xref: /freebsd/sys/net/rtsock.c (revision 5ca8e32633c4ffbbcd6762e5888b6a4ba0708c6c)
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 #include "opt_ddb.h"
32 #include "opt_route.h"
33 #include "opt_inet.h"
34 #include "opt_inet6.h"
35 
36 #include <sys/param.h>
37 #include <sys/jail.h>
38 #include <sys/kernel.h>
39 #include <sys/eventhandler.h>
40 #include <sys/domain.h>
41 #include <sys/lock.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/priv.h>
45 #include <sys/proc.h>
46 #include <sys/protosw.h>
47 #include <sys/rmlock.h>
48 #include <sys/rwlock.h>
49 #include <sys/signalvar.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/systm.h>
54 
55 #include <net/if.h>
56 #include <net/if_var.h>
57 #include <net/if_private.h>
58 #include <net/if_dl.h>
59 #include <net/if_llatbl.h>
60 #include <net/if_types.h>
61 #include <net/netisr.h>
62 #include <net/route.h>
63 #include <net/route/route_ctl.h>
64 #include <net/route/route_var.h>
65 #include <net/vnet.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/if_ether.h>
69 #include <netinet/ip_carp.h>
70 #ifdef INET6
71 #include <netinet6/in6_var.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/scope6_var.h>
74 #endif
75 #include <net/route/nhop.h>
76 
77 #define	DEBUG_MOD_NAME	rtsock
78 #define	DEBUG_MAX_LEVEL	LOG_DEBUG
79 #include <net/route/route_debug.h>
80 _DECLARE_DEBUG(LOG_INFO);
81 
82 #ifdef COMPAT_FREEBSD32
83 #include <sys/mount.h>
84 #include <compat/freebsd32/freebsd32.h>
85 
86 struct if_msghdr32 {
87 	uint16_t ifm_msglen;
88 	uint8_t	ifm_version;
89 	uint8_t	ifm_type;
90 	int32_t	ifm_addrs;
91 	int32_t	ifm_flags;
92 	uint16_t ifm_index;
93 	uint16_t _ifm_spare1;
94 	struct	if_data ifm_data;
95 };
96 
97 struct if_msghdrl32 {
98 	uint16_t ifm_msglen;
99 	uint8_t	ifm_version;
100 	uint8_t	ifm_type;
101 	int32_t	ifm_addrs;
102 	int32_t	ifm_flags;
103 	uint16_t ifm_index;
104 	uint16_t _ifm_spare1;
105 	uint16_t ifm_len;
106 	uint16_t ifm_data_off;
107 	uint32_t _ifm_spare2;
108 	struct	if_data ifm_data;
109 };
110 
111 struct ifa_msghdrl32 {
112 	uint16_t ifam_msglen;
113 	uint8_t	ifam_version;
114 	uint8_t	ifam_type;
115 	int32_t	ifam_addrs;
116 	int32_t	ifam_flags;
117 	uint16_t ifam_index;
118 	uint16_t _ifam_spare1;
119 	uint16_t ifam_len;
120 	uint16_t ifam_data_off;
121 	int32_t	ifam_metric;
122 	struct	if_data ifam_data;
123 };
124 
125 #define SA_SIZE32(sa)						\
126     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
127 	sizeof(int)		:				\
128 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
129 
130 #endif /* COMPAT_FREEBSD32 */
131 
132 struct linear_buffer {
133 	char		*base;	/* Base allocated memory pointer */
134 	uint32_t	offset;	/* Currently used offset */
135 	uint32_t	size;	/* Total buffer size */
136 };
137 #define	SCRATCH_BUFFER_SIZE	1024
138 
139 #define	RTS_PID_LOG(_l, _fmt, ...)	RT_LOG_##_l(_l, "PID %d: " _fmt, curproc ? curproc->p_pid : 0, ## __VA_ARGS__)
140 
141 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
142 
143 /* NB: these are not modified */
144 static struct	sockaddr route_src = { 2, PF_ROUTE, };
145 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
146 
147 /* These are external hooks for CARP. */
148 int	(*carp_get_vhid_p)(struct ifaddr *);
149 
150 /*
151  * Used by rtsock callback code to decide whether to filter the update
152  * notification to a socket bound to a particular FIB.
153  */
154 #define	RTS_FILTER_FIB	M_PROTO8
155 /*
156  * Used to store address family of the notification.
157  */
158 #define	m_rtsock_family	m_pkthdr.PH_loc.eight[0]
159 
160 struct rcb {
161 	LIST_ENTRY(rcb) list;
162 	struct socket	*rcb_socket;
163 	sa_family_t	rcb_family;
164 };
165 
166 typedef struct {
167 	LIST_HEAD(, rcb)	cblist;
168 	int	ip_count;	/* attached w/ AF_INET */
169 	int	ip6_count;	/* attached w/ AF_INET6 */
170 	int	any_count;	/* total attached */
171 } route_cb_t;
172 VNET_DEFINE_STATIC(route_cb_t, route_cb);
173 #define	V_route_cb VNET(route_cb)
174 
175 struct mtx rtsock_mtx;
176 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
177 
178 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
179 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
180 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
181 
182 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
183 
184 struct walkarg {
185 	int	family;
186 	int	w_tmemsize;
187 	int	w_op, w_arg;
188 	caddr_t	w_tmem;
189 	struct sysctl_req *w_req;
190 	struct sockaddr *dst;
191 	struct sockaddr *mask;
192 };
193 
194 static void	rts_input(struct mbuf *m);
195 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
196 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
197 			struct walkarg *w, int *plen);
198 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
199 			struct rt_addrinfo *rtinfo);
200 static int	cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb);
201 static int	sysctl_dumpentry(struct rtentry *rt, void *vw);
202 static int	sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
203 			uint32_t weight, struct walkarg *w);
204 static int	sysctl_iflist(int af, struct walkarg *w);
205 static int	sysctl_ifmalist(int af, struct walkarg *w);
206 static void	rt_getmetrics(const struct rtentry *rt,
207 			const struct nhop_object *nh, struct rt_metrics *out);
208 static void	rt_dispatch(struct mbuf *, sa_family_t);
209 static void	rt_ifannouncemsg(struct ifnet *ifp, int what);
210 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
211 			struct rt_msghdr *rtm, struct rib_cmd_info *rc);
212 static int	update_rtm_from_rc(struct rt_addrinfo *info,
213 			struct rt_msghdr **prtm, int alloc_len,
214 			struct rib_cmd_info *rc, struct nhop_object *nh);
215 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
216 			struct mbuf *m, sa_family_t saf, u_int fibnum,
217 			int rtm_errno);
218 static void	rtsock_notify_event(uint32_t fibnum, const struct rib_cmd_info *rc);
219 static void	rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask);
220 
221 static struct netisr_handler rtsock_nh = {
222 	.nh_name = "rtsock",
223 	.nh_handler = rts_input,
224 	.nh_proto = NETISR_ROUTE,
225 	.nh_policy = NETISR_POLICY_SOURCE,
226 };
227 
228 static int
229 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
230 {
231 	int error, qlimit;
232 
233 	netisr_getqlimit(&rtsock_nh, &qlimit);
234 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
235         if (error || !req->newptr)
236                 return (error);
237 	if (qlimit < 1)
238 		return (EINVAL);
239 	return (netisr_setqlimit(&rtsock_nh, qlimit));
240 }
241 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
242     CTLTYPE_INT | CTLFLAG_RWTUN | CTLFLAG_NOFETCH | CTLFLAG_MPSAFE,
243     0, 0, sysctl_route_netisr_maxqlen, "I",
244     "maximum routing socket dispatch queue length");
245 
246 static void
247 vnet_rts_init(void)
248 {
249 	int tmp;
250 
251 	if (IS_DEFAULT_VNET(curvnet)) {
252 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
253 			rtsock_nh.nh_qlimit = tmp;
254 		netisr_register(&rtsock_nh);
255 	}
256 #ifdef VIMAGE
257 	 else
258 		netisr_register_vnet(&rtsock_nh);
259 #endif
260 }
261 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
262     vnet_rts_init, 0);
263 
264 #ifdef VIMAGE
265 static void
266 vnet_rts_uninit(void)
267 {
268 
269 	netisr_unregister_vnet(&rtsock_nh);
270 }
271 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
272     vnet_rts_uninit, 0);
273 #endif
274 
275 static void
276 report_route_event(const struct rib_cmd_info *rc, void *_cbdata)
277 {
278 	uint32_t fibnum = (uint32_t)(uintptr_t)_cbdata;
279 	struct nhop_object *nh;
280 
281 	nh = rc->rc_cmd == RTM_DELETE ? rc->rc_nh_old : rc->rc_nh_new;
282 	rt_routemsg(rc->rc_cmd, rc->rc_rt, nh, fibnum);
283 }
284 
285 static void
286 rts_handle_route_event(uint32_t fibnum, const struct rib_cmd_info *rc)
287 {
288 #ifdef ROUTE_MPATH
289 	if ((rc->rc_nh_new && NH_IS_NHGRP(rc->rc_nh_new)) ||
290 	    (rc->rc_nh_old && NH_IS_NHGRP(rc->rc_nh_old))) {
291 		rib_decompose_notification(rc, report_route_event,
292 		    (void *)(uintptr_t)fibnum);
293 	} else
294 #endif
295 		report_route_event(rc, (void *)(uintptr_t)fibnum);
296 }
297 static struct rtbridge rtsbridge = {
298 	.route_f = rts_handle_route_event,
299 	.ifmsg_f = rtsock_ifmsg,
300 };
301 static struct rtbridge *rtsbridge_orig_p;
302 
303 static void
304 rtsock_notify_event(uint32_t fibnum, const struct rib_cmd_info *rc)
305 {
306 	netlink_callback_p->route_f(fibnum, rc);
307 }
308 
309 static void
310 rtsock_init(void)
311 {
312 	rtsbridge_orig_p = rtsock_callback_p;
313 	rtsock_callback_p = &rtsbridge;
314 }
315 SYSINIT(rtsock_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rtsock_init, NULL);
316 
317 static void
318 rts_handle_ifnet_arrival(void *arg __unused, struct ifnet *ifp)
319 {
320 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
321 }
322 EVENTHANDLER_DEFINE(ifnet_arrival_event, rts_handle_ifnet_arrival, NULL, 0);
323 
324 static void
325 rts_handle_ifnet_departure(void *arg __unused, struct ifnet *ifp)
326 {
327 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
328 }
329 EVENTHANDLER_DEFINE(ifnet_departure_event, rts_handle_ifnet_departure, NULL, 0);
330 
331 static void
332 rts_append_data(struct socket *so, struct mbuf *m)
333 {
334 
335 	if (sbappendaddr(&so->so_rcv, &route_src, m, NULL) == 0) {
336 		soroverflow(so);
337 		m_freem(m);
338 	} else
339 		sorwakeup(so);
340 }
341 
342 static void
343 rts_input(struct mbuf *m)
344 {
345 	struct rcb *rcb;
346 	struct socket *last;
347 
348 	last = NULL;
349 	RTSOCK_LOCK();
350 	LIST_FOREACH(rcb, &V_route_cb.cblist, list) {
351 		if (rcb->rcb_family != AF_UNSPEC &&
352 		    rcb->rcb_family != m->m_rtsock_family)
353 			continue;
354 		if ((m->m_flags & RTS_FILTER_FIB) &&
355 		    M_GETFIB(m) != rcb->rcb_socket->so_fibnum)
356 			continue;
357 		if (last != NULL) {
358 			struct mbuf *n;
359 
360 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
361 			if (n != NULL)
362 				rts_append_data(last, n);
363 		}
364 		last = rcb->rcb_socket;
365 	}
366 	if (last != NULL)
367 		rts_append_data(last, m);
368 	else
369 		m_freem(m);
370 	RTSOCK_UNLOCK();
371 }
372 
373 static void
374 rts_close(struct socket *so)
375 {
376 
377 	soisdisconnected(so);
378 }
379 
380 static SYSCTL_NODE(_net, OID_AUTO, rtsock, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
381     "Routing socket infrastructure");
382 static u_long rts_sendspace = 8192;
383 SYSCTL_ULONG(_net_rtsock, OID_AUTO, sendspace, CTLFLAG_RW, &rts_sendspace, 0,
384     "Default routing socket send space");
385 static u_long rts_recvspace = 8192;
386 SYSCTL_ULONG(_net_rtsock, OID_AUTO, recvspace, CTLFLAG_RW, &rts_recvspace, 0,
387     "Default routing socket receive space");
388 
389 static int
390 rts_attach(struct socket *so, int proto, struct thread *td)
391 {
392 	struct rcb *rcb;
393 	int error;
394 
395 	error = soreserve(so, rts_sendspace, rts_recvspace);
396 	if (error)
397 		return (error);
398 
399 	rcb = malloc(sizeof(*rcb), M_PCB, M_WAITOK);
400 	rcb->rcb_socket = so;
401 	rcb->rcb_family = proto;
402 
403 	so->so_pcb = rcb;
404 	so->so_fibnum = td->td_proc->p_fibnum;
405 	so->so_options |= SO_USELOOPBACK;
406 
407 	RTSOCK_LOCK();
408 	LIST_INSERT_HEAD(&V_route_cb.cblist, rcb, list);
409 	switch (proto) {
410 	case AF_INET:
411 		V_route_cb.ip_count++;
412 		break;
413 	case AF_INET6:
414 		V_route_cb.ip6_count++;
415 		break;
416 	}
417 	V_route_cb.any_count++;
418 	RTSOCK_UNLOCK();
419 	soisconnected(so);
420 
421 	return (0);
422 }
423 
424 static void
425 rts_detach(struct socket *so)
426 {
427 	struct rcb *rcb = so->so_pcb;
428 
429 	RTSOCK_LOCK();
430 	LIST_REMOVE(rcb, list);
431 	switch(rcb->rcb_family) {
432 	case AF_INET:
433 		V_route_cb.ip_count--;
434 		break;
435 	case AF_INET6:
436 		V_route_cb.ip6_count--;
437 		break;
438 	}
439 	V_route_cb.any_count--;
440 	RTSOCK_UNLOCK();
441 	free(rcb, M_PCB);
442 	so->so_pcb = NULL;
443 }
444 
445 static int
446 rts_disconnect(struct socket *so)
447 {
448 
449 	return (ENOTCONN);
450 }
451 
452 static int
453 rts_shutdown(struct socket *so, enum shutdown_how how)
454 {
455 	/*
456 	 * Note: route socket marks itself as connected through its lifetime.
457 	 */
458 	switch (how) {
459 	case SHUT_RD:
460 		sorflush(so);
461 		break;
462 	case SHUT_RDWR:
463 		sorflush(so);
464 		/* FALLTHROUGH */
465 	case SHUT_WR:
466 		socantsendmore(so);
467 	}
468 
469 	return (0);
470 }
471 
472 #ifndef _SOCKADDR_UNION_DEFINED
473 #define	_SOCKADDR_UNION_DEFINED
474 /*
475  * The union of all possible address formats we handle.
476  */
477 union sockaddr_union {
478 	struct sockaddr		sa;
479 	struct sockaddr_in	sin;
480 	struct sockaddr_in6	sin6;
481 };
482 #endif /* _SOCKADDR_UNION_DEFINED */
483 
484 static int
485 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
486     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
487 {
488 #if defined(INET) || defined(INET6)
489 	struct epoch_tracker et;
490 #endif
491 
492 	/* First, see if the returned address is part of the jail. */
493 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
494 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
495 		return (0);
496 	}
497 
498 	switch (info->rti_info[RTAX_DST]->sa_family) {
499 #ifdef INET
500 	case AF_INET:
501 	{
502 		struct in_addr ia;
503 		struct ifaddr *ifa;
504 		int found;
505 
506 		found = 0;
507 		/*
508 		 * Try to find an address on the given outgoing interface
509 		 * that belongs to the jail.
510 		 */
511 		NET_EPOCH_ENTER(et);
512 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
513 			struct sockaddr *sa;
514 			sa = ifa->ifa_addr;
515 			if (sa->sa_family != AF_INET)
516 				continue;
517 			ia = ((struct sockaddr_in *)sa)->sin_addr;
518 			if (prison_check_ip4(cred, &ia) == 0) {
519 				found = 1;
520 				break;
521 			}
522 		}
523 		NET_EPOCH_EXIT(et);
524 		if (!found) {
525 			/*
526 			 * As a last resort return the 'default' jail address.
527 			 */
528 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
529 			    sin_addr;
530 			if (prison_get_ip4(cred, &ia) != 0)
531 				return (ESRCH);
532 		}
533 		bzero(&saun->sin, sizeof(struct sockaddr_in));
534 		saun->sin.sin_len = sizeof(struct sockaddr_in);
535 		saun->sin.sin_family = AF_INET;
536 		saun->sin.sin_addr.s_addr = ia.s_addr;
537 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
538 		break;
539 	}
540 #endif
541 #ifdef INET6
542 	case AF_INET6:
543 	{
544 		struct in6_addr ia6;
545 		struct ifaddr *ifa;
546 		int found;
547 
548 		found = 0;
549 		/*
550 		 * Try to find an address on the given outgoing interface
551 		 * that belongs to the jail.
552 		 */
553 		NET_EPOCH_ENTER(et);
554 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
555 			struct sockaddr *sa;
556 			sa = ifa->ifa_addr;
557 			if (sa->sa_family != AF_INET6)
558 				continue;
559 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
560 			    &ia6, sizeof(struct in6_addr));
561 			if (prison_check_ip6(cred, &ia6) == 0) {
562 				found = 1;
563 				break;
564 			}
565 		}
566 		NET_EPOCH_EXIT(et);
567 		if (!found) {
568 			/*
569 			 * As a last resort return the 'default' jail address.
570 			 */
571 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
572 			    sin6_addr;
573 			if (prison_get_ip6(cred, &ia6) != 0)
574 				return (ESRCH);
575 		}
576 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
577 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
578 		saun->sin6.sin6_family = AF_INET6;
579 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
580 		if (sa6_recoverscope(&saun->sin6) != 0)
581 			return (ESRCH);
582 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
583 		break;
584 	}
585 #endif
586 	default:
587 		return (ESRCH);
588 	}
589 	return (0);
590 }
591 
592 static int
593 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
594 {
595 	struct ifaddr *ifa;
596 	sa_family_t saf;
597 
598 	if (V_loif == NULL) {
599 		RTS_PID_LOG(LOG_INFO, "Unable to add blackhole/reject nhop without loopback");
600 		return (ENOTSUP);
601 	}
602 	info->rti_ifp = V_loif;
603 
604 	saf = info->rti_info[RTAX_DST]->sa_family;
605 
606 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
607 		if (ifa->ifa_addr->sa_family == saf) {
608 			info->rti_ifa = ifa;
609 			break;
610 		}
611 	}
612 	if (info->rti_ifa == NULL) {
613 		RTS_PID_LOG(LOG_INFO, "Unable to find ifa for blackhole/reject nhop");
614 		return (ENOTSUP);
615 	}
616 
617 	bzero(saun, sizeof(union sockaddr_union));
618 	switch (saf) {
619 #ifdef INET
620 	case AF_INET:
621 		saun->sin.sin_family = AF_INET;
622 		saun->sin.sin_len = sizeof(struct sockaddr_in);
623 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
624 		break;
625 #endif
626 #ifdef INET6
627 	case AF_INET6:
628 		saun->sin6.sin6_family = AF_INET6;
629 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
630 		saun->sin6.sin6_addr = in6addr_loopback;
631 		break;
632 #endif
633 	default:
634 		RTS_PID_LOG(LOG_INFO, "unsupported family: %d", saf);
635 		return (ENOTSUP);
636 	}
637 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
638 	info->rti_flags |= RTF_GATEWAY;
639 
640 	return (0);
641 }
642 
643 /*
644  * Fills in @info based on userland-provided @rtm message.
645  *
646  * Returns 0 on success.
647  */
648 static int
649 fill_addrinfo(struct rt_msghdr *rtm, int len, struct linear_buffer *lb, u_int fibnum,
650     struct rt_addrinfo *info)
651 {
652 	int error;
653 
654 	rtm->rtm_pid = curproc->p_pid;
655 	info->rti_addrs = rtm->rtm_addrs;
656 
657 	info->rti_mflags = rtm->rtm_inits;
658 	info->rti_rmx = &rtm->rtm_rmx;
659 
660 	/*
661 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
662 	 * link-local address because rtrequest requires addresses with
663 	 * embedded scope id.
664 	 */
665 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
666 		return (EINVAL);
667 
668 	info->rti_flags = rtm->rtm_flags;
669 	error = cleanup_xaddrs(info, lb);
670 	if (error != 0)
671 		return (error);
672 	/*
673 	 * Verify that the caller has the appropriate privilege; RTM_GET
674 	 * is the only operation the non-superuser is allowed.
675 	 */
676 	if (rtm->rtm_type != RTM_GET) {
677 		error = priv_check(curthread, PRIV_NET_ROUTE);
678 		if (error != 0)
679 			return (error);
680 	}
681 
682 	/*
683 	 * The given gateway address may be an interface address.
684 	 * For example, issuing a "route change" command on a route
685 	 * entry that was created from a tunnel, and the gateway
686 	 * address given is the local end point. In this case the
687 	 * RTF_GATEWAY flag must be cleared or the destination will
688 	 * not be reachable even though there is no error message.
689 	 */
690 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
691 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
692 		struct nhop_object *nh;
693 
694 		/*
695 		 * A host route through the loopback interface is
696 		 * installed for each interface address. In pre 8.0
697 		 * releases the interface address of a PPP link type
698 		 * is not reachable locally. This behavior is fixed as
699 		 * part of the new L2/L3 redesign and rewrite work. The
700 		 * signature of this interface address route is the
701 		 * AF_LINK sa_family type of the gateway, and the
702 		 * rt_ifp has the IFF_LOOPBACK flag set.
703 		 */
704 		nh = rib_lookup(fibnum, info->rti_info[RTAX_GATEWAY], NHR_NONE, 0);
705 		if (nh != NULL && nh->gw_sa.sa_family == AF_LINK &&
706 		    nh->nh_ifp->if_flags & IFF_LOOPBACK) {
707 				info->rti_flags &= ~RTF_GATEWAY;
708 				info->rti_flags |= RTF_GWFLAG_COMPAT;
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(const 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(const 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 static int
1029 rts_send(struct socket *so, int flags, struct mbuf *m,
1030     struct sockaddr *nam, struct mbuf *control, struct thread *td)
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 	if ((flags & PRUS_OOB) || control != NULL) {
1046 		m_freem(m);
1047 		if (control != NULL)
1048 			m_freem(control);
1049 		return (EOPNOTSUPP);
1050 	}
1051 
1052 	fibnum = so->so_fibnum;
1053 #define senderr(e) { error = e; goto flush;}
1054 	if (m == NULL || ((m->m_len < sizeof(long)) &&
1055 		       (m = m_pullup(m, sizeof(long))) == NULL))
1056 		return (ENOBUFS);
1057 	if ((m->m_flags & M_PKTHDR) == 0)
1058 		panic("route_output");
1059 	NET_EPOCH_ENTER(et);
1060 	len = m->m_pkthdr.len;
1061 	if (len < sizeof(*rtm) ||
1062 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1063 		senderr(EINVAL);
1064 
1065 	/*
1066 	 * Most of current messages are in range 200-240 bytes,
1067 	 * minimize possible re-allocation on reply using larger size
1068 	 * buffer aligned on 1k boundaty.
1069 	 */
1070 	alloc_len = roundup2(len, 1024);
1071 	int total_len = alloc_len + SCRATCH_BUFFER_SIZE;
1072 	if ((rtm = malloc(total_len, M_TEMP, M_NOWAIT)) == NULL)
1073 		senderr(ENOBUFS);
1074 
1075 	m_copydata(m, 0, len, (caddr_t)rtm);
1076 	bzero(&info, sizeof(info));
1077 	nh = NULL;
1078 	struct linear_buffer lb = {
1079 		.base = (char *)rtm + alloc_len,
1080 		.size = SCRATCH_BUFFER_SIZE,
1081 	};
1082 
1083 	if (rtm->rtm_version != RTM_VERSION) {
1084 		/* Do not touch message since format is unknown */
1085 		free(rtm, M_TEMP);
1086 		rtm = NULL;
1087 		senderr(EPROTONOSUPPORT);
1088 	}
1089 
1090 	/*
1091 	 * Starting from here, it is possible
1092 	 * to alter original message and insert
1093 	 * caller PID and error value.
1094 	 */
1095 
1096 	if ((error = fill_addrinfo(rtm, len, &lb, fibnum, &info)) != 0) {
1097 		senderr(error);
1098 	}
1099 	/* fill_addringo() embeds scope into IPv6 addresses */
1100 #ifdef INET6
1101 	rti_need_deembed = 1;
1102 #endif
1103 
1104 	saf = info.rti_info[RTAX_DST]->sa_family;
1105 
1106 	/* support for new ARP code */
1107 	if (rtm->rtm_flags & RTF_LLDATA) {
1108 		error = lla_rt_output(rtm, &info);
1109 		goto flush;
1110 	}
1111 
1112 	union sockaddr_union gw_saun;
1113 	int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1114 	if (blackhole_flags != 0) {
1115 		if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1116 			error = fill_blackholeinfo(&info, &gw_saun);
1117 		else {
1118 			RTS_PID_LOG(LOG_DEBUG, "both BLACKHOLE and REJECT flags specifiied");
1119 			error = EINVAL;
1120 		}
1121 		if (error != 0)
1122 			senderr(error);
1123 	}
1124 
1125 	switch (rtm->rtm_type) {
1126 	case RTM_ADD:
1127 	case RTM_CHANGE:
1128 		if (rtm->rtm_type == RTM_ADD) {
1129 			if (info.rti_info[RTAX_GATEWAY] == NULL) {
1130 				RTS_PID_LOG(LOG_DEBUG, "RTM_ADD w/o gateway");
1131 				senderr(EINVAL);
1132 			}
1133 		}
1134 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1135 		if (error == 0) {
1136 			rtsock_notify_event(fibnum, &rc);
1137 #ifdef ROUTE_MPATH
1138 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
1139 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1140 				struct rib_cmd_info rc_simple = {};
1141 				rib_decompose_notification(&rc,
1142 				    save_add_notification, (void *)&rc_simple);
1143 				rc = rc_simple;
1144 			}
1145 #endif
1146 			/* nh MAY be empty if RTM_CHANGE request is no-op */
1147 			nh = rc.rc_nh_new;
1148 			if (nh != NULL) {
1149 				rtm->rtm_index = nh->nh_ifp->if_index;
1150 				rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1151 			}
1152 		}
1153 		break;
1154 
1155 	case RTM_DELETE:
1156 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1157 		if (error == 0) {
1158 			rtsock_notify_event(fibnum, &rc);
1159 #ifdef ROUTE_MPATH
1160 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1161 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1162 				struct rib_cmd_info rc_simple = {};
1163 				rib_decompose_notification(&rc,
1164 				    save_del_notification, (void *)&rc_simple);
1165 				rc = rc_simple;
1166 			}
1167 #endif
1168 			nh = rc.rc_nh_old;
1169 		}
1170 		break;
1171 
1172 	case RTM_GET:
1173 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1174 		if (error != 0)
1175 			senderr(error);
1176 		nh = rc.rc_nh_new;
1177 
1178 		if (!rt_is_exportable(rc.rc_rt, curthread->td_ucred))
1179 			senderr(ESRCH);
1180 		break;
1181 
1182 	default:
1183 		senderr(EOPNOTSUPP);
1184 	}
1185 
1186 	if (error == 0 && nh != NULL) {
1187 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1188 		/*
1189 		 * Note that some sockaddr pointers may have changed to
1190 		 * point to memory outsize @rtm. Some may be pointing
1191 		 * to the on-stack variables.
1192 		 * Given that, any pointer in @info CANNOT BE USED.
1193 		 */
1194 
1195 		/*
1196 		 * scopeid deembedding has been performed while
1197 		 * writing updated rtm in rtsock_msg_buffer().
1198 		 * With that in mind, skip deembedding procedure below.
1199 		 */
1200 #ifdef INET6
1201 		rti_need_deembed = 0;
1202 #endif
1203 	}
1204 
1205 flush:
1206 	NET_EPOCH_EXIT(et);
1207 
1208 #ifdef INET6
1209 	if (rtm != NULL) {
1210 		if (rti_need_deembed) {
1211 			/* sin6_scope_id is recovered before sending rtm. */
1212 			sin6 = (struct sockaddr_in6 *)&ss;
1213 			for (i = 0; i < RTAX_MAX; i++) {
1214 				if (info.rti_info[i] == NULL)
1215 					continue;
1216 				if (info.rti_info[i]->sa_family != AF_INET6)
1217 					continue;
1218 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1219 				if (sa6_recoverscope(sin6) == 0)
1220 					bcopy(sin6, info.rti_info[i],
1221 						    sizeof(*sin6));
1222 			}
1223 			if (update_rtm_from_info(&info, &rtm, alloc_len) != 0) {
1224 				if (error != 0)
1225 					error = ENOBUFS;
1226 			}
1227 		}
1228 	}
1229 #endif
1230 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1231 
1232 	return (error);
1233 }
1234 
1235 /*
1236  * Sends the prepared reply message in @rtm to all rtsock clients.
1237  * Frees @m and @rtm.
1238  *
1239  */
1240 static void
1241 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1242     sa_family_t saf, u_int fibnum, int rtm_errno)
1243 {
1244 	struct rcb *rcb = NULL;
1245 
1246 	/*
1247 	 * Check to see if we don't want our own messages.
1248 	 */
1249 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1250 		if (V_route_cb.any_count <= 1) {
1251 			if (rtm != NULL)
1252 				free(rtm, M_TEMP);
1253 			m_freem(m);
1254 			return;
1255 		}
1256 		/* There is another listener, so construct message */
1257 		rcb = so->so_pcb;
1258 	}
1259 
1260 	if (rtm != NULL) {
1261 		if (rtm_errno!= 0)
1262 			rtm->rtm_errno = rtm_errno;
1263 		else
1264 			rtm->rtm_flags |= RTF_DONE;
1265 
1266 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1267 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1268 			m_freem(m);
1269 			m = NULL;
1270 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1271 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1272 
1273 		free(rtm, M_TEMP);
1274 	}
1275 	if (m != NULL) {
1276 		M_SETFIB(m, fibnum);
1277 		m->m_flags |= RTS_FILTER_FIB;
1278 		if (rcb) {
1279 			/*
1280 			 * XXX insure we don't get a copy by
1281 			 * invalidating our protocol
1282 			 */
1283 			sa_family_t family = rcb->rcb_family;
1284 			rcb->rcb_family = AF_UNSPEC;
1285 			rt_dispatch(m, saf);
1286 			rcb->rcb_family = family;
1287 		} else
1288 			rt_dispatch(m, saf);
1289 	}
1290 }
1291 
1292 static void
1293 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1294     struct rt_metrics *out)
1295 {
1296 
1297 	bzero(out, sizeof(*out));
1298 	out->rmx_mtu = nh->nh_mtu;
1299 	out->rmx_weight = rt->rt_weight;
1300 	out->rmx_nhidx = nhop_get_idx(nh);
1301 	/* Kernel -> userland timebase conversion. */
1302 	out->rmx_expire = nhop_get_expire(nh) ?
1303 	    nhop_get_expire(nh) - time_uptime + time_second : 0;
1304 }
1305 
1306 /*
1307  * Extract the addresses of the passed sockaddrs.
1308  * Do a little sanity checking so as to avoid bad memory references.
1309  * This data is derived straight from userland.
1310  */
1311 static int
1312 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1313 {
1314 	struct sockaddr *sa;
1315 	int i;
1316 
1317 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1318 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1319 			continue;
1320 		sa = (struct sockaddr *)cp;
1321 		/*
1322 		 * It won't fit.
1323 		 */
1324 		if (cp + sa->sa_len > cplim) {
1325 			RTS_PID_LOG(LOG_DEBUG, "sa_len too big for sa type %d", i);
1326 			return (EINVAL);
1327 		}
1328 		/*
1329 		 * there are no more.. quit now
1330 		 * If there are more bits, they are in error.
1331 		 * I've seen this. route(1) can evidently generate these.
1332 		 * This causes kernel to core dump.
1333 		 * for compatibility, If we see this, point to a safe address.
1334 		 */
1335 		if (sa->sa_len == 0) {
1336 			rtinfo->rti_info[i] = &sa_zero;
1337 			return (0); /* should be EINVAL but for compat */
1338 		}
1339 		/* accept it */
1340 #ifdef INET6
1341 		if (sa->sa_family == AF_INET6)
1342 			sa6_embedscope((struct sockaddr_in6 *)sa,
1343 			    V_ip6_use_defzone);
1344 #endif
1345 		rtinfo->rti_info[i] = sa;
1346 		cp += SA_SIZE(sa);
1347 	}
1348 	return (0);
1349 }
1350 
1351 #ifdef INET
1352 static inline void
1353 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1354 {
1355 
1356 	const struct sockaddr_in nsin = {
1357 		.sin_family = AF_INET,
1358 		.sin_len = sizeof(struct sockaddr_in),
1359 		.sin_addr = addr,
1360 	};
1361 	*sin = nsin;
1362 }
1363 #endif
1364 
1365 #ifdef INET6
1366 static inline void
1367 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1368     uint32_t scopeid)
1369 {
1370 
1371 	const struct sockaddr_in6 nsin6 = {
1372 		.sin6_family = AF_INET6,
1373 		.sin6_len = sizeof(struct sockaddr_in6),
1374 		.sin6_addr = *addr6,
1375 		.sin6_scope_id = scopeid,
1376 	};
1377 	*sin6 = nsin6;
1378 }
1379 #endif
1380 
1381 #if defined(INET6) || defined(INET)
1382 /*
1383  * Checks if gateway is suitable for lltable operations.
1384  * Lltable code requires AF_LINK gateway with ifindex
1385  *  and mac address specified.
1386  * Returns 0 on success.
1387  */
1388 static int
1389 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1390 {
1391 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1392 
1393 	if (sdl->sdl_family != AF_LINK)
1394 		return (EINVAL);
1395 
1396 	if (sdl->sdl_index == 0) {
1397 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gateway w/o ifindex");
1398 		return (EINVAL);
1399 	}
1400 
1401 	if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len) {
1402 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gw: sdl_nlen/sdl_alen too large");
1403 		return (EINVAL);
1404 	}
1405 
1406 	return (0);
1407 }
1408 
1409 static int
1410 cleanup_xaddrs_gateway(struct rt_addrinfo *info, struct linear_buffer *lb)
1411 {
1412 	struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1413 	struct sockaddr *sa;
1414 
1415 	if (info->rti_flags & RTF_LLDATA)
1416 		return (cleanup_xaddrs_lladdr(info));
1417 
1418 	switch (gw->sa_family) {
1419 #ifdef INET
1420 	case AF_INET:
1421 		{
1422 			struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw;
1423 
1424 			/* Ensure reads do not go beyoud SA boundary */
1425 			if (SA_SIZE(gw) < offsetof(struct sockaddr_in, sin_zero)) {
1426 				RTS_PID_LOG(LOG_DEBUG, "gateway sin_len too small: %d",
1427 				    gw->sa_len);
1428 				return (EINVAL);
1429 			}
1430 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_in));
1431 			if (sa == NULL)
1432 				return (ENOBUFS);
1433 			fill_sockaddr_inet((struct sockaddr_in *)sa, gw_sin->sin_addr);
1434 			info->rti_info[RTAX_GATEWAY] = sa;
1435 		}
1436 		break;
1437 #endif
1438 #ifdef INET6
1439 	case AF_INET6:
1440 		{
1441 			struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1442 			if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1443 				RTS_PID_LOG(LOG_DEBUG, "gateway sin6_len too small: %d",
1444 				    gw->sa_len);
1445 				return (EINVAL);
1446 			}
1447 			fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1448 			break;
1449 		}
1450 #endif
1451 	case AF_LINK:
1452 		{
1453 			struct sockaddr_dl *gw_sdl;
1454 
1455 			size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data);
1456 			gw_sdl = (struct sockaddr_dl *)gw;
1457 			if (gw_sdl->sdl_len < sdl_min_len) {
1458 				RTS_PID_LOG(LOG_DEBUG, "gateway sdl_len too small: %d",
1459 				    gw_sdl->sdl_len);
1460 				return (EINVAL);
1461 			}
1462 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_dl_short));
1463 			if (sa == NULL)
1464 				return (ENOBUFS);
1465 
1466 			const struct sockaddr_dl_short sdl = {
1467 				.sdl_family = AF_LINK,
1468 				.sdl_len = sizeof(struct sockaddr_dl_short),
1469 				.sdl_index = gw_sdl->sdl_index,
1470 			};
1471 			*((struct sockaddr_dl_short *)sa) = sdl;
1472 			info->rti_info[RTAX_GATEWAY] = sa;
1473 			break;
1474 		}
1475 	}
1476 
1477 	return (0);
1478 }
1479 #endif
1480 
1481 static void
1482 remove_netmask(struct rt_addrinfo *info)
1483 {
1484 	info->rti_info[RTAX_NETMASK] = NULL;
1485 	info->rti_flags |= RTF_HOST;
1486 	info->rti_addrs &= ~RTA_NETMASK;
1487 }
1488 
1489 #ifdef INET
1490 static int
1491 cleanup_xaddrs_inet(struct rt_addrinfo *info, struct linear_buffer *lb)
1492 {
1493 	struct sockaddr_in *dst_sa, *mask_sa;
1494 	const int sa_len = sizeof(struct sockaddr_in);
1495 	struct in_addr dst, mask;
1496 
1497 	/* Check & fixup dst/netmask combination first */
1498 	dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1499 	mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1500 
1501 	/* Ensure reads do not go beyound the buffer size */
1502 	if (SA_SIZE(dst_sa) < offsetof(struct sockaddr_in, sin_zero)) {
1503 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin_len too small: %d",
1504 		    dst_sa->sin_len);
1505 		return (EINVAL);
1506 	}
1507 
1508 	if ((mask_sa != NULL) && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1509 		/*
1510 		 * Some older routing software encode mask length into the
1511 		 * sin_len, thus resulting in "truncated" sockaddr.
1512 		 */
1513 		int len = mask_sa->sin_len - offsetof(struct sockaddr_in, sin_addr);
1514 		if (len >= 0) {
1515 			mask.s_addr = 0;
1516 			if (len > sizeof(struct in_addr))
1517 				len = sizeof(struct in_addr);
1518 			memcpy(&mask, &mask_sa->sin_addr, len);
1519 		} else {
1520 			RTS_PID_LOG(LOG_DEBUG, "prefix mask sin_len too small: %d",
1521 			    mask_sa->sin_len);
1522 			return (EINVAL);
1523 		}
1524 	} else
1525 		mask.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST;
1526 
1527 	dst.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr));
1528 
1529 	/* Construct new "clean" dst/mask sockaddresses */
1530 	if ((dst_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1531 		return (ENOBUFS);
1532 	fill_sockaddr_inet(dst_sa, dst);
1533 	info->rti_info[RTAX_DST] = (struct sockaddr *)dst_sa;
1534 
1535 	if (mask.s_addr != INADDR_BROADCAST) {
1536 		if ((mask_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1537 			return (ENOBUFS);
1538 		fill_sockaddr_inet(mask_sa, mask);
1539 		info->rti_info[RTAX_NETMASK] = (struct sockaddr *)mask_sa;
1540 		info->rti_flags &= ~RTF_HOST;
1541 	} else
1542 		remove_netmask(info);
1543 
1544 	/* Check gateway */
1545 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1546 		return (cleanup_xaddrs_gateway(info, lb));
1547 
1548 	return (0);
1549 }
1550 #endif
1551 
1552 #ifdef INET6
1553 static int
1554 cleanup_xaddrs_inet6(struct rt_addrinfo *info, struct linear_buffer *lb)
1555 {
1556 	struct sockaddr *sa;
1557 	struct sockaddr_in6 *dst_sa, *mask_sa;
1558 	struct in6_addr mask, *dst;
1559 	const int sa_len = sizeof(struct sockaddr_in6);
1560 
1561 	/* Check & fixup dst/netmask combination first */
1562 	dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1563 	mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1564 
1565 	if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1566 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin6_len too small: %d",
1567 		    dst_sa->sin6_len);
1568 		return (EINVAL);
1569 	}
1570 
1571 	if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1572 		/*
1573 		 * Some older routing software encode mask length into the
1574 		 * sin6_len, thus resulting in "truncated" sockaddr.
1575 		 */
1576 		int len = mask_sa->sin6_len - offsetof(struct sockaddr_in6, sin6_addr);
1577 		if (len >= 0) {
1578 			bzero(&mask, sizeof(mask));
1579 			if (len > sizeof(struct in6_addr))
1580 				len = sizeof(struct in6_addr);
1581 			memcpy(&mask, &mask_sa->sin6_addr, len);
1582 		} else {
1583 			RTS_PID_LOG(LOG_DEBUG, "rtsock: prefix mask sin6_len too small: %d",
1584 			    mask_sa->sin6_len);
1585 			return (EINVAL);
1586 		}
1587 	} else
1588 		mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1589 
1590 	dst = &dst_sa->sin6_addr;
1591 	IN6_MASK_ADDR(dst, &mask);
1592 
1593 	if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1594 		return (ENOBUFS);
1595 	fill_sockaddr_inet6((struct sockaddr_in6 *)sa, dst, 0);
1596 	info->rti_info[RTAX_DST] = sa;
1597 
1598 	if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128)) {
1599 		if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1600 			return (ENOBUFS);
1601 		fill_sockaddr_inet6((struct sockaddr_in6 *)sa, &mask, 0);
1602 		info->rti_info[RTAX_NETMASK] = sa;
1603 		info->rti_flags &= ~RTF_HOST;
1604 	} else
1605 		remove_netmask(info);
1606 
1607 	/* Check gateway */
1608 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1609 		return (cleanup_xaddrs_gateway(info, lb));
1610 
1611 	return (0);
1612 }
1613 #endif
1614 
1615 static int
1616 cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb)
1617 {
1618 	int error = EAFNOSUPPORT;
1619 
1620 	if (info->rti_info[RTAX_DST] == NULL) {
1621 		RTS_PID_LOG(LOG_DEBUG, "prefix dst is not set");
1622 		return (EINVAL);
1623 	}
1624 
1625 	if (info->rti_flags & RTF_LLDATA) {
1626 		/*
1627 		 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1628 		 * prefix along with the actual address in RTA_DST.
1629 		 * Remove netmask to avoid unnecessary address masking.
1630 		 */
1631 		remove_netmask(info);
1632 	}
1633 
1634 	switch (info->rti_info[RTAX_DST]->sa_family) {
1635 #ifdef INET
1636 	case AF_INET:
1637 		error = cleanup_xaddrs_inet(info, lb);
1638 		break;
1639 #endif
1640 #ifdef INET6
1641 	case AF_INET6:
1642 		error = cleanup_xaddrs_inet6(info, lb);
1643 		break;
1644 #endif
1645 	}
1646 
1647 	return (error);
1648 }
1649 
1650 /*
1651  * Fill in @dmask with valid netmask leaving original @smask
1652  * intact. Mostly used with radix netmasks.
1653  */
1654 struct sockaddr *
1655 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1656     struct sockaddr_storage *dmask)
1657 {
1658 	if (dst == NULL || smask == NULL)
1659 		return (NULL);
1660 
1661 	memset(dmask, 0, dst->sa_len);
1662 	memcpy(dmask, smask, smask->sa_len);
1663 	dmask->ss_len = dst->sa_len;
1664 	dmask->ss_family = dst->sa_family;
1665 
1666 	return ((struct sockaddr *)dmask);
1667 }
1668 
1669 /*
1670  * Writes information related to @rtinfo object to newly-allocated mbuf.
1671  * Assumes MCLBYTES is enough to construct any message.
1672  * Used for OS notifications of vaious events (if/ifa announces,etc)
1673  *
1674  * Returns allocated mbuf or NULL on failure.
1675  */
1676 static struct mbuf *
1677 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1678 {
1679 	struct sockaddr_storage ss;
1680 	struct rt_msghdr *rtm;
1681 	struct mbuf *m;
1682 	int i;
1683 	struct sockaddr *sa;
1684 #ifdef INET6
1685 	struct sockaddr_in6 *sin6;
1686 #endif
1687 	int len, dlen;
1688 
1689 	switch (type) {
1690 	case RTM_DELADDR:
1691 	case RTM_NEWADDR:
1692 		len = sizeof(struct ifa_msghdr);
1693 		break;
1694 
1695 	case RTM_DELMADDR:
1696 	case RTM_NEWMADDR:
1697 		len = sizeof(struct ifma_msghdr);
1698 		break;
1699 
1700 	case RTM_IFINFO:
1701 		len = sizeof(struct if_msghdr);
1702 		break;
1703 
1704 	case RTM_IFANNOUNCE:
1705 	case RTM_IEEE80211:
1706 		len = sizeof(struct if_announcemsghdr);
1707 		break;
1708 
1709 	default:
1710 		len = sizeof(struct rt_msghdr);
1711 	}
1712 
1713 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1714 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1715 	if (len > MHLEN)
1716 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1717 	else
1718 		m = m_gethdr(M_NOWAIT, MT_DATA);
1719 	if (m == NULL)
1720 		return (m);
1721 
1722 	m->m_pkthdr.len = m->m_len = len;
1723 	rtm = mtod(m, struct rt_msghdr *);
1724 	bzero((caddr_t)rtm, len);
1725 	for (i = 0; i < RTAX_MAX; i++) {
1726 		if ((sa = rtinfo->rti_info[i]) == NULL)
1727 			continue;
1728 		rtinfo->rti_addrs |= (1 << i);
1729 
1730 		dlen = SA_SIZE(sa);
1731 		KASSERT(dlen <= sizeof(ss),
1732 		    ("%s: sockaddr size overflow", __func__));
1733 		bzero(&ss, sizeof(ss));
1734 		bcopy(sa, &ss, sa->sa_len);
1735 		sa = (struct sockaddr *)&ss;
1736 #ifdef INET6
1737 		if (sa->sa_family == AF_INET6) {
1738 			sin6 = (struct sockaddr_in6 *)sa;
1739 			(void)sa6_recoverscope(sin6);
1740 		}
1741 #endif
1742 		m_copyback(m, len, dlen, (caddr_t)sa);
1743 		len += dlen;
1744 	}
1745 	if (m->m_pkthdr.len != len) {
1746 		m_freem(m);
1747 		return (NULL);
1748 	}
1749 	rtm->rtm_msglen = len;
1750 	rtm->rtm_version = RTM_VERSION;
1751 	rtm->rtm_type = type;
1752 	return (m);
1753 }
1754 
1755 /*
1756  * Writes information related to @rtinfo object to preallocated buffer.
1757  * Stores needed size in @plen. If @w is NULL, calculates size without
1758  * writing.
1759  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1760  *
1761  * Returns 0 on success.
1762  *
1763  */
1764 static int
1765 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1766 {
1767 	struct sockaddr_storage ss;
1768 	int len, buflen = 0, dlen, i;
1769 	caddr_t cp = NULL;
1770 	struct rt_msghdr *rtm = NULL;
1771 #ifdef INET6
1772 	struct sockaddr_in6 *sin6;
1773 #endif
1774 #ifdef COMPAT_FREEBSD32
1775 	bool compat32 = false;
1776 #endif
1777 
1778 	switch (type) {
1779 	case RTM_DELADDR:
1780 	case RTM_NEWADDR:
1781 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1782 #ifdef COMPAT_FREEBSD32
1783 			if (w->w_req->flags & SCTL_MASK32) {
1784 				len = sizeof(struct ifa_msghdrl32);
1785 				compat32 = true;
1786 			} else
1787 #endif
1788 				len = sizeof(struct ifa_msghdrl);
1789 		} else
1790 			len = sizeof(struct ifa_msghdr);
1791 		break;
1792 
1793 	case RTM_IFINFO:
1794 #ifdef COMPAT_FREEBSD32
1795 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1796 			if (w->w_op == NET_RT_IFLISTL)
1797 				len = sizeof(struct if_msghdrl32);
1798 			else
1799 				len = sizeof(struct if_msghdr32);
1800 			compat32 = true;
1801 			break;
1802 		}
1803 #endif
1804 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1805 			len = sizeof(struct if_msghdrl);
1806 		else
1807 			len = sizeof(struct if_msghdr);
1808 		break;
1809 
1810 	case RTM_NEWMADDR:
1811 		len = sizeof(struct ifma_msghdr);
1812 		break;
1813 
1814 	default:
1815 		len = sizeof(struct rt_msghdr);
1816 	}
1817 
1818 	if (w != NULL) {
1819 		rtm = (struct rt_msghdr *)w->w_tmem;
1820 		buflen = w->w_tmemsize - len;
1821 		cp = (caddr_t)w->w_tmem + len;
1822 	}
1823 
1824 	rtinfo->rti_addrs = 0;
1825 	for (i = 0; i < RTAX_MAX; i++) {
1826 		struct sockaddr *sa;
1827 
1828 		if ((sa = rtinfo->rti_info[i]) == NULL)
1829 			continue;
1830 		rtinfo->rti_addrs |= (1 << i);
1831 #ifdef COMPAT_FREEBSD32
1832 		if (compat32)
1833 			dlen = SA_SIZE32(sa);
1834 		else
1835 #endif
1836 			dlen = SA_SIZE(sa);
1837 		if (cp != NULL && buflen >= dlen) {
1838 			KASSERT(dlen <= sizeof(ss),
1839 			    ("%s: sockaddr size overflow", __func__));
1840 			bzero(&ss, sizeof(ss));
1841 			bcopy(sa, &ss, sa->sa_len);
1842 			sa = (struct sockaddr *)&ss;
1843 #ifdef INET6
1844 			if (sa->sa_family == AF_INET6) {
1845 				sin6 = (struct sockaddr_in6 *)sa;
1846 				(void)sa6_recoverscope(sin6);
1847 			}
1848 #endif
1849 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1850 			cp += dlen;
1851 			buflen -= dlen;
1852 		} else if (cp != NULL) {
1853 			/*
1854 			 * Buffer too small. Count needed size
1855 			 * and return with error.
1856 			 */
1857 			cp = NULL;
1858 		}
1859 
1860 		len += dlen;
1861 	}
1862 
1863 	if (cp != NULL) {
1864 		dlen = ALIGN(len) - len;
1865 		if (buflen < dlen)
1866 			cp = NULL;
1867 		else {
1868 			bzero(cp, dlen);
1869 			cp += dlen;
1870 			buflen -= dlen;
1871 		}
1872 	}
1873 	len = ALIGN(len);
1874 
1875 	if (cp != NULL) {
1876 		/* fill header iff buffer is large enough */
1877 		rtm->rtm_version = RTM_VERSION;
1878 		rtm->rtm_type = type;
1879 		rtm->rtm_msglen = len;
1880 	}
1881 
1882 	*plen = len;
1883 
1884 	if (w != NULL && cp == NULL)
1885 		return (ENOBUFS);
1886 
1887 	return (0);
1888 }
1889 
1890 /*
1891  * This routine is called to generate a message from the routing
1892  * socket indicating that a redirect has occurred, a routing lookup
1893  * has failed, or that a protocol has detected timeouts to a particular
1894  * destination.
1895  */
1896 void
1897 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1898     int fibnum)
1899 {
1900 	struct rt_msghdr *rtm;
1901 	struct mbuf *m;
1902 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1903 
1904 	if (V_route_cb.any_count == 0)
1905 		return;
1906 	m = rtsock_msg_mbuf(type, rtinfo);
1907 	if (m == NULL)
1908 		return;
1909 
1910 	if (fibnum != RT_ALL_FIBS) {
1911 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1912 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1913 		M_SETFIB(m, fibnum);
1914 		m->m_flags |= RTS_FILTER_FIB;
1915 	}
1916 
1917 	rtm = mtod(m, struct rt_msghdr *);
1918 	rtm->rtm_flags = RTF_DONE | flags;
1919 	rtm->rtm_errno = error;
1920 	rtm->rtm_addrs = rtinfo->rti_addrs;
1921 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1922 }
1923 
1924 void
1925 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1926 {
1927 
1928 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1929 }
1930 
1931 /*
1932  * This routine is called to generate a message from the routing
1933  * socket indicating that the status of a network interface has changed.
1934  */
1935 static void
1936 rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask __unused)
1937 {
1938 	struct if_msghdr *ifm;
1939 	struct mbuf *m;
1940 	struct rt_addrinfo info;
1941 
1942 	if (V_route_cb.any_count == 0)
1943 		return;
1944 	bzero((caddr_t)&info, sizeof(info));
1945 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1946 	if (m == NULL)
1947 		return;
1948 	ifm = mtod(m, struct if_msghdr *);
1949 	ifm->ifm_index = ifp->if_index;
1950 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1951 	if_data_copy(ifp, &ifm->ifm_data);
1952 	ifm->ifm_addrs = 0;
1953 	rt_dispatch(m, AF_UNSPEC);
1954 }
1955 
1956 /*
1957  * Announce interface address arrival/withdraw.
1958  * Please do not call directly, use rt_addrmsg().
1959  * Assume input data to be valid.
1960  * Returns 0 on success.
1961  */
1962 int
1963 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1964 {
1965 	struct rt_addrinfo info;
1966 	struct sockaddr *sa;
1967 	int ncmd;
1968 	struct mbuf *m;
1969 	struct ifa_msghdr *ifam;
1970 	struct ifnet *ifp = ifa->ifa_ifp;
1971 	struct sockaddr_storage ss;
1972 
1973 	if (V_route_cb.any_count == 0)
1974 		return (0);
1975 
1976 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1977 
1978 	bzero((caddr_t)&info, sizeof(info));
1979 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1980 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1981 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1982 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1983 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1984 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1985 		return (ENOBUFS);
1986 	ifam = mtod(m, struct ifa_msghdr *);
1987 	ifam->ifam_index = ifp->if_index;
1988 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1989 	ifam->ifam_flags = ifa->ifa_flags;
1990 	ifam->ifam_addrs = info.rti_addrs;
1991 
1992 	if (fibnum != RT_ALL_FIBS) {
1993 		M_SETFIB(m, fibnum);
1994 		m->m_flags |= RTS_FILTER_FIB;
1995 	}
1996 
1997 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1998 
1999 	return (0);
2000 }
2001 
2002 /*
2003  * Announce route addition/removal to rtsock based on @rt data.
2004  * Callers are advives to use rt_routemsg() instead of using this
2005  *  function directly.
2006  * Assume @rt data is consistent.
2007  *
2008  * Returns 0 on success.
2009  */
2010 int
2011 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
2012     int fibnum)
2013 {
2014 	union sockaddr_union dst, mask;
2015 	struct rt_addrinfo info;
2016 
2017 	if (V_route_cb.any_count == 0)
2018 		return (0);
2019 
2020 	int family = rt_get_family(rt);
2021 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
2022 	export_rtaddrs(rt, &dst.sa, &mask.sa);
2023 
2024 	bzero((caddr_t)&info, sizeof(info));
2025 	info.rti_info[RTAX_DST] = &dst.sa;
2026 	info.rti_info[RTAX_NETMASK] = &mask.sa;
2027 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2028 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
2029 	info.rti_ifp = nh->nh_ifp;
2030 
2031 	return (rtsock_routemsg_info(cmd, &info, fibnum));
2032 }
2033 
2034 int
2035 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
2036 {
2037 	struct rt_msghdr *rtm;
2038 	struct sockaddr *sa;
2039 	struct mbuf *m;
2040 
2041 	if (V_route_cb.any_count == 0)
2042 		return (0);
2043 
2044 	if (info->rti_flags & RTF_HOST)
2045 		info->rti_info[RTAX_NETMASK] = NULL;
2046 
2047 	m = rtsock_msg_mbuf(cmd, info);
2048 	if (m == NULL)
2049 		return (ENOBUFS);
2050 
2051 	if (fibnum != RT_ALL_FIBS) {
2052 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
2053 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
2054 		M_SETFIB(m, fibnum);
2055 		m->m_flags |= RTS_FILTER_FIB;
2056 	}
2057 
2058 	rtm = mtod(m, struct rt_msghdr *);
2059 	rtm->rtm_addrs = info->rti_addrs;
2060 	if (info->rti_ifp != NULL)
2061 		rtm->rtm_index = info->rti_ifp->if_index;
2062 	/* Add RTF_DONE to indicate command 'completion' required by API */
2063 	info->rti_flags |= RTF_DONE;
2064 	/* Reported routes has to be up */
2065 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
2066 		info->rti_flags |= RTF_UP;
2067 	rtm->rtm_flags = info->rti_flags;
2068 
2069 	sa = info->rti_info[RTAX_DST];
2070 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2071 
2072 	return (0);
2073 }
2074 
2075 /*
2076  * This is the analogue to the rt_newaddrmsg which performs the same
2077  * function but for multicast group memberhips.  This is easier since
2078  * there is no route state to worry about.
2079  */
2080 void
2081 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
2082 {
2083 	struct rt_addrinfo info;
2084 	struct mbuf *m = NULL;
2085 	struct ifnet *ifp = ifma->ifma_ifp;
2086 	struct ifma_msghdr *ifmam;
2087 
2088 	if (V_route_cb.any_count == 0)
2089 		return;
2090 
2091 	bzero((caddr_t)&info, sizeof(info));
2092 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2093 	if (ifp && ifp->if_addr)
2094 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2095 	else
2096 		info.rti_info[RTAX_IFP] = NULL;
2097 	/*
2098 	 * If a link-layer address is present, present it as a ``gateway''
2099 	 * (similarly to how ARP entries, e.g., are presented).
2100 	 */
2101 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
2102 	m = rtsock_msg_mbuf(cmd, &info);
2103 	if (m == NULL)
2104 		return;
2105 	ifmam = mtod(m, struct ifma_msghdr *);
2106 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
2107 	    __func__));
2108 	ifmam->ifmam_index = ifp->if_index;
2109 	ifmam->ifmam_addrs = info.rti_addrs;
2110 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
2111 }
2112 
2113 static struct mbuf *
2114 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
2115 	struct rt_addrinfo *info)
2116 {
2117 	struct if_announcemsghdr *ifan;
2118 	struct mbuf *m;
2119 
2120 	if (V_route_cb.any_count == 0)
2121 		return NULL;
2122 	bzero((caddr_t)info, sizeof(*info));
2123 	m = rtsock_msg_mbuf(type, info);
2124 	if (m != NULL) {
2125 		ifan = mtod(m, struct if_announcemsghdr *);
2126 		ifan->ifan_index = ifp->if_index;
2127 		strlcpy(ifan->ifan_name, ifp->if_xname,
2128 			sizeof(ifan->ifan_name));
2129 		ifan->ifan_what = what;
2130 	}
2131 	return m;
2132 }
2133 
2134 /*
2135  * This is called to generate routing socket messages indicating
2136  * IEEE80211 wireless events.
2137  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2138  */
2139 void
2140 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2141 {
2142 	struct mbuf *m;
2143 	struct rt_addrinfo info;
2144 
2145 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2146 	if (m != NULL) {
2147 		/*
2148 		 * Append the ieee80211 data.  Try to stick it in the
2149 		 * mbuf containing the ifannounce msg; otherwise allocate
2150 		 * a new mbuf and append.
2151 		 *
2152 		 * NB: we assume m is a single mbuf.
2153 		 */
2154 		if (data_len > M_TRAILINGSPACE(m)) {
2155 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2156 			if (n == NULL) {
2157 				m_freem(m);
2158 				return;
2159 			}
2160 			bcopy(data, mtod(n, void *), data_len);
2161 			n->m_len = data_len;
2162 			m->m_next = n;
2163 		} else if (data_len > 0) {
2164 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2165 			m->m_len += data_len;
2166 		}
2167 		if (m->m_flags & M_PKTHDR)
2168 			m->m_pkthdr.len += data_len;
2169 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2170 		rt_dispatch(m, AF_UNSPEC);
2171 	}
2172 }
2173 
2174 /*
2175  * This is called to generate routing socket messages indicating
2176  * network interface arrival and departure.
2177  */
2178 static void
2179 rt_ifannouncemsg(struct ifnet *ifp, int what)
2180 {
2181 	struct mbuf *m;
2182 	struct rt_addrinfo info;
2183 
2184 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2185 	if (m != NULL)
2186 		rt_dispatch(m, AF_UNSPEC);
2187 }
2188 
2189 static void
2190 rt_dispatch(struct mbuf *m, sa_family_t saf)
2191 {
2192 
2193 	M_ASSERTPKTHDR(m);
2194 
2195 	m->m_rtsock_family = saf;
2196 	if (V_loif)
2197 		m->m_pkthdr.rcvif = V_loif;
2198 	else {
2199 		m_freem(m);
2200 		return;
2201 	}
2202 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
2203 }
2204 
2205 /*
2206  * This is used in dumping the kernel table via sysctl().
2207  */
2208 static int
2209 sysctl_dumpentry(struct rtentry *rt, void *vw)
2210 {
2211 	struct walkarg *w = vw;
2212 	struct nhop_object *nh;
2213 
2214 	NET_EPOCH_ASSERT();
2215 
2216 	if (!rt_is_exportable(rt, w->w_req->td->td_ucred))
2217 		return (0);
2218 
2219 	export_rtaddrs(rt, w->dst, w->mask);
2220 	nh = rt_get_raw_nhop(rt);
2221 #ifdef ROUTE_MPATH
2222 	if (NH_IS_NHGRP(nh)) {
2223 		const struct weightened_nhop *wn;
2224 		uint32_t num_nhops;
2225 		int error;
2226 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2227 		for (int i = 0; i < num_nhops; i++) {
2228 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2229 			if (error != 0)
2230 				return (error);
2231 		}
2232 	} else
2233 #endif
2234 		sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2235 
2236 	return (0);
2237 }
2238 
2239 
2240 static int
2241 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2242     struct walkarg *w)
2243 {
2244 	struct rt_addrinfo info;
2245 	int error = 0, size;
2246 	uint32_t rtflags;
2247 
2248 	rtflags = nhop_get_rtflags(nh);
2249 
2250 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2251 		return (0);
2252 
2253 	bzero((caddr_t)&info, sizeof(info));
2254 	info.rti_info[RTAX_DST] = w->dst;
2255 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2256 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2257 	info.rti_info[RTAX_GENMASK] = 0;
2258 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2259 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2260 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2261 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2262 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2263 	}
2264 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2265 		return (error);
2266 	if (w->w_req && w->w_tmem) {
2267 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2268 
2269 		bzero(&rtm->rtm_index,
2270 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2271 
2272 		/*
2273 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2274 		 * and RTF_UP (if entry is linked, which is always true here).
2275 		 * Given that, use nhop rtflags & add RTF_UP.
2276 		 */
2277 		rtm->rtm_flags = rtflags | RTF_UP;
2278 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2279 			rtm->rtm_flags = RTF_GATEWAY |
2280 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2281 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2282 		rtm->rtm_rmx.rmx_weight = weight;
2283 		rtm->rtm_index = nh->nh_ifp->if_index;
2284 		rtm->rtm_addrs = info.rti_addrs;
2285 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2286 		return (error);
2287 	}
2288 	return (error);
2289 }
2290 
2291 static int
2292 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2293     struct rt_addrinfo *info, struct walkarg *w, int len)
2294 {
2295 	struct if_msghdrl *ifm;
2296 	struct if_data *ifd;
2297 
2298 	ifm = (struct if_msghdrl *)w->w_tmem;
2299 
2300 #ifdef COMPAT_FREEBSD32
2301 	if (w->w_req->flags & SCTL_MASK32) {
2302 		struct if_msghdrl32 *ifm32;
2303 
2304 		ifm32 = (struct if_msghdrl32 *)ifm;
2305 		ifm32->ifm_addrs = info->rti_addrs;
2306 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2307 		ifm32->ifm_index = ifp->if_index;
2308 		ifm32->_ifm_spare1 = 0;
2309 		ifm32->ifm_len = sizeof(*ifm32);
2310 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2311 		ifm32->_ifm_spare2 = 0;
2312 		ifd = &ifm32->ifm_data;
2313 	} else
2314 #endif
2315 	{
2316 		ifm->ifm_addrs = info->rti_addrs;
2317 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2318 		ifm->ifm_index = ifp->if_index;
2319 		ifm->_ifm_spare1 = 0;
2320 		ifm->ifm_len = sizeof(*ifm);
2321 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2322 		ifm->_ifm_spare2 = 0;
2323 		ifd = &ifm->ifm_data;
2324 	}
2325 
2326 	memcpy(ifd, src_ifd, sizeof(*ifd));
2327 
2328 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2329 }
2330 
2331 static int
2332 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2333     struct rt_addrinfo *info, struct walkarg *w, int len)
2334 {
2335 	struct if_msghdr *ifm;
2336 	struct if_data *ifd;
2337 
2338 	ifm = (struct if_msghdr *)w->w_tmem;
2339 
2340 #ifdef COMPAT_FREEBSD32
2341 	if (w->w_req->flags & SCTL_MASK32) {
2342 		struct if_msghdr32 *ifm32;
2343 
2344 		ifm32 = (struct if_msghdr32 *)ifm;
2345 		ifm32->ifm_addrs = info->rti_addrs;
2346 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2347 		ifm32->ifm_index = ifp->if_index;
2348 		ifm32->_ifm_spare1 = 0;
2349 		ifd = &ifm32->ifm_data;
2350 	} else
2351 #endif
2352 	{
2353 		ifm->ifm_addrs = info->rti_addrs;
2354 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2355 		ifm->ifm_index = ifp->if_index;
2356 		ifm->_ifm_spare1 = 0;
2357 		ifd = &ifm->ifm_data;
2358 	}
2359 
2360 	memcpy(ifd, src_ifd, sizeof(*ifd));
2361 
2362 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2363 }
2364 
2365 static int
2366 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2367     struct walkarg *w, int len)
2368 {
2369 	struct ifa_msghdrl *ifam;
2370 	struct if_data *ifd;
2371 
2372 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2373 
2374 #ifdef COMPAT_FREEBSD32
2375 	if (w->w_req->flags & SCTL_MASK32) {
2376 		struct ifa_msghdrl32 *ifam32;
2377 
2378 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2379 		ifam32->ifam_addrs = info->rti_addrs;
2380 		ifam32->ifam_flags = ifa->ifa_flags;
2381 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2382 		ifam32->_ifam_spare1 = 0;
2383 		ifam32->ifam_len = sizeof(*ifam32);
2384 		ifam32->ifam_data_off =
2385 		    offsetof(struct ifa_msghdrl32, ifam_data);
2386 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2387 		ifd = &ifam32->ifam_data;
2388 	} else
2389 #endif
2390 	{
2391 		ifam->ifam_addrs = info->rti_addrs;
2392 		ifam->ifam_flags = ifa->ifa_flags;
2393 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2394 		ifam->_ifam_spare1 = 0;
2395 		ifam->ifam_len = sizeof(*ifam);
2396 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2397 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2398 		ifd = &ifam->ifam_data;
2399 	}
2400 
2401 	bzero(ifd, sizeof(*ifd));
2402 	ifd->ifi_datalen = sizeof(struct if_data);
2403 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2404 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2405 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2406 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2407 
2408 	/* Fixup if_data carp(4) vhid. */
2409 	if (carp_get_vhid_p != NULL)
2410 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2411 
2412 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2413 }
2414 
2415 static int
2416 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2417     struct walkarg *w, int len)
2418 {
2419 	struct ifa_msghdr *ifam;
2420 
2421 	ifam = (struct ifa_msghdr *)w->w_tmem;
2422 	ifam->ifam_addrs = info->rti_addrs;
2423 	ifam->ifam_flags = ifa->ifa_flags;
2424 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2425 	ifam->_ifam_spare1 = 0;
2426 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2427 
2428 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2429 }
2430 
2431 static int
2432 sysctl_iflist(int af, struct walkarg *w)
2433 {
2434 	struct ifnet *ifp;
2435 	struct ifaddr *ifa;
2436 	struct if_data ifd;
2437 	struct rt_addrinfo info;
2438 	int len, error = 0;
2439 	struct sockaddr_storage ss;
2440 
2441 	bzero((caddr_t)&info, sizeof(info));
2442 	bzero(&ifd, sizeof(ifd));
2443 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2444 		if (w->w_arg && w->w_arg != ifp->if_index)
2445 			continue;
2446 		if_data_copy(ifp, &ifd);
2447 		ifa = ifp->if_addr;
2448 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2449 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2450 		if (error != 0)
2451 			goto done;
2452 		info.rti_info[RTAX_IFP] = NULL;
2453 		if (w->w_req && w->w_tmem) {
2454 			if (w->w_op == NET_RT_IFLISTL)
2455 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2456 				    len);
2457 			else
2458 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2459 				    len);
2460 			if (error)
2461 				goto done;
2462 		}
2463 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2464 			if (af && af != ifa->ifa_addr->sa_family)
2465 				continue;
2466 			if (prison_if(w->w_req->td->td_ucred,
2467 			    ifa->ifa_addr) != 0)
2468 				continue;
2469 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2470 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2471 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2472 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2473 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2474 			if (error != 0)
2475 				goto done;
2476 			if (w->w_req && w->w_tmem) {
2477 				if (w->w_op == NET_RT_IFLISTL)
2478 					error = sysctl_iflist_ifaml(ifa, &info,
2479 					    w, len);
2480 				else
2481 					error = sysctl_iflist_ifam(ifa, &info,
2482 					    w, len);
2483 				if (error)
2484 					goto done;
2485 			}
2486 		}
2487 		info.rti_info[RTAX_IFA] = NULL;
2488 		info.rti_info[RTAX_NETMASK] = NULL;
2489 		info.rti_info[RTAX_BRD] = NULL;
2490 	}
2491 done:
2492 	return (error);
2493 }
2494 
2495 static int
2496 sysctl_ifmalist(int af, struct walkarg *w)
2497 {
2498 	struct rt_addrinfo info;
2499 	struct ifaddr *ifa;
2500 	struct ifmultiaddr *ifma;
2501 	struct ifnet *ifp;
2502 	int error, len;
2503 
2504 	NET_EPOCH_ASSERT();
2505 
2506 	error = 0;
2507 	bzero((caddr_t)&info, sizeof(info));
2508 
2509 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2510 		if (w->w_arg && w->w_arg != ifp->if_index)
2511 			continue;
2512 		ifa = ifp->if_addr;
2513 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2514 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2515 			if (af && af != ifma->ifma_addr->sa_family)
2516 				continue;
2517 			if (prison_if(w->w_req->td->td_ucred,
2518 			    ifma->ifma_addr) != 0)
2519 				continue;
2520 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2521 			info.rti_info[RTAX_GATEWAY] =
2522 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2523 			    ifma->ifma_lladdr : NULL;
2524 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2525 			if (error != 0)
2526 				break;
2527 			if (w->w_req && w->w_tmem) {
2528 				struct ifma_msghdr *ifmam;
2529 
2530 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2531 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2532 				ifmam->ifmam_flags = 0;
2533 				ifmam->ifmam_addrs = info.rti_addrs;
2534 				ifmam->_ifmam_spare1 = 0;
2535 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2536 				if (error != 0)
2537 					break;
2538 			}
2539 		}
2540 		if (error != 0)
2541 			break;
2542 	}
2543 	return (error);
2544 }
2545 
2546 static void
2547 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2548 {
2549 	union sockaddr_union sa_dst, sa_mask;
2550 
2551 	w->family = family;
2552 	w->dst = (struct sockaddr *)&sa_dst;
2553 	w->mask = (struct sockaddr *)&sa_mask;
2554 
2555 	init_sockaddrs_family(family, w->dst, w->mask);
2556 
2557 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2558 }
2559 
2560 static int
2561 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2562 {
2563 	struct epoch_tracker et;
2564 	int	*name = (int *)arg1;
2565 	u_int	namelen = arg2;
2566 	struct rib_head *rnh = NULL; /* silence compiler. */
2567 	int	i, lim, error = EINVAL;
2568 	int	fib = 0;
2569 	u_char	af;
2570 	struct	walkarg w;
2571 
2572 	if (namelen < 3)
2573 		return (EINVAL);
2574 
2575 	name++;
2576 	namelen--;
2577 	if (req->newptr)
2578 		return (EPERM);
2579 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2580 		if (namelen == 3)
2581 			fib = req->td->td_proc->p_fibnum;
2582 		else if (namelen == 4)
2583 			fib = (name[3] == RT_ALL_FIBS) ?
2584 			    req->td->td_proc->p_fibnum : name[3];
2585 		else
2586 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2587 		if (fib < 0 || fib >= rt_numfibs)
2588 			return (EINVAL);
2589 	} else if (namelen != 3)
2590 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2591 	af = name[0];
2592 	if (af > AF_MAX)
2593 		return (EINVAL);
2594 	bzero(&w, sizeof(w));
2595 	w.w_op = name[1];
2596 	w.w_arg = name[2];
2597 	w.w_req = req;
2598 
2599 	error = sysctl_wire_old_buffer(req, 0);
2600 	if (error)
2601 		return (error);
2602 
2603 	/*
2604 	 * Allocate reply buffer in advance.
2605 	 * All rtsock messages has maximum length of u_short.
2606 	 */
2607 	w.w_tmemsize = 65536;
2608 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2609 
2610 	NET_EPOCH_ENTER(et);
2611 	switch (w.w_op) {
2612 	case NET_RT_DUMP:
2613 	case NET_RT_FLAGS:
2614 		if (af == 0) {			/* dump all tables */
2615 			i = 1;
2616 			lim = AF_MAX;
2617 		} else				/* dump only one table */
2618 			i = lim = af;
2619 
2620 		/*
2621 		 * take care of llinfo entries, the caller must
2622 		 * specify an AF
2623 		 */
2624 		if (w.w_op == NET_RT_FLAGS &&
2625 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2626 			if (af != 0)
2627 				error = lltable_sysctl_dumparp(af, w.w_req);
2628 			else
2629 				error = EINVAL;
2630 			break;
2631 		}
2632 		/*
2633 		 * take care of routing entries
2634 		 */
2635 		for (error = 0; error == 0 && i <= lim; i++) {
2636 			rnh = rt_tables_get_rnh(fib, i);
2637 			if (rnh != NULL) {
2638 				rtable_sysctl_dump(fib, i, &w);
2639 			} else if (af != 0)
2640 				error = EAFNOSUPPORT;
2641 		}
2642 		break;
2643 	case NET_RT_NHOP:
2644 	case NET_RT_NHGRP:
2645 		/* Allow dumping one specific af/fib at a time */
2646 		if (namelen < 4) {
2647 			error = EINVAL;
2648 			break;
2649 		}
2650 		fib = name[3];
2651 		if (fib < 0 || fib > rt_numfibs) {
2652 			error = EINVAL;
2653 			break;
2654 		}
2655 		rnh = rt_tables_get_rnh(fib, af);
2656 		if (rnh == NULL) {
2657 			error = EAFNOSUPPORT;
2658 			break;
2659 		}
2660 		if (w.w_op == NET_RT_NHOP)
2661 			error = nhops_dump_sysctl(rnh, w.w_req);
2662 		else
2663 #ifdef ROUTE_MPATH
2664 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2665 #else
2666 			error = ENOTSUP;
2667 #endif
2668 		break;
2669 	case NET_RT_IFLIST:
2670 	case NET_RT_IFLISTL:
2671 		error = sysctl_iflist(af, &w);
2672 		break;
2673 
2674 	case NET_RT_IFMALIST:
2675 		error = sysctl_ifmalist(af, &w);
2676 		break;
2677 	}
2678 	NET_EPOCH_EXIT(et);
2679 
2680 	free(w.w_tmem, M_TEMP);
2681 	return (error);
2682 }
2683 
2684 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2685     sysctl_rtsock, "Return route tables and interface/address lists");
2686 
2687 /*
2688  * Definitions of protocols supported in the ROUTE domain.
2689  */
2690 
2691 static struct domain routedomain;		/* or at least forward */
2692 
2693 static struct protosw routesw = {
2694 	.pr_type =		SOCK_RAW,
2695 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2696 	.pr_abort =		rts_close,
2697 	.pr_attach =		rts_attach,
2698 	.pr_detach =		rts_detach,
2699 	.pr_send =		rts_send,
2700 	.pr_shutdown =		rts_shutdown,
2701 	.pr_disconnect =	rts_disconnect,
2702 	.pr_close =		rts_close,
2703 };
2704 
2705 static struct domain routedomain = {
2706 	.dom_family =		PF_ROUTE,
2707 	.dom_name =		"route",
2708 	.dom_nprotosw =		1,
2709 	.dom_protosw =		{ &routesw },
2710 };
2711 
2712 DOMAIN_SET(route);
2713