xref: /freebsd/sys/net/rtsock.c (revision 1165fc9a526630487a1feb63daef65c5aee1a583)
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 #define	DEBUG_MOD_NAME	rtsock
80 #define	DEBUG_MAX_LEVEL	LOG_DEBUG
81 #include <net/route/route_debug.h>
82 _DECLARE_DEBUG(LOG_INFO);
83 
84 #ifdef COMPAT_FREEBSD32
85 #include <sys/mount.h>
86 #include <compat/freebsd32/freebsd32.h>
87 
88 struct if_msghdr32 {
89 	uint16_t ifm_msglen;
90 	uint8_t	ifm_version;
91 	uint8_t	ifm_type;
92 	int32_t	ifm_addrs;
93 	int32_t	ifm_flags;
94 	uint16_t ifm_index;
95 	uint16_t _ifm_spare1;
96 	struct	if_data ifm_data;
97 };
98 
99 struct if_msghdrl32 {
100 	uint16_t ifm_msglen;
101 	uint8_t	ifm_version;
102 	uint8_t	ifm_type;
103 	int32_t	ifm_addrs;
104 	int32_t	ifm_flags;
105 	uint16_t ifm_index;
106 	uint16_t _ifm_spare1;
107 	uint16_t ifm_len;
108 	uint16_t ifm_data_off;
109 	uint32_t _ifm_spare2;
110 	struct	if_data ifm_data;
111 };
112 
113 struct ifa_msghdrl32 {
114 	uint16_t ifam_msglen;
115 	uint8_t	ifam_version;
116 	uint8_t	ifam_type;
117 	int32_t	ifam_addrs;
118 	int32_t	ifam_flags;
119 	uint16_t ifam_index;
120 	uint16_t _ifam_spare1;
121 	uint16_t ifam_len;
122 	uint16_t ifam_data_off;
123 	int32_t	ifam_metric;
124 	struct	if_data ifam_data;
125 };
126 
127 #define SA_SIZE32(sa)						\
128     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
129 	sizeof(int)		:				\
130 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
131 
132 #endif /* COMPAT_FREEBSD32 */
133 
134 struct linear_buffer {
135 	char		*base;	/* Base allocated memory pointer */
136 	uint32_t	offset;	/* Currently used offset */
137 	uint32_t	size;	/* Total buffer size */
138 };
139 #define	SCRATCH_BUFFER_SIZE	1024
140 
141 #define	RTS_PID_LOG(_l, _fmt, ...)	RT_LOG_##_l(_l, "PID %d: " _fmt, curproc ? curproc->p_pid : 0, ## __VA_ARGS__)
142 
143 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
144 
145 /* NB: these are not modified */
146 static struct	sockaddr route_src = { 2, PF_ROUTE, };
147 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
148 
149 /* These are external hooks for CARP. */
150 int	(*carp_get_vhid_p)(struct ifaddr *);
151 
152 /*
153  * Used by rtsock/raw_input callback code to decide whether to filter the update
154  * notification to a socket bound to a particular FIB.
155  */
156 #define	RTS_FILTER_FIB	M_PROTO8
157 
158 typedef struct {
159 	int	ip_count;	/* attached w/ AF_INET */
160 	int	ip6_count;	/* attached w/ AF_INET6 */
161 	int	any_count;	/* total attached */
162 } route_cb_t;
163 VNET_DEFINE_STATIC(route_cb_t, route_cb);
164 #define	V_route_cb VNET(route_cb)
165 
166 struct mtx rtsock_mtx;
167 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
168 
169 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
170 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
171 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
172 
173 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
174 
175 struct walkarg {
176 	int	family;
177 	int	w_tmemsize;
178 	int	w_op, w_arg;
179 	caddr_t	w_tmem;
180 	struct sysctl_req *w_req;
181 	struct sockaddr *dst;
182 	struct sockaddr *mask;
183 };
184 
185 static void	rts_input(struct mbuf *m);
186 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
187 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
188 			struct walkarg *w, int *plen);
189 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
190 			struct rt_addrinfo *rtinfo);
191 static int	cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb);
192 static int	sysctl_dumpentry(struct rtentry *rt, void *vw);
193 static int	sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
194 			uint32_t weight, struct walkarg *w);
195 static int	sysctl_iflist(int af, struct walkarg *w);
196 static int	sysctl_ifmalist(int af, struct walkarg *w);
197 static int	route_output(struct mbuf *m, struct socket *so, ...);
198 static void	rt_getmetrics(const struct rtentry *rt,
199 			const struct nhop_object *nh, struct rt_metrics *out);
200 static void	rt_dispatch(struct mbuf *, sa_family_t);
201 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
202 			struct rt_msghdr *rtm, struct rib_cmd_info *rc);
203 static int	update_rtm_from_rc(struct rt_addrinfo *info,
204 			struct rt_msghdr **prtm, int alloc_len,
205 			struct rib_cmd_info *rc, struct nhop_object *nh);
206 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
207 			struct mbuf *m, sa_family_t saf, u_int fibnum,
208 			int rtm_errno);
209 static bool	can_export_rte(struct ucred *td_ucred, bool rt_is_host,
210 			const struct sockaddr *rt_dst);
211 
212 static struct netisr_handler rtsock_nh = {
213 	.nh_name = "rtsock",
214 	.nh_handler = rts_input,
215 	.nh_proto = NETISR_ROUTE,
216 	.nh_policy = NETISR_POLICY_SOURCE,
217 };
218 
219 static int
220 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
221 {
222 	int error, qlimit;
223 
224 	netisr_getqlimit(&rtsock_nh, &qlimit);
225 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
226         if (error || !req->newptr)
227                 return (error);
228 	if (qlimit < 1)
229 		return (EINVAL);
230 	return (netisr_setqlimit(&rtsock_nh, qlimit));
231 }
232 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
233     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
234     0, 0, sysctl_route_netisr_maxqlen, "I",
235     "maximum routing socket dispatch queue length");
236 
237 static void
238 vnet_rts_init(void)
239 {
240 	int tmp;
241 
242 	if (IS_DEFAULT_VNET(curvnet)) {
243 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
244 			rtsock_nh.nh_qlimit = tmp;
245 		netisr_register(&rtsock_nh);
246 	}
247 #ifdef VIMAGE
248 	 else
249 		netisr_register_vnet(&rtsock_nh);
250 #endif
251 }
252 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
253     vnet_rts_init, 0);
254 
255 #ifdef VIMAGE
256 static void
257 vnet_rts_uninit(void)
258 {
259 
260 	netisr_unregister_vnet(&rtsock_nh);
261 }
262 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
263     vnet_rts_uninit, 0);
264 #endif
265 
266 static int
267 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
268     struct rawcb *rp)
269 {
270 	int fibnum;
271 
272 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
273 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
274 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
275 
276 	/* No filtering requested. */
277 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
278 		return (0);
279 
280 	/* Check if it is a rts and the fib matches the one of the socket. */
281 	fibnum = M_GETFIB(m);
282 	if (proto->sp_family != PF_ROUTE ||
283 	    rp->rcb_socket == NULL ||
284 	    rp->rcb_socket->so_fibnum == fibnum)
285 		return (0);
286 
287 	/* Filtering requested and no match, the socket shall be skipped. */
288 	return (1);
289 }
290 
291 static void
292 rts_input(struct mbuf *m)
293 {
294 	struct sockproto route_proto;
295 	unsigned short *family;
296 	struct m_tag *tag;
297 
298 	route_proto.sp_family = PF_ROUTE;
299 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
300 	if (tag != NULL) {
301 		family = (unsigned short *)(tag + 1);
302 		route_proto.sp_protocol = *family;
303 		m_tag_delete(m, tag);
304 	} else
305 		route_proto.sp_protocol = 0;
306 
307 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
308 }
309 
310 /*
311  * It really doesn't make any sense at all for this code to share much
312  * with raw_usrreq.c, since its functionality is so restricted.  XXX
313  */
314 static void
315 rts_abort(struct socket *so)
316 {
317 
318 	raw_usrreqs.pru_abort(so);
319 }
320 
321 static void
322 rts_close(struct socket *so)
323 {
324 
325 	raw_usrreqs.pru_close(so);
326 }
327 
328 /* pru_accept is EOPNOTSUPP */
329 
330 static int
331 rts_attach(struct socket *so, int proto, struct thread *td)
332 {
333 	struct rawcb *rp;
334 	int error;
335 
336 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
337 
338 	/* XXX */
339 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
340 
341 	so->so_pcb = (caddr_t)rp;
342 	so->so_fibnum = td->td_proc->p_fibnum;
343 	error = raw_attach(so, proto);
344 	rp = sotorawcb(so);
345 	if (error) {
346 		so->so_pcb = NULL;
347 		free(rp, M_PCB);
348 		return error;
349 	}
350 	RTSOCK_LOCK();
351 	switch(rp->rcb_proto.sp_protocol) {
352 	case AF_INET:
353 		V_route_cb.ip_count++;
354 		break;
355 	case AF_INET6:
356 		V_route_cb.ip6_count++;
357 		break;
358 	}
359 	V_route_cb.any_count++;
360 	RTSOCK_UNLOCK();
361 	soisconnected(so);
362 	so->so_options |= SO_USELOOPBACK;
363 	return 0;
364 }
365 
366 static int
367 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
368 {
369 
370 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
371 }
372 
373 static int
374 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
375 {
376 
377 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
378 }
379 
380 /* pru_connect2 is EOPNOTSUPP */
381 /* pru_control is EOPNOTSUPP */
382 
383 static void
384 rts_detach(struct socket *so)
385 {
386 	struct rawcb *rp = sotorawcb(so);
387 
388 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
389 
390 	RTSOCK_LOCK();
391 	switch(rp->rcb_proto.sp_protocol) {
392 	case AF_INET:
393 		V_route_cb.ip_count--;
394 		break;
395 	case AF_INET6:
396 		V_route_cb.ip6_count--;
397 		break;
398 	}
399 	V_route_cb.any_count--;
400 	RTSOCK_UNLOCK();
401 	raw_usrreqs.pru_detach(so);
402 }
403 
404 static int
405 rts_disconnect(struct socket *so)
406 {
407 
408 	return (raw_usrreqs.pru_disconnect(so));
409 }
410 
411 /* pru_listen is EOPNOTSUPP */
412 
413 static int
414 rts_peeraddr(struct socket *so, struct sockaddr **nam)
415 {
416 
417 	return (raw_usrreqs.pru_peeraddr(so, nam));
418 }
419 
420 /* pru_rcvd is EOPNOTSUPP */
421 /* pru_rcvoob is EOPNOTSUPP */
422 
423 static int
424 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
425 	 struct mbuf *control, struct thread *td)
426 {
427 
428 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
429 }
430 
431 /* pru_sense is null */
432 
433 static int
434 rts_shutdown(struct socket *so)
435 {
436 
437 	return (raw_usrreqs.pru_shutdown(so));
438 }
439 
440 static int
441 rts_sockaddr(struct socket *so, struct sockaddr **nam)
442 {
443 
444 	return (raw_usrreqs.pru_sockaddr(so, nam));
445 }
446 
447 static struct pr_usrreqs route_usrreqs = {
448 	.pru_abort =		rts_abort,
449 	.pru_attach =		rts_attach,
450 	.pru_bind =		rts_bind,
451 	.pru_connect =		rts_connect,
452 	.pru_detach =		rts_detach,
453 	.pru_disconnect =	rts_disconnect,
454 	.pru_peeraddr =		rts_peeraddr,
455 	.pru_send =		rts_send,
456 	.pru_shutdown =		rts_shutdown,
457 	.pru_sockaddr =		rts_sockaddr,
458 	.pru_close =		rts_close,
459 };
460 
461 #ifndef _SOCKADDR_UNION_DEFINED
462 #define	_SOCKADDR_UNION_DEFINED
463 /*
464  * The union of all possible address formats we handle.
465  */
466 union sockaddr_union {
467 	struct sockaddr		sa;
468 	struct sockaddr_in	sin;
469 	struct sockaddr_in6	sin6;
470 };
471 #endif /* _SOCKADDR_UNION_DEFINED */
472 
473 static int
474 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
475     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
476 {
477 #if defined(INET) || defined(INET6)
478 	struct epoch_tracker et;
479 #endif
480 
481 	/* First, see if the returned address is part of the jail. */
482 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
483 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
484 		return (0);
485 	}
486 
487 	switch (info->rti_info[RTAX_DST]->sa_family) {
488 #ifdef INET
489 	case AF_INET:
490 	{
491 		struct in_addr ia;
492 		struct ifaddr *ifa;
493 		int found;
494 
495 		found = 0;
496 		/*
497 		 * Try to find an address on the given outgoing interface
498 		 * that belongs to the jail.
499 		 */
500 		NET_EPOCH_ENTER(et);
501 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
502 			struct sockaddr *sa;
503 			sa = ifa->ifa_addr;
504 			if (sa->sa_family != AF_INET)
505 				continue;
506 			ia = ((struct sockaddr_in *)sa)->sin_addr;
507 			if (prison_check_ip4(cred, &ia) == 0) {
508 				found = 1;
509 				break;
510 			}
511 		}
512 		NET_EPOCH_EXIT(et);
513 		if (!found) {
514 			/*
515 			 * As a last resort return the 'default' jail address.
516 			 */
517 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
518 			    sin_addr;
519 			if (prison_get_ip4(cred, &ia) != 0)
520 				return (ESRCH);
521 		}
522 		bzero(&saun->sin, sizeof(struct sockaddr_in));
523 		saun->sin.sin_len = sizeof(struct sockaddr_in);
524 		saun->sin.sin_family = AF_INET;
525 		saun->sin.sin_addr.s_addr = ia.s_addr;
526 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
527 		break;
528 	}
529 #endif
530 #ifdef INET6
531 	case AF_INET6:
532 	{
533 		struct in6_addr ia6;
534 		struct ifaddr *ifa;
535 		int found;
536 
537 		found = 0;
538 		/*
539 		 * Try to find an address on the given outgoing interface
540 		 * that belongs to the jail.
541 		 */
542 		NET_EPOCH_ENTER(et);
543 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
544 			struct sockaddr *sa;
545 			sa = ifa->ifa_addr;
546 			if (sa->sa_family != AF_INET6)
547 				continue;
548 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
549 			    &ia6, sizeof(struct in6_addr));
550 			if (prison_check_ip6(cred, &ia6) == 0) {
551 				found = 1;
552 				break;
553 			}
554 		}
555 		NET_EPOCH_EXIT(et);
556 		if (!found) {
557 			/*
558 			 * As a last resort return the 'default' jail address.
559 			 */
560 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
561 			    sin6_addr;
562 			if (prison_get_ip6(cred, &ia6) != 0)
563 				return (ESRCH);
564 		}
565 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
566 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
567 		saun->sin6.sin6_family = AF_INET6;
568 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
569 		if (sa6_recoverscope(&saun->sin6) != 0)
570 			return (ESRCH);
571 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
572 		break;
573 	}
574 #endif
575 	default:
576 		return (ESRCH);
577 	}
578 	return (0);
579 }
580 
581 static int
582 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
583 {
584 	struct ifaddr *ifa;
585 	sa_family_t saf;
586 
587 	if (V_loif == NULL) {
588 		RTS_PID_LOG(LOG_INFO, "Unable to add blackhole/reject nhop without loopback");
589 		return (ENOTSUP);
590 	}
591 	info->rti_ifp = V_loif;
592 
593 	saf = info->rti_info[RTAX_DST]->sa_family;
594 
595 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
596 		if (ifa->ifa_addr->sa_family == saf) {
597 			info->rti_ifa = ifa;
598 			break;
599 		}
600 	}
601 	if (info->rti_ifa == NULL) {
602 		RTS_PID_LOG(LOG_INFO, "Unable to find ifa for blackhole/reject nhop");
603 		return (ENOTSUP);
604 	}
605 
606 	bzero(saun, sizeof(union sockaddr_union));
607 	switch (saf) {
608 #ifdef INET
609 	case AF_INET:
610 		saun->sin.sin_family = AF_INET;
611 		saun->sin.sin_len = sizeof(struct sockaddr_in);
612 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
613 		break;
614 #endif
615 #ifdef INET6
616 	case AF_INET6:
617 		saun->sin6.sin6_family = AF_INET6;
618 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
619 		saun->sin6.sin6_addr = in6addr_loopback;
620 		break;
621 #endif
622 	default:
623 		RTS_PID_LOG(LOG_INFO, "unsupported family: %d", saf);
624 		return (ENOTSUP);
625 	}
626 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
627 	info->rti_flags |= RTF_GATEWAY;
628 
629 	return (0);
630 }
631 
632 /*
633  * Fills in @info based on userland-provided @rtm message.
634  *
635  * Returns 0 on success.
636  */
637 static int
638 fill_addrinfo(struct rt_msghdr *rtm, int len, struct linear_buffer *lb, u_int fibnum,
639     struct rt_addrinfo *info)
640 {
641 	int error;
642 
643 	rtm->rtm_pid = curproc->p_pid;
644 	info->rti_addrs = rtm->rtm_addrs;
645 
646 	info->rti_mflags = rtm->rtm_inits;
647 	info->rti_rmx = &rtm->rtm_rmx;
648 
649 	/*
650 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
651 	 * link-local address because rtrequest requires addresses with
652 	 * embedded scope id.
653 	 */
654 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
655 		return (EINVAL);
656 
657 	info->rti_flags = rtm->rtm_flags;
658 	error = cleanup_xaddrs(info, lb);
659 	if (error != 0)
660 		return (error);
661 	/*
662 	 * Verify that the caller has the appropriate privilege; RTM_GET
663 	 * is the only operation the non-superuser is allowed.
664 	 */
665 	if (rtm->rtm_type != RTM_GET) {
666 		error = priv_check(curthread, PRIV_NET_ROUTE);
667 		if (error != 0)
668 			return (error);
669 	}
670 
671 	/*
672 	 * The given gateway address may be an interface address.
673 	 * For example, issuing a "route change" command on a route
674 	 * entry that was created from a tunnel, and the gateway
675 	 * address given is the local end point. In this case the
676 	 * RTF_GATEWAY flag must be cleared or the destination will
677 	 * not be reachable even though there is no error message.
678 	 */
679 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
680 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
681 		struct rt_addrinfo ginfo;
682 		struct sockaddr *gdst;
683 		struct sockaddr_storage ss;
684 
685 		bzero(&ginfo, sizeof(ginfo));
686 		bzero(&ss, sizeof(ss));
687 		ss.ss_len = sizeof(ss);
688 
689 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
690 		gdst = info->rti_info[RTAX_GATEWAY];
691 
692 		/*
693 		 * A host route through the loopback interface is
694 		 * installed for each interface adddress. In pre 8.0
695 		 * releases the interface address of a PPP link type
696 		 * is not reachable locally. This behavior is fixed as
697 		 * part of the new L2/L3 redesign and rewrite work. The
698 		 * signature of this interface address route is the
699 		 * AF_LINK sa_family type of the gateway, and the
700 		 * rt_ifp has the IFF_LOOPBACK flag set.
701 		 */
702 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
703 			if (ss.ss_family == AF_LINK &&
704 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
705 				info->rti_flags &= ~RTF_GATEWAY;
706 				info->rti_flags |= RTF_GWFLAG_COMPAT;
707 			}
708 			rib_free_info(&ginfo);
709 		}
710 	}
711 
712 	return (0);
713 }
714 
715 static struct nhop_object *
716 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
717 {
718 	if (!NH_IS_NHGRP(nh))
719 		return (nh);
720 #ifdef ROUTE_MPATH
721 	const struct weightened_nhop *wn;
722 	uint32_t num_nhops;
723 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
724 	if (gw == NULL)
725 		return (wn[0].nh);
726 	for (int i = 0; i < num_nhops; i++) {
727 		if (match_nhop_gw(wn[i].nh, gw))
728 			return (wn[i].nh);
729 	}
730 #endif
731 	return (NULL);
732 }
733 
734 /*
735  * Handles RTM_GET message from routing socket, returning matching rt.
736  *
737  * Returns:
738  * 0 on success, with locked and referenced matching rt in @rt_nrt
739  * errno of failure
740  */
741 static int
742 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
743     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
744 {
745 	RIB_RLOCK_TRACKER;
746 	struct rib_head *rnh;
747 	struct nhop_object *nh;
748 	sa_family_t saf;
749 
750 	saf = info->rti_info[RTAX_DST]->sa_family;
751 
752 	rnh = rt_tables_get_rnh(fibnum, saf);
753 	if (rnh == NULL)
754 		return (EAFNOSUPPORT);
755 
756 	RIB_RLOCK(rnh);
757 
758 	/*
759 	 * By (implicit) convention host route (one without netmask)
760 	 * means longest-prefix-match request and the route with netmask
761 	 * means exact-match lookup.
762 	 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128
763 	 * prefixes, use original data to check for the netmask presence.
764 	 */
765 	if ((rtm->rtm_addrs & RTA_NETMASK) == 0) {
766 		/*
767 		 * Provide longest prefix match for
768 		 * address lookup (no mask).
769 		 * 'route -n get addr'
770 		 */
771 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
772 		    info->rti_info[RTAX_DST], &rnh->head);
773 	} else
774 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
775 		    info->rti_info[RTAX_DST],
776 		    info->rti_info[RTAX_NETMASK], &rnh->head);
777 
778 	if (rc->rc_rt == NULL) {
779 		RIB_RUNLOCK(rnh);
780 		return (ESRCH);
781 	}
782 
783 	nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
784 	if (nh == NULL) {
785 		RIB_RUNLOCK(rnh);
786 		return (ESRCH);
787 	}
788 	/*
789 	 * If performing proxied L2 entry insertion, and
790 	 * the actual PPP host entry is found, perform
791 	 * another search to retrieve the prefix route of
792 	 * the local end point of the PPP link.
793 	 * TODO: move this logic to userland.
794 	 */
795 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
796 		struct sockaddr_storage laddr;
797 
798 		if (nh->nh_ifp != NULL &&
799 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
800 			struct ifaddr *ifa;
801 
802 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
803 					RT_ALL_FIBS);
804 			if (ifa != NULL)
805 				rt_maskedcopy(ifa->ifa_addr,
806 					      (struct sockaddr *)&laddr,
807 					      ifa->ifa_netmask);
808 		} else
809 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
810 				      (struct sockaddr *)&laddr,
811 				      nh->nh_ifa->ifa_netmask);
812 		/*
813 		 * refactor rt and no lock operation necessary
814 		 */
815 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(
816 		    (struct sockaddr *)&laddr, &rnh->head);
817 		if (rc->rc_rt == NULL) {
818 			RIB_RUNLOCK(rnh);
819 			return (ESRCH);
820 		}
821 		nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
822 		if (nh == NULL) {
823 			RIB_RUNLOCK(rnh);
824 			return (ESRCH);
825 		}
826 	}
827 	rc->rc_nh_new = nh;
828 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
829 	RIB_RUNLOCK(rnh);
830 
831 	return (0);
832 }
833 
834 static void
835 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
836 {
837 #ifdef INET
838 	if (family == AF_INET) {
839 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
840 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
841 
842 		bzero(dst4, sizeof(struct sockaddr_in));
843 		bzero(mask4, sizeof(struct sockaddr_in));
844 
845 		dst4->sin_family = AF_INET;
846 		dst4->sin_len = sizeof(struct sockaddr_in);
847 		mask4->sin_family = AF_INET;
848 		mask4->sin_len = sizeof(struct sockaddr_in);
849 	}
850 #endif
851 #ifdef INET6
852 	if (family == AF_INET6) {
853 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
854 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
855 
856 		bzero(dst6, sizeof(struct sockaddr_in6));
857 		bzero(mask6, sizeof(struct sockaddr_in6));
858 
859 		dst6->sin6_family = AF_INET6;
860 		dst6->sin6_len = sizeof(struct sockaddr_in6);
861 		mask6->sin6_family = AF_INET6;
862 		mask6->sin6_len = sizeof(struct sockaddr_in6);
863 	}
864 #endif
865 }
866 
867 static void
868 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
869     struct sockaddr *mask)
870 {
871 #ifdef INET
872 	if (dst->sa_family == AF_INET) {
873 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
874 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
875 		uint32_t scopeid = 0;
876 		rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
877 		    &scopeid);
878 		return;
879 	}
880 #endif
881 #ifdef INET6
882 	if (dst->sa_family == AF_INET6) {
883 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
884 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
885 		uint32_t scopeid = 0;
886 		rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
887 		    &mask6->sin6_addr, &scopeid);
888 		dst6->sin6_scope_id = scopeid;
889 		return;
890 	}
891 #endif
892 }
893 
894 static int
895 update_rtm_from_info(struct rt_addrinfo *info, struct rt_msghdr **prtm,
896     int alloc_len)
897 {
898 	struct rt_msghdr *rtm, *orig_rtm = NULL;
899 	struct walkarg w;
900 	int len;
901 
902 	rtm = *prtm;
903 	/* Check if we need to realloc storage */
904 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
905 	if (len > alloc_len) {
906 		struct rt_msghdr *tmp_rtm;
907 
908 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
909 		if (tmp_rtm == NULL)
910 			return (ENOBUFS);
911 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
912 		orig_rtm = rtm;
913 		rtm = tmp_rtm;
914 		alloc_len = len;
915 
916 		/*
917 		 * Delay freeing original rtm as info contains
918 		 * data referencing it.
919 		 */
920 	}
921 
922 	w.w_tmem = (caddr_t)rtm;
923 	w.w_tmemsize = alloc_len;
924 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
925 	rtm->rtm_addrs = info->rti_addrs;
926 
927 	if (orig_rtm != NULL)
928 		free(orig_rtm, M_TEMP);
929 	*prtm = rtm;
930 	return (0);
931 }
932 
933 
934 /*
935  * Update sockaddrs, flags, etc in @prtm based on @rc data.
936  * rtm can be reallocated.
937  *
938  * Returns 0 on success, along with pointer to (potentially reallocated)
939  *  rtm.
940  *
941  */
942 static int
943 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
944     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
945 {
946 	union sockaddr_union saun;
947 	struct rt_msghdr *rtm;
948 	struct ifnet *ifp;
949 	int error;
950 
951 	rtm = *prtm;
952 	union sockaddr_union sa_dst, sa_mask;
953 	int family = info->rti_info[RTAX_DST]->sa_family;
954 	init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
955 	export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
956 
957 	info->rti_info[RTAX_DST] = &sa_dst.sa;
958 	info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
959 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
960 	info->rti_info[RTAX_GENMASK] = 0;
961 	ifp = nh->nh_ifp;
962 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
963 		if (ifp) {
964 			info->rti_info[RTAX_IFP] =
965 			    ifp->if_addr->ifa_addr;
966 			error = rtm_get_jailed(info, ifp, nh,
967 			    &saun, curthread->td_ucred);
968 			if (error != 0)
969 				return (error);
970 			if (ifp->if_flags & IFF_POINTOPOINT)
971 				info->rti_info[RTAX_BRD] =
972 				    nh->nh_ifa->ifa_dstaddr;
973 			rtm->rtm_index = ifp->if_index;
974 		} else {
975 			info->rti_info[RTAX_IFP] = NULL;
976 			info->rti_info[RTAX_IFA] = NULL;
977 		}
978 	} else if (ifp != NULL)
979 		rtm->rtm_index = ifp->if_index;
980 
981 	if ((error = update_rtm_from_info(info, prtm, alloc_len)) != 0)
982 		return (error);
983 
984 	rtm = *prtm;
985 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
986 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
987 		rtm->rtm_flags = RTF_GATEWAY |
988 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
989 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
990 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
991 
992 	return (0);
993 }
994 
995 #ifdef ROUTE_MPATH
996 static void
997 save_del_notification(struct rib_cmd_info *rc, void *_cbdata)
998 {
999 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1000 
1001 	if (rc->rc_cmd == RTM_DELETE)
1002 		*rc_new = *rc;
1003 }
1004 
1005 static void
1006 save_add_notification(struct rib_cmd_info *rc, void *_cbdata)
1007 {
1008 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1009 
1010 	if (rc->rc_cmd == RTM_ADD)
1011 		*rc_new = *rc;
1012 }
1013 #endif
1014 
1015 #if defined(INET6) || defined(INET)
1016 static struct sockaddr *
1017 alloc_sockaddr_aligned(struct linear_buffer *lb, int len)
1018 {
1019 	len = roundup2(len, sizeof(uint64_t));
1020 	if (lb->offset + len > lb->size)
1021 		return (NULL);
1022 	struct sockaddr *sa = (struct sockaddr *)(lb->base + lb->offset);
1023 	lb->offset += len;
1024 	return (sa);
1025 }
1026 #endif
1027 
1028 /*ARGSUSED*/
1029 static int
1030 route_output(struct mbuf *m, struct socket *so, ...)
1031 {
1032 	struct rt_msghdr *rtm = NULL;
1033 	struct rt_addrinfo info;
1034 	struct epoch_tracker et;
1035 #ifdef INET6
1036 	struct sockaddr_storage ss;
1037 	struct sockaddr_in6 *sin6;
1038 	int i, rti_need_deembed = 0;
1039 #endif
1040 	int alloc_len = 0, len, error = 0, fibnum;
1041 	sa_family_t saf = AF_UNSPEC;
1042 	struct rib_cmd_info rc;
1043 	struct nhop_object *nh;
1044 
1045 	fibnum = so->so_fibnum;
1046 #define senderr(e) { error = e; goto flush;}
1047 	if (m == NULL || ((m->m_len < sizeof(long)) &&
1048 		       (m = m_pullup(m, sizeof(long))) == NULL))
1049 		return (ENOBUFS);
1050 	if ((m->m_flags & M_PKTHDR) == 0)
1051 		panic("route_output");
1052 	NET_EPOCH_ENTER(et);
1053 	len = m->m_pkthdr.len;
1054 	if (len < sizeof(*rtm) ||
1055 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1056 		senderr(EINVAL);
1057 
1058 	/*
1059 	 * Most of current messages are in range 200-240 bytes,
1060 	 * minimize possible re-allocation on reply using larger size
1061 	 * buffer aligned on 1k boundaty.
1062 	 */
1063 	alloc_len = roundup2(len, 1024);
1064 	int total_len = alloc_len + SCRATCH_BUFFER_SIZE;
1065 	if ((rtm = malloc(total_len, M_TEMP, M_NOWAIT)) == NULL)
1066 		senderr(ENOBUFS);
1067 
1068 	m_copydata(m, 0, len, (caddr_t)rtm);
1069 	bzero(&info, sizeof(info));
1070 	nh = NULL;
1071 	struct linear_buffer lb = {
1072 		.base = (char *)rtm + alloc_len,
1073 		.size = SCRATCH_BUFFER_SIZE,
1074 	};
1075 
1076 	if (rtm->rtm_version != RTM_VERSION) {
1077 		/* Do not touch message since format is unknown */
1078 		free(rtm, M_TEMP);
1079 		rtm = NULL;
1080 		senderr(EPROTONOSUPPORT);
1081 	}
1082 
1083 	/*
1084 	 * Starting from here, it is possible
1085 	 * to alter original message and insert
1086 	 * caller PID and error value.
1087 	 */
1088 
1089 	if ((error = fill_addrinfo(rtm, len, &lb, fibnum, &info)) != 0) {
1090 		senderr(error);
1091 	}
1092 	/* fill_addringo() embeds scope into IPv6 addresses */
1093 #ifdef INET6
1094 	rti_need_deembed = 1;
1095 #endif
1096 
1097 	saf = info.rti_info[RTAX_DST]->sa_family;
1098 
1099 	/* support for new ARP code */
1100 	if (rtm->rtm_flags & RTF_LLDATA) {
1101 		error = lla_rt_output(rtm, &info);
1102 		goto flush;
1103 	}
1104 
1105 	union sockaddr_union gw_saun;
1106 	int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1107 	if (blackhole_flags != 0) {
1108 		if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1109 			error = fill_blackholeinfo(&info, &gw_saun);
1110 		else {
1111 			RTS_PID_LOG(LOG_DEBUG, "both BLACKHOLE and REJECT flags specifiied");
1112 			error = EINVAL;
1113 		}
1114 		if (error != 0)
1115 			senderr(error);
1116 	}
1117 
1118 	switch (rtm->rtm_type) {
1119 	case RTM_ADD:
1120 	case RTM_CHANGE:
1121 		if (rtm->rtm_type == RTM_ADD) {
1122 			if (info.rti_info[RTAX_GATEWAY] == NULL) {
1123 				RTS_PID_LOG(LOG_DEBUG, "RTM_ADD w/o gateway");
1124 				senderr(EINVAL);
1125 			}
1126 		}
1127 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1128 		if (error == 0) {
1129 #ifdef ROUTE_MPATH
1130 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
1131 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1132 				struct rib_cmd_info rc_simple = {};
1133 				rib_decompose_notification(&rc,
1134 				    save_add_notification, (void *)&rc_simple);
1135 				rc = rc_simple;
1136 			}
1137 #endif
1138 			/* nh MAY be empty if RTM_CHANGE request is no-op */
1139 			nh = rc.rc_nh_new;
1140 			if (nh != NULL) {
1141 				rtm->rtm_index = nh->nh_ifp->if_index;
1142 				rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1143 			}
1144 		}
1145 		break;
1146 
1147 	case RTM_DELETE:
1148 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1149 		if (error == 0) {
1150 #ifdef ROUTE_MPATH
1151 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1152 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1153 				struct rib_cmd_info rc_simple = {};
1154 				rib_decompose_notification(&rc,
1155 				    save_del_notification, (void *)&rc_simple);
1156 				rc = rc_simple;
1157 			}
1158 #endif
1159 			nh = rc.rc_nh_old;
1160 		}
1161 		break;
1162 
1163 	case RTM_GET:
1164 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1165 		if (error != 0)
1166 			senderr(error);
1167 		nh = rc.rc_nh_new;
1168 
1169 		if (!can_export_rte(curthread->td_ucred,
1170 		    info.rti_info[RTAX_NETMASK] == NULL,
1171 		    info.rti_info[RTAX_DST])) {
1172 			senderr(ESRCH);
1173 		}
1174 		break;
1175 
1176 	default:
1177 		senderr(EOPNOTSUPP);
1178 	}
1179 
1180 	if (error == 0 && nh != NULL) {
1181 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1182 		/*
1183 		 * Note that some sockaddr pointers may have changed to
1184 		 * point to memory outsize @rtm. Some may be pointing
1185 		 * to the on-stack variables.
1186 		 * Given that, any pointer in @info CANNOT BE USED.
1187 		 */
1188 
1189 		/*
1190 		 * scopeid deembedding has been performed while
1191 		 * writing updated rtm in rtsock_msg_buffer().
1192 		 * With that in mind, skip deembedding procedure below.
1193 		 */
1194 #ifdef INET6
1195 		rti_need_deembed = 0;
1196 #endif
1197 	}
1198 
1199 flush:
1200 	NET_EPOCH_EXIT(et);
1201 
1202 #ifdef INET6
1203 	if (rtm != NULL) {
1204 		if (rti_need_deembed) {
1205 			/* sin6_scope_id is recovered before sending rtm. */
1206 			sin6 = (struct sockaddr_in6 *)&ss;
1207 			for (i = 0; i < RTAX_MAX; i++) {
1208 				if (info.rti_info[i] == NULL)
1209 					continue;
1210 				if (info.rti_info[i]->sa_family != AF_INET6)
1211 					continue;
1212 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1213 				if (sa6_recoverscope(sin6) == 0)
1214 					bcopy(sin6, info.rti_info[i],
1215 						    sizeof(*sin6));
1216 			}
1217 			if (update_rtm_from_info(&info, &rtm, alloc_len) != 0) {
1218 				if (error != 0)
1219 					error = ENOBUFS;
1220 			}
1221 		}
1222 	}
1223 #endif
1224 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1225 
1226 	return (error);
1227 }
1228 
1229 /*
1230  * Sends the prepared reply message in @rtm to all rtsock clients.
1231  * Frees @m and @rtm.
1232  *
1233  */
1234 static void
1235 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1236     sa_family_t saf, u_int fibnum, int rtm_errno)
1237 {
1238 	struct rawcb *rp = NULL;
1239 
1240 	/*
1241 	 * Check to see if we don't want our own messages.
1242 	 */
1243 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1244 		if (V_route_cb.any_count <= 1) {
1245 			if (rtm != NULL)
1246 				free(rtm, M_TEMP);
1247 			m_freem(m);
1248 			return;
1249 		}
1250 		/* There is another listener, so construct message */
1251 		rp = sotorawcb(so);
1252 	}
1253 
1254 	if (rtm != NULL) {
1255 		if (rtm_errno!= 0)
1256 			rtm->rtm_errno = rtm_errno;
1257 		else
1258 			rtm->rtm_flags |= RTF_DONE;
1259 
1260 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1261 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1262 			m_freem(m);
1263 			m = NULL;
1264 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1265 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1266 
1267 		free(rtm, M_TEMP);
1268 	}
1269 	if (m != NULL) {
1270 		M_SETFIB(m, fibnum);
1271 		m->m_flags |= RTS_FILTER_FIB;
1272 		if (rp) {
1273 			/*
1274 			 * XXX insure we don't get a copy by
1275 			 * invalidating our protocol
1276 			 */
1277 			unsigned short family = rp->rcb_proto.sp_family;
1278 			rp->rcb_proto.sp_family = 0;
1279 			rt_dispatch(m, saf);
1280 			rp->rcb_proto.sp_family = family;
1281 		} else
1282 			rt_dispatch(m, saf);
1283 	}
1284 }
1285 
1286 static void
1287 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1288     struct rt_metrics *out)
1289 {
1290 
1291 	bzero(out, sizeof(*out));
1292 	out->rmx_mtu = nh->nh_mtu;
1293 	out->rmx_weight = rt->rt_weight;
1294 	out->rmx_nhidx = nhop_get_idx(nh);
1295 	/* Kernel -> userland timebase conversion. */
1296 	out->rmx_expire = nhop_get_expire(nh) ?
1297 	    nhop_get_expire(nh) - time_uptime + time_second : 0;
1298 }
1299 
1300 /*
1301  * Extract the addresses of the passed sockaddrs.
1302  * Do a little sanity checking so as to avoid bad memory references.
1303  * This data is derived straight from userland.
1304  */
1305 static int
1306 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1307 {
1308 	struct sockaddr *sa;
1309 	int i;
1310 
1311 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1312 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1313 			continue;
1314 		sa = (struct sockaddr *)cp;
1315 		/*
1316 		 * It won't fit.
1317 		 */
1318 		if (cp + sa->sa_len > cplim) {
1319 			RTS_PID_LOG(LOG_DEBUG, "sa_len too big for sa type %d", i);
1320 			return (EINVAL);
1321 		}
1322 		/*
1323 		 * there are no more.. quit now
1324 		 * If there are more bits, they are in error.
1325 		 * I've seen this. route(1) can evidently generate these.
1326 		 * This causes kernel to core dump.
1327 		 * for compatibility, If we see this, point to a safe address.
1328 		 */
1329 		if (sa->sa_len == 0) {
1330 			rtinfo->rti_info[i] = &sa_zero;
1331 			return (0); /* should be EINVAL but for compat */
1332 		}
1333 		/* accept it */
1334 #ifdef INET6
1335 		if (sa->sa_family == AF_INET6)
1336 			sa6_embedscope((struct sockaddr_in6 *)sa,
1337 			    V_ip6_use_defzone);
1338 #endif
1339 		rtinfo->rti_info[i] = sa;
1340 		cp += SA_SIZE(sa);
1341 	}
1342 	return (0);
1343 }
1344 
1345 #ifdef INET
1346 static inline void
1347 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1348 {
1349 
1350 	const struct sockaddr_in nsin = {
1351 		.sin_family = AF_INET,
1352 		.sin_len = sizeof(struct sockaddr_in),
1353 		.sin_addr = addr,
1354 	};
1355 	*sin = nsin;
1356 }
1357 #endif
1358 
1359 #ifdef INET6
1360 static inline void
1361 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1362     uint32_t scopeid)
1363 {
1364 
1365 	const struct sockaddr_in6 nsin6 = {
1366 		.sin6_family = AF_INET6,
1367 		.sin6_len = sizeof(struct sockaddr_in6),
1368 		.sin6_addr = *addr6,
1369 		.sin6_scope_id = scopeid,
1370 	};
1371 	*sin6 = nsin6;
1372 }
1373 #endif
1374 
1375 #if defined(INET6) || defined(INET)
1376 /*
1377  * Checks if gateway is suitable for lltable operations.
1378  * Lltable code requires AF_LINK gateway with ifindex
1379  *  and mac address specified.
1380  * Returns 0 on success.
1381  */
1382 static int
1383 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1384 {
1385 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1386 
1387 	if (sdl->sdl_family != AF_LINK)
1388 		return (EINVAL);
1389 
1390 	if (sdl->sdl_index == 0) {
1391 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gateway w/o ifindex");
1392 		return (EINVAL);
1393 	}
1394 
1395 	if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len) {
1396 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gw: sdl_nlen/sdl_alen too large");
1397 		return (EINVAL);
1398 	}
1399 
1400 	return (0);
1401 }
1402 
1403 static int
1404 cleanup_xaddrs_gateway(struct rt_addrinfo *info, struct linear_buffer *lb)
1405 {
1406 	struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1407 	struct sockaddr *sa;
1408 
1409 	if (info->rti_flags & RTF_LLDATA)
1410 		return (cleanup_xaddrs_lladdr(info));
1411 
1412 	switch (gw->sa_family) {
1413 #ifdef INET
1414 	case AF_INET:
1415 		{
1416 			struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw;
1417 
1418 			/* Ensure reads do not go beyoud SA boundary */
1419 			if (SA_SIZE(gw) < offsetof(struct sockaddr_in, sin_zero)) {
1420 				RTS_PID_LOG(LOG_DEBUG, "gateway sin_len too small: %d",
1421 				    gw->sa_len);
1422 				return (EINVAL);
1423 			}
1424 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_in));
1425 			if (sa == NULL)
1426 				return (ENOBUFS);
1427 			fill_sockaddr_inet((struct sockaddr_in *)sa, gw_sin->sin_addr);
1428 			info->rti_info[RTAX_GATEWAY] = sa;
1429 		}
1430 		break;
1431 #endif
1432 #ifdef INET6
1433 	case AF_INET6:
1434 		{
1435 			struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1436 			if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1437 				RTS_PID_LOG(LOG_DEBUG, "gateway sin6_len too small: %d",
1438 				    gw->sa_len);
1439 				return (EINVAL);
1440 			}
1441 			fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1442 			break;
1443 		}
1444 #endif
1445 	case AF_LINK:
1446 		{
1447 			struct sockaddr_dl *gw_sdl;
1448 
1449 			size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data);
1450 			gw_sdl = (struct sockaddr_dl *)gw;
1451 			if (gw_sdl->sdl_len < sdl_min_len) {
1452 				RTS_PID_LOG(LOG_DEBUG, "gateway sdl_len too small: %d",
1453 				    gw_sdl->sdl_len);
1454 				return (EINVAL);
1455 			}
1456 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_dl_short));
1457 			if (sa == NULL)
1458 				return (ENOBUFS);
1459 
1460 			const struct sockaddr_dl_short sdl = {
1461 				.sdl_family = AF_LINK,
1462 				.sdl_len = sizeof(struct sockaddr_dl_short),
1463 				.sdl_index = gw_sdl->sdl_index,
1464 			};
1465 			*((struct sockaddr_dl_short *)sa) = sdl;
1466 			info->rti_info[RTAX_GATEWAY] = sa;
1467 			break;
1468 		}
1469 	}
1470 
1471 	return (0);
1472 }
1473 #endif
1474 
1475 static void
1476 remove_netmask(struct rt_addrinfo *info)
1477 {
1478 	info->rti_info[RTAX_NETMASK] = NULL;
1479 	info->rti_flags |= RTF_HOST;
1480 	info->rti_addrs &= ~RTA_NETMASK;
1481 }
1482 
1483 #ifdef INET
1484 static int
1485 cleanup_xaddrs_inet(struct rt_addrinfo *info, struct linear_buffer *lb)
1486 {
1487 	struct sockaddr_in *dst_sa, *mask_sa;
1488 	const int sa_len = sizeof(struct sockaddr_in);
1489 	struct in_addr dst, mask;
1490 
1491 	/* Check & fixup dst/netmask combination first */
1492 	dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1493 	mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1494 
1495 	/* Ensure reads do not go beyound the buffer size */
1496 	if (SA_SIZE(dst_sa) < offsetof(struct sockaddr_in, sin_zero)) {
1497 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin_len too small: %d",
1498 		    dst_sa->sin_len);
1499 		return (EINVAL);
1500 	}
1501 
1502 	if ((mask_sa != NULL) && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1503 		/*
1504 		 * Some older routing software encode mask length into the
1505 		 * sin_len, thus resulting in "truncated" sockaddr.
1506 		 */
1507 		int len = mask_sa->sin_len - offsetof(struct sockaddr_in, sin_addr);
1508 		if (len >= 0) {
1509 			mask.s_addr = 0;
1510 			if (len > sizeof(struct in_addr))
1511 				len = sizeof(struct in_addr);
1512 			memcpy(&mask, &mask_sa->sin_addr, len);
1513 		} else {
1514 			RTS_PID_LOG(LOG_DEBUG, "prefix mask sin_len too small: %d",
1515 			    mask_sa->sin_len);
1516 			return (EINVAL);
1517 		}
1518 	} else
1519 		mask.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST;
1520 
1521 	dst.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr));
1522 
1523 	/* Construct new "clean" dst/mask sockaddresses */
1524 	if ((dst_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1525 		return (ENOBUFS);
1526 	fill_sockaddr_inet(dst_sa, dst);
1527 	info->rti_info[RTAX_DST] = (struct sockaddr *)dst_sa;
1528 
1529 	if (mask.s_addr != INADDR_BROADCAST) {
1530 		if ((mask_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1531 			return (ENOBUFS);
1532 		fill_sockaddr_inet(mask_sa, mask);
1533 		info->rti_info[RTAX_NETMASK] = (struct sockaddr *)mask_sa;
1534 		info->rti_flags &= ~RTF_HOST;
1535 	} else
1536 		remove_netmask(info);
1537 
1538 	/* Check gateway */
1539 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1540 		return (cleanup_xaddrs_gateway(info, lb));
1541 
1542 	return (0);
1543 }
1544 #endif
1545 
1546 #ifdef INET6
1547 static int
1548 cleanup_xaddrs_inet6(struct rt_addrinfo *info, struct linear_buffer *lb)
1549 {
1550 	struct sockaddr *sa;
1551 	struct sockaddr_in6 *dst_sa, *mask_sa;
1552 	struct in6_addr mask, *dst;
1553 	const int sa_len = sizeof(struct sockaddr_in6);
1554 
1555 	/* Check & fixup dst/netmask combination first */
1556 	dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1557 	mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1558 
1559 	if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1560 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin6_len too small: %d",
1561 		    dst_sa->sin6_len);
1562 		return (EINVAL);
1563 	}
1564 
1565 	if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1566 		/*
1567 		 * Some older routing software encode mask length into the
1568 		 * sin6_len, thus resulting in "truncated" sockaddr.
1569 		 */
1570 		int len = mask_sa->sin6_len - offsetof(struct sockaddr_in6, sin6_addr);
1571 		if (len >= 0) {
1572 			bzero(&mask, sizeof(mask));
1573 			if (len > sizeof(struct in6_addr))
1574 				len = sizeof(struct in6_addr);
1575 			memcpy(&mask, &mask_sa->sin6_addr, len);
1576 		} else {
1577 			RTS_PID_LOG(LOG_DEBUG, "rtsock: prefix mask sin6_len too small: %d",
1578 			    mask_sa->sin6_len);
1579 			return (EINVAL);
1580 		}
1581 	} else
1582 		mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1583 
1584 	dst = &dst_sa->sin6_addr;
1585 	IN6_MASK_ADDR(dst, &mask);
1586 
1587 	if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1588 		return (ENOBUFS);
1589 	fill_sockaddr_inet6((struct sockaddr_in6 *)sa, dst, 0);
1590 	info->rti_info[RTAX_DST] = sa;
1591 
1592 	if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128)) {
1593 		if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1594 			return (ENOBUFS);
1595 		fill_sockaddr_inet6((struct sockaddr_in6 *)sa, &mask, 0);
1596 		info->rti_info[RTAX_NETMASK] = sa;
1597 		info->rti_flags &= ~RTF_HOST;
1598 	} else
1599 		remove_netmask(info);
1600 
1601 	/* Check gateway */
1602 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1603 		return (cleanup_xaddrs_gateway(info, lb));
1604 
1605 	return (0);
1606 }
1607 #endif
1608 
1609 static int
1610 cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb)
1611 {
1612 	int error = EAFNOSUPPORT;
1613 
1614 	if (info->rti_info[RTAX_DST] == NULL) {
1615 		RTS_PID_LOG(LOG_DEBUG, "prefix dst is not set");
1616 		return (EINVAL);
1617 	}
1618 
1619 	if (info->rti_flags & RTF_LLDATA) {
1620 		/*
1621 		 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1622 		 * prefix along with the actual address in RTA_DST.
1623 		 * Remove netmask to avoid unnecessary address masking.
1624 		 */
1625 		remove_netmask(info);
1626 	}
1627 
1628 	switch (info->rti_info[RTAX_DST]->sa_family) {
1629 #ifdef INET
1630 	case AF_INET:
1631 		error = cleanup_xaddrs_inet(info, lb);
1632 		break;
1633 #endif
1634 #ifdef INET6
1635 	case AF_INET6:
1636 		error = cleanup_xaddrs_inet6(info, lb);
1637 		break;
1638 #endif
1639 	}
1640 
1641 	return (error);
1642 }
1643 
1644 /*
1645  * Fill in @dmask with valid netmask leaving original @smask
1646  * intact. Mostly used with radix netmasks.
1647  */
1648 struct sockaddr *
1649 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1650     struct sockaddr_storage *dmask)
1651 {
1652 	if (dst == NULL || smask == NULL)
1653 		return (NULL);
1654 
1655 	memset(dmask, 0, dst->sa_len);
1656 	memcpy(dmask, smask, smask->sa_len);
1657 	dmask->ss_len = dst->sa_len;
1658 	dmask->ss_family = dst->sa_family;
1659 
1660 	return ((struct sockaddr *)dmask);
1661 }
1662 
1663 /*
1664  * Writes information related to @rtinfo object to newly-allocated mbuf.
1665  * Assumes MCLBYTES is enough to construct any message.
1666  * Used for OS notifications of vaious events (if/ifa announces,etc)
1667  *
1668  * Returns allocated mbuf or NULL on failure.
1669  */
1670 static struct mbuf *
1671 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1672 {
1673 	struct sockaddr_storage ss;
1674 	struct rt_msghdr *rtm;
1675 	struct mbuf *m;
1676 	int i;
1677 	struct sockaddr *sa;
1678 #ifdef INET6
1679 	struct sockaddr_in6 *sin6;
1680 #endif
1681 	int len, dlen;
1682 
1683 	switch (type) {
1684 	case RTM_DELADDR:
1685 	case RTM_NEWADDR:
1686 		len = sizeof(struct ifa_msghdr);
1687 		break;
1688 
1689 	case RTM_DELMADDR:
1690 	case RTM_NEWMADDR:
1691 		len = sizeof(struct ifma_msghdr);
1692 		break;
1693 
1694 	case RTM_IFINFO:
1695 		len = sizeof(struct if_msghdr);
1696 		break;
1697 
1698 	case RTM_IFANNOUNCE:
1699 	case RTM_IEEE80211:
1700 		len = sizeof(struct if_announcemsghdr);
1701 		break;
1702 
1703 	default:
1704 		len = sizeof(struct rt_msghdr);
1705 	}
1706 
1707 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1708 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1709 	if (len > MHLEN)
1710 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1711 	else
1712 		m = m_gethdr(M_NOWAIT, MT_DATA);
1713 	if (m == NULL)
1714 		return (m);
1715 
1716 	m->m_pkthdr.len = m->m_len = len;
1717 	rtm = mtod(m, struct rt_msghdr *);
1718 	bzero((caddr_t)rtm, len);
1719 	for (i = 0; i < RTAX_MAX; i++) {
1720 		if ((sa = rtinfo->rti_info[i]) == NULL)
1721 			continue;
1722 		rtinfo->rti_addrs |= (1 << i);
1723 
1724 		dlen = SA_SIZE(sa);
1725 		KASSERT(dlen <= sizeof(ss),
1726 		    ("%s: sockaddr size overflow", __func__));
1727 		bzero(&ss, sizeof(ss));
1728 		bcopy(sa, &ss, sa->sa_len);
1729 		sa = (struct sockaddr *)&ss;
1730 #ifdef INET6
1731 		if (sa->sa_family == AF_INET6) {
1732 			sin6 = (struct sockaddr_in6 *)sa;
1733 			(void)sa6_recoverscope(sin6);
1734 		}
1735 #endif
1736 		m_copyback(m, len, dlen, (caddr_t)sa);
1737 		len += dlen;
1738 	}
1739 	if (m->m_pkthdr.len != len) {
1740 		m_freem(m);
1741 		return (NULL);
1742 	}
1743 	rtm->rtm_msglen = len;
1744 	rtm->rtm_version = RTM_VERSION;
1745 	rtm->rtm_type = type;
1746 	return (m);
1747 }
1748 
1749 /*
1750  * Writes information related to @rtinfo object to preallocated buffer.
1751  * Stores needed size in @plen. If @w is NULL, calculates size without
1752  * writing.
1753  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1754  *
1755  * Returns 0 on success.
1756  *
1757  */
1758 static int
1759 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1760 {
1761 	struct sockaddr_storage ss;
1762 	int len, buflen = 0, dlen, i;
1763 	caddr_t cp = NULL;
1764 	struct rt_msghdr *rtm = NULL;
1765 #ifdef INET6
1766 	struct sockaddr_in6 *sin6;
1767 #endif
1768 #ifdef COMPAT_FREEBSD32
1769 	bool compat32 = false;
1770 #endif
1771 
1772 	switch (type) {
1773 	case RTM_DELADDR:
1774 	case RTM_NEWADDR:
1775 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1776 #ifdef COMPAT_FREEBSD32
1777 			if (w->w_req->flags & SCTL_MASK32) {
1778 				len = sizeof(struct ifa_msghdrl32);
1779 				compat32 = true;
1780 			} else
1781 #endif
1782 				len = sizeof(struct ifa_msghdrl);
1783 		} else
1784 			len = sizeof(struct ifa_msghdr);
1785 		break;
1786 
1787 	case RTM_IFINFO:
1788 #ifdef COMPAT_FREEBSD32
1789 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1790 			if (w->w_op == NET_RT_IFLISTL)
1791 				len = sizeof(struct if_msghdrl32);
1792 			else
1793 				len = sizeof(struct if_msghdr32);
1794 			compat32 = true;
1795 			break;
1796 		}
1797 #endif
1798 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1799 			len = sizeof(struct if_msghdrl);
1800 		else
1801 			len = sizeof(struct if_msghdr);
1802 		break;
1803 
1804 	case RTM_NEWMADDR:
1805 		len = sizeof(struct ifma_msghdr);
1806 		break;
1807 
1808 	default:
1809 		len = sizeof(struct rt_msghdr);
1810 	}
1811 
1812 	if (w != NULL) {
1813 		rtm = (struct rt_msghdr *)w->w_tmem;
1814 		buflen = w->w_tmemsize - len;
1815 		cp = (caddr_t)w->w_tmem + len;
1816 	}
1817 
1818 	rtinfo->rti_addrs = 0;
1819 	for (i = 0; i < RTAX_MAX; i++) {
1820 		struct sockaddr *sa;
1821 
1822 		if ((sa = rtinfo->rti_info[i]) == NULL)
1823 			continue;
1824 		rtinfo->rti_addrs |= (1 << i);
1825 #ifdef COMPAT_FREEBSD32
1826 		if (compat32)
1827 			dlen = SA_SIZE32(sa);
1828 		else
1829 #endif
1830 			dlen = SA_SIZE(sa);
1831 		if (cp != NULL && buflen >= dlen) {
1832 			KASSERT(dlen <= sizeof(ss),
1833 			    ("%s: sockaddr size overflow", __func__));
1834 			bzero(&ss, sizeof(ss));
1835 			bcopy(sa, &ss, sa->sa_len);
1836 			sa = (struct sockaddr *)&ss;
1837 #ifdef INET6
1838 			if (sa->sa_family == AF_INET6) {
1839 				sin6 = (struct sockaddr_in6 *)sa;
1840 				(void)sa6_recoverscope(sin6);
1841 			}
1842 #endif
1843 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1844 			cp += dlen;
1845 			buflen -= dlen;
1846 		} else if (cp != NULL) {
1847 			/*
1848 			 * Buffer too small. Count needed size
1849 			 * and return with error.
1850 			 */
1851 			cp = NULL;
1852 		}
1853 
1854 		len += dlen;
1855 	}
1856 
1857 	if (cp != NULL) {
1858 		dlen = ALIGN(len) - len;
1859 		if (buflen < dlen)
1860 			cp = NULL;
1861 		else {
1862 			bzero(cp, dlen);
1863 			cp += dlen;
1864 			buflen -= dlen;
1865 		}
1866 	}
1867 	len = ALIGN(len);
1868 
1869 	if (cp != NULL) {
1870 		/* fill header iff buffer is large enough */
1871 		rtm->rtm_version = RTM_VERSION;
1872 		rtm->rtm_type = type;
1873 		rtm->rtm_msglen = len;
1874 	}
1875 
1876 	*plen = len;
1877 
1878 	if (w != NULL && cp == NULL)
1879 		return (ENOBUFS);
1880 
1881 	return (0);
1882 }
1883 
1884 /*
1885  * This routine is called to generate a message from the routing
1886  * socket indicating that a redirect has occurred, a routing lookup
1887  * has failed, or that a protocol has detected timeouts to a particular
1888  * destination.
1889  */
1890 void
1891 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1892     int fibnum)
1893 {
1894 	struct rt_msghdr *rtm;
1895 	struct mbuf *m;
1896 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1897 
1898 	if (V_route_cb.any_count == 0)
1899 		return;
1900 	m = rtsock_msg_mbuf(type, rtinfo);
1901 	if (m == NULL)
1902 		return;
1903 
1904 	if (fibnum != RT_ALL_FIBS) {
1905 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1906 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1907 		M_SETFIB(m, fibnum);
1908 		m->m_flags |= RTS_FILTER_FIB;
1909 	}
1910 
1911 	rtm = mtod(m, struct rt_msghdr *);
1912 	rtm->rtm_flags = RTF_DONE | flags;
1913 	rtm->rtm_errno = error;
1914 	rtm->rtm_addrs = rtinfo->rti_addrs;
1915 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1916 }
1917 
1918 void
1919 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1920 {
1921 
1922 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1923 }
1924 
1925 /*
1926  * This routine is called to generate a message from the routing
1927  * socket indicating that the status of a network interface has changed.
1928  */
1929 void
1930 rt_ifmsg(struct ifnet *ifp)
1931 {
1932 	struct if_msghdr *ifm;
1933 	struct mbuf *m;
1934 	struct rt_addrinfo info;
1935 
1936 	if (V_route_cb.any_count == 0)
1937 		return;
1938 	bzero((caddr_t)&info, sizeof(info));
1939 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1940 	if (m == NULL)
1941 		return;
1942 	ifm = mtod(m, struct if_msghdr *);
1943 	ifm->ifm_index = ifp->if_index;
1944 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1945 	if_data_copy(ifp, &ifm->ifm_data);
1946 	ifm->ifm_addrs = 0;
1947 	rt_dispatch(m, AF_UNSPEC);
1948 }
1949 
1950 /*
1951  * Announce interface address arrival/withdraw.
1952  * Please do not call directly, use rt_addrmsg().
1953  * Assume input data to be valid.
1954  * Returns 0 on success.
1955  */
1956 int
1957 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1958 {
1959 	struct rt_addrinfo info;
1960 	struct sockaddr *sa;
1961 	int ncmd;
1962 	struct mbuf *m;
1963 	struct ifa_msghdr *ifam;
1964 	struct ifnet *ifp = ifa->ifa_ifp;
1965 	struct sockaddr_storage ss;
1966 
1967 	if (V_route_cb.any_count == 0)
1968 		return (0);
1969 
1970 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1971 
1972 	bzero((caddr_t)&info, sizeof(info));
1973 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1974 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1975 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1976 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1977 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1978 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1979 		return (ENOBUFS);
1980 	ifam = mtod(m, struct ifa_msghdr *);
1981 	ifam->ifam_index = ifp->if_index;
1982 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1983 	ifam->ifam_flags = ifa->ifa_flags;
1984 	ifam->ifam_addrs = info.rti_addrs;
1985 
1986 	if (fibnum != RT_ALL_FIBS) {
1987 		M_SETFIB(m, fibnum);
1988 		m->m_flags |= RTS_FILTER_FIB;
1989 	}
1990 
1991 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1992 
1993 	return (0);
1994 }
1995 
1996 /*
1997  * Announce route addition/removal to rtsock based on @rt data.
1998  * Callers are advives to use rt_routemsg() instead of using this
1999  *  function directly.
2000  * Assume @rt data is consistent.
2001  *
2002  * Returns 0 on success.
2003  */
2004 int
2005 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
2006     int fibnum)
2007 {
2008 	union sockaddr_union dst, mask;
2009 	struct rt_addrinfo info;
2010 
2011 	if (V_route_cb.any_count == 0)
2012 		return (0);
2013 
2014 	int family = rt_get_family(rt);
2015 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
2016 	export_rtaddrs(rt, &dst.sa, &mask.sa);
2017 
2018 	bzero((caddr_t)&info, sizeof(info));
2019 	info.rti_info[RTAX_DST] = &dst.sa;
2020 	info.rti_info[RTAX_NETMASK] = &mask.sa;
2021 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2022 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
2023 	info.rti_ifp = nh->nh_ifp;
2024 
2025 	return (rtsock_routemsg_info(cmd, &info, fibnum));
2026 }
2027 
2028 int
2029 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
2030 {
2031 	struct rt_msghdr *rtm;
2032 	struct sockaddr *sa;
2033 	struct mbuf *m;
2034 
2035 	if (V_route_cb.any_count == 0)
2036 		return (0);
2037 
2038 	if (info->rti_flags & RTF_HOST)
2039 		info->rti_info[RTAX_NETMASK] = NULL;
2040 
2041 	m = rtsock_msg_mbuf(cmd, info);
2042 	if (m == NULL)
2043 		return (ENOBUFS);
2044 
2045 	if (fibnum != RT_ALL_FIBS) {
2046 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
2047 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
2048 		M_SETFIB(m, fibnum);
2049 		m->m_flags |= RTS_FILTER_FIB;
2050 	}
2051 
2052 	rtm = mtod(m, struct rt_msghdr *);
2053 	rtm->rtm_addrs = info->rti_addrs;
2054 	if (info->rti_ifp != NULL)
2055 		rtm->rtm_index = info->rti_ifp->if_index;
2056 	/* Add RTF_DONE to indicate command 'completion' required by API */
2057 	info->rti_flags |= RTF_DONE;
2058 	/* Reported routes has to be up */
2059 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
2060 		info->rti_flags |= RTF_UP;
2061 	rtm->rtm_flags = info->rti_flags;
2062 
2063 	sa = info->rti_info[RTAX_DST];
2064 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2065 
2066 	return (0);
2067 }
2068 
2069 /*
2070  * This is the analogue to the rt_newaddrmsg which performs the same
2071  * function but for multicast group memberhips.  This is easier since
2072  * there is no route state to worry about.
2073  */
2074 void
2075 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
2076 {
2077 	struct rt_addrinfo info;
2078 	struct mbuf *m = NULL;
2079 	struct ifnet *ifp = ifma->ifma_ifp;
2080 	struct ifma_msghdr *ifmam;
2081 
2082 	if (V_route_cb.any_count == 0)
2083 		return;
2084 
2085 	bzero((caddr_t)&info, sizeof(info));
2086 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2087 	if (ifp && ifp->if_addr)
2088 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2089 	else
2090 		info.rti_info[RTAX_IFP] = NULL;
2091 	/*
2092 	 * If a link-layer address is present, present it as a ``gateway''
2093 	 * (similarly to how ARP entries, e.g., are presented).
2094 	 */
2095 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
2096 	m = rtsock_msg_mbuf(cmd, &info);
2097 	if (m == NULL)
2098 		return;
2099 	ifmam = mtod(m, struct ifma_msghdr *);
2100 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
2101 	    __func__));
2102 	ifmam->ifmam_index = ifp->if_index;
2103 	ifmam->ifmam_addrs = info.rti_addrs;
2104 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
2105 }
2106 
2107 static struct mbuf *
2108 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
2109 	struct rt_addrinfo *info)
2110 {
2111 	struct if_announcemsghdr *ifan;
2112 	struct mbuf *m;
2113 
2114 	if (V_route_cb.any_count == 0)
2115 		return NULL;
2116 	bzero((caddr_t)info, sizeof(*info));
2117 	m = rtsock_msg_mbuf(type, info);
2118 	if (m != NULL) {
2119 		ifan = mtod(m, struct if_announcemsghdr *);
2120 		ifan->ifan_index = ifp->if_index;
2121 		strlcpy(ifan->ifan_name, ifp->if_xname,
2122 			sizeof(ifan->ifan_name));
2123 		ifan->ifan_what = what;
2124 	}
2125 	return m;
2126 }
2127 
2128 /*
2129  * This is called to generate routing socket messages indicating
2130  * IEEE80211 wireless events.
2131  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2132  */
2133 void
2134 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2135 {
2136 	struct mbuf *m;
2137 	struct rt_addrinfo info;
2138 
2139 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2140 	if (m != NULL) {
2141 		/*
2142 		 * Append the ieee80211 data.  Try to stick it in the
2143 		 * mbuf containing the ifannounce msg; otherwise allocate
2144 		 * a new mbuf and append.
2145 		 *
2146 		 * NB: we assume m is a single mbuf.
2147 		 */
2148 		if (data_len > M_TRAILINGSPACE(m)) {
2149 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2150 			if (n == NULL) {
2151 				m_freem(m);
2152 				return;
2153 			}
2154 			bcopy(data, mtod(n, void *), data_len);
2155 			n->m_len = data_len;
2156 			m->m_next = n;
2157 		} else if (data_len > 0) {
2158 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2159 			m->m_len += data_len;
2160 		}
2161 		if (m->m_flags & M_PKTHDR)
2162 			m->m_pkthdr.len += data_len;
2163 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2164 		rt_dispatch(m, AF_UNSPEC);
2165 	}
2166 }
2167 
2168 /*
2169  * This is called to generate routing socket messages indicating
2170  * network interface arrival and departure.
2171  */
2172 void
2173 rt_ifannouncemsg(struct ifnet *ifp, int what)
2174 {
2175 	struct mbuf *m;
2176 	struct rt_addrinfo info;
2177 
2178 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2179 	if (m != NULL)
2180 		rt_dispatch(m, AF_UNSPEC);
2181 }
2182 
2183 static void
2184 rt_dispatch(struct mbuf *m, sa_family_t saf)
2185 {
2186 	struct m_tag *tag;
2187 
2188 	/*
2189 	 * Preserve the family from the sockaddr, if any, in an m_tag for
2190 	 * use when injecting the mbuf into the routing socket buffer from
2191 	 * the netisr.
2192 	 */
2193 	if (saf != AF_UNSPEC) {
2194 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
2195 		    M_NOWAIT);
2196 		if (tag == NULL) {
2197 			m_freem(m);
2198 			return;
2199 		}
2200 		*(unsigned short *)(tag + 1) = saf;
2201 		m_tag_prepend(m, tag);
2202 	}
2203 	if (V_loif)
2204 		m->m_pkthdr.rcvif = V_loif;
2205 	else {
2206 		m_freem(m);
2207 		return;
2208 	}
2209 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
2210 }
2211 
2212 /*
2213  * Checks if rte can be exported w.r.t jails/vnets.
2214  *
2215  * Returns true if it can, false otherwise.
2216  */
2217 static bool
2218 can_export_rte(struct ucred *td_ucred, bool rt_is_host,
2219     const struct sockaddr *rt_dst)
2220 {
2221 
2222 	if ((!rt_is_host) ? jailed_without_vnet(td_ucred)
2223 	    : prison_if(td_ucred, rt_dst) != 0)
2224 		return (false);
2225 	return (true);
2226 }
2227 
2228 
2229 /*
2230  * This is used in dumping the kernel table via sysctl().
2231  */
2232 static int
2233 sysctl_dumpentry(struct rtentry *rt, void *vw)
2234 {
2235 	struct walkarg *w = vw;
2236 	struct nhop_object *nh;
2237 
2238 	NET_EPOCH_ASSERT();
2239 
2240 	export_rtaddrs(rt, w->dst, w->mask);
2241 	if (!can_export_rte(w->w_req->td->td_ucred, rt_is_host(rt), w->dst))
2242 		return (0);
2243 	nh = rt_get_raw_nhop(rt);
2244 #ifdef ROUTE_MPATH
2245 	if (NH_IS_NHGRP(nh)) {
2246 		const struct weightened_nhop *wn;
2247 		uint32_t num_nhops;
2248 		int error;
2249 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2250 		for (int i = 0; i < num_nhops; i++) {
2251 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2252 			if (error != 0)
2253 				return (error);
2254 		}
2255 	} else
2256 #endif
2257 		sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2258 
2259 	return (0);
2260 }
2261 
2262 
2263 static int
2264 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2265     struct walkarg *w)
2266 {
2267 	struct rt_addrinfo info;
2268 	int error = 0, size;
2269 	uint32_t rtflags;
2270 
2271 	rtflags = nhop_get_rtflags(nh);
2272 
2273 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2274 		return (0);
2275 
2276 	bzero((caddr_t)&info, sizeof(info));
2277 	info.rti_info[RTAX_DST] = w->dst;
2278 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2279 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2280 	info.rti_info[RTAX_GENMASK] = 0;
2281 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2282 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2283 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2284 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2285 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2286 	}
2287 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2288 		return (error);
2289 	if (w->w_req && w->w_tmem) {
2290 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2291 
2292 		bzero(&rtm->rtm_index,
2293 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2294 
2295 		/*
2296 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2297 		 * and RTF_UP (if entry is linked, which is always true here).
2298 		 * Given that, use nhop rtflags & add RTF_UP.
2299 		 */
2300 		rtm->rtm_flags = rtflags | RTF_UP;
2301 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2302 			rtm->rtm_flags = RTF_GATEWAY |
2303 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2304 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2305 		rtm->rtm_rmx.rmx_weight = weight;
2306 		rtm->rtm_index = nh->nh_ifp->if_index;
2307 		rtm->rtm_addrs = info.rti_addrs;
2308 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2309 		return (error);
2310 	}
2311 	return (error);
2312 }
2313 
2314 static int
2315 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2316     struct rt_addrinfo *info, struct walkarg *w, int len)
2317 {
2318 	struct if_msghdrl *ifm;
2319 	struct if_data *ifd;
2320 
2321 	ifm = (struct if_msghdrl *)w->w_tmem;
2322 
2323 #ifdef COMPAT_FREEBSD32
2324 	if (w->w_req->flags & SCTL_MASK32) {
2325 		struct if_msghdrl32 *ifm32;
2326 
2327 		ifm32 = (struct if_msghdrl32 *)ifm;
2328 		ifm32->ifm_addrs = info->rti_addrs;
2329 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2330 		ifm32->ifm_index = ifp->if_index;
2331 		ifm32->_ifm_spare1 = 0;
2332 		ifm32->ifm_len = sizeof(*ifm32);
2333 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2334 		ifm32->_ifm_spare2 = 0;
2335 		ifd = &ifm32->ifm_data;
2336 	} else
2337 #endif
2338 	{
2339 		ifm->ifm_addrs = info->rti_addrs;
2340 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2341 		ifm->ifm_index = ifp->if_index;
2342 		ifm->_ifm_spare1 = 0;
2343 		ifm->ifm_len = sizeof(*ifm);
2344 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2345 		ifm->_ifm_spare2 = 0;
2346 		ifd = &ifm->ifm_data;
2347 	}
2348 
2349 	memcpy(ifd, src_ifd, sizeof(*ifd));
2350 
2351 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2352 }
2353 
2354 static int
2355 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2356     struct rt_addrinfo *info, struct walkarg *w, int len)
2357 {
2358 	struct if_msghdr *ifm;
2359 	struct if_data *ifd;
2360 
2361 	ifm = (struct if_msghdr *)w->w_tmem;
2362 
2363 #ifdef COMPAT_FREEBSD32
2364 	if (w->w_req->flags & SCTL_MASK32) {
2365 		struct if_msghdr32 *ifm32;
2366 
2367 		ifm32 = (struct if_msghdr32 *)ifm;
2368 		ifm32->ifm_addrs = info->rti_addrs;
2369 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2370 		ifm32->ifm_index = ifp->if_index;
2371 		ifm32->_ifm_spare1 = 0;
2372 		ifd = &ifm32->ifm_data;
2373 	} else
2374 #endif
2375 	{
2376 		ifm->ifm_addrs = info->rti_addrs;
2377 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2378 		ifm->ifm_index = ifp->if_index;
2379 		ifm->_ifm_spare1 = 0;
2380 		ifd = &ifm->ifm_data;
2381 	}
2382 
2383 	memcpy(ifd, src_ifd, sizeof(*ifd));
2384 
2385 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2386 }
2387 
2388 static int
2389 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2390     struct walkarg *w, int len)
2391 {
2392 	struct ifa_msghdrl *ifam;
2393 	struct if_data *ifd;
2394 
2395 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2396 
2397 #ifdef COMPAT_FREEBSD32
2398 	if (w->w_req->flags & SCTL_MASK32) {
2399 		struct ifa_msghdrl32 *ifam32;
2400 
2401 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2402 		ifam32->ifam_addrs = info->rti_addrs;
2403 		ifam32->ifam_flags = ifa->ifa_flags;
2404 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2405 		ifam32->_ifam_spare1 = 0;
2406 		ifam32->ifam_len = sizeof(*ifam32);
2407 		ifam32->ifam_data_off =
2408 		    offsetof(struct ifa_msghdrl32, ifam_data);
2409 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2410 		ifd = &ifam32->ifam_data;
2411 	} else
2412 #endif
2413 	{
2414 		ifam->ifam_addrs = info->rti_addrs;
2415 		ifam->ifam_flags = ifa->ifa_flags;
2416 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2417 		ifam->_ifam_spare1 = 0;
2418 		ifam->ifam_len = sizeof(*ifam);
2419 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2420 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2421 		ifd = &ifam->ifam_data;
2422 	}
2423 
2424 	bzero(ifd, sizeof(*ifd));
2425 	ifd->ifi_datalen = sizeof(struct if_data);
2426 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2427 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2428 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2429 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2430 
2431 	/* Fixup if_data carp(4) vhid. */
2432 	if (carp_get_vhid_p != NULL)
2433 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2434 
2435 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2436 }
2437 
2438 static int
2439 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2440     struct walkarg *w, int len)
2441 {
2442 	struct ifa_msghdr *ifam;
2443 
2444 	ifam = (struct ifa_msghdr *)w->w_tmem;
2445 	ifam->ifam_addrs = info->rti_addrs;
2446 	ifam->ifam_flags = ifa->ifa_flags;
2447 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2448 	ifam->_ifam_spare1 = 0;
2449 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2450 
2451 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2452 }
2453 
2454 static int
2455 sysctl_iflist(int af, struct walkarg *w)
2456 {
2457 	struct ifnet *ifp;
2458 	struct ifaddr *ifa;
2459 	struct if_data ifd;
2460 	struct rt_addrinfo info;
2461 	int len, error = 0;
2462 	struct sockaddr_storage ss;
2463 
2464 	bzero((caddr_t)&info, sizeof(info));
2465 	bzero(&ifd, sizeof(ifd));
2466 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2467 		if (w->w_arg && w->w_arg != ifp->if_index)
2468 			continue;
2469 		if_data_copy(ifp, &ifd);
2470 		ifa = ifp->if_addr;
2471 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2472 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2473 		if (error != 0)
2474 			goto done;
2475 		info.rti_info[RTAX_IFP] = NULL;
2476 		if (w->w_req && w->w_tmem) {
2477 			if (w->w_op == NET_RT_IFLISTL)
2478 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2479 				    len);
2480 			else
2481 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2482 				    len);
2483 			if (error)
2484 				goto done;
2485 		}
2486 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2487 			if (af && af != ifa->ifa_addr->sa_family)
2488 				continue;
2489 			if (prison_if(w->w_req->td->td_ucred,
2490 			    ifa->ifa_addr) != 0)
2491 				continue;
2492 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2493 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2494 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2495 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2496 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2497 			if (error != 0)
2498 				goto done;
2499 			if (w->w_req && w->w_tmem) {
2500 				if (w->w_op == NET_RT_IFLISTL)
2501 					error = sysctl_iflist_ifaml(ifa, &info,
2502 					    w, len);
2503 				else
2504 					error = sysctl_iflist_ifam(ifa, &info,
2505 					    w, len);
2506 				if (error)
2507 					goto done;
2508 			}
2509 		}
2510 		info.rti_info[RTAX_IFA] = NULL;
2511 		info.rti_info[RTAX_NETMASK] = NULL;
2512 		info.rti_info[RTAX_BRD] = NULL;
2513 	}
2514 done:
2515 	return (error);
2516 }
2517 
2518 static int
2519 sysctl_ifmalist(int af, struct walkarg *w)
2520 {
2521 	struct rt_addrinfo info;
2522 	struct ifaddr *ifa;
2523 	struct ifmultiaddr *ifma;
2524 	struct ifnet *ifp;
2525 	int error, len;
2526 
2527 	NET_EPOCH_ASSERT();
2528 
2529 	error = 0;
2530 	bzero((caddr_t)&info, sizeof(info));
2531 
2532 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2533 		if (w->w_arg && w->w_arg != ifp->if_index)
2534 			continue;
2535 		ifa = ifp->if_addr;
2536 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2537 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2538 			if (af && af != ifma->ifma_addr->sa_family)
2539 				continue;
2540 			if (prison_if(w->w_req->td->td_ucred,
2541 			    ifma->ifma_addr) != 0)
2542 				continue;
2543 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2544 			info.rti_info[RTAX_GATEWAY] =
2545 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2546 			    ifma->ifma_lladdr : NULL;
2547 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2548 			if (error != 0)
2549 				break;
2550 			if (w->w_req && w->w_tmem) {
2551 				struct ifma_msghdr *ifmam;
2552 
2553 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2554 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2555 				ifmam->ifmam_flags = 0;
2556 				ifmam->ifmam_addrs = info.rti_addrs;
2557 				ifmam->_ifmam_spare1 = 0;
2558 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2559 				if (error != 0)
2560 					break;
2561 			}
2562 		}
2563 		if (error != 0)
2564 			break;
2565 	}
2566 	return (error);
2567 }
2568 
2569 static void
2570 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2571 {
2572 	union sockaddr_union sa_dst, sa_mask;
2573 
2574 	w->family = family;
2575 	w->dst = (struct sockaddr *)&sa_dst;
2576 	w->mask = (struct sockaddr *)&sa_mask;
2577 
2578 	init_sockaddrs_family(family, w->dst, w->mask);
2579 
2580 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2581 }
2582 
2583 static int
2584 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2585 {
2586 	struct epoch_tracker et;
2587 	int	*name = (int *)arg1;
2588 	u_int	namelen = arg2;
2589 	struct rib_head *rnh = NULL; /* silence compiler. */
2590 	int	i, lim, error = EINVAL;
2591 	int	fib = 0;
2592 	u_char	af;
2593 	struct	walkarg w;
2594 
2595 	if (namelen < 3)
2596 		return (EINVAL);
2597 
2598 	name++;
2599 	namelen--;
2600 	if (req->newptr)
2601 		return (EPERM);
2602 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2603 		if (namelen == 3)
2604 			fib = req->td->td_proc->p_fibnum;
2605 		else if (namelen == 4)
2606 			fib = (name[3] == RT_ALL_FIBS) ?
2607 			    req->td->td_proc->p_fibnum : name[3];
2608 		else
2609 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2610 		if (fib < 0 || fib >= rt_numfibs)
2611 			return (EINVAL);
2612 	} else if (namelen != 3)
2613 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2614 	af = name[0];
2615 	if (af > AF_MAX)
2616 		return (EINVAL);
2617 	bzero(&w, sizeof(w));
2618 	w.w_op = name[1];
2619 	w.w_arg = name[2];
2620 	w.w_req = req;
2621 
2622 	error = sysctl_wire_old_buffer(req, 0);
2623 	if (error)
2624 		return (error);
2625 
2626 	/*
2627 	 * Allocate reply buffer in advance.
2628 	 * All rtsock messages has maximum length of u_short.
2629 	 */
2630 	w.w_tmemsize = 65536;
2631 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2632 
2633 	NET_EPOCH_ENTER(et);
2634 	switch (w.w_op) {
2635 	case NET_RT_DUMP:
2636 	case NET_RT_FLAGS:
2637 		if (af == 0) {			/* dump all tables */
2638 			i = 1;
2639 			lim = AF_MAX;
2640 		} else				/* dump only one table */
2641 			i = lim = af;
2642 
2643 		/*
2644 		 * take care of llinfo entries, the caller must
2645 		 * specify an AF
2646 		 */
2647 		if (w.w_op == NET_RT_FLAGS &&
2648 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2649 			if (af != 0)
2650 				error = lltable_sysctl_dumparp(af, w.w_req);
2651 			else
2652 				error = EINVAL;
2653 			break;
2654 		}
2655 		/*
2656 		 * take care of routing entries
2657 		 */
2658 		for (error = 0; error == 0 && i <= lim; i++) {
2659 			rnh = rt_tables_get_rnh(fib, i);
2660 			if (rnh != NULL) {
2661 				rtable_sysctl_dump(fib, i, &w);
2662 			} else if (af != 0)
2663 				error = EAFNOSUPPORT;
2664 		}
2665 		break;
2666 	case NET_RT_NHOP:
2667 	case NET_RT_NHGRP:
2668 		/* Allow dumping one specific af/fib at a time */
2669 		if (namelen < 4) {
2670 			error = EINVAL;
2671 			break;
2672 		}
2673 		fib = name[3];
2674 		if (fib < 0 || fib > rt_numfibs) {
2675 			error = EINVAL;
2676 			break;
2677 		}
2678 		rnh = rt_tables_get_rnh(fib, af);
2679 		if (rnh == NULL) {
2680 			error = EAFNOSUPPORT;
2681 			break;
2682 		}
2683 		if (w.w_op == NET_RT_NHOP)
2684 			error = nhops_dump_sysctl(rnh, w.w_req);
2685 		else
2686 #ifdef ROUTE_MPATH
2687 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2688 #else
2689 			error = ENOTSUP;
2690 #endif
2691 		break;
2692 	case NET_RT_IFLIST:
2693 	case NET_RT_IFLISTL:
2694 		error = sysctl_iflist(af, &w);
2695 		break;
2696 
2697 	case NET_RT_IFMALIST:
2698 		error = sysctl_ifmalist(af, &w);
2699 		break;
2700 	}
2701 	NET_EPOCH_EXIT(et);
2702 
2703 	free(w.w_tmem, M_TEMP);
2704 	return (error);
2705 }
2706 
2707 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2708     sysctl_rtsock, "Return route tables and interface/address lists");
2709 
2710 /*
2711  * Definitions of protocols supported in the ROUTE domain.
2712  */
2713 
2714 static struct domain routedomain;		/* or at least forward */
2715 
2716 static struct protosw routesw[] = {
2717 {
2718 	.pr_type =		SOCK_RAW,
2719 	.pr_domain =		&routedomain,
2720 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2721 	.pr_output =		route_output,
2722 	.pr_ctlinput =		raw_ctlinput,
2723 	.pr_usrreqs =		&route_usrreqs
2724 }
2725 };
2726 
2727 static struct domain routedomain = {
2728 	.dom_family =		PF_ROUTE,
2729 	.dom_name =		"route",
2730 	.dom_protosw =		routesw,
2731 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
2732 };
2733 
2734 DOMAIN_SET(route);
2735