xref: /freebsd/sys/net/rtsock.c (revision b59851e99c20f3a72c34bdf9919e3bf49b894e4e)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1988, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
32  * $FreeBSD$
33  */
34 #include "opt_ddb.h"
35 #include "opt_route.h"
36 #include "opt_inet.h"
37 #include "opt_inet6.h"
38 
39 #include <sys/param.h>
40 #include <sys/jail.h>
41 #include <sys/kernel.h>
42 #include <sys/domain.h>
43 #include <sys/lock.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/priv.h>
47 #include <sys/proc.h>
48 #include <sys/protosw.h>
49 #include <sys/rmlock.h>
50 #include <sys/rwlock.h>
51 #include <sys/signalvar.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/sysctl.h>
55 #include <sys/systm.h>
56 
57 #include <net/if.h>
58 #include <net/if_var.h>
59 #include <net/if_dl.h>
60 #include <net/if_llatbl.h>
61 #include <net/if_types.h>
62 #include <net/netisr.h>
63 #include <net/raw_cb.h>
64 #include <net/route.h>
65 #include <net/route/route_ctl.h>
66 #include <net/route/route_var.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/if_ether.h>
71 #include <netinet/ip_carp.h>
72 #ifdef INET6
73 #include <netinet6/in6_var.h>
74 #include <netinet6/ip6_var.h>
75 #include <netinet6/scope6_var.h>
76 #endif
77 #include <net/route/nhop.h>
78 
79 #ifdef COMPAT_FREEBSD32
80 #include <sys/mount.h>
81 #include <compat/freebsd32/freebsd32.h>
82 
83 struct if_msghdr32 {
84 	uint16_t ifm_msglen;
85 	uint8_t	ifm_version;
86 	uint8_t	ifm_type;
87 	int32_t	ifm_addrs;
88 	int32_t	ifm_flags;
89 	uint16_t ifm_index;
90 	uint16_t _ifm_spare1;
91 	struct	if_data ifm_data;
92 };
93 
94 struct if_msghdrl32 {
95 	uint16_t ifm_msglen;
96 	uint8_t	ifm_version;
97 	uint8_t	ifm_type;
98 	int32_t	ifm_addrs;
99 	int32_t	ifm_flags;
100 	uint16_t ifm_index;
101 	uint16_t _ifm_spare1;
102 	uint16_t ifm_len;
103 	uint16_t ifm_data_off;
104 	uint32_t _ifm_spare2;
105 	struct	if_data ifm_data;
106 };
107 
108 struct ifa_msghdrl32 {
109 	uint16_t ifam_msglen;
110 	uint8_t	ifam_version;
111 	uint8_t	ifam_type;
112 	int32_t	ifam_addrs;
113 	int32_t	ifam_flags;
114 	uint16_t ifam_index;
115 	uint16_t _ifam_spare1;
116 	uint16_t ifam_len;
117 	uint16_t ifam_data_off;
118 	int32_t	ifam_metric;
119 	struct	if_data ifam_data;
120 };
121 
122 #define SA_SIZE32(sa)						\
123     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
124 	sizeof(int)		:				\
125 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
126 
127 #endif /* COMPAT_FREEBSD32 */
128 
129 struct linear_buffer {
130 	char		*base;	/* Base allocated memory pointer */
131 	uint32_t	offset;	/* Currently used offset */
132 	uint32_t	size;	/* Total buffer size */
133 };
134 #define	SCRATCH_BUFFER_SIZE	1024
135 
136 #define	RTS_PID_PRINTF(_fmt, ...) \
137     printf("rtsock:%s(): PID %d: " _fmt "\n", __func__, curproc->p_pid, ## __VA_ARGS__)
138 
139 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
140 
141 /* NB: these are not modified */
142 static struct	sockaddr route_src = { 2, PF_ROUTE, };
143 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
144 
145 /* These are external hooks for CARP. */
146 int	(*carp_get_vhid_p)(struct ifaddr *);
147 
148 /*
149  * Used by rtsock/raw_input callback code to decide whether to filter the update
150  * notification to a socket bound to a particular FIB.
151  */
152 #define	RTS_FILTER_FIB	M_PROTO8
153 
154 typedef struct {
155 	int	ip_count;	/* attached w/ AF_INET */
156 	int	ip6_count;	/* attached w/ AF_INET6 */
157 	int	any_count;	/* total attached */
158 } route_cb_t;
159 VNET_DEFINE_STATIC(route_cb_t, route_cb);
160 #define	V_route_cb VNET(route_cb)
161 
162 struct mtx rtsock_mtx;
163 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
164 
165 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
166 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
167 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
168 
169 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
170 
171 struct walkarg {
172 	int	family;
173 	int	w_tmemsize;
174 	int	w_op, w_arg;
175 	caddr_t	w_tmem;
176 	struct sysctl_req *w_req;
177 	struct sockaddr *dst;
178 	struct sockaddr *mask;
179 };
180 
181 static void	rts_input(struct mbuf *m);
182 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
183 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
184 			struct walkarg *w, int *plen);
185 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
186 			struct rt_addrinfo *rtinfo);
187 static int	cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb);
188 static int	sysctl_dumpentry(struct rtentry *rt, void *vw);
189 static int	sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
190 			uint32_t weight, struct walkarg *w);
191 static int	sysctl_iflist(int af, struct walkarg *w);
192 static int	sysctl_ifmalist(int af, struct walkarg *w);
193 static int	route_output(struct mbuf *m, struct socket *so, ...);
194 static void	rt_getmetrics(const struct rtentry *rt,
195 			const struct nhop_object *nh, struct rt_metrics *out);
196 static void	rt_dispatch(struct mbuf *, sa_family_t);
197 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
198 			struct rt_msghdr *rtm, struct rib_cmd_info *rc);
199 static int	update_rtm_from_rc(struct rt_addrinfo *info,
200 			struct rt_msghdr **prtm, int alloc_len,
201 			struct rib_cmd_info *rc, struct nhop_object *nh);
202 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
203 			struct mbuf *m, sa_family_t saf, u_int fibnum,
204 			int rtm_errno);
205 static bool	can_export_rte(struct ucred *td_ucred, bool rt_is_host,
206 			const struct sockaddr *rt_dst);
207 
208 static struct netisr_handler rtsock_nh = {
209 	.nh_name = "rtsock",
210 	.nh_handler = rts_input,
211 	.nh_proto = NETISR_ROUTE,
212 	.nh_policy = NETISR_POLICY_SOURCE,
213 };
214 
215 static int
216 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
217 {
218 	int error, qlimit;
219 
220 	netisr_getqlimit(&rtsock_nh, &qlimit);
221 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
222         if (error || !req->newptr)
223                 return (error);
224 	if (qlimit < 1)
225 		return (EINVAL);
226 	return (netisr_setqlimit(&rtsock_nh, qlimit));
227 }
228 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
229     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
230     0, 0, sysctl_route_netisr_maxqlen, "I",
231     "maximum routing socket dispatch queue length");
232 
233 static void
234 vnet_rts_init(void)
235 {
236 	int tmp;
237 
238 	if (IS_DEFAULT_VNET(curvnet)) {
239 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
240 			rtsock_nh.nh_qlimit = tmp;
241 		netisr_register(&rtsock_nh);
242 	}
243 #ifdef VIMAGE
244 	 else
245 		netisr_register_vnet(&rtsock_nh);
246 #endif
247 }
248 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
249     vnet_rts_init, 0);
250 
251 #ifdef VIMAGE
252 static void
253 vnet_rts_uninit(void)
254 {
255 
256 	netisr_unregister_vnet(&rtsock_nh);
257 }
258 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
259     vnet_rts_uninit, 0);
260 #endif
261 
262 static int
263 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
264     struct rawcb *rp)
265 {
266 	int fibnum;
267 
268 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
269 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
270 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
271 
272 	/* No filtering requested. */
273 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
274 		return (0);
275 
276 	/* Check if it is a rts and the fib matches the one of the socket. */
277 	fibnum = M_GETFIB(m);
278 	if (proto->sp_family != PF_ROUTE ||
279 	    rp->rcb_socket == NULL ||
280 	    rp->rcb_socket->so_fibnum == fibnum)
281 		return (0);
282 
283 	/* Filtering requested and no match, the socket shall be skipped. */
284 	return (1);
285 }
286 
287 static void
288 rts_input(struct mbuf *m)
289 {
290 	struct sockproto route_proto;
291 	unsigned short *family;
292 	struct m_tag *tag;
293 
294 	route_proto.sp_family = PF_ROUTE;
295 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
296 	if (tag != NULL) {
297 		family = (unsigned short *)(tag + 1);
298 		route_proto.sp_protocol = *family;
299 		m_tag_delete(m, tag);
300 	} else
301 		route_proto.sp_protocol = 0;
302 
303 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
304 }
305 
306 /*
307  * It really doesn't make any sense at all for this code to share much
308  * with raw_usrreq.c, since its functionality is so restricted.  XXX
309  */
310 static void
311 rts_abort(struct socket *so)
312 {
313 
314 	raw_usrreqs.pru_abort(so);
315 }
316 
317 static void
318 rts_close(struct socket *so)
319 {
320 
321 	raw_usrreqs.pru_close(so);
322 }
323 
324 /* pru_accept is EOPNOTSUPP */
325 
326 static int
327 rts_attach(struct socket *so, int proto, struct thread *td)
328 {
329 	struct rawcb *rp;
330 	int error;
331 
332 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
333 
334 	/* XXX */
335 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
336 
337 	so->so_pcb = (caddr_t)rp;
338 	so->so_fibnum = td->td_proc->p_fibnum;
339 	error = raw_attach(so, proto);
340 	rp = sotorawcb(so);
341 	if (error) {
342 		so->so_pcb = NULL;
343 		free(rp, M_PCB);
344 		return error;
345 	}
346 	RTSOCK_LOCK();
347 	switch(rp->rcb_proto.sp_protocol) {
348 	case AF_INET:
349 		V_route_cb.ip_count++;
350 		break;
351 	case AF_INET6:
352 		V_route_cb.ip6_count++;
353 		break;
354 	}
355 	V_route_cb.any_count++;
356 	RTSOCK_UNLOCK();
357 	soisconnected(so);
358 	so->so_options |= SO_USELOOPBACK;
359 	return 0;
360 }
361 
362 static int
363 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
364 {
365 
366 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
367 }
368 
369 static int
370 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
371 {
372 
373 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
374 }
375 
376 /* pru_connect2 is EOPNOTSUPP */
377 /* pru_control is EOPNOTSUPP */
378 
379 static void
380 rts_detach(struct socket *so)
381 {
382 	struct rawcb *rp = sotorawcb(so);
383 
384 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
385 
386 	RTSOCK_LOCK();
387 	switch(rp->rcb_proto.sp_protocol) {
388 	case AF_INET:
389 		V_route_cb.ip_count--;
390 		break;
391 	case AF_INET6:
392 		V_route_cb.ip6_count--;
393 		break;
394 	}
395 	V_route_cb.any_count--;
396 	RTSOCK_UNLOCK();
397 	raw_usrreqs.pru_detach(so);
398 }
399 
400 static int
401 rts_disconnect(struct socket *so)
402 {
403 
404 	return (raw_usrreqs.pru_disconnect(so));
405 }
406 
407 /* pru_listen is EOPNOTSUPP */
408 
409 static int
410 rts_peeraddr(struct socket *so, struct sockaddr **nam)
411 {
412 
413 	return (raw_usrreqs.pru_peeraddr(so, nam));
414 }
415 
416 /* pru_rcvd is EOPNOTSUPP */
417 /* pru_rcvoob is EOPNOTSUPP */
418 
419 static int
420 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
421 	 struct mbuf *control, struct thread *td)
422 {
423 
424 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
425 }
426 
427 /* pru_sense is null */
428 
429 static int
430 rts_shutdown(struct socket *so)
431 {
432 
433 	return (raw_usrreqs.pru_shutdown(so));
434 }
435 
436 static int
437 rts_sockaddr(struct socket *so, struct sockaddr **nam)
438 {
439 
440 	return (raw_usrreqs.pru_sockaddr(so, nam));
441 }
442 
443 static struct pr_usrreqs route_usrreqs = {
444 	.pru_abort =		rts_abort,
445 	.pru_attach =		rts_attach,
446 	.pru_bind =		rts_bind,
447 	.pru_connect =		rts_connect,
448 	.pru_detach =		rts_detach,
449 	.pru_disconnect =	rts_disconnect,
450 	.pru_peeraddr =		rts_peeraddr,
451 	.pru_send =		rts_send,
452 	.pru_shutdown =		rts_shutdown,
453 	.pru_sockaddr =		rts_sockaddr,
454 	.pru_close =		rts_close,
455 };
456 
457 #ifndef _SOCKADDR_UNION_DEFINED
458 #define	_SOCKADDR_UNION_DEFINED
459 /*
460  * The union of all possible address formats we handle.
461  */
462 union sockaddr_union {
463 	struct sockaddr		sa;
464 	struct sockaddr_in	sin;
465 	struct sockaddr_in6	sin6;
466 };
467 #endif /* _SOCKADDR_UNION_DEFINED */
468 
469 static int
470 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
471     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
472 {
473 #if defined(INET) || defined(INET6)
474 	struct epoch_tracker et;
475 #endif
476 
477 	/* First, see if the returned address is part of the jail. */
478 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
479 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
480 		return (0);
481 	}
482 
483 	switch (info->rti_info[RTAX_DST]->sa_family) {
484 #ifdef INET
485 	case AF_INET:
486 	{
487 		struct in_addr ia;
488 		struct ifaddr *ifa;
489 		int found;
490 
491 		found = 0;
492 		/*
493 		 * Try to find an address on the given outgoing interface
494 		 * that belongs to the jail.
495 		 */
496 		NET_EPOCH_ENTER(et);
497 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
498 			struct sockaddr *sa;
499 			sa = ifa->ifa_addr;
500 			if (sa->sa_family != AF_INET)
501 				continue;
502 			ia = ((struct sockaddr_in *)sa)->sin_addr;
503 			if (prison_check_ip4(cred, &ia) == 0) {
504 				found = 1;
505 				break;
506 			}
507 		}
508 		NET_EPOCH_EXIT(et);
509 		if (!found) {
510 			/*
511 			 * As a last resort return the 'default' jail address.
512 			 */
513 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
514 			    sin_addr;
515 			if (prison_get_ip4(cred, &ia) != 0)
516 				return (ESRCH);
517 		}
518 		bzero(&saun->sin, sizeof(struct sockaddr_in));
519 		saun->sin.sin_len = sizeof(struct sockaddr_in);
520 		saun->sin.sin_family = AF_INET;
521 		saun->sin.sin_addr.s_addr = ia.s_addr;
522 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
523 		break;
524 	}
525 #endif
526 #ifdef INET6
527 	case AF_INET6:
528 	{
529 		struct in6_addr ia6;
530 		struct ifaddr *ifa;
531 		int found;
532 
533 		found = 0;
534 		/*
535 		 * Try to find an address on the given outgoing interface
536 		 * that belongs to the jail.
537 		 */
538 		NET_EPOCH_ENTER(et);
539 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
540 			struct sockaddr *sa;
541 			sa = ifa->ifa_addr;
542 			if (sa->sa_family != AF_INET6)
543 				continue;
544 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
545 			    &ia6, sizeof(struct in6_addr));
546 			if (prison_check_ip6(cred, &ia6) == 0) {
547 				found = 1;
548 				break;
549 			}
550 		}
551 		NET_EPOCH_EXIT(et);
552 		if (!found) {
553 			/*
554 			 * As a last resort return the 'default' jail address.
555 			 */
556 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
557 			    sin6_addr;
558 			if (prison_get_ip6(cred, &ia6) != 0)
559 				return (ESRCH);
560 		}
561 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
562 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
563 		saun->sin6.sin6_family = AF_INET6;
564 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
565 		if (sa6_recoverscope(&saun->sin6) != 0)
566 			return (ESRCH);
567 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
568 		break;
569 	}
570 #endif
571 	default:
572 		return (ESRCH);
573 	}
574 	return (0);
575 }
576 
577 static int
578 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
579 {
580 	struct ifaddr *ifa;
581 	sa_family_t saf;
582 
583 	if (V_loif == NULL) {
584 		RTS_PID_PRINTF("Unable to add blackhole/reject nhop without loopback");
585 		return (ENOTSUP);
586 	}
587 	info->rti_ifp = V_loif;
588 
589 	saf = info->rti_info[RTAX_DST]->sa_family;
590 
591 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
592 		if (ifa->ifa_addr->sa_family == saf) {
593 			info->rti_ifa = ifa;
594 			break;
595 		}
596 	}
597 	if (info->rti_ifa == NULL)
598 		return (ENOTSUP);
599 
600 	bzero(saun, sizeof(union sockaddr_union));
601 	switch (saf) {
602 #ifdef INET
603 	case AF_INET:
604 		saun->sin.sin_family = AF_INET;
605 		saun->sin.sin_len = sizeof(struct sockaddr_in);
606 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
607 		break;
608 #endif
609 #ifdef INET6
610 	case AF_INET6:
611 		saun->sin6.sin6_family = AF_INET6;
612 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
613 		saun->sin6.sin6_addr = in6addr_loopback;
614 		break;
615 #endif
616 	default:
617 		return (ENOTSUP);
618 	}
619 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
620 	info->rti_flags |= RTF_GATEWAY;
621 
622 	return (0);
623 }
624 
625 /*
626  * Fills in @info based on userland-provided @rtm message.
627  *
628  * Returns 0 on success.
629  */
630 static int
631 fill_addrinfo(struct rt_msghdr *rtm, int len, struct linear_buffer *lb, u_int fibnum,
632     struct rt_addrinfo *info)
633 {
634 	int error;
635 	sa_family_t saf;
636 
637 	rtm->rtm_pid = curproc->p_pid;
638 	info->rti_addrs = rtm->rtm_addrs;
639 
640 	info->rti_mflags = rtm->rtm_inits;
641 	info->rti_rmx = &rtm->rtm_rmx;
642 
643 	/*
644 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
645 	 * link-local address because rtrequest requires addresses with
646 	 * embedded scope id.
647 	 */
648 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
649 		return (EINVAL);
650 
651 	info->rti_flags = rtm->rtm_flags;
652 	error = cleanup_xaddrs(info, lb);
653 	if (error != 0)
654 		return (error);
655 	saf = info->rti_info[RTAX_DST]->sa_family;
656 	/*
657 	 * Verify that the caller has the appropriate privilege; RTM_GET
658 	 * is the only operation the non-superuser is allowed.
659 	 */
660 	if (rtm->rtm_type != RTM_GET) {
661 		error = priv_check(curthread, PRIV_NET_ROUTE);
662 		if (error != 0)
663 			return (error);
664 	}
665 
666 	/*
667 	 * The given gateway address may be an interface address.
668 	 * For example, issuing a "route change" command on a route
669 	 * entry that was created from a tunnel, and the gateway
670 	 * address given is the local end point. In this case the
671 	 * RTF_GATEWAY flag must be cleared or the destination will
672 	 * not be reachable even though there is no error message.
673 	 */
674 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
675 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
676 		struct rt_addrinfo ginfo;
677 		struct sockaddr *gdst;
678 		struct sockaddr_storage ss;
679 
680 		bzero(&ginfo, sizeof(ginfo));
681 		bzero(&ss, sizeof(ss));
682 		ss.ss_len = sizeof(ss);
683 
684 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
685 		gdst = info->rti_info[RTAX_GATEWAY];
686 
687 		/*
688 		 * A host route through the loopback interface is
689 		 * installed for each interface adddress. In pre 8.0
690 		 * releases the interface address of a PPP link type
691 		 * is not reachable locally. This behavior is fixed as
692 		 * part of the new L2/L3 redesign and rewrite work. The
693 		 * signature of this interface address route is the
694 		 * AF_LINK sa_family type of the gateway, and the
695 		 * rt_ifp has the IFF_LOOPBACK flag set.
696 		 */
697 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
698 			if (ss.ss_family == AF_LINK &&
699 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
700 				info->rti_flags &= ~RTF_GATEWAY;
701 				info->rti_flags |= RTF_GWFLAG_COMPAT;
702 			}
703 			rib_free_info(&ginfo);
704 		}
705 	}
706 
707 	return (0);
708 }
709 
710 static struct nhop_object *
711 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
712 {
713 	if (!NH_IS_NHGRP(nh))
714 		return (nh);
715 #ifdef ROUTE_MPATH
716 	struct weightened_nhop *wn;
717 	uint32_t num_nhops;
718 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
719 	if (gw == NULL)
720 		return (wn[0].nh);
721 	for (int i = 0; i < num_nhops; i++) {
722 		if (match_nhop_gw(wn[i].nh, gw))
723 			return (wn[i].nh);
724 	}
725 #endif
726 	return (NULL);
727 }
728 
729 /*
730  * Handles RTM_GET message from routing socket, returning matching rt.
731  *
732  * Returns:
733  * 0 on success, with locked and referenced matching rt in @rt_nrt
734  * errno of failure
735  */
736 static int
737 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
738     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
739 {
740 	RIB_RLOCK_TRACKER;
741 	struct rib_head *rnh;
742 	struct nhop_object *nh;
743 	sa_family_t saf;
744 
745 	saf = info->rti_info[RTAX_DST]->sa_family;
746 
747 	rnh = rt_tables_get_rnh(fibnum, saf);
748 	if (rnh == NULL)
749 		return (EAFNOSUPPORT);
750 
751 	RIB_RLOCK(rnh);
752 
753 	/*
754 	 * By (implicit) convention host route (one without netmask)
755 	 * means longest-prefix-match request and the route with netmask
756 	 * means exact-match lookup.
757 	 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128
758 	 * prefixes, use original data to check for the netmask presence.
759 	 */
760 	if ((rtm->rtm_addrs & RTA_NETMASK) == 0) {
761 		/*
762 		 * Provide longest prefix match for
763 		 * address lookup (no mask).
764 		 * 'route -n get addr'
765 		 */
766 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
767 		    info->rti_info[RTAX_DST], &rnh->head);
768 	} else
769 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
770 		    info->rti_info[RTAX_DST],
771 		    info->rti_info[RTAX_NETMASK], &rnh->head);
772 
773 	if (rc->rc_rt == NULL) {
774 		RIB_RUNLOCK(rnh);
775 		return (ESRCH);
776 	}
777 
778 	nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
779 	if (nh == NULL) {
780 		RIB_RUNLOCK(rnh);
781 		return (ESRCH);
782 	}
783 	/*
784 	 * If performing proxied L2 entry insertion, and
785 	 * the actual PPP host entry is found, perform
786 	 * another search to retrieve the prefix route of
787 	 * the local end point of the PPP link.
788 	 * TODO: move this logic to userland.
789 	 */
790 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
791 		struct sockaddr laddr;
792 
793 		if (nh->nh_ifp != NULL &&
794 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
795 			struct ifaddr *ifa;
796 
797 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
798 					RT_ALL_FIBS);
799 			if (ifa != NULL)
800 				rt_maskedcopy(ifa->ifa_addr,
801 					      &laddr,
802 					      ifa->ifa_netmask);
803 		} else
804 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
805 				      &laddr,
806 				      nh->nh_ifa->ifa_netmask);
807 		/*
808 		 * refactor rt and no lock operation necessary
809 		 */
810 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
811 		    &rnh->head);
812 		if (rc->rc_rt == NULL) {
813 			RIB_RUNLOCK(rnh);
814 			return (ESRCH);
815 		}
816 		nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
817 		if (nh == NULL) {
818 			RIB_RUNLOCK(rnh);
819 			return (ESRCH);
820 		}
821 	}
822 	rc->rc_nh_new = nh;
823 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
824 	RIB_RUNLOCK(rnh);
825 
826 	return (0);
827 }
828 
829 static void
830 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
831 {
832 #ifdef INET
833 	if (family == AF_INET) {
834 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
835 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
836 
837 		bzero(dst4, sizeof(struct sockaddr_in));
838 		bzero(mask4, sizeof(struct sockaddr_in));
839 
840 		dst4->sin_family = AF_INET;
841 		dst4->sin_len = sizeof(struct sockaddr_in);
842 		mask4->sin_family = AF_INET;
843 		mask4->sin_len = sizeof(struct sockaddr_in);
844 	}
845 #endif
846 #ifdef INET6
847 	if (family == AF_INET6) {
848 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
849 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
850 
851 		bzero(dst6, sizeof(struct sockaddr_in6));
852 		bzero(mask6, sizeof(struct sockaddr_in6));
853 
854 		dst6->sin6_family = AF_INET6;
855 		dst6->sin6_len = sizeof(struct sockaddr_in6);
856 		mask6->sin6_family = AF_INET6;
857 		mask6->sin6_len = sizeof(struct sockaddr_in6);
858 	}
859 #endif
860 }
861 
862 static void
863 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
864     struct sockaddr *mask)
865 {
866 #ifdef INET
867 	if (dst->sa_family == AF_INET) {
868 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
869 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
870 		uint32_t scopeid = 0;
871 		rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
872 		    &scopeid);
873 		return;
874 	}
875 #endif
876 #ifdef INET6
877 	if (dst->sa_family == AF_INET6) {
878 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
879 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
880 		uint32_t scopeid = 0;
881 		rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
882 		    &mask6->sin6_addr, &scopeid);
883 		dst6->sin6_scope_id = scopeid;
884 		return;
885 	}
886 #endif
887 }
888 
889 static int
890 update_rtm_from_info(struct rt_addrinfo *info, struct rt_msghdr **prtm,
891     int alloc_len)
892 {
893 	struct rt_msghdr *rtm, *orig_rtm = NULL;
894 	struct walkarg w;
895 	int len;
896 
897 	rtm = *prtm;
898 	/* Check if we need to realloc storage */
899 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
900 	if (len > alloc_len) {
901 		struct rt_msghdr *tmp_rtm;
902 
903 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
904 		if (tmp_rtm == NULL)
905 			return (ENOBUFS);
906 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
907 		orig_rtm = rtm;
908 		rtm = tmp_rtm;
909 		alloc_len = len;
910 
911 		/*
912 		 * Delay freeing original rtm as info contains
913 		 * data referencing it.
914 		 */
915 	}
916 
917 	w.w_tmem = (caddr_t)rtm;
918 	w.w_tmemsize = alloc_len;
919 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
920 	rtm->rtm_addrs = info->rti_addrs;
921 
922 	if (orig_rtm != NULL)
923 		free(orig_rtm, M_TEMP);
924 	*prtm = rtm;
925 	return (0);
926 }
927 
928 
929 /*
930  * Update sockaddrs, flags, etc in @prtm based on @rc data.
931  * rtm can be reallocated.
932  *
933  * Returns 0 on success, along with pointer to (potentially reallocated)
934  *  rtm.
935  *
936  */
937 static int
938 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
939     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
940 {
941 	union sockaddr_union saun;
942 	struct rt_msghdr *rtm;
943 	struct ifnet *ifp;
944 	int error;
945 
946 	rtm = *prtm;
947 	union sockaddr_union sa_dst, sa_mask;
948 	int family = info->rti_info[RTAX_DST]->sa_family;
949 	init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
950 	export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
951 
952 	info->rti_info[RTAX_DST] = &sa_dst.sa;
953 	info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
954 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
955 	info->rti_info[RTAX_GENMASK] = 0;
956 	ifp = nh->nh_ifp;
957 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
958 		if (ifp) {
959 			info->rti_info[RTAX_IFP] =
960 			    ifp->if_addr->ifa_addr;
961 			error = rtm_get_jailed(info, ifp, nh,
962 			    &saun, curthread->td_ucred);
963 			if (error != 0)
964 				return (error);
965 			if (ifp->if_flags & IFF_POINTOPOINT)
966 				info->rti_info[RTAX_BRD] =
967 				    nh->nh_ifa->ifa_dstaddr;
968 			rtm->rtm_index = ifp->if_index;
969 		} else {
970 			info->rti_info[RTAX_IFP] = NULL;
971 			info->rti_info[RTAX_IFA] = NULL;
972 		}
973 	} else if (ifp != NULL)
974 		rtm->rtm_index = ifp->if_index;
975 
976 	if ((error = update_rtm_from_info(info, prtm, alloc_len)) != 0)
977 		return (error);
978 
979 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
980 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
981 		rtm->rtm_flags = RTF_GATEWAY |
982 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
983 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
984 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
985 
986 	return (0);
987 }
988 
989 #ifdef ROUTE_MPATH
990 static void
991 save_del_notification(struct rib_cmd_info *rc, void *_cbdata)
992 {
993 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
994 
995 	if (rc->rc_cmd == RTM_DELETE)
996 		*rc_new = *rc;
997 }
998 
999 static void
1000 save_add_notification(struct rib_cmd_info *rc, void *_cbdata)
1001 {
1002 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1003 
1004 	if (rc->rc_cmd == RTM_ADD)
1005 		*rc_new = *rc;
1006 }
1007 #endif
1008 
1009 static struct sockaddr *
1010 alloc_sockaddr_aligned(struct linear_buffer *lb, int len)
1011 {
1012 	len = roundup2(len, sizeof(uint64_t));
1013 	if (lb->offset + len > lb->size)
1014 		return (NULL);
1015 	struct sockaddr *sa = (struct sockaddr *)(lb->base + lb->offset);
1016 	lb->offset += len;
1017 	return (sa);
1018 }
1019 
1020 /*ARGSUSED*/
1021 static int
1022 route_output(struct mbuf *m, struct socket *so, ...)
1023 {
1024 	struct rt_msghdr *rtm = NULL;
1025 	struct rtentry *rt = NULL;
1026 	struct rt_addrinfo info;
1027 	struct epoch_tracker et;
1028 #ifdef INET6
1029 	struct sockaddr_storage ss;
1030 	struct sockaddr_in6 *sin6;
1031 	int i, rti_need_deembed = 0;
1032 #endif
1033 	int alloc_len = 0, len, error = 0, fibnum;
1034 	sa_family_t saf = AF_UNSPEC;
1035 	struct rib_cmd_info rc;
1036 	struct nhop_object *nh;
1037 
1038 	fibnum = so->so_fibnum;
1039 #define senderr(e) { error = e; goto flush;}
1040 	if (m == NULL || ((m->m_len < sizeof(long)) &&
1041 		       (m = m_pullup(m, sizeof(long))) == NULL))
1042 		return (ENOBUFS);
1043 	if ((m->m_flags & M_PKTHDR) == 0)
1044 		panic("route_output");
1045 	NET_EPOCH_ENTER(et);
1046 	len = m->m_pkthdr.len;
1047 	if (len < sizeof(*rtm) ||
1048 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1049 		senderr(EINVAL);
1050 
1051 	/*
1052 	 * Most of current messages are in range 200-240 bytes,
1053 	 * minimize possible re-allocation on reply using larger size
1054 	 * buffer aligned on 1k boundaty.
1055 	 */
1056 	alloc_len = roundup2(len, 1024);
1057 	int total_len = alloc_len + SCRATCH_BUFFER_SIZE;
1058 	if ((rtm = malloc(total_len, M_TEMP, M_NOWAIT)) == NULL)
1059 		senderr(ENOBUFS);
1060 
1061 	m_copydata(m, 0, len, (caddr_t)rtm);
1062 	bzero(&info, sizeof(info));
1063 	nh = NULL;
1064 	struct linear_buffer lb = {
1065 		.base = (char *)rtm + alloc_len,
1066 		.size = SCRATCH_BUFFER_SIZE,
1067 	};
1068 
1069 	if (rtm->rtm_version != RTM_VERSION) {
1070 		/* Do not touch message since format is unknown */
1071 		free(rtm, M_TEMP);
1072 		rtm = NULL;
1073 		senderr(EPROTONOSUPPORT);
1074 	}
1075 
1076 	/*
1077 	 * Starting from here, it is possible
1078 	 * to alter original message and insert
1079 	 * caller PID and error value.
1080 	 */
1081 
1082 	if ((error = fill_addrinfo(rtm, len, &lb, fibnum, &info)) != 0) {
1083 		senderr(error);
1084 	}
1085 	/* fill_addringo() embeds scope into IPv6 addresses */
1086 #ifdef INET6
1087 	rti_need_deembed = 1;
1088 #endif
1089 
1090 	saf = info.rti_info[RTAX_DST]->sa_family;
1091 
1092 	/* support for new ARP code */
1093 	if (rtm->rtm_flags & RTF_LLDATA) {
1094 		error = lla_rt_output(rtm, &info);
1095 		goto flush;
1096 	}
1097 
1098 	union sockaddr_union gw_saun;
1099 	int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1100 	if (blackhole_flags != 0) {
1101 		if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1102 			error = fill_blackholeinfo(&info, &gw_saun);
1103 		else
1104 			error = EINVAL;
1105 		if (error != 0)
1106 			senderr(error);
1107 	}
1108 
1109 	switch (rtm->rtm_type) {
1110 	case RTM_ADD:
1111 	case RTM_CHANGE:
1112 		if (rtm->rtm_type == RTM_ADD) {
1113 			if (info.rti_info[RTAX_GATEWAY] == NULL)
1114 				senderr(EINVAL);
1115 		}
1116 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1117 		if (error == 0) {
1118 #ifdef ROUTE_MPATH
1119 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
1120 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1121 				struct rib_cmd_info rc_simple = {};
1122 				rib_decompose_notification(&rc,
1123 				    save_add_notification, (void *)&rc_simple);
1124 				rc = rc_simple;
1125 			}
1126 #endif
1127 			nh = rc.rc_nh_new;
1128 			rtm->rtm_index = nh->nh_ifp->if_index;
1129 			rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1130 		}
1131 		break;
1132 
1133 	case RTM_DELETE:
1134 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1135 		if (error == 0) {
1136 #ifdef ROUTE_MPATH
1137 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1138 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1139 				struct rib_cmd_info rc_simple = {};
1140 				rib_decompose_notification(&rc,
1141 				    save_del_notification, (void *)&rc_simple);
1142 				rc = rc_simple;
1143 			}
1144 #endif
1145 			nh = rc.rc_nh_old;
1146 		}
1147 		break;
1148 
1149 	case RTM_GET:
1150 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1151 		if (error != 0)
1152 			senderr(error);
1153 		nh = rc.rc_nh_new;
1154 
1155 		if (!can_export_rte(curthread->td_ucred,
1156 		    info.rti_info[RTAX_NETMASK] == NULL,
1157 		    info.rti_info[RTAX_DST])) {
1158 			senderr(ESRCH);
1159 		}
1160 		break;
1161 
1162 	default:
1163 		senderr(EOPNOTSUPP);
1164 	}
1165 
1166 	if (error == 0) {
1167 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1168 		/*
1169 		 * Note that some sockaddr pointers may have changed to
1170 		 * point to memory outsize @rtm. Some may be pointing
1171 		 * to the on-stack variables.
1172 		 * Given that, any pointer in @info CANNOT BE USED.
1173 		 */
1174 
1175 		/*
1176 		 * scopeid deembedding has been performed while
1177 		 * writing updated rtm in rtsock_msg_buffer().
1178 		 * With that in mind, skip deembedding procedure below.
1179 		 */
1180 #ifdef INET6
1181 		rti_need_deembed = 0;
1182 #endif
1183 	}
1184 
1185 flush:
1186 	NET_EPOCH_EXIT(et);
1187 	rt = NULL;
1188 
1189 #ifdef INET6
1190 	if (rtm != NULL) {
1191 		if (rti_need_deembed) {
1192 			/* sin6_scope_id is recovered before sending rtm. */
1193 			sin6 = (struct sockaddr_in6 *)&ss;
1194 			for (i = 0; i < RTAX_MAX; i++) {
1195 				if (info.rti_info[i] == NULL)
1196 					continue;
1197 				if (info.rti_info[i]->sa_family != AF_INET6)
1198 					continue;
1199 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1200 				if (sa6_recoverscope(sin6) == 0)
1201 					bcopy(sin6, info.rti_info[i],
1202 						    sizeof(*sin6));
1203 			}
1204 			if (update_rtm_from_info(&info, &rtm, alloc_len) != 0) {
1205 				if (error != 0)
1206 					error = ENOBUFS;
1207 			}
1208 		}
1209 	}
1210 #endif
1211 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1212 
1213 	return (error);
1214 }
1215 
1216 /*
1217  * Sends the prepared reply message in @rtm to all rtsock clients.
1218  * Frees @m and @rtm.
1219  *
1220  */
1221 static void
1222 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1223     sa_family_t saf, u_int fibnum, int rtm_errno)
1224 {
1225 	struct rawcb *rp = NULL;
1226 
1227 	/*
1228 	 * Check to see if we don't want our own messages.
1229 	 */
1230 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1231 		if (V_route_cb.any_count <= 1) {
1232 			if (rtm != NULL)
1233 				free(rtm, M_TEMP);
1234 			m_freem(m);
1235 			return;
1236 		}
1237 		/* There is another listener, so construct message */
1238 		rp = sotorawcb(so);
1239 	}
1240 
1241 	if (rtm != NULL) {
1242 		if (rtm_errno!= 0)
1243 			rtm->rtm_errno = rtm_errno;
1244 		else
1245 			rtm->rtm_flags |= RTF_DONE;
1246 
1247 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1248 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1249 			m_freem(m);
1250 			m = NULL;
1251 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1252 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1253 
1254 		free(rtm, M_TEMP);
1255 	}
1256 	if (m != NULL) {
1257 		M_SETFIB(m, fibnum);
1258 		m->m_flags |= RTS_FILTER_FIB;
1259 		if (rp) {
1260 			/*
1261 			 * XXX insure we don't get a copy by
1262 			 * invalidating our protocol
1263 			 */
1264 			unsigned short family = rp->rcb_proto.sp_family;
1265 			rp->rcb_proto.sp_family = 0;
1266 			rt_dispatch(m, saf);
1267 			rp->rcb_proto.sp_family = family;
1268 		} else
1269 			rt_dispatch(m, saf);
1270 	}
1271 }
1272 
1273 static void
1274 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1275     struct rt_metrics *out)
1276 {
1277 
1278 	bzero(out, sizeof(*out));
1279 	out->rmx_mtu = nh->nh_mtu;
1280 	out->rmx_weight = rt->rt_weight;
1281 	out->rmx_nhidx = nhop_get_idx(nh);
1282 	/* Kernel -> userland timebase conversion. */
1283 	out->rmx_expire = rt->rt_expire ?
1284 	    rt->rt_expire - time_uptime + time_second : 0;
1285 }
1286 
1287 /*
1288  * Extract the addresses of the passed sockaddrs.
1289  * Do a little sanity checking so as to avoid bad memory references.
1290  * This data is derived straight from userland.
1291  */
1292 static int
1293 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1294 {
1295 	struct sockaddr *sa;
1296 	int i;
1297 
1298 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1299 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1300 			continue;
1301 		sa = (struct sockaddr *)cp;
1302 		/*
1303 		 * It won't fit.
1304 		 */
1305 		if (cp + sa->sa_len > cplim)
1306 			return (EINVAL);
1307 		/*
1308 		 * there are no more.. quit now
1309 		 * If there are more bits, they are in error.
1310 		 * I've seen this. route(1) can evidently generate these.
1311 		 * This causes kernel to core dump.
1312 		 * for compatibility, If we see this, point to a safe address.
1313 		 */
1314 		if (sa->sa_len == 0) {
1315 			rtinfo->rti_info[i] = &sa_zero;
1316 			return (0); /* should be EINVAL but for compat */
1317 		}
1318 		/* accept it */
1319 #ifdef INET6
1320 		if (sa->sa_family == AF_INET6)
1321 			sa6_embedscope((struct sockaddr_in6 *)sa,
1322 			    V_ip6_use_defzone);
1323 #endif
1324 		rtinfo->rti_info[i] = sa;
1325 		cp += SA_SIZE(sa);
1326 	}
1327 	return (0);
1328 }
1329 
1330 #ifdef INET
1331 static inline void
1332 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1333 {
1334 
1335 	const struct sockaddr_in nsin = {
1336 		.sin_family = AF_INET,
1337 		.sin_len = sizeof(struct sockaddr_in),
1338 		.sin_addr = addr,
1339 	};
1340 	*sin = nsin;
1341 }
1342 #endif
1343 
1344 #ifdef INET6
1345 static inline void
1346 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1347     uint32_t scopeid)
1348 {
1349 
1350 	const struct sockaddr_in6 nsin6 = {
1351 		.sin6_family = AF_INET6,
1352 		.sin6_len = sizeof(struct sockaddr_in6),
1353 		.sin6_addr = *addr6,
1354 		.sin6_scope_id = scopeid,
1355 	};
1356 	*sin6 = nsin6;
1357 }
1358 #endif
1359 
1360 /*
1361  * Checks if gateway is suitable for lltable operations.
1362  * Lltable code requires AF_LINK gateway with ifindex
1363  *  and mac address specified.
1364  * Returns 0 on success.
1365  */
1366 static int
1367 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1368 {
1369 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1370 
1371 	if (sdl->sdl_family != AF_LINK)
1372 		return (EINVAL);
1373 
1374 	if (sdl->sdl_index == 0)
1375 		return (EINVAL);
1376 
1377 	if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len)
1378 		return (EINVAL);
1379 
1380 	return (0);
1381 }
1382 
1383 static int
1384 cleanup_xaddrs_gateway(struct rt_addrinfo *info, struct linear_buffer *lb)
1385 {
1386 	struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1387 	struct sockaddr *sa;
1388 
1389 	if (info->rti_flags & RTF_LLDATA)
1390 		return (cleanup_xaddrs_lladdr(info));
1391 
1392 	switch (gw->sa_family) {
1393 #ifdef INET
1394 	case AF_INET:
1395 		{
1396 			struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw;
1397 
1398 			/* Ensure reads do not go beyoud SA boundary */
1399 			if (SA_SIZE(gw) < offsetof(struct sockaddr_in, sin_zero)) {
1400 				RTS_PID_PRINTF("gateway sin_len too small: %d", gw->sa_len);
1401 				return (EINVAL);
1402 			}
1403 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_in));
1404 			if (sa == NULL)
1405 				return (ENOBUFS);
1406 			fill_sockaddr_inet((struct sockaddr_in *)sa, gw_sin->sin_addr);
1407 			info->rti_info[RTAX_GATEWAY] = sa;
1408 		}
1409 		break;
1410 #endif
1411 #ifdef INET6
1412 	case AF_INET6:
1413 		{
1414 			struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1415 			if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1416 				RTS_PID_PRINTF("gateway sin6_len too small: %d", gw->sa_len);
1417 				return (EINVAL);
1418 			}
1419 			fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1420 			break;
1421 		}
1422 #endif
1423 	case AF_LINK:
1424 		{
1425 			struct sockaddr_dl *gw_sdl;
1426 
1427 			size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data);
1428 			gw_sdl = (struct sockaddr_dl *)gw;
1429 			if (gw_sdl->sdl_len < sdl_min_len) {
1430 				RTS_PID_PRINTF("gateway sdl_len too small: %d", gw_sdl->sdl_len);
1431 				return (EINVAL);
1432 			}
1433 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_dl_short));
1434 			if (sa == NULL)
1435 				return (ENOBUFS);
1436 
1437 			const struct sockaddr_dl_short sdl = {
1438 				.sdl_family = AF_LINK,
1439 				.sdl_len = sizeof(struct sockaddr_dl_short),
1440 				.sdl_index = gw_sdl->sdl_index,
1441 			};
1442 			*((struct sockaddr_dl_short *)sa) = sdl;
1443 			info->rti_info[RTAX_GATEWAY] = sa;
1444 			break;
1445 		}
1446 	}
1447 
1448 	return (0);
1449 }
1450 
1451 static void
1452 remove_netmask(struct rt_addrinfo *info)
1453 {
1454 	info->rti_info[RTAX_NETMASK] = NULL;
1455 	info->rti_flags |= RTF_HOST;
1456 	info->rti_addrs &= ~RTA_NETMASK;
1457 }
1458 
1459 #ifdef INET
1460 static int
1461 cleanup_xaddrs_inet(struct rt_addrinfo *info, struct linear_buffer *lb)
1462 {
1463 	struct sockaddr_in *dst_sa, *mask_sa;
1464 	const int sa_len = sizeof(struct sockaddr_in);
1465 	struct in_addr dst, mask;
1466 
1467 	/* Check & fixup dst/netmask combination first */
1468 	dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1469 	mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1470 
1471 	/* Ensure reads do not go beyound the buffer size */
1472 	if (SA_SIZE(dst_sa) < offsetof(struct sockaddr_in, sin_zero))
1473 		return (EINVAL);
1474 
1475 	if ((mask_sa != NULL) && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1476 		/*
1477 		 * Some older routing software encode mask length into the
1478 		 * sin_len, thus resulting in "truncated" sockaddr.
1479 		 */
1480 		int len = mask_sa->sin_len - offsetof(struct sockaddr_in, sin_addr);
1481 		if (len >= 0) {
1482 			mask.s_addr = 0;
1483 			if (len > sizeof(struct in_addr))
1484 				len = sizeof(struct in_addr);
1485 			memcpy(&mask, &mask_sa->sin_addr, len);
1486 		} else {
1487 			RTS_PID_PRINTF("prefix mask sin_len too small: %d", mask_sa->sin_len);
1488 			return (EINVAL);
1489 		}
1490 	} else
1491 		mask.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST;
1492 
1493 	dst.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr));
1494 
1495 	/* Construct new "clean" dst/mask sockaddresses */
1496 	if ((dst_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1497 		return (ENOBUFS);
1498 	fill_sockaddr_inet(dst_sa, dst);
1499 	info->rti_info[RTAX_DST] = (struct sockaddr *)dst_sa;
1500 
1501 	if (mask.s_addr != INADDR_BROADCAST) {
1502 		if ((mask_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1503 			return (ENOBUFS);
1504 		fill_sockaddr_inet(mask_sa, mask);
1505 		info->rti_info[RTAX_NETMASK] = (struct sockaddr *)mask_sa;
1506 		info->rti_flags &= ~RTF_HOST;
1507 	} else
1508 		remove_netmask(info);
1509 
1510 	/* Check gateway */
1511 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1512 		return (cleanup_xaddrs_gateway(info, lb));
1513 
1514 	return (0);
1515 }
1516 #endif
1517 
1518 #ifdef INET6
1519 static int
1520 cleanup_xaddrs_inet6(struct rt_addrinfo *info, struct linear_buffer *lb)
1521 {
1522 	struct sockaddr *sa;
1523 	struct sockaddr_in6 *dst_sa, *mask_sa;
1524 	struct in6_addr mask, *dst;
1525 	const int sa_len = sizeof(struct sockaddr_in6);
1526 
1527 	/* Check & fixup dst/netmask combination first */
1528 	dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1529 	mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1530 
1531 	if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1532 		RTS_PID_PRINTF("prefix dst sin6_len too small: %d", dst_sa->sin6_len);
1533 		return (EINVAL);
1534 	}
1535 
1536 	if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1537 		/*
1538 		 * Some older routing software encode mask length into the
1539 		 * sin6_len, thus resulting in "truncated" sockaddr.
1540 		 */
1541 		int len = mask_sa->sin6_len - offsetof(struct sockaddr_in6, sin6_addr);
1542 		if (len >= 0) {
1543 			bzero(&mask, sizeof(mask));
1544 			if (len > sizeof(struct in6_addr))
1545 				len = sizeof(struct in6_addr);
1546 			memcpy(&mask, &mask_sa->sin6_addr, len);
1547 		} else {
1548 			RTS_PID_PRINTF("rtsock: prefix mask sin6_len too small: %d", mask_sa->sin6_len);
1549 			return (EINVAL);
1550 		}
1551 	} else
1552 		mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1553 
1554 	dst = &dst_sa->sin6_addr;
1555 	IN6_MASK_ADDR(dst, &mask);
1556 
1557 	if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1558 		return (ENOBUFS);
1559 	fill_sockaddr_inet6((struct sockaddr_in6 *)sa, dst, 0);
1560 	info->rti_info[RTAX_DST] = sa;
1561 
1562 	if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128)) {
1563 		if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1564 			return (ENOBUFS);
1565 		fill_sockaddr_inet6((struct sockaddr_in6 *)sa, &mask, 0);
1566 		info->rti_info[RTAX_NETMASK] = sa;
1567 		info->rti_flags &= ~RTF_HOST;
1568 	} else
1569 		remove_netmask(info);
1570 
1571 	/* Check gateway */
1572 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1573 		return (cleanup_xaddrs_gateway(info, lb));
1574 
1575 	return (0);
1576 }
1577 #endif
1578 
1579 static int
1580 cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb)
1581 {
1582 	int error = EAFNOSUPPORT;
1583 
1584 	if (info->rti_info[RTAX_DST] == NULL)
1585 		return (EINVAL);
1586 
1587 	if (info->rti_flags & RTF_LLDATA) {
1588 		/*
1589 		 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1590 		 * prefix along with the actual address in RTA_DST.
1591 		 * Remove netmask to avoid unnecessary address masking.
1592 		 */
1593 		remove_netmask(info);
1594 	}
1595 
1596 	switch (info->rti_info[RTAX_DST]->sa_family) {
1597 #ifdef INET
1598 	case AF_INET:
1599 		error = cleanup_xaddrs_inet(info, lb);
1600 		break;
1601 #endif
1602 #ifdef INET6
1603 	case AF_INET6:
1604 		error = cleanup_xaddrs_inet6(info, lb);
1605 		break;
1606 #endif
1607 	}
1608 
1609 	return (error);
1610 }
1611 
1612 /*
1613  * Fill in @dmask with valid netmask leaving original @smask
1614  * intact. Mostly used with radix netmasks.
1615  */
1616 struct sockaddr *
1617 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1618     struct sockaddr_storage *dmask)
1619 {
1620 	if (dst == NULL || smask == NULL)
1621 		return (NULL);
1622 
1623 	memset(dmask, 0, dst->sa_len);
1624 	memcpy(dmask, smask, smask->sa_len);
1625 	dmask->ss_len = dst->sa_len;
1626 	dmask->ss_family = dst->sa_family;
1627 
1628 	return ((struct sockaddr *)dmask);
1629 }
1630 
1631 /*
1632  * Writes information related to @rtinfo object to newly-allocated mbuf.
1633  * Assumes MCLBYTES is enough to construct any message.
1634  * Used for OS notifications of vaious events (if/ifa announces,etc)
1635  *
1636  * Returns allocated mbuf or NULL on failure.
1637  */
1638 static struct mbuf *
1639 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1640 {
1641 	struct sockaddr_storage ss;
1642 	struct rt_msghdr *rtm;
1643 	struct mbuf *m;
1644 	int i;
1645 	struct sockaddr *sa;
1646 #ifdef INET6
1647 	struct sockaddr_in6 *sin6;
1648 #endif
1649 	int len, dlen;
1650 
1651 	switch (type) {
1652 	case RTM_DELADDR:
1653 	case RTM_NEWADDR:
1654 		len = sizeof(struct ifa_msghdr);
1655 		break;
1656 
1657 	case RTM_DELMADDR:
1658 	case RTM_NEWMADDR:
1659 		len = sizeof(struct ifma_msghdr);
1660 		break;
1661 
1662 	case RTM_IFINFO:
1663 		len = sizeof(struct if_msghdr);
1664 		break;
1665 
1666 	case RTM_IFANNOUNCE:
1667 	case RTM_IEEE80211:
1668 		len = sizeof(struct if_announcemsghdr);
1669 		break;
1670 
1671 	default:
1672 		len = sizeof(struct rt_msghdr);
1673 	}
1674 
1675 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1676 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1677 	if (len > MHLEN)
1678 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1679 	else
1680 		m = m_gethdr(M_NOWAIT, MT_DATA);
1681 	if (m == NULL)
1682 		return (m);
1683 
1684 	m->m_pkthdr.len = m->m_len = len;
1685 	rtm = mtod(m, struct rt_msghdr *);
1686 	bzero((caddr_t)rtm, len);
1687 	for (i = 0; i < RTAX_MAX; i++) {
1688 		if ((sa = rtinfo->rti_info[i]) == NULL)
1689 			continue;
1690 		rtinfo->rti_addrs |= (1 << i);
1691 
1692 		dlen = SA_SIZE(sa);
1693 		KASSERT(dlen <= sizeof(ss),
1694 		    ("%s: sockaddr size overflow", __func__));
1695 		bzero(&ss, sizeof(ss));
1696 		bcopy(sa, &ss, sa->sa_len);
1697 		sa = (struct sockaddr *)&ss;
1698 #ifdef INET6
1699 		if (sa->sa_family == AF_INET6) {
1700 			sin6 = (struct sockaddr_in6 *)sa;
1701 			(void)sa6_recoverscope(sin6);
1702 		}
1703 #endif
1704 		m_copyback(m, len, dlen, (caddr_t)sa);
1705 		len += dlen;
1706 	}
1707 	if (m->m_pkthdr.len != len) {
1708 		m_freem(m);
1709 		return (NULL);
1710 	}
1711 	rtm->rtm_msglen = len;
1712 	rtm->rtm_version = RTM_VERSION;
1713 	rtm->rtm_type = type;
1714 	return (m);
1715 }
1716 
1717 /*
1718  * Writes information related to @rtinfo object to preallocated buffer.
1719  * Stores needed size in @plen. If @w is NULL, calculates size without
1720  * writing.
1721  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1722  *
1723  * Returns 0 on success.
1724  *
1725  */
1726 static int
1727 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1728 {
1729 	struct sockaddr_storage ss;
1730 	int len, buflen = 0, dlen, i;
1731 	caddr_t cp = NULL;
1732 	struct rt_msghdr *rtm = NULL;
1733 #ifdef INET6
1734 	struct sockaddr_in6 *sin6;
1735 #endif
1736 #ifdef COMPAT_FREEBSD32
1737 	bool compat32 = false;
1738 #endif
1739 
1740 	switch (type) {
1741 	case RTM_DELADDR:
1742 	case RTM_NEWADDR:
1743 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1744 #ifdef COMPAT_FREEBSD32
1745 			if (w->w_req->flags & SCTL_MASK32) {
1746 				len = sizeof(struct ifa_msghdrl32);
1747 				compat32 = true;
1748 			} else
1749 #endif
1750 				len = sizeof(struct ifa_msghdrl);
1751 		} else
1752 			len = sizeof(struct ifa_msghdr);
1753 		break;
1754 
1755 	case RTM_IFINFO:
1756 #ifdef COMPAT_FREEBSD32
1757 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1758 			if (w->w_op == NET_RT_IFLISTL)
1759 				len = sizeof(struct if_msghdrl32);
1760 			else
1761 				len = sizeof(struct if_msghdr32);
1762 			compat32 = true;
1763 			break;
1764 		}
1765 #endif
1766 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1767 			len = sizeof(struct if_msghdrl);
1768 		else
1769 			len = sizeof(struct if_msghdr);
1770 		break;
1771 
1772 	case RTM_NEWMADDR:
1773 		len = sizeof(struct ifma_msghdr);
1774 		break;
1775 
1776 	default:
1777 		len = sizeof(struct rt_msghdr);
1778 	}
1779 
1780 	if (w != NULL) {
1781 		rtm = (struct rt_msghdr *)w->w_tmem;
1782 		buflen = w->w_tmemsize - len;
1783 		cp = (caddr_t)w->w_tmem + len;
1784 	}
1785 
1786 	rtinfo->rti_addrs = 0;
1787 	for (i = 0; i < RTAX_MAX; i++) {
1788 		struct sockaddr *sa;
1789 
1790 		if ((sa = rtinfo->rti_info[i]) == NULL)
1791 			continue;
1792 		rtinfo->rti_addrs |= (1 << i);
1793 #ifdef COMPAT_FREEBSD32
1794 		if (compat32)
1795 			dlen = SA_SIZE32(sa);
1796 		else
1797 #endif
1798 			dlen = SA_SIZE(sa);
1799 		if (cp != NULL && buflen >= dlen) {
1800 			KASSERT(dlen <= sizeof(ss),
1801 			    ("%s: sockaddr size overflow", __func__));
1802 			bzero(&ss, sizeof(ss));
1803 			bcopy(sa, &ss, sa->sa_len);
1804 			sa = (struct sockaddr *)&ss;
1805 #ifdef INET6
1806 			if (sa->sa_family == AF_INET6) {
1807 				sin6 = (struct sockaddr_in6 *)sa;
1808 				(void)sa6_recoverscope(sin6);
1809 			}
1810 #endif
1811 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1812 			cp += dlen;
1813 			buflen -= dlen;
1814 		} else if (cp != NULL) {
1815 			/*
1816 			 * Buffer too small. Count needed size
1817 			 * and return with error.
1818 			 */
1819 			cp = NULL;
1820 		}
1821 
1822 		len += dlen;
1823 	}
1824 
1825 	if (cp != NULL) {
1826 		dlen = ALIGN(len) - len;
1827 		if (buflen < dlen)
1828 			cp = NULL;
1829 		else {
1830 			bzero(cp, dlen);
1831 			cp += dlen;
1832 			buflen -= dlen;
1833 		}
1834 	}
1835 	len = ALIGN(len);
1836 
1837 	if (cp != NULL) {
1838 		/* fill header iff buffer is large enough */
1839 		rtm->rtm_version = RTM_VERSION;
1840 		rtm->rtm_type = type;
1841 		rtm->rtm_msglen = len;
1842 	}
1843 
1844 	*plen = len;
1845 
1846 	if (w != NULL && cp == NULL)
1847 		return (ENOBUFS);
1848 
1849 	return (0);
1850 }
1851 
1852 /*
1853  * This routine is called to generate a message from the routing
1854  * socket indicating that a redirect has occurred, a routing lookup
1855  * has failed, or that a protocol has detected timeouts to a particular
1856  * destination.
1857  */
1858 void
1859 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1860     int fibnum)
1861 {
1862 	struct rt_msghdr *rtm;
1863 	struct mbuf *m;
1864 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1865 
1866 	if (V_route_cb.any_count == 0)
1867 		return;
1868 	m = rtsock_msg_mbuf(type, rtinfo);
1869 	if (m == NULL)
1870 		return;
1871 
1872 	if (fibnum != RT_ALL_FIBS) {
1873 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1874 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1875 		M_SETFIB(m, fibnum);
1876 		m->m_flags |= RTS_FILTER_FIB;
1877 	}
1878 
1879 	rtm = mtod(m, struct rt_msghdr *);
1880 	rtm->rtm_flags = RTF_DONE | flags;
1881 	rtm->rtm_errno = error;
1882 	rtm->rtm_addrs = rtinfo->rti_addrs;
1883 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1884 }
1885 
1886 void
1887 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1888 {
1889 
1890 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1891 }
1892 
1893 /*
1894  * This routine is called to generate a message from the routing
1895  * socket indicating that the status of a network interface has changed.
1896  */
1897 void
1898 rt_ifmsg(struct ifnet *ifp)
1899 {
1900 	struct if_msghdr *ifm;
1901 	struct mbuf *m;
1902 	struct rt_addrinfo info;
1903 
1904 	if (V_route_cb.any_count == 0)
1905 		return;
1906 	bzero((caddr_t)&info, sizeof(info));
1907 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1908 	if (m == NULL)
1909 		return;
1910 	ifm = mtod(m, struct if_msghdr *);
1911 	ifm->ifm_index = ifp->if_index;
1912 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1913 	if_data_copy(ifp, &ifm->ifm_data);
1914 	ifm->ifm_addrs = 0;
1915 	rt_dispatch(m, AF_UNSPEC);
1916 }
1917 
1918 /*
1919  * Announce interface address arrival/withdraw.
1920  * Please do not call directly, use rt_addrmsg().
1921  * Assume input data to be valid.
1922  * Returns 0 on success.
1923  */
1924 int
1925 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1926 {
1927 	struct rt_addrinfo info;
1928 	struct sockaddr *sa;
1929 	int ncmd;
1930 	struct mbuf *m;
1931 	struct ifa_msghdr *ifam;
1932 	struct ifnet *ifp = ifa->ifa_ifp;
1933 	struct sockaddr_storage ss;
1934 
1935 	if (V_route_cb.any_count == 0)
1936 		return (0);
1937 
1938 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1939 
1940 	bzero((caddr_t)&info, sizeof(info));
1941 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1942 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1943 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1944 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1945 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1946 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1947 		return (ENOBUFS);
1948 	ifam = mtod(m, struct ifa_msghdr *);
1949 	ifam->ifam_index = ifp->if_index;
1950 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1951 	ifam->ifam_flags = ifa->ifa_flags;
1952 	ifam->ifam_addrs = info.rti_addrs;
1953 
1954 	if (fibnum != RT_ALL_FIBS) {
1955 		M_SETFIB(m, fibnum);
1956 		m->m_flags |= RTS_FILTER_FIB;
1957 	}
1958 
1959 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1960 
1961 	return (0);
1962 }
1963 
1964 /*
1965  * Announce route addition/removal to rtsock based on @rt data.
1966  * Callers are advives to use rt_routemsg() instead of using this
1967  *  function directly.
1968  * Assume @rt data is consistent.
1969  *
1970  * Returns 0 on success.
1971  */
1972 int
1973 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
1974     int fibnum)
1975 {
1976 	union sockaddr_union dst, mask;
1977 	struct rt_addrinfo info;
1978 
1979 	if (V_route_cb.any_count == 0)
1980 		return (0);
1981 
1982 	int family = rt_get_family(rt);
1983 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
1984 	export_rtaddrs(rt, &dst.sa, &mask.sa);
1985 
1986 	bzero((caddr_t)&info, sizeof(info));
1987 	info.rti_info[RTAX_DST] = &dst.sa;
1988 	info.rti_info[RTAX_NETMASK] = &mask.sa;
1989 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1990 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
1991 	info.rti_ifp = nh->nh_ifp;
1992 
1993 	return (rtsock_routemsg_info(cmd, &info, fibnum));
1994 }
1995 
1996 int
1997 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1998 {
1999 	struct rt_msghdr *rtm;
2000 	struct sockaddr *sa;
2001 	struct mbuf *m;
2002 
2003 	if (V_route_cb.any_count == 0)
2004 		return (0);
2005 
2006 	if (info->rti_flags & RTF_HOST)
2007 		info->rti_info[RTAX_NETMASK] = NULL;
2008 
2009 	m = rtsock_msg_mbuf(cmd, info);
2010 	if (m == NULL)
2011 		return (ENOBUFS);
2012 
2013 	if (fibnum != RT_ALL_FIBS) {
2014 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
2015 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
2016 		M_SETFIB(m, fibnum);
2017 		m->m_flags |= RTS_FILTER_FIB;
2018 	}
2019 
2020 	rtm = mtod(m, struct rt_msghdr *);
2021 	rtm->rtm_addrs = info->rti_addrs;
2022 	if (info->rti_ifp != NULL)
2023 		rtm->rtm_index = info->rti_ifp->if_index;
2024 	/* Add RTF_DONE to indicate command 'completion' required by API */
2025 	info->rti_flags |= RTF_DONE;
2026 	/* Reported routes has to be up */
2027 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
2028 		info->rti_flags |= RTF_UP;
2029 	rtm->rtm_flags = info->rti_flags;
2030 
2031 	sa = info->rti_info[RTAX_DST];
2032 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2033 
2034 	return (0);
2035 }
2036 
2037 /*
2038  * This is the analogue to the rt_newaddrmsg which performs the same
2039  * function but for multicast group memberhips.  This is easier since
2040  * there is no route state to worry about.
2041  */
2042 void
2043 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
2044 {
2045 	struct rt_addrinfo info;
2046 	struct mbuf *m = NULL;
2047 	struct ifnet *ifp = ifma->ifma_ifp;
2048 	struct ifma_msghdr *ifmam;
2049 
2050 	if (V_route_cb.any_count == 0)
2051 		return;
2052 
2053 	bzero((caddr_t)&info, sizeof(info));
2054 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2055 	if (ifp && ifp->if_addr)
2056 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2057 	else
2058 		info.rti_info[RTAX_IFP] = NULL;
2059 	/*
2060 	 * If a link-layer address is present, present it as a ``gateway''
2061 	 * (similarly to how ARP entries, e.g., are presented).
2062 	 */
2063 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
2064 	m = rtsock_msg_mbuf(cmd, &info);
2065 	if (m == NULL)
2066 		return;
2067 	ifmam = mtod(m, struct ifma_msghdr *);
2068 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
2069 	    __func__));
2070 	ifmam->ifmam_index = ifp->if_index;
2071 	ifmam->ifmam_addrs = info.rti_addrs;
2072 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
2073 }
2074 
2075 static struct mbuf *
2076 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
2077 	struct rt_addrinfo *info)
2078 {
2079 	struct if_announcemsghdr *ifan;
2080 	struct mbuf *m;
2081 
2082 	if (V_route_cb.any_count == 0)
2083 		return NULL;
2084 	bzero((caddr_t)info, sizeof(*info));
2085 	m = rtsock_msg_mbuf(type, info);
2086 	if (m != NULL) {
2087 		ifan = mtod(m, struct if_announcemsghdr *);
2088 		ifan->ifan_index = ifp->if_index;
2089 		strlcpy(ifan->ifan_name, ifp->if_xname,
2090 			sizeof(ifan->ifan_name));
2091 		ifan->ifan_what = what;
2092 	}
2093 	return m;
2094 }
2095 
2096 /*
2097  * This is called to generate routing socket messages indicating
2098  * IEEE80211 wireless events.
2099  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2100  */
2101 void
2102 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2103 {
2104 	struct mbuf *m;
2105 	struct rt_addrinfo info;
2106 
2107 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2108 	if (m != NULL) {
2109 		/*
2110 		 * Append the ieee80211 data.  Try to stick it in the
2111 		 * mbuf containing the ifannounce msg; otherwise allocate
2112 		 * a new mbuf and append.
2113 		 *
2114 		 * NB: we assume m is a single mbuf.
2115 		 */
2116 		if (data_len > M_TRAILINGSPACE(m)) {
2117 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2118 			if (n == NULL) {
2119 				m_freem(m);
2120 				return;
2121 			}
2122 			bcopy(data, mtod(n, void *), data_len);
2123 			n->m_len = data_len;
2124 			m->m_next = n;
2125 		} else if (data_len > 0) {
2126 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2127 			m->m_len += data_len;
2128 		}
2129 		if (m->m_flags & M_PKTHDR)
2130 			m->m_pkthdr.len += data_len;
2131 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2132 		rt_dispatch(m, AF_UNSPEC);
2133 	}
2134 }
2135 
2136 /*
2137  * This is called to generate routing socket messages indicating
2138  * network interface arrival and departure.
2139  */
2140 void
2141 rt_ifannouncemsg(struct ifnet *ifp, int what)
2142 {
2143 	struct mbuf *m;
2144 	struct rt_addrinfo info;
2145 
2146 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2147 	if (m != NULL)
2148 		rt_dispatch(m, AF_UNSPEC);
2149 }
2150 
2151 static void
2152 rt_dispatch(struct mbuf *m, sa_family_t saf)
2153 {
2154 	struct m_tag *tag;
2155 
2156 	/*
2157 	 * Preserve the family from the sockaddr, if any, in an m_tag for
2158 	 * use when injecting the mbuf into the routing socket buffer from
2159 	 * the netisr.
2160 	 */
2161 	if (saf != AF_UNSPEC) {
2162 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
2163 		    M_NOWAIT);
2164 		if (tag == NULL) {
2165 			m_freem(m);
2166 			return;
2167 		}
2168 		*(unsigned short *)(tag + 1) = saf;
2169 		m_tag_prepend(m, tag);
2170 	}
2171 #ifdef VIMAGE
2172 	if (V_loif)
2173 		m->m_pkthdr.rcvif = V_loif;
2174 	else {
2175 		m_freem(m);
2176 		return;
2177 	}
2178 #endif
2179 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
2180 }
2181 
2182 /*
2183  * Checks if rte can be exported w.r.t jails/vnets.
2184  *
2185  * Returns true if it can, false otherwise.
2186  */
2187 static bool
2188 can_export_rte(struct ucred *td_ucred, bool rt_is_host,
2189     const struct sockaddr *rt_dst)
2190 {
2191 
2192 	if ((!rt_is_host) ? jailed_without_vnet(td_ucred)
2193 	    : prison_if(td_ucred, rt_dst) != 0)
2194 		return (false);
2195 	return (true);
2196 }
2197 
2198 
2199 /*
2200  * This is used in dumping the kernel table via sysctl().
2201  */
2202 static int
2203 sysctl_dumpentry(struct rtentry *rt, void *vw)
2204 {
2205 	struct walkarg *w = vw;
2206 	struct nhop_object *nh;
2207 	int error = 0;
2208 
2209 	NET_EPOCH_ASSERT();
2210 
2211 	export_rtaddrs(rt, w->dst, w->mask);
2212 	if (!can_export_rte(w->w_req->td->td_ucred, rt_is_host(rt), w->dst))
2213 		return (0);
2214 	nh = rt_get_raw_nhop(rt);
2215 #ifdef ROUTE_MPATH
2216 	if (NH_IS_NHGRP(nh)) {
2217 		struct weightened_nhop *wn;
2218 		uint32_t num_nhops;
2219 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2220 		for (int i = 0; i < num_nhops; i++) {
2221 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2222 			if (error != 0)
2223 				return (error);
2224 		}
2225 	} else
2226 #endif
2227 		error = sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2228 
2229 	return (0);
2230 }
2231 
2232 
2233 static int
2234 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2235     struct walkarg *w)
2236 {
2237 	struct rt_addrinfo info;
2238 	int error = 0, size;
2239 	uint32_t rtflags;
2240 
2241 	rtflags = nhop_get_rtflags(nh);
2242 
2243 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2244 		return (0);
2245 
2246 	bzero((caddr_t)&info, sizeof(info));
2247 	info.rti_info[RTAX_DST] = w->dst;
2248 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2249 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2250 	info.rti_info[RTAX_GENMASK] = 0;
2251 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2252 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2253 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2254 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2255 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2256 	}
2257 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2258 		return (error);
2259 	if (w->w_req && w->w_tmem) {
2260 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2261 
2262 		bzero(&rtm->rtm_index,
2263 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2264 
2265 		/*
2266 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2267 		 * and RTF_UP (if entry is linked, which is always true here).
2268 		 * Given that, use nhop rtflags & add RTF_UP.
2269 		 */
2270 		rtm->rtm_flags = rtflags | RTF_UP;
2271 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2272 			rtm->rtm_flags = RTF_GATEWAY |
2273 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2274 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2275 		rtm->rtm_rmx.rmx_weight = weight;
2276 		rtm->rtm_index = nh->nh_ifp->if_index;
2277 		rtm->rtm_addrs = info.rti_addrs;
2278 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2279 		return (error);
2280 	}
2281 	return (error);
2282 }
2283 
2284 static int
2285 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2286     struct rt_addrinfo *info, struct walkarg *w, int len)
2287 {
2288 	struct if_msghdrl *ifm;
2289 	struct if_data *ifd;
2290 
2291 	ifm = (struct if_msghdrl *)w->w_tmem;
2292 
2293 #ifdef COMPAT_FREEBSD32
2294 	if (w->w_req->flags & SCTL_MASK32) {
2295 		struct if_msghdrl32 *ifm32;
2296 
2297 		ifm32 = (struct if_msghdrl32 *)ifm;
2298 		ifm32->ifm_addrs = info->rti_addrs;
2299 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2300 		ifm32->ifm_index = ifp->if_index;
2301 		ifm32->_ifm_spare1 = 0;
2302 		ifm32->ifm_len = sizeof(*ifm32);
2303 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2304 		ifm32->_ifm_spare2 = 0;
2305 		ifd = &ifm32->ifm_data;
2306 	} else
2307 #endif
2308 	{
2309 		ifm->ifm_addrs = info->rti_addrs;
2310 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2311 		ifm->ifm_index = ifp->if_index;
2312 		ifm->_ifm_spare1 = 0;
2313 		ifm->ifm_len = sizeof(*ifm);
2314 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2315 		ifm->_ifm_spare2 = 0;
2316 		ifd = &ifm->ifm_data;
2317 	}
2318 
2319 	memcpy(ifd, src_ifd, sizeof(*ifd));
2320 
2321 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2322 }
2323 
2324 static int
2325 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2326     struct rt_addrinfo *info, struct walkarg *w, int len)
2327 {
2328 	struct if_msghdr *ifm;
2329 	struct if_data *ifd;
2330 
2331 	ifm = (struct if_msghdr *)w->w_tmem;
2332 
2333 #ifdef COMPAT_FREEBSD32
2334 	if (w->w_req->flags & SCTL_MASK32) {
2335 		struct if_msghdr32 *ifm32;
2336 
2337 		ifm32 = (struct if_msghdr32 *)ifm;
2338 		ifm32->ifm_addrs = info->rti_addrs;
2339 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2340 		ifm32->ifm_index = ifp->if_index;
2341 		ifm32->_ifm_spare1 = 0;
2342 		ifd = &ifm32->ifm_data;
2343 	} else
2344 #endif
2345 	{
2346 		ifm->ifm_addrs = info->rti_addrs;
2347 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2348 		ifm->ifm_index = ifp->if_index;
2349 		ifm->_ifm_spare1 = 0;
2350 		ifd = &ifm->ifm_data;
2351 	}
2352 
2353 	memcpy(ifd, src_ifd, sizeof(*ifd));
2354 
2355 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2356 }
2357 
2358 static int
2359 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2360     struct walkarg *w, int len)
2361 {
2362 	struct ifa_msghdrl *ifam;
2363 	struct if_data *ifd;
2364 
2365 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2366 
2367 #ifdef COMPAT_FREEBSD32
2368 	if (w->w_req->flags & SCTL_MASK32) {
2369 		struct ifa_msghdrl32 *ifam32;
2370 
2371 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2372 		ifam32->ifam_addrs = info->rti_addrs;
2373 		ifam32->ifam_flags = ifa->ifa_flags;
2374 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2375 		ifam32->_ifam_spare1 = 0;
2376 		ifam32->ifam_len = sizeof(*ifam32);
2377 		ifam32->ifam_data_off =
2378 		    offsetof(struct ifa_msghdrl32, ifam_data);
2379 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2380 		ifd = &ifam32->ifam_data;
2381 	} else
2382 #endif
2383 	{
2384 		ifam->ifam_addrs = info->rti_addrs;
2385 		ifam->ifam_flags = ifa->ifa_flags;
2386 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2387 		ifam->_ifam_spare1 = 0;
2388 		ifam->ifam_len = sizeof(*ifam);
2389 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2390 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2391 		ifd = &ifam->ifam_data;
2392 	}
2393 
2394 	bzero(ifd, sizeof(*ifd));
2395 	ifd->ifi_datalen = sizeof(struct if_data);
2396 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2397 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2398 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2399 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2400 
2401 	/* Fixup if_data carp(4) vhid. */
2402 	if (carp_get_vhid_p != NULL)
2403 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2404 
2405 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2406 }
2407 
2408 static int
2409 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2410     struct walkarg *w, int len)
2411 {
2412 	struct ifa_msghdr *ifam;
2413 
2414 	ifam = (struct ifa_msghdr *)w->w_tmem;
2415 	ifam->ifam_addrs = info->rti_addrs;
2416 	ifam->ifam_flags = ifa->ifa_flags;
2417 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2418 	ifam->_ifam_spare1 = 0;
2419 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2420 
2421 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2422 }
2423 
2424 static int
2425 sysctl_iflist(int af, struct walkarg *w)
2426 {
2427 	struct ifnet *ifp;
2428 	struct ifaddr *ifa;
2429 	struct if_data ifd;
2430 	struct rt_addrinfo info;
2431 	int len, error = 0;
2432 	struct sockaddr_storage ss;
2433 
2434 	bzero((caddr_t)&info, sizeof(info));
2435 	bzero(&ifd, sizeof(ifd));
2436 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2437 		if (w->w_arg && w->w_arg != ifp->if_index)
2438 			continue;
2439 		if_data_copy(ifp, &ifd);
2440 		ifa = ifp->if_addr;
2441 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2442 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2443 		if (error != 0)
2444 			goto done;
2445 		info.rti_info[RTAX_IFP] = NULL;
2446 		if (w->w_req && w->w_tmem) {
2447 			if (w->w_op == NET_RT_IFLISTL)
2448 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2449 				    len);
2450 			else
2451 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2452 				    len);
2453 			if (error)
2454 				goto done;
2455 		}
2456 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2457 			if (af && af != ifa->ifa_addr->sa_family)
2458 				continue;
2459 			if (prison_if(w->w_req->td->td_ucred,
2460 			    ifa->ifa_addr) != 0)
2461 				continue;
2462 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2463 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2464 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2465 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2466 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2467 			if (error != 0)
2468 				goto done;
2469 			if (w->w_req && w->w_tmem) {
2470 				if (w->w_op == NET_RT_IFLISTL)
2471 					error = sysctl_iflist_ifaml(ifa, &info,
2472 					    w, len);
2473 				else
2474 					error = sysctl_iflist_ifam(ifa, &info,
2475 					    w, len);
2476 				if (error)
2477 					goto done;
2478 			}
2479 		}
2480 		info.rti_info[RTAX_IFA] = NULL;
2481 		info.rti_info[RTAX_NETMASK] = NULL;
2482 		info.rti_info[RTAX_BRD] = NULL;
2483 	}
2484 done:
2485 	return (error);
2486 }
2487 
2488 static int
2489 sysctl_ifmalist(int af, struct walkarg *w)
2490 {
2491 	struct rt_addrinfo info;
2492 	struct ifaddr *ifa;
2493 	struct ifmultiaddr *ifma;
2494 	struct ifnet *ifp;
2495 	int error, len;
2496 
2497 	NET_EPOCH_ASSERT();
2498 
2499 	error = 0;
2500 	bzero((caddr_t)&info, sizeof(info));
2501 
2502 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2503 		if (w->w_arg && w->w_arg != ifp->if_index)
2504 			continue;
2505 		ifa = ifp->if_addr;
2506 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2507 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2508 			if (af && af != ifma->ifma_addr->sa_family)
2509 				continue;
2510 			if (prison_if(w->w_req->td->td_ucred,
2511 			    ifma->ifma_addr) != 0)
2512 				continue;
2513 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2514 			info.rti_info[RTAX_GATEWAY] =
2515 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2516 			    ifma->ifma_lladdr : NULL;
2517 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2518 			if (error != 0)
2519 				break;
2520 			if (w->w_req && w->w_tmem) {
2521 				struct ifma_msghdr *ifmam;
2522 
2523 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2524 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2525 				ifmam->ifmam_flags = 0;
2526 				ifmam->ifmam_addrs = info.rti_addrs;
2527 				ifmam->_ifmam_spare1 = 0;
2528 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2529 				if (error != 0)
2530 					break;
2531 			}
2532 		}
2533 		if (error != 0)
2534 			break;
2535 	}
2536 	return (error);
2537 }
2538 
2539 static void
2540 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2541 {
2542 	union sockaddr_union sa_dst, sa_mask;
2543 
2544 	w->family = family;
2545 	w->dst = (struct sockaddr *)&sa_dst;
2546 	w->mask = (struct sockaddr *)&sa_mask;
2547 
2548 	init_sockaddrs_family(family, w->dst, w->mask);
2549 
2550 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2551 }
2552 
2553 static int
2554 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2555 {
2556 	struct epoch_tracker et;
2557 	int	*name = (int *)arg1;
2558 	u_int	namelen = arg2;
2559 	struct rib_head *rnh = NULL; /* silence compiler. */
2560 	int	i, lim, error = EINVAL;
2561 	int	fib = 0;
2562 	u_char	af;
2563 	struct	walkarg w;
2564 
2565 	name ++;
2566 	namelen--;
2567 	if (req->newptr)
2568 		return (EPERM);
2569 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2570 		if (namelen == 3)
2571 			fib = req->td->td_proc->p_fibnum;
2572 		else if (namelen == 4)
2573 			fib = (name[3] == RT_ALL_FIBS) ?
2574 			    req->td->td_proc->p_fibnum : name[3];
2575 		else
2576 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2577 		if (fib < 0 || fib >= rt_numfibs)
2578 			return (EINVAL);
2579 	} else if (namelen != 3)
2580 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2581 	af = name[0];
2582 	if (af > AF_MAX)
2583 		return (EINVAL);
2584 	bzero(&w, sizeof(w));
2585 	w.w_op = name[1];
2586 	w.w_arg = name[2];
2587 	w.w_req = req;
2588 
2589 	error = sysctl_wire_old_buffer(req, 0);
2590 	if (error)
2591 		return (error);
2592 
2593 	/*
2594 	 * Allocate reply buffer in advance.
2595 	 * All rtsock messages has maximum length of u_short.
2596 	 */
2597 	w.w_tmemsize = 65536;
2598 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2599 
2600 	NET_EPOCH_ENTER(et);
2601 	switch (w.w_op) {
2602 	case NET_RT_DUMP:
2603 	case NET_RT_FLAGS:
2604 		if (af == 0) {			/* dump all tables */
2605 			i = 1;
2606 			lim = AF_MAX;
2607 		} else				/* dump only one table */
2608 			i = lim = af;
2609 
2610 		/*
2611 		 * take care of llinfo entries, the caller must
2612 		 * specify an AF
2613 		 */
2614 		if (w.w_op == NET_RT_FLAGS &&
2615 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2616 			if (af != 0)
2617 				error = lltable_sysctl_dumparp(af, w.w_req);
2618 			else
2619 				error = EINVAL;
2620 			break;
2621 		}
2622 		/*
2623 		 * take care of routing entries
2624 		 */
2625 		for (error = 0; error == 0 && i <= lim; i++) {
2626 			rnh = rt_tables_get_rnh(fib, i);
2627 			if (rnh != NULL) {
2628 				rtable_sysctl_dump(fib, i, &w);
2629 			} else if (af != 0)
2630 				error = EAFNOSUPPORT;
2631 		}
2632 		break;
2633 	case NET_RT_NHOP:
2634 	case NET_RT_NHGRP:
2635 		/* Allow dumping one specific af/fib at a time */
2636 		if (namelen < 4) {
2637 			error = EINVAL;
2638 			break;
2639 		}
2640 		fib = name[3];
2641 		if (fib < 0 || fib > rt_numfibs) {
2642 			error = EINVAL;
2643 			break;
2644 		}
2645 		rnh = rt_tables_get_rnh(fib, af);
2646 		if (rnh == NULL) {
2647 			error = EAFNOSUPPORT;
2648 			break;
2649 		}
2650 		if (w.w_op == NET_RT_NHOP)
2651 			error = nhops_dump_sysctl(rnh, w.w_req);
2652 		else
2653 #ifdef ROUTE_MPATH
2654 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2655 #else
2656 			error = ENOTSUP;
2657 #endif
2658 		break;
2659 	case NET_RT_IFLIST:
2660 	case NET_RT_IFLISTL:
2661 		error = sysctl_iflist(af, &w);
2662 		break;
2663 
2664 	case NET_RT_IFMALIST:
2665 		error = sysctl_ifmalist(af, &w);
2666 		break;
2667 	}
2668 	NET_EPOCH_EXIT(et);
2669 
2670 	free(w.w_tmem, M_TEMP);
2671 	return (error);
2672 }
2673 
2674 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2675     sysctl_rtsock, "Return route tables and interface/address lists");
2676 
2677 /*
2678  * Definitions of protocols supported in the ROUTE domain.
2679  */
2680 
2681 static struct domain routedomain;		/* or at least forward */
2682 
2683 static struct protosw routesw[] = {
2684 {
2685 	.pr_type =		SOCK_RAW,
2686 	.pr_domain =		&routedomain,
2687 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2688 	.pr_output =		route_output,
2689 	.pr_ctlinput =		raw_ctlinput,
2690 	.pr_init =		raw_init,
2691 	.pr_usrreqs =		&route_usrreqs
2692 }
2693 };
2694 
2695 static struct domain routedomain = {
2696 	.dom_family =		PF_ROUTE,
2697 	.dom_name =		"route",
2698 	.dom_protosw =		routesw,
2699 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
2700 };
2701 
2702 VNET_DOMAIN_SET(route);
2703