 	if (V_loif)
1485 		m->m_pkthdr.rcvif = V_loif;
1486 	else {
1487 		m_freem(m);
1488 		return;
1489 	}
1490 #endif
1491 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1492 }
1493 
1494 /*
1495  * This is used in dumping the kernel table via sysctl().
1496  */
1497 static int
1498 sysctl_dumpentry(struct radix_node *rn, void *vw)
1499 {
1500 	struct walkarg *w = vw;
1501 	struct rtentry *rt = (struct rtentry *)rn;
1502 	int error = 0, size;
1503 	struct rt_addrinfo info;
1504 	struct sockaddr_storage ss;
1505 
1506 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1507 		return 0;
1508 	if ((rt->rt_flags & RTF_HOST) == 0
1509 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1510 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1511 		return (0);
1512 	bzero((caddr_t)&info, sizeof(info));
1513 	info.rti_info[RTAX_DST] = rt_key(rt);
1514 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1515 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1516 	    rt_mask(rt), &ss);
1517 	info.rti_info[RTAX_GENMASK] = 0;
1518 	if (rt->rt_ifp) {
1519 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1520 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1521 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1522 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1523 	}
1524 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1525 		return (error);
1526 	if (w->w_req && w->w_tmem) {
1527 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1528 
1529 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1530 			rtm->rtm_flags = RTF_GATEWAY |
1531 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1532 		else
1533 			rtm->rtm_flags = rt->rt_flags;
1534 		rt_getmetrics(rt, &rtm->rtm_rmx);
1535 		rtm->rtm_index = rt->rt_ifp->if_index;
1536 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1537 		rtm->rtm_addrs = info.rti_addrs;
1538 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1539 		return (error);
1540 	}
1541 	return (error);
1542 }
1543 
1544 static int
1545 sysctl_iflist_ifml(struct ifnet *ifp, struct rt_addrinfo *info,
1546     struct walkarg *w, int len)
1547 {
1548 	struct if_msghdrl *ifm;
1549 	struct if_data *ifd;
1550 
1551 	ifm = (struct if_msghdrl *)w->w_tmem;
1552 
1553 #ifdef COMPAT_FREEBSD32
1554 	if (w->w_req->flags & SCTL_MASK32) {
1555 		struct if_msghdrl32 *ifm32;
1556 
1557 		ifm32 = (struct if_msghdrl32 *)ifm;
1558 		ifm32->ifm_addrs = info->rti_addrs;
1559 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1560 		ifm32->ifm_index = ifp->if_index;
1561 		ifm32->_ifm_spare1 = 0;
1562 		ifm32->ifm_len = sizeof(*ifm32);
1563 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1564 		ifd = &ifm32->ifm_data;
1565 	} else
1566 #endif
1567 	{
1568 		ifm->ifm_addrs = info->rti_addrs;
1569 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1570 		ifm->ifm_index = ifp->if_index;
1571 		ifm->_ifm_spare1 = 0;
1572 		ifm->ifm_len = sizeof(*ifm);
1573 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1574 		ifd = &ifm->ifm_data;
1575 	}
1576 
1577 	*ifd = ifp->if_data;
1578 
1579 	/* Some drivers still use ifqueue(9), add its stats. */
1580 	ifd->ifi_oqdrops += ifp->if_snd.ifq_drops;
1581 
1582 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1583 }
1584 
1585 static int
1586 sysctl_iflist_ifm(struct ifnet *ifp, struct rt_addrinfo *info,
1587     struct walkarg *w, int len)
1588 {
1589 	struct if_msghdr *ifm;
1590 	struct if_data *ifd;
1591 
1592 	ifm = (struct if_msghdr *)w->w_tmem;
1593 
1594 #ifdef COMPAT_FREEBSD32
1595 	if (w->w_req->flags & SCTL_MASK32) {
1596 		struct if_msghdr32 *ifm32;
1597 
1598 		ifm32 = (struct if_msghdr32 *)ifm;
1599 		ifm32->ifm_addrs = info->rti_addrs;
1600 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1601 		ifm32->ifm_index = ifp->if_index;
1602 		ifd = &ifm32->ifm_data;
1603 	} else
1604 #endif
1605 	{
1606 		ifm->ifm_addrs = info->rti_addrs;
1607 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1608 		ifm->ifm_index = ifp->if_index;
1609 		ifd = &ifm->ifm_data;
1610 	}
1611 
1612 	*ifd = ifp->if_data;
1613 
1614 	/* Some drivers still use ifqueue(9), add its stats. */
1615 	ifd->ifi_oqdrops += ifp->if_snd.ifq_drops;
1616 
1617 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1618 }
1619 
1620 static int
1621 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1622     struct walkarg *w, int len)
1623 {
1624 	struct ifa_msghdrl *ifam;
1625 	struct if_data *ifd;
1626 
1627 	ifam = (struct ifa_msghdrl *)w->w_tmem;
1628 
1629 #ifdef COMPAT_FREEBSD32
1630 	if (w->w_req->flags & SCTL_MASK32) {
1631 		struct ifa_msghdrl32 *ifam32;
1632 
1633 		ifam32 = (struct ifa_msghdrl32 *)ifam;
1634 		ifam32->ifam_addrs = info->rti_addrs;
1635 		ifam32->ifam_flags = ifa->ifa_flags;
1636 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
1637 		ifam32->_ifam_spare1 = 0;
1638 		ifam32->ifam_len = sizeof(*ifam32);
1639 		ifam32->ifam_data_off =
1640 		    offsetof(struct ifa_msghdrl32, ifam_data);
1641 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1642 		ifd = &ifam32->ifam_data;
1643 	} else
1644 #endif
1645 	{
1646 		ifam->ifam_addrs = info->rti_addrs;
1647 		ifam->ifam_flags = ifa->ifa_flags;
1648 		ifam->ifam_index = ifa->ifa_ifp->if_index;
1649 		ifam->_ifam_spare1 = 0;
1650 		ifam->ifam_len = sizeof(*ifam);
1651 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1652 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1653 		ifd = &ifam->ifam_data;
1654 	}
1655 
1656 	bzero(ifd, sizeof(*ifd));
1657 	ifd->ifi_datalen = sizeof(struct if_data);
1658 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1659 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1660 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1661 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1662 
1663 	/* Fixup if_data carp(4) vhid. */
1664 	if (carp_get_vhid_p != NULL)
1665 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1666 
1667 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1668 }
1669 
1670 static int
1671 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1672     struct walkarg *w, int len)
1673 {
1674 	struct ifa_msghdr *ifam;
1675 
1676 	ifam = (struct ifa_msghdr *)w->w_tmem;
1677 	ifam->ifam_addrs = info->rti_addrs;
1678 	ifam->ifam_flags = ifa->ifa_flags;
1679 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1680 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1681 
1682 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1683 }
1684 
1685 static int
1686 sysctl_iflist(int af, struct walkarg *w)
1687 {
1688 	struct ifnet *ifp;
1689 	struct ifaddr *ifa;
1690 	struct rt_addrinfo info;
1691 	int len, error = 0;
1692 	struct sockaddr_storage ss;
1693 
1694 	bzero((caddr_t)&info, sizeof(info));
1695 	IFNET_RLOCK_NOSLEEP();
1696 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1697 		if (w->w_arg && w->w_arg != ifp->if_index)
1698 			continue;
1699 		IF_ADDR_RLOCK(ifp);
1700 		ifa = ifp->if_addr;
1701 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1702 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1703 		if (error != 0)
1704 			goto done;
1705 		info.rti_info[RTAX_IFP] = NULL;
1706 		if (w->w_req && w->w_tmem) {
1707 			if (w->w_op == NET_RT_IFLISTL)
1708 				error = sysctl_iflist_ifml(ifp, &info, w, len);
1709 			else
1710 				error = sysctl_iflist_ifm(ifp, &info, w, len);
1711 			if (error)
1712 				goto done;
1713 		}
1714 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1715 			if (af && af != ifa->ifa_addr->sa_family)
1716 				continue;
1717 			if (prison_if(w->w_req->td->td_ucred,
1718 			    ifa->ifa_addr) != 0)
1719 				continue;
1720 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1721 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1722 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
1723 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1724 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1725 			if (error != 0)
1726 				goto done;
1727 			if (w->w_req && w->w_tmem) {
1728 				if (w->w_op == NET_RT_IFLISTL)
1729 					error = sysctl_iflist_ifaml(ifa, &info,
1730 					    w, len);
1731 				else
1732 					error = sysctl_iflist_ifam(ifa, &info,
1733 					    w, len);
1734 				if (error)
1735 					goto done;
1736 			}
1737 		}
1738 		IF_ADDR_RUNLOCK(ifp);
1739 		info.rti_info[RTAX_IFA] = NULL;
1740 		info.rti_info[RTAX_NETMASK] = NULL;
1741 		info.rti_info[RTAX_BRD] = NULL;
1742 	}
1743 done:
1744 	if (ifp != NULL)
1745 		IF_ADDR_RUNLOCK(ifp);
1746 	IFNET_RUNLOCK_NOSLEEP();
1747 	return (error);
1748 }
1749 
1750 static int
1751 sysctl_ifmalist(int af, struct walkarg *w)
1752 {
1753 	struct ifnet *ifp;
1754 	struct ifmultiaddr *ifma;
1755 	struct	rt_addrinfo info;
1756 	int	len, error = 0;
1757 	struct ifaddr *ifa;
1758 
1759 	bzero((caddr_t)&info, sizeof(info));
1760 	IFNET_RLOCK_NOSLEEP();
1761 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1762 		if (w->w_arg && w->w_arg != ifp->if_index)
1763 			continue;
1764 		ifa = ifp->if_addr;
1765 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1766 		IF_ADDR_RLOCK(ifp);
1767 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1768 			if (af && af != ifma->ifma_addr->sa_family)
1769 				continue;
1770 			if (prison_if(w->w_req->td->td_ucred,
1771 			    ifma->ifma_addr) != 0)
1772 				continue;
1773 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1774 			info.rti_info[RTAX_GATEWAY] =
1775 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1776 			    ifma->ifma_lladdr : NULL;
1777 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1778 			if (error != 0)
1779 				goto done;
1780 			if (w->w_req && w->w_tmem) {
1781 				struct ifma_msghdr *ifmam;
1782 
1783 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1784 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1785 				ifmam->ifmam_flags = 0;
1786 				ifmam->ifmam_addrs = info.rti_addrs;
1787 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1788 				if (error) {
1789 					IF_ADDR_RUNLOCK(ifp);
1790 					goto done;
1791 				}
1792 			}
1793 		}
1794 		IF_ADDR_RUNLOCK(ifp);
1795 	}
1796 done:
1797 	IFNET_RUNLOCK_NOSLEEP();
1798 	return (error);
1799 }
1800 
1801 static int
1802 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1803 {
1804 	int	*name = (int *)arg1;
1805 	u_int	namelen = arg2;
1806 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1807 	int	i, lim, error = EINVAL;
1808 	int	fib = 0;
1809 	u_char	af;
1810 	struct	walkarg w;
1811 
1812 	name ++;
1813 	namelen--;
1814 	if (req->newptr)
1815 		return (EPERM);
1816 	if (name[1] == NET_RT_DUMP) {
1817 		if (namelen == 3)
1818 			fib = req->td->td_proc->p_fibnum;
1819 		else if (namelen == 4)
1820 			fib = (name[3] == RT_ALL_FIBS) ?
1821 			    req->td->td_proc->p_fibnum : name[3];
1822 		else
1823 			return ((namelen < 3) ? EISDIR : ENOTDIR);
1824 		if (fib < 0 || fib >= rt_numfibs)
1825 			return (EINVAL);
1826 	} else if (namelen != 3)
1827 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1828 	af = name[0];
1829 	if (af > AF_MAX)
1830 		return (EINVAL);
1831 	bzero(&w, sizeof(w));
1832 	w.w_op = name[1];
1833 	w.w_arg = name[2];
1834 	w.w_req = req;
1835 
1836 	error = sysctl_wire_old_buffer(req, 0);
1837 	if (error)
1838 		return (error);
1839 
1840 	/*
1841 	 * Allocate reply buffer in advance.
1842 	 * All rtsock messages has maximum length of u_short.
1843 	 */
1844 	w.w_tmemsize = 65536;
1845 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
1846 
1847 	switch (w.w_op) {
1848 
1849 	case NET_RT_DUMP:
1850 	case NET_RT_FLAGS:
1851 		if (af == 0) {			/* dump all tables */
1852 			i = 1;
1853 			lim = AF_MAX;
1854 		} else				/* dump only one table */
1855 			i = lim = af;
1856 
1857 		/*
1858 		 * take care of llinfo entries, the caller must
1859 		 * specify an AF
1860 		 */
1861 		if (w.w_op == NET_RT_FLAGS &&
1862 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1863 			if (af != 0)
1864 				error = lltable_sysctl_dumparp(af, w.w_req);
1865 			else
1866 				error = EINVAL;
1867 			break;
1868 		}
1869 		/*
1870 		 * take care of routing entries
1871 		 */
1872 		for (error = 0; error == 0 && i <= lim; i++) {
1873 			rnh = rt_tables_get_rnh(fib, i);
1874 			if (rnh != NULL) {
1875 				RADIX_NODE_HEAD_RLOCK(rnh);
1876 			    	error = rnh->rnh_walktree(rnh,
1877 				    sysctl_dumpentry, &w);
1878 				RADIX_NODE_HEAD_RUNLOCK(rnh);
1879 			} else if (af != 0)
1880 				error = EAFNOSUPPORT;
1881 		}
1882 		break;
1883 
1884 	case NET_RT_IFLIST:
1885 	case NET_RT_IFLISTL:
1886 		error = sysctl_iflist(af, &w);
1887 		break;
1888 
1889 	case NET_RT_IFMALIST:
1890 		error = sysctl_ifmalist(af, &w);
1891 		break;
1892 	}
1893 
1894 	free(w.w_tmem, M_TEMP);
1895 	return (error);
1896 }
1897 
1898 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1899 
1900 /*
1901  * Definitions of protocols supported in the ROUTE domain.
1902  */
1903 
1904 static struct domain routedomain;		/* or at least forward */
1905 
1906 static struct protosw routesw[] = {
1907 {
1908 	.pr_type =		SOCK_RAW,
1909 	.pr_domain =		&routedomain,
1910 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1911 	.pr_output =		route_output,
1912 	.pr_ctlinput =		raw_ctlinput,
1913 	.pr_init =		raw_init,
1914 	.pr_usrreqs =		&route_usrreqs
1915 }
1916 };
1917 
1918 static struct domain routedomain = {
1919 	.dom_family =		PF_ROUTE,
1920 	.dom_name =		 "route",
1921 	.dom_protosw =		routesw,
1922 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1923 };
1924 
1925 VNET_DOMAIN_SET(route);
1926