xref: /freebsd/sys/net/rtsock.c (revision d0b2dbfa0ecf2bbc9709efc5e20baf8e4b44bbbf)
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  */
33 #include "opt_ddb.h"
34 #include "opt_route.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/jail.h>
40 #include <sys/kernel.h>
41 #include <sys/eventhandler.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_private.h>
60 #include <net/if_dl.h>
61 #include <net/if_llatbl.h>
62 #include <net/if_types.h>
63 #include <net/netisr.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 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  * Used to store address family of the notification.
159  */
160 #define	m_rtsock_family	m_pkthdr.PH_loc.eight[0]
161 
162 struct rcb {
163 	LIST_ENTRY(rcb) list;
164 	struct socket	*rcb_socket;
165 	sa_family_t	rcb_family;
166 };
167 
168 typedef struct {
169 	LIST_HEAD(, rcb)	cblist;
170 	int	ip_count;	/* attached w/ AF_INET */
171 	int	ip6_count;	/* attached w/ AF_INET6 */
172 	int	any_count;	/* total attached */
173 } route_cb_t;
174 VNET_DEFINE_STATIC(route_cb_t, route_cb);
175 #define	V_route_cb VNET(route_cb)
176 
177 struct mtx rtsock_mtx;
178 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
179 
180 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
181 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
182 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
183 
184 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
185 
186 struct walkarg {
187 	int	family;
188 	int	w_tmemsize;
189 	int	w_op, w_arg;
190 	caddr_t	w_tmem;
191 	struct sysctl_req *w_req;
192 	struct sockaddr *dst;
193 	struct sockaddr *mask;
194 };
195 
196 static void	rts_input(struct mbuf *m);
197 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
198 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
199 			struct walkarg *w, int *plen);
200 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
201 			struct rt_addrinfo *rtinfo);
202 static int	cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb);
203 static int	sysctl_dumpentry(struct rtentry *rt, void *vw);
204 static int	sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
205 			uint32_t weight, struct walkarg *w);
206 static int	sysctl_iflist(int af, struct walkarg *w);
207 static int	sysctl_ifmalist(int af, struct walkarg *w);
208 static void	rt_getmetrics(const struct rtentry *rt,
209 			const struct nhop_object *nh, struct rt_metrics *out);
210 static void	rt_dispatch(struct mbuf *, sa_family_t);
211 static void	rt_ifannouncemsg(struct ifnet *ifp, int what);
212 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
213 			struct rt_msghdr *rtm, struct rib_cmd_info *rc);
214 static int	update_rtm_from_rc(struct rt_addrinfo *info,
215 			struct rt_msghdr **prtm, int alloc_len,
216 			struct rib_cmd_info *rc, struct nhop_object *nh);
217 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
218 			struct mbuf *m, sa_family_t saf, u_int fibnum,
219 			int rtm_errno);
220 static void	rtsock_notify_event(uint32_t fibnum, const struct rib_cmd_info *rc);
221 static void	rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask);
222 
223 static struct netisr_handler rtsock_nh = {
224 	.nh_name = "rtsock",
225 	.nh_handler = rts_input,
226 	.nh_proto = NETISR_ROUTE,
227 	.nh_policy = NETISR_POLICY_SOURCE,
228 };
229 
230 static int
231 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
232 {
233 	int error, qlimit;
234 
235 	netisr_getqlimit(&rtsock_nh, &qlimit);
236 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
237         if (error || !req->newptr)
238                 return (error);
239 	if (qlimit < 1)
240 		return (EINVAL);
241 	return (netisr_setqlimit(&rtsock_nh, qlimit));
242 }
243 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
244     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
245     0, 0, sysctl_route_netisr_maxqlen, "I",
246     "maximum routing socket dispatch queue length");
247 
248 static void
249 vnet_rts_init(void)
250 {
251 	int tmp;
252 
253 	if (IS_DEFAULT_VNET(curvnet)) {
254 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
255 			rtsock_nh.nh_qlimit = tmp;
256 		netisr_register(&rtsock_nh);
257 	}
258 #ifdef VIMAGE
259 	 else
260 		netisr_register_vnet(&rtsock_nh);
261 #endif
262 }
263 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
264     vnet_rts_init, 0);
265 
266 #ifdef VIMAGE
267 static void
268 vnet_rts_uninit(void)
269 {
270 
271 	netisr_unregister_vnet(&rtsock_nh);
272 }
273 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
274     vnet_rts_uninit, 0);
275 #endif
276 
277 static void
278 report_route_event(const struct rib_cmd_info *rc, void *_cbdata)
279 {
280 	uint32_t fibnum = (uint32_t)(uintptr_t)_cbdata;
281 	struct nhop_object *nh;
282 
283 	nh = rc->rc_cmd == RTM_DELETE ? rc->rc_nh_old : rc->rc_nh_new;
284 	rt_routemsg(rc->rc_cmd, rc->rc_rt, nh, fibnum);
285 }
286 
287 static void
288 rts_handle_route_event(uint32_t fibnum, const struct rib_cmd_info *rc)
289 {
290 #ifdef ROUTE_MPATH
291 	if ((rc->rc_nh_new && NH_IS_NHGRP(rc->rc_nh_new)) ||
292 	    (rc->rc_nh_old && NH_IS_NHGRP(rc->rc_nh_old))) {
293 		rib_decompose_notification(rc, report_route_event,
294 		    (void *)(uintptr_t)fibnum);
295 	} else
296 #endif
297 		report_route_event(rc, (void *)(uintptr_t)fibnum);
298 }
299 static struct rtbridge rtsbridge = {
300 	.route_f = rts_handle_route_event,
301 	.ifmsg_f = rtsock_ifmsg,
302 };
303 static struct rtbridge *rtsbridge_orig_p;
304 
305 static void
306 rtsock_notify_event(uint32_t fibnum, const struct rib_cmd_info *rc)
307 {
308 	netlink_callback_p->route_f(fibnum, rc);
309 }
310 
311 static void
312 rtsock_init(void)
313 {
314 	rtsbridge_orig_p = rtsock_callback_p;
315 	rtsock_callback_p = &rtsbridge;
316 }
317 SYSINIT(rtsock_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rtsock_init, NULL);
318 
319 static void
320 rts_handle_ifnet_arrival(void *arg __unused, struct ifnet *ifp)
321 {
322 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
323 }
324 EVENTHANDLER_DEFINE(ifnet_arrival_event, rts_handle_ifnet_arrival, NULL, 0);
325 
326 static void
327 rts_handle_ifnet_departure(void *arg __unused, struct ifnet *ifp)
328 {
329 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
330 }
331 EVENTHANDLER_DEFINE(ifnet_departure_event, rts_handle_ifnet_departure, NULL, 0);
332 
333 static void
334 rts_append_data(struct socket *so, struct mbuf *m)
335 {
336 
337 	if (sbappendaddr(&so->so_rcv, &route_src, m, NULL) == 0) {
338 		soroverflow(so);
339 		m_freem(m);
340 	} else
341 		sorwakeup(so);
342 }
343 
344 static void
345 rts_input(struct mbuf *m)
346 {
347 	struct rcb *rcb;
348 	struct socket *last;
349 
350 	last = NULL;
351 	RTSOCK_LOCK();
352 	LIST_FOREACH(rcb, &V_route_cb.cblist, list) {
353 		if (rcb->rcb_family != AF_UNSPEC &&
354 		    rcb->rcb_family != m->m_rtsock_family)
355 			continue;
356 		if ((m->m_flags & RTS_FILTER_FIB) &&
357 		    M_GETFIB(m) != rcb->rcb_socket->so_fibnum)
358 			continue;
359 		if (last != NULL) {
360 			struct mbuf *n;
361 
362 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
363 			if (n != NULL)
364 				rts_append_data(last, n);
365 		}
366 		last = rcb->rcb_socket;
367 	}
368 	if (last != NULL)
369 		rts_append_data(last, m);
370 	else
371 		m_freem(m);
372 	RTSOCK_UNLOCK();
373 }
374 
375 static void
376 rts_close(struct socket *so)
377 {
378 
379 	soisdisconnected(so);
380 }
381 
382 static SYSCTL_NODE(_net, OID_AUTO, rtsock, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
383     "Routing socket infrastructure");
384 static u_long rts_sendspace = 8192;
385 SYSCTL_ULONG(_net_rtsock, OID_AUTO, sendspace, CTLFLAG_RW, &rts_sendspace, 0,
386     "Default routing socket send space");
387 static u_long rts_recvspace = 8192;
388 SYSCTL_ULONG(_net_rtsock, OID_AUTO, recvspace, CTLFLAG_RW, &rts_recvspace, 0,
389     "Default routing socket receive space");
390 
391 static int
392 rts_attach(struct socket *so, int proto, struct thread *td)
393 {
394 	struct rcb *rcb;
395 	int error;
396 
397 	error = soreserve(so, rts_sendspace, rts_recvspace);
398 	if (error)
399 		return (error);
400 
401 	rcb = malloc(sizeof(*rcb), M_PCB, M_WAITOK);
402 	rcb->rcb_socket = so;
403 	rcb->rcb_family = proto;
404 
405 	so->so_pcb = rcb;
406 	so->so_fibnum = td->td_proc->p_fibnum;
407 	so->so_options |= SO_USELOOPBACK;
408 
409 	RTSOCK_LOCK();
410 	LIST_INSERT_HEAD(&V_route_cb.cblist, rcb, list);
411 	switch (proto) {
412 	case AF_INET:
413 		V_route_cb.ip_count++;
414 		break;
415 	case AF_INET6:
416 		V_route_cb.ip6_count++;
417 		break;
418 	}
419 	V_route_cb.any_count++;
420 	RTSOCK_UNLOCK();
421 	soisconnected(so);
422 
423 	return (0);
424 }
425 
426 static void
427 rts_detach(struct socket *so)
428 {
429 	struct rcb *rcb = so->so_pcb;
430 
431 	RTSOCK_LOCK();
432 	LIST_REMOVE(rcb, list);
433 	switch(rcb->rcb_family) {
434 	case AF_INET:
435 		V_route_cb.ip_count--;
436 		break;
437 	case AF_INET6:
438 		V_route_cb.ip6_count--;
439 		break;
440 	}
441 	V_route_cb.any_count--;
442 	RTSOCK_UNLOCK();
443 	free(rcb, M_PCB);
444 	so->so_pcb = NULL;
445 }
446 
447 static int
448 rts_disconnect(struct socket *so)
449 {
450 
451 	return (ENOTCONN);
452 }
453 
454 static int
455 rts_shutdown(struct socket *so)
456 {
457 
458 	socantsendmore(so);
459 	return (0);
460 }
461 
462 #ifndef _SOCKADDR_UNION_DEFINED
463 #define	_SOCKADDR_UNION_DEFINED
464 /*
465  * The union of all possible address formats we handle.
466  */
467 union sockaddr_union {
468 	struct sockaddr		sa;
469 	struct sockaddr_in	sin;
470 	struct sockaddr_in6	sin6;
471 };
472 #endif /* _SOCKADDR_UNION_DEFINED */
473 
474 static int
475 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
476     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
477 {
478 #if defined(INET) || defined(INET6)
479 	struct epoch_tracker et;
480 #endif
481 
482 	/* First, see if the returned address is part of the jail. */
483 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
484 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
485 		return (0);
486 	}
487 
488 	switch (info->rti_info[RTAX_DST]->sa_family) {
489 #ifdef INET
490 	case AF_INET:
491 	{
492 		struct in_addr ia;
493 		struct ifaddr *ifa;
494 		int found;
495 
496 		found = 0;
497 		/*
498 		 * Try to find an address on the given outgoing interface
499 		 * that belongs to the jail.
500 		 */
501 		NET_EPOCH_ENTER(et);
502 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
503 			struct sockaddr *sa;
504 			sa = ifa->ifa_addr;
505 			if (sa->sa_family != AF_INET)
506 				continue;
507 			ia = ((struct sockaddr_in *)sa)->sin_addr;
508 			if (prison_check_ip4(cred, &ia) == 0) {
509 				found = 1;
510 				break;
511 			}
512 		}
513 		NET_EPOCH_EXIT(et);
514 		if (!found) {
515 			/*
516 			 * As a last resort return the 'default' jail address.
517 			 */
518 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
519 			    sin_addr;
520 			if (prison_get_ip4(cred, &ia) != 0)
521 				return (ESRCH);
522 		}
523 		bzero(&saun->sin, sizeof(struct sockaddr_in));
524 		saun->sin.sin_len = sizeof(struct sockaddr_in);
525 		saun->sin.sin_family = AF_INET;
526 		saun->sin.sin_addr.s_addr = ia.s_addr;
527 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
528 		break;
529 	}
530 #endif
531 #ifdef INET6
532 	case AF_INET6:
533 	{
534 		struct in6_addr ia6;
535 		struct ifaddr *ifa;
536 		int found;
537 
538 		found = 0;
539 		/*
540 		 * Try to find an address on the given outgoing interface
541 		 * that belongs to the jail.
542 		 */
543 		NET_EPOCH_ENTER(et);
544 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
545 			struct sockaddr *sa;
546 			sa = ifa->ifa_addr;
547 			if (sa->sa_family != AF_INET6)
548 				continue;
549 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
550 			    &ia6, sizeof(struct in6_addr));
551 			if (prison_check_ip6(cred, &ia6) == 0) {
552 				found = 1;
553 				break;
554 			}
555 		}
556 		NET_EPOCH_EXIT(et);
557 		if (!found) {
558 			/*
559 			 * As a last resort return the 'default' jail address.
560 			 */
561 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
562 			    sin6_addr;
563 			if (prison_get_ip6(cred, &ia6) != 0)
564 				return (ESRCH);
565 		}
566 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
567 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
568 		saun->sin6.sin6_family = AF_INET6;
569 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
570 		if (sa6_recoverscope(&saun->sin6) != 0)
571 			return (ESRCH);
572 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
573 		break;
574 	}
575 #endif
576 	default:
577 		return (ESRCH);
578 	}
579 	return (0);
580 }
581 
582 static int
583 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
584 {
585 	struct ifaddr *ifa;
586 	sa_family_t saf;
587 
588 	if (V_loif == NULL) {
589 		RTS_PID_LOG(LOG_INFO, "Unable to add blackhole/reject nhop without loopback");
590 		return (ENOTSUP);
591 	}
592 	info->rti_ifp = V_loif;
593 
594 	saf = info->rti_info[RTAX_DST]->sa_family;
595 
596 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
597 		if (ifa->ifa_addr->sa_family == saf) {
598 			info->rti_ifa = ifa;
599 			break;
600 		}
601 	}
602 	if (info->rti_ifa == NULL) {
603 		RTS_PID_LOG(LOG_INFO, "Unable to find ifa for blackhole/reject nhop");
604 		return (ENOTSUP);
605 	}
606 
607 	bzero(saun, sizeof(union sockaddr_union));
608 	switch (saf) {
609 #ifdef INET
610 	case AF_INET:
611 		saun->sin.sin_family = AF_INET;
612 		saun->sin.sin_len = sizeof(struct sockaddr_in);
613 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
614 		break;
615 #endif
616 #ifdef INET6
617 	case AF_INET6:
618 		saun->sin6.sin6_family = AF_INET6;
619 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
620 		saun->sin6.sin6_addr = in6addr_loopback;
621 		break;
622 #endif
623 	default:
624 		RTS_PID_LOG(LOG_INFO, "unsupported family: %d", saf);
625 		return (ENOTSUP);
626 	}
627 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
628 	info->rti_flags |= RTF_GATEWAY;
629 
630 	return (0);
631 }
632 
633 /*
634  * Fills in @info based on userland-provided @rtm message.
635  *
636  * Returns 0 on success.
637  */
638 static int
639 fill_addrinfo(struct rt_msghdr *rtm, int len, struct linear_buffer *lb, u_int fibnum,
640     struct rt_addrinfo *info)
641 {
642 	int error;
643 
644 	rtm->rtm_pid = curproc->p_pid;
645 	info->rti_addrs = rtm->rtm_addrs;
646 
647 	info->rti_mflags = rtm->rtm_inits;
648 	info->rti_rmx = &rtm->rtm_rmx;
649 
650 	/*
651 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
652 	 * link-local address because rtrequest requires addresses with
653 	 * embedded scope id.
654 	 */
655 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
656 		return (EINVAL);
657 
658 	info->rti_flags = rtm->rtm_flags;
659 	error = cleanup_xaddrs(info, lb);
660 	if (error != 0)
661 		return (error);
662 	/*
663 	 * Verify that the caller has the appropriate privilege; RTM_GET
664 	 * is the only operation the non-superuser is allowed.
665 	 */
666 	if (rtm->rtm_type != RTM_GET) {
667 		error = priv_check(curthread, PRIV_NET_ROUTE);
668 		if (error != 0)
669 			return (error);
670 	}
671 
672 	/*
673 	 * The given gateway address may be an interface address.
674 	 * For example, issuing a "route change" command on a route
675 	 * entry that was created from a tunnel, and the gateway
676 	 * address given is the local end point. In this case the
677 	 * RTF_GATEWAY flag must be cleared or the destination will
678 	 * not be reachable even though there is no error message.
679 	 */
680 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
681 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
682 		struct nhop_object *nh;
683 
684 		/*
685 		 * A host route through the loopback interface is
686 		 * installed for each interface adddress. In pre 8.0
687 		 * releases the interface address of a PPP link type
688 		 * is not reachable locally. This behavior is fixed as
689 		 * part of the new L2/L3 redesign and rewrite work. The
690 		 * signature of this interface address route is the
691 		 * AF_LINK sa_family type of the gateway, and the
692 		 * rt_ifp has the IFF_LOOPBACK flag set.
693 		 */
694 		nh = rib_lookup(fibnum, info->rti_info[RTAX_GATEWAY], NHR_NONE, 0);
695 		if (nh != NULL && nh->gw_sa.sa_family == AF_LINK &&
696 		    nh->nh_ifp->if_flags & IFF_LOOPBACK) {
697 				info->rti_flags &= ~RTF_GATEWAY;
698 				info->rti_flags |= RTF_GWFLAG_COMPAT;
699 		}
700 	}
701 
702 	return (0);
703 }
704 
705 static struct nhop_object *
706 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
707 {
708 	if (!NH_IS_NHGRP(nh))
709 		return (nh);
710 #ifdef ROUTE_MPATH
711 	const struct weightened_nhop *wn;
712 	uint32_t num_nhops;
713 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
714 	if (gw == NULL)
715 		return (wn[0].nh);
716 	for (int i = 0; i < num_nhops; i++) {
717 		if (match_nhop_gw(wn[i].nh, gw))
718 			return (wn[i].nh);
719 	}
720 #endif
721 	return (NULL);
722 }
723 
724 /*
725  * Handles RTM_GET message from routing socket, returning matching rt.
726  *
727  * Returns:
728  * 0 on success, with locked and referenced matching rt in @rt_nrt
729  * errno of failure
730  */
731 static int
732 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
733     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
734 {
735 	RIB_RLOCK_TRACKER;
736 	struct rib_head *rnh;
737 	struct nhop_object *nh;
738 	sa_family_t saf;
739 
740 	saf = info->rti_info[RTAX_DST]->sa_family;
741 
742 	rnh = rt_tables_get_rnh(fibnum, saf);
743 	if (rnh == NULL)
744 		return (EAFNOSUPPORT);
745 
746 	RIB_RLOCK(rnh);
747 
748 	/*
749 	 * By (implicit) convention host route (one without netmask)
750 	 * means longest-prefix-match request and the route with netmask
751 	 * means exact-match lookup.
752 	 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128
753 	 * prefixes, use original data to check for the netmask presence.
754 	 */
755 	if ((rtm->rtm_addrs & RTA_NETMASK) == 0) {
756 		/*
757 		 * Provide longest prefix match for
758 		 * address lookup (no mask).
759 		 * 'route -n get addr'
760 		 */
761 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
762 		    info->rti_info[RTAX_DST], &rnh->head);
763 	} else
764 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
765 		    info->rti_info[RTAX_DST],
766 		    info->rti_info[RTAX_NETMASK], &rnh->head);
767 
768 	if (rc->rc_rt == NULL) {
769 		RIB_RUNLOCK(rnh);
770 		return (ESRCH);
771 	}
772 
773 	nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
774 	if (nh == NULL) {
775 		RIB_RUNLOCK(rnh);
776 		return (ESRCH);
777 	}
778 	/*
779 	 * If performing proxied L2 entry insertion, and
780 	 * the actual PPP host entry is found, perform
781 	 * another search to retrieve the prefix route of
782 	 * the local end point of the PPP link.
783 	 * TODO: move this logic to userland.
784 	 */
785 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
786 		struct sockaddr_storage laddr;
787 
788 		if (nh->nh_ifp != NULL &&
789 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
790 			struct ifaddr *ifa;
791 
792 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
793 					RT_ALL_FIBS);
794 			if (ifa != NULL)
795 				rt_maskedcopy(ifa->ifa_addr,
796 					      (struct sockaddr *)&laddr,
797 					      ifa->ifa_netmask);
798 		} else
799 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
800 				      (struct sockaddr *)&laddr,
801 				      nh->nh_ifa->ifa_netmask);
802 		/*
803 		 * refactor rt and no lock operation necessary
804 		 */
805 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(
806 		    (struct sockaddr *)&laddr, &rnh->head);
807 		if (rc->rc_rt == NULL) {
808 			RIB_RUNLOCK(rnh);
809 			return (ESRCH);
810 		}
811 		nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
812 		if (nh == NULL) {
813 			RIB_RUNLOCK(rnh);
814 			return (ESRCH);
815 		}
816 	}
817 	rc->rc_nh_new = nh;
818 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
819 	RIB_RUNLOCK(rnh);
820 
821 	return (0);
822 }
823 
824 static void
825 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
826 {
827 #ifdef INET
828 	if (family == AF_INET) {
829 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
830 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
831 
832 		bzero(dst4, sizeof(struct sockaddr_in));
833 		bzero(mask4, sizeof(struct sockaddr_in));
834 
835 		dst4->sin_family = AF_INET;
836 		dst4->sin_len = sizeof(struct sockaddr_in);
837 		mask4->sin_family = AF_INET;
838 		mask4->sin_len = sizeof(struct sockaddr_in);
839 	}
840 #endif
841 #ifdef INET6
842 	if (family == AF_INET6) {
843 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
844 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
845 
846 		bzero(dst6, sizeof(struct sockaddr_in6));
847 		bzero(mask6, sizeof(struct sockaddr_in6));
848 
849 		dst6->sin6_family = AF_INET6;
850 		dst6->sin6_len = sizeof(struct sockaddr_in6);
851 		mask6->sin6_family = AF_INET6;
852 		mask6->sin6_len = sizeof(struct sockaddr_in6);
853 	}
854 #endif
855 }
856 
857 static void
858 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
859     struct sockaddr *mask)
860 {
861 #ifdef INET
862 	if (dst->sa_family == AF_INET) {
863 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
864 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
865 		uint32_t scopeid = 0;
866 		rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
867 		    &scopeid);
868 		return;
869 	}
870 #endif
871 #ifdef INET6
872 	if (dst->sa_family == AF_INET6) {
873 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
874 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
875 		uint32_t scopeid = 0;
876 		rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
877 		    &mask6->sin6_addr, &scopeid);
878 		dst6->sin6_scope_id = scopeid;
879 		return;
880 	}
881 #endif
882 }
883 
884 static int
885 update_rtm_from_info(struct rt_addrinfo *info, struct rt_msghdr **prtm,
886     int alloc_len)
887 {
888 	struct rt_msghdr *rtm, *orig_rtm = NULL;
889 	struct walkarg w;
890 	int len;
891 
892 	rtm = *prtm;
893 	/* Check if we need to realloc storage */
894 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
895 	if (len > alloc_len) {
896 		struct rt_msghdr *tmp_rtm;
897 
898 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
899 		if (tmp_rtm == NULL)
900 			return (ENOBUFS);
901 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
902 		orig_rtm = rtm;
903 		rtm = tmp_rtm;
904 		alloc_len = len;
905 
906 		/*
907 		 * Delay freeing original rtm as info contains
908 		 * data referencing it.
909 		 */
910 	}
911 
912 	w.w_tmem = (caddr_t)rtm;
913 	w.w_tmemsize = alloc_len;
914 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
915 	rtm->rtm_addrs = info->rti_addrs;
916 
917 	if (orig_rtm != NULL)
918 		free(orig_rtm, M_TEMP);
919 	*prtm = rtm;
920 	return (0);
921 }
922 
923 
924 /*
925  * Update sockaddrs, flags, etc in @prtm based on @rc data.
926  * rtm can be reallocated.
927  *
928  * Returns 0 on success, along with pointer to (potentially reallocated)
929  *  rtm.
930  *
931  */
932 static int
933 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
934     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
935 {
936 	union sockaddr_union saun;
937 	struct rt_msghdr *rtm;
938 	struct ifnet *ifp;
939 	int error;
940 
941 	rtm = *prtm;
942 	union sockaddr_union sa_dst, sa_mask;
943 	int family = info->rti_info[RTAX_DST]->sa_family;
944 	init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
945 	export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
946 
947 	info->rti_info[RTAX_DST] = &sa_dst.sa;
948 	info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
949 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
950 	info->rti_info[RTAX_GENMASK] = 0;
951 	ifp = nh->nh_ifp;
952 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
953 		if (ifp) {
954 			info->rti_info[RTAX_IFP] =
955 			    ifp->if_addr->ifa_addr;
956 			error = rtm_get_jailed(info, ifp, nh,
957 			    &saun, curthread->td_ucred);
958 			if (error != 0)
959 				return (error);
960 			if (ifp->if_flags & IFF_POINTOPOINT)
961 				info->rti_info[RTAX_BRD] =
962 				    nh->nh_ifa->ifa_dstaddr;
963 			rtm->rtm_index = ifp->if_index;
964 		} else {
965 			info->rti_info[RTAX_IFP] = NULL;
966 			info->rti_info[RTAX_IFA] = NULL;
967 		}
968 	} else if (ifp != NULL)
969 		rtm->rtm_index = ifp->if_index;
970 
971 	if ((error = update_rtm_from_info(info, prtm, alloc_len)) != 0)
972 		return (error);
973 
974 	rtm = *prtm;
975 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
976 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
977 		rtm->rtm_flags = RTF_GATEWAY |
978 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
979 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
980 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
981 
982 	return (0);
983 }
984 
985 #ifdef ROUTE_MPATH
986 static void
987 save_del_notification(const struct rib_cmd_info *rc, void *_cbdata)
988 {
989 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
990 
991 	if (rc->rc_cmd == RTM_DELETE)
992 		*rc_new = *rc;
993 }
994 
995 static void
996 save_add_notification(const struct rib_cmd_info *rc, void *_cbdata)
997 {
998 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
999 
1000 	if (rc->rc_cmd == RTM_ADD)
1001 		*rc_new = *rc;
1002 }
1003 #endif
1004 
1005 #if defined(INET6) || defined(INET)
1006 static struct sockaddr *
1007 alloc_sockaddr_aligned(struct linear_buffer *lb, int len)
1008 {
1009 	len = roundup2(len, sizeof(uint64_t));
1010 	if (lb->offset + len > lb->size)
1011 		return (NULL);
1012 	struct sockaddr *sa = (struct sockaddr *)(lb->base + lb->offset);
1013 	lb->offset += len;
1014 	return (sa);
1015 }
1016 #endif
1017 
1018 static int
1019 rts_send(struct socket *so, int flags, struct mbuf *m,
1020     struct sockaddr *nam, struct mbuf *control, struct thread *td)
1021 {
1022 	struct rt_msghdr *rtm = NULL;
1023 	struct rt_addrinfo info;
1024 	struct epoch_tracker et;
1025 #ifdef INET6
1026 	struct sockaddr_storage ss;
1027 	struct sockaddr_in6 *sin6;
1028 	int i, rti_need_deembed = 0;
1029 #endif
1030 	int alloc_len = 0, len, error = 0, fibnum;
1031 	sa_family_t saf = AF_UNSPEC;
1032 	struct rib_cmd_info rc;
1033 	struct nhop_object *nh;
1034 
1035 	if ((flags & PRUS_OOB) || control != NULL) {
1036 		m_freem(m);
1037 		if (control != NULL)
1038 			m_freem(control);
1039 		return (EOPNOTSUPP);
1040 	}
1041 
1042 	fibnum = so->so_fibnum;
1043 #define senderr(e) { error = e; goto flush;}
1044 	if (m == NULL || ((m->m_len < sizeof(long)) &&
1045 		       (m = m_pullup(m, sizeof(long))) == NULL))
1046 		return (ENOBUFS);
1047 	if ((m->m_flags & M_PKTHDR) == 0)
1048 		panic("route_output");
1049 	NET_EPOCH_ENTER(et);
1050 	len = m->m_pkthdr.len;
1051 	if (len < sizeof(*rtm) ||
1052 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1053 		senderr(EINVAL);
1054 
1055 	/*
1056 	 * Most of current messages are in range 200-240 bytes,
1057 	 * minimize possible re-allocation on reply using larger size
1058 	 * buffer aligned on 1k boundaty.
1059 	 */
1060 	alloc_len = roundup2(len, 1024);
1061 	int total_len = alloc_len + SCRATCH_BUFFER_SIZE;
1062 	if ((rtm = malloc(total_len, M_TEMP, M_NOWAIT)) == NULL)
1063 		senderr(ENOBUFS);
1064 
1065 	m_copydata(m, 0, len, (caddr_t)rtm);
1066 	bzero(&info, sizeof(info));
1067 	nh = NULL;
1068 	struct linear_buffer lb = {
1069 		.base = (char *)rtm + alloc_len,
1070 		.size = SCRATCH_BUFFER_SIZE,
1071 	};
1072 
1073 	if (rtm->rtm_version != RTM_VERSION) {
1074 		/* Do not touch message since format is unknown */
1075 		free(rtm, M_TEMP);
1076 		rtm = NULL;
1077 		senderr(EPROTONOSUPPORT);
1078 	}
1079 
1080 	/*
1081 	 * Starting from here, it is possible
1082 	 * to alter original message and insert
1083 	 * caller PID and error value.
1084 	 */
1085 
1086 	if ((error = fill_addrinfo(rtm, len, &lb, fibnum, &info)) != 0) {
1087 		senderr(error);
1088 	}
1089 	/* fill_addringo() embeds scope into IPv6 addresses */
1090 #ifdef INET6
1091 	rti_need_deembed = 1;
1092 #endif
1093 
1094 	saf = info.rti_info[RTAX_DST]->sa_family;
1095 
1096 	/* support for new ARP code */
1097 	if (rtm->rtm_flags & RTF_LLDATA) {
1098 		error = lla_rt_output(rtm, &info);
1099 		goto flush;
1100 	}
1101 
1102 	union sockaddr_union gw_saun;
1103 	int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1104 	if (blackhole_flags != 0) {
1105 		if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1106 			error = fill_blackholeinfo(&info, &gw_saun);
1107 		else {
1108 			RTS_PID_LOG(LOG_DEBUG, "both BLACKHOLE and REJECT flags specifiied");
1109 			error = EINVAL;
1110 		}
1111 		if (error != 0)
1112 			senderr(error);
1113 	}
1114 
1115 	switch (rtm->rtm_type) {
1116 	case RTM_ADD:
1117 	case RTM_CHANGE:
1118 		if (rtm->rtm_type == RTM_ADD) {
1119 			if (info.rti_info[RTAX_GATEWAY] == NULL) {
1120 				RTS_PID_LOG(LOG_DEBUG, "RTM_ADD w/o gateway");
1121 				senderr(EINVAL);
1122 			}
1123 		}
1124 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1125 		if (error == 0) {
1126 			rtsock_notify_event(fibnum, &rc);
1127 #ifdef ROUTE_MPATH
1128 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
1129 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1130 				struct rib_cmd_info rc_simple = {};
1131 				rib_decompose_notification(&rc,
1132 				    save_add_notification, (void *)&rc_simple);
1133 				rc = rc_simple;
1134 			}
1135 #endif
1136 			/* nh MAY be empty if RTM_CHANGE request is no-op */
1137 			nh = rc.rc_nh_new;
1138 			if (nh != NULL) {
1139 				rtm->rtm_index = nh->nh_ifp->if_index;
1140 				rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1141 			}
1142 		}
1143 		break;
1144 
1145 	case RTM_DELETE:
1146 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1147 		if (error == 0) {
1148 			rtsock_notify_event(fibnum, &rc);
1149 #ifdef ROUTE_MPATH
1150 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1151 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1152 				struct rib_cmd_info rc_simple = {};
1153 				rib_decompose_notification(&rc,
1154 				    save_del_notification, (void *)&rc_simple);
1155 				rc = rc_simple;
1156 			}
1157 #endif
1158 			nh = rc.rc_nh_old;
1159 		}
1160 		break;
1161 
1162 	case RTM_GET:
1163 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1164 		if (error != 0)
1165 			senderr(error);
1166 		nh = rc.rc_nh_new;
1167 
1168 		if (!rt_is_exportable(rc.rc_rt, curthread->td_ucred))
1169 			senderr(ESRCH);
1170 		break;
1171 
1172 	default:
1173 		senderr(EOPNOTSUPP);
1174 	}
1175 
1176 	if (error == 0 && nh != NULL) {
1177 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1178 		/*
1179 		 * Note that some sockaddr pointers may have changed to
1180 		 * point to memory outsize @rtm. Some may be pointing
1181 		 * to the on-stack variables.
1182 		 * Given that, any pointer in @info CANNOT BE USED.
1183 		 */
1184 
1185 		/*
1186 		 * scopeid deembedding has been performed while
1187 		 * writing updated rtm in rtsock_msg_buffer().
1188 		 * With that in mind, skip deembedding procedure below.
1189 		 */
1190 #ifdef INET6
1191 		rti_need_deembed = 0;
1192 #endif
1193 	}
1194 
1195 flush:
1196 	NET_EPOCH_EXIT(et);
1197 
1198 #ifdef INET6
1199 	if (rtm != NULL) {
1200 		if (rti_need_deembed) {
1201 			/* sin6_scope_id is recovered before sending rtm. */
1202 			sin6 = (struct sockaddr_in6 *)&ss;
1203 			for (i = 0; i < RTAX_MAX; i++) {
1204 				if (info.rti_info[i] == NULL)
1205 					continue;
1206 				if (info.rti_info[i]->sa_family != AF_INET6)
1207 					continue;
1208 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1209 				if (sa6_recoverscope(sin6) == 0)
1210 					bcopy(sin6, info.rti_info[i],
1211 						    sizeof(*sin6));
1212 			}
1213 			if (update_rtm_from_info(&info, &rtm, alloc_len) != 0) {
1214 				if (error != 0)
1215 					error = ENOBUFS;
1216 			}
1217 		}
1218 	}
1219 #endif
1220 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1221 
1222 	return (error);
1223 }
1224 
1225 /*
1226  * Sends the prepared reply message in @rtm to all rtsock clients.
1227  * Frees @m and @rtm.
1228  *
1229  */
1230 static void
1231 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1232     sa_family_t saf, u_int fibnum, int rtm_errno)
1233 {
1234 	struct rcb *rcb = NULL;
1235 
1236 	/*
1237 	 * Check to see if we don't want our own messages.
1238 	 */
1239 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1240 		if (V_route_cb.any_count <= 1) {
1241 			if (rtm != NULL)
1242 				free(rtm, M_TEMP);
1243 			m_freem(m);
1244 			return;
1245 		}
1246 		/* There is another listener, so construct message */
1247 		rcb = so->so_pcb;
1248 	}
1249 
1250 	if (rtm != NULL) {
1251 		if (rtm_errno!= 0)
1252 			rtm->rtm_errno = rtm_errno;
1253 		else
1254 			rtm->rtm_flags |= RTF_DONE;
1255 
1256 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1257 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1258 			m_freem(m);
1259 			m = NULL;
1260 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1261 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1262 
1263 		free(rtm, M_TEMP);
1264 	}
1265 	if (m != NULL) {
1266 		M_SETFIB(m, fibnum);
1267 		m->m_flags |= RTS_FILTER_FIB;
1268 		if (rcb) {
1269 			/*
1270 			 * XXX insure we don't get a copy by
1271 			 * invalidating our protocol
1272 			 */
1273 			sa_family_t family = rcb->rcb_family;
1274 			rcb->rcb_family = AF_UNSPEC;
1275 			rt_dispatch(m, saf);
1276 			rcb->rcb_family = family;
1277 		} else
1278 			rt_dispatch(m, saf);
1279 	}
1280 }
1281 
1282 static void
1283 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1284     struct rt_metrics *out)
1285 {
1286 
1287 	bzero(out, sizeof(*out));
1288 	out->rmx_mtu = nh->nh_mtu;
1289 	out->rmx_weight = rt->rt_weight;
1290 	out->rmx_nhidx = nhop_get_idx(nh);
1291 	/* Kernel -> userland timebase conversion. */
1292 	out->rmx_expire = nhop_get_expire(nh) ?
1293 	    nhop_get_expire(nh) - time_uptime + time_second : 0;
1294 }
1295 
1296 /*
1297  * Extract the addresses of the passed sockaddrs.
1298  * Do a little sanity checking so as to avoid bad memory references.
1299  * This data is derived straight from userland.
1300  */
1301 static int
1302 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1303 {
1304 	struct sockaddr *sa;
1305 	int i;
1306 
1307 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1308 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1309 			continue;
1310 		sa = (struct sockaddr *)cp;
1311 		/*
1312 		 * It won't fit.
1313 		 */
1314 		if (cp + sa->sa_len > cplim) {
1315 			RTS_PID_LOG(LOG_DEBUG, "sa_len too big for sa type %d", i);
1316 			return (EINVAL);
1317 		}
1318 		/*
1319 		 * there are no more.. quit now
1320 		 * If there are more bits, they are in error.
1321 		 * I've seen this. route(1) can evidently generate these.
1322 		 * This causes kernel to core dump.
1323 		 * for compatibility, If we see this, point to a safe address.
1324 		 */
1325 		if (sa->sa_len == 0) {
1326 			rtinfo->rti_info[i] = &sa_zero;
1327 			return (0); /* should be EINVAL but for compat */
1328 		}
1329 		/* accept it */
1330 #ifdef INET6
1331 		if (sa->sa_family == AF_INET6)
1332 			sa6_embedscope((struct sockaddr_in6 *)sa,
1333 			    V_ip6_use_defzone);
1334 #endif
1335 		rtinfo->rti_info[i] = sa;
1336 		cp += SA_SIZE(sa);
1337 	}
1338 	return (0);
1339 }
1340 
1341 #ifdef INET
1342 static inline void
1343 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1344 {
1345 
1346 	const struct sockaddr_in nsin = {
1347 		.sin_family = AF_INET,
1348 		.sin_len = sizeof(struct sockaddr_in),
1349 		.sin_addr = addr,
1350 	};
1351 	*sin = nsin;
1352 }
1353 #endif
1354 
1355 #ifdef INET6
1356 static inline void
1357 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1358     uint32_t scopeid)
1359 {
1360 
1361 	const struct sockaddr_in6 nsin6 = {
1362 		.sin6_family = AF_INET6,
1363 		.sin6_len = sizeof(struct sockaddr_in6),
1364 		.sin6_addr = *addr6,
1365 		.sin6_scope_id = scopeid,
1366 	};
1367 	*sin6 = nsin6;
1368 }
1369 #endif
1370 
1371 #if defined(INET6) || defined(INET)
1372 /*
1373  * Checks if gateway is suitable for lltable operations.
1374  * Lltable code requires AF_LINK gateway with ifindex
1375  *  and mac address specified.
1376  * Returns 0 on success.
1377  */
1378 static int
1379 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1380 {
1381 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1382 
1383 	if (sdl->sdl_family != AF_LINK)
1384 		return (EINVAL);
1385 
1386 	if (sdl->sdl_index == 0) {
1387 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gateway w/o ifindex");
1388 		return (EINVAL);
1389 	}
1390 
1391 	if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len) {
1392 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gw: sdl_nlen/sdl_alen too large");
1393 		return (EINVAL);
1394 	}
1395 
1396 	return (0);
1397 }
1398 
1399 static int
1400 cleanup_xaddrs_gateway(struct rt_addrinfo *info, struct linear_buffer *lb)
1401 {
1402 	struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1403 	struct sockaddr *sa;
1404 
1405 	if (info->rti_flags & RTF_LLDATA)
1406 		return (cleanup_xaddrs_lladdr(info));
1407 
1408 	switch (gw->sa_family) {
1409 #ifdef INET
1410 	case AF_INET:
1411 		{
1412 			struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw;
1413 
1414 			/* Ensure reads do not go beyoud SA boundary */
1415 			if (SA_SIZE(gw) < offsetof(struct sockaddr_in, sin_zero)) {
1416 				RTS_PID_LOG(LOG_DEBUG, "gateway sin_len too small: %d",
1417 				    gw->sa_len);
1418 				return (EINVAL);
1419 			}
1420 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_in));
1421 			if (sa == NULL)
1422 				return (ENOBUFS);
1423 			fill_sockaddr_inet((struct sockaddr_in *)sa, gw_sin->sin_addr);
1424 			info->rti_info[RTAX_GATEWAY] = sa;
1425 		}
1426 		break;
1427 #endif
1428 #ifdef INET6
1429 	case AF_INET6:
1430 		{
1431 			struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1432 			if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1433 				RTS_PID_LOG(LOG_DEBUG, "gateway sin6_len too small: %d",
1434 				    gw->sa_len);
1435 				return (EINVAL);
1436 			}
1437 			fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1438 			break;
1439 		}
1440 #endif
1441 	case AF_LINK:
1442 		{
1443 			struct sockaddr_dl *gw_sdl;
1444 
1445 			size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data);
1446 			gw_sdl = (struct sockaddr_dl *)gw;
1447 			if (gw_sdl->sdl_len < sdl_min_len) {
1448 				RTS_PID_LOG(LOG_DEBUG, "gateway sdl_len too small: %d",
1449 				    gw_sdl->sdl_len);
1450 				return (EINVAL);
1451 			}
1452 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_dl_short));
1453 			if (sa == NULL)
1454 				return (ENOBUFS);
1455 
1456 			const struct sockaddr_dl_short sdl = {
1457 				.sdl_family = AF_LINK,
1458 				.sdl_len = sizeof(struct sockaddr_dl_short),
1459 				.sdl_index = gw_sdl->sdl_index,
1460 			};
1461 			*((struct sockaddr_dl_short *)sa) = sdl;
1462 			info->rti_info[RTAX_GATEWAY] = sa;
1463 			break;
1464 		}
1465 	}
1466 
1467 	return (0);
1468 }
1469 #endif
1470 
1471 static void
1472 remove_netmask(struct rt_addrinfo *info)
1473 {
1474 	info->rti_info[RTAX_NETMASK] = NULL;
1475 	info->rti_flags |= RTF_HOST;
1476 	info->rti_addrs &= ~RTA_NETMASK;
1477 }
1478 
1479 #ifdef INET
1480 static int
1481 cleanup_xaddrs_inet(struct rt_addrinfo *info, struct linear_buffer *lb)
1482 {
1483 	struct sockaddr_in *dst_sa, *mask_sa;
1484 	const int sa_len = sizeof(struct sockaddr_in);
1485 	struct in_addr dst, mask;
1486 
1487 	/* Check & fixup dst/netmask combination first */
1488 	dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1489 	mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1490 
1491 	/* Ensure reads do not go beyound the buffer size */
1492 	if (SA_SIZE(dst_sa) < offsetof(struct sockaddr_in, sin_zero)) {
1493 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin_len too small: %d",
1494 		    dst_sa->sin_len);
1495 		return (EINVAL);
1496 	}
1497 
1498 	if ((mask_sa != NULL) && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1499 		/*
1500 		 * Some older routing software encode mask length into the
1501 		 * sin_len, thus resulting in "truncated" sockaddr.
1502 		 */
1503 		int len = mask_sa->sin_len - offsetof(struct sockaddr_in, sin_addr);
1504 		if (len >= 0) {
1505 			mask.s_addr = 0;
1506 			if (len > sizeof(struct in_addr))
1507 				len = sizeof(struct in_addr);
1508 			memcpy(&mask, &mask_sa->sin_addr, len);
1509 		} else {
1510 			RTS_PID_LOG(LOG_DEBUG, "prefix mask sin_len too small: %d",
1511 			    mask_sa->sin_len);
1512 			return (EINVAL);
1513 		}
1514 	} else
1515 		mask.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST;
1516 
1517 	dst.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr));
1518 
1519 	/* Construct new "clean" dst/mask sockaddresses */
1520 	if ((dst_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1521 		return (ENOBUFS);
1522 	fill_sockaddr_inet(dst_sa, dst);
1523 	info->rti_info[RTAX_DST] = (struct sockaddr *)dst_sa;
1524 
1525 	if (mask.s_addr != INADDR_BROADCAST) {
1526 		if ((mask_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1527 			return (ENOBUFS);
1528 		fill_sockaddr_inet(mask_sa, mask);
1529 		info->rti_info[RTAX_NETMASK] = (struct sockaddr *)mask_sa;
1530 		info->rti_flags &= ~RTF_HOST;
1531 	} else
1532 		remove_netmask(info);
1533 
1534 	/* Check gateway */
1535 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1536 		return (cleanup_xaddrs_gateway(info, lb));
1537 
1538 	return (0);
1539 }
1540 #endif
1541 
1542 #ifdef INET6
1543 static int
1544 cleanup_xaddrs_inet6(struct rt_addrinfo *info, struct linear_buffer *lb)
1545 {
1546 	struct sockaddr *sa;
1547 	struct sockaddr_in6 *dst_sa, *mask_sa;
1548 	struct in6_addr mask, *dst;
1549 	const int sa_len = sizeof(struct sockaddr_in6);
1550 
1551 	/* Check & fixup dst/netmask combination first */
1552 	dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1553 	mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1554 
1555 	if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1556 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin6_len too small: %d",
1557 		    dst_sa->sin6_len);
1558 		return (EINVAL);
1559 	}
1560 
1561 	if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1562 		/*
1563 		 * Some older routing software encode mask length into the
1564 		 * sin6_len, thus resulting in "truncated" sockaddr.
1565 		 */
1566 		int len = mask_sa->sin6_len - offsetof(struct sockaddr_in6, sin6_addr);
1567 		if (len >= 0) {
1568 			bzero(&mask, sizeof(mask));
1569 			if (len > sizeof(struct in6_addr))
1570 				len = sizeof(struct in6_addr);
1571 			memcpy(&mask, &mask_sa->sin6_addr, len);
1572 		} else {
1573 			RTS_PID_LOG(LOG_DEBUG, "rtsock: prefix mask sin6_len too small: %d",
1574 			    mask_sa->sin6_len);
1575 			return (EINVAL);
1576 		}
1577 	} else
1578 		mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1579 
1580 	dst = &dst_sa->sin6_addr;
1581 	IN6_MASK_ADDR(dst, &mask);
1582 
1583 	if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1584 		return (ENOBUFS);
1585 	fill_sockaddr_inet6((struct sockaddr_in6 *)sa, dst, 0);
1586 	info->rti_info[RTAX_DST] = sa;
1587 
1588 	if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128)) {
1589 		if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1590 			return (ENOBUFS);
1591 		fill_sockaddr_inet6((struct sockaddr_in6 *)sa, &mask, 0);
1592 		info->rti_info[RTAX_NETMASK] = sa;
1593 		info->rti_flags &= ~RTF_HOST;
1594 	} else
1595 		remove_netmask(info);
1596 
1597 	/* Check gateway */
1598 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1599 		return (cleanup_xaddrs_gateway(info, lb));
1600 
1601 	return (0);
1602 }
1603 #endif
1604 
1605 static int
1606 cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb)
1607 {
1608 	int error = EAFNOSUPPORT;
1609 
1610 	if (info->rti_info[RTAX_DST] == NULL) {
1611 		RTS_PID_LOG(LOG_DEBUG, "prefix dst is not set");
1612 		return (EINVAL);
1613 	}
1614 
1615 	if (info->rti_flags & RTF_LLDATA) {
1616 		/*
1617 		 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1618 		 * prefix along with the actual address in RTA_DST.
1619 		 * Remove netmask to avoid unnecessary address masking.
1620 		 */
1621 		remove_netmask(info);
1622 	}
1623 
1624 	switch (info->rti_info[RTAX_DST]->sa_family) {
1625 #ifdef INET
1626 	case AF_INET:
1627 		error = cleanup_xaddrs_inet(info, lb);
1628 		break;
1629 #endif
1630 #ifdef INET6
1631 	case AF_INET6:
1632 		error = cleanup_xaddrs_inet6(info, lb);
1633 		break;
1634 #endif
1635 	}
1636 
1637 	return (error);
1638 }
1639 
1640 /*
1641  * Fill in @dmask with valid netmask leaving original @smask
1642  * intact. Mostly used with radix netmasks.
1643  */
1644 struct sockaddr *
1645 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1646     struct sockaddr_storage *dmask)
1647 {
1648 	if (dst == NULL || smask == NULL)
1649 		return (NULL);
1650 
1651 	memset(dmask, 0, dst->sa_len);
1652 	memcpy(dmask, smask, smask->sa_len);
1653 	dmask->ss_len = dst->sa_len;
1654 	dmask->ss_family = dst->sa_family;
1655 
1656 	return ((struct sockaddr *)dmask);
1657 }
1658 
1659 /*
1660  * Writes information related to @rtinfo object to newly-allocated mbuf.
1661  * Assumes MCLBYTES is enough to construct any message.
1662  * Used for OS notifications of vaious events (if/ifa announces,etc)
1663  *
1664  * Returns allocated mbuf or NULL on failure.
1665  */
1666 static struct mbuf *
1667 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1668 {
1669 	struct sockaddr_storage ss;
1670 	struct rt_msghdr *rtm;
1671 	struct mbuf *m;
1672 	int i;
1673 	struct sockaddr *sa;
1674 #ifdef INET6
1675 	struct sockaddr_in6 *sin6;
1676 #endif
1677 	int len, dlen;
1678 
1679 	switch (type) {
1680 	case RTM_DELADDR:
1681 	case RTM_NEWADDR:
1682 		len = sizeof(struct ifa_msghdr);
1683 		break;
1684 
1685 	case RTM_DELMADDR:
1686 	case RTM_NEWMADDR:
1687 		len = sizeof(struct ifma_msghdr);
1688 		break;
1689 
1690 	case RTM_IFINFO:
1691 		len = sizeof(struct if_msghdr);
1692 		break;
1693 
1694 	case RTM_IFANNOUNCE:
1695 	case RTM_IEEE80211:
1696 		len = sizeof(struct if_announcemsghdr);
1697 		break;
1698 
1699 	default:
1700 		len = sizeof(struct rt_msghdr);
1701 	}
1702 
1703 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1704 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1705 	if (len > MHLEN)
1706 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1707 	else
1708 		m = m_gethdr(M_NOWAIT, MT_DATA);
1709 	if (m == NULL)
1710 		return (m);
1711 
1712 	m->m_pkthdr.len = m->m_len = len;
1713 	rtm = mtod(m, struct rt_msghdr *);
1714 	bzero((caddr_t)rtm, len);
1715 	for (i = 0; i < RTAX_MAX; i++) {
1716 		if ((sa = rtinfo->rti_info[i]) == NULL)
1717 			continue;
1718 		rtinfo->rti_addrs |= (1 << i);
1719 
1720 		dlen = SA_SIZE(sa);
1721 		KASSERT(dlen <= sizeof(ss),
1722 		    ("%s: sockaddr size overflow", __func__));
1723 		bzero(&ss, sizeof(ss));
1724 		bcopy(sa, &ss, sa->sa_len);
1725 		sa = (struct sockaddr *)&ss;
1726 #ifdef INET6
1727 		if (sa->sa_family == AF_INET6) {
1728 			sin6 = (struct sockaddr_in6 *)sa;
1729 			(void)sa6_recoverscope(sin6);
1730 		}
1731 #endif
1732 		m_copyback(m, len, dlen, (caddr_t)sa);
1733 		len += dlen;
1734 	}
1735 	if (m->m_pkthdr.len != len) {
1736 		m_freem(m);
1737 		return (NULL);
1738 	}
1739 	rtm->rtm_msglen = len;
1740 	rtm->rtm_version = RTM_VERSION;
1741 	rtm->rtm_type = type;
1742 	return (m);
1743 }
1744 
1745 /*
1746  * Writes information related to @rtinfo object to preallocated buffer.
1747  * Stores needed size in @plen. If @w is NULL, calculates size without
1748  * writing.
1749  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1750  *
1751  * Returns 0 on success.
1752  *
1753  */
1754 static int
1755 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1756 {
1757 	struct sockaddr_storage ss;
1758 	int len, buflen = 0, dlen, i;
1759 	caddr_t cp = NULL;
1760 	struct rt_msghdr *rtm = NULL;
1761 #ifdef INET6
1762 	struct sockaddr_in6 *sin6;
1763 #endif
1764 #ifdef COMPAT_FREEBSD32
1765 	bool compat32 = false;
1766 #endif
1767 
1768 	switch (type) {
1769 	case RTM_DELADDR:
1770 	case RTM_NEWADDR:
1771 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1772 #ifdef COMPAT_FREEBSD32
1773 			if (w->w_req->flags & SCTL_MASK32) {
1774 				len = sizeof(struct ifa_msghdrl32);
1775 				compat32 = true;
1776 			} else
1777 #endif
1778 				len = sizeof(struct ifa_msghdrl);
1779 		} else
1780 			len = sizeof(struct ifa_msghdr);
1781 		break;
1782 
1783 	case RTM_IFINFO:
1784 #ifdef COMPAT_FREEBSD32
1785 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1786 			if (w->w_op == NET_RT_IFLISTL)
1787 				len = sizeof(struct if_msghdrl32);
1788 			else
1789 				len = sizeof(struct if_msghdr32);
1790 			compat32 = true;
1791 			break;
1792 		}
1793 #endif
1794 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1795 			len = sizeof(struct if_msghdrl);
1796 		else
1797 			len = sizeof(struct if_msghdr);
1798 		break;
1799 
1800 	case RTM_NEWMADDR:
1801 		len = sizeof(struct ifma_msghdr);
1802 		break;
1803 
1804 	default:
1805 		len = sizeof(struct rt_msghdr);
1806 	}
1807 
1808 	if (w != NULL) {
1809 		rtm = (struct rt_msghdr *)w->w_tmem;
1810 		buflen = w->w_tmemsize - len;
1811 		cp = (caddr_t)w->w_tmem + len;
1812 	}
1813 
1814 	rtinfo->rti_addrs = 0;
1815 	for (i = 0; i < RTAX_MAX; i++) {
1816 		struct sockaddr *sa;
1817 
1818 		if ((sa = rtinfo->rti_info[i]) == NULL)
1819 			continue;
1820 		rtinfo->rti_addrs |= (1 << i);
1821 #ifdef COMPAT_FREEBSD32
1822 		if (compat32)
1823 			dlen = SA_SIZE32(sa);
1824 		else
1825 #endif
1826 			dlen = SA_SIZE(sa);
1827 		if (cp != NULL && buflen >= dlen) {
1828 			KASSERT(dlen <= sizeof(ss),
1829 			    ("%s: sockaddr size overflow", __func__));
1830 			bzero(&ss, sizeof(ss));
1831 			bcopy(sa, &ss, sa->sa_len);
1832 			sa = (struct sockaddr *)&ss;
1833 #ifdef INET6
1834 			if (sa->sa_family == AF_INET6) {
1835 				sin6 = (struct sockaddr_in6 *)sa;
1836 				(void)sa6_recoverscope(sin6);
1837 			}
1838 #endif
1839 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1840 			cp += dlen;
1841 			buflen -= dlen;
1842 		} else if (cp != NULL) {
1843 			/*
1844 			 * Buffer too small. Count needed size
1845 			 * and return with error.
1846 			 */
1847 			cp = NULL;
1848 		}
1849 
1850 		len += dlen;
1851 	}
1852 
1853 	if (cp != NULL) {
1854 		dlen = ALIGN(len) - len;
1855 		if (buflen < dlen)
1856 			cp = NULL;
1857 		else {
1858 			bzero(cp, dlen);
1859 			cp += dlen;
1860 			buflen -= dlen;
1861 		}
1862 	}
1863 	len = ALIGN(len);
1864 
1865 	if (cp != NULL) {
1866 		/* fill header iff buffer is large enough */
1867 		rtm->rtm_version = RTM_VERSION;
1868 		rtm->rtm_type = type;
1869 		rtm->rtm_msglen = len;
1870 	}
1871 
1872 	*plen = len;
1873 
1874 	if (w != NULL && cp == NULL)
1875 		return (ENOBUFS);
1876 
1877 	return (0);
1878 }
1879 
1880 /*
1881  * This routine is called to generate a message from the routing
1882  * socket indicating that a redirect has occurred, a routing lookup
1883  * has failed, or that a protocol has detected timeouts to a particular
1884  * destination.
1885  */
1886 void
1887 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1888     int fibnum)
1889 {
1890 	struct rt_msghdr *rtm;
1891 	struct mbuf *m;
1892 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1893 
1894 	if (V_route_cb.any_count == 0)
1895 		return;
1896 	m = rtsock_msg_mbuf(type, rtinfo);
1897 	if (m == NULL)
1898 		return;
1899 
1900 	if (fibnum != RT_ALL_FIBS) {
1901 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1902 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1903 		M_SETFIB(m, fibnum);
1904 		m->m_flags |= RTS_FILTER_FIB;
1905 	}
1906 
1907 	rtm = mtod(m, struct rt_msghdr *);
1908 	rtm->rtm_flags = RTF_DONE | flags;
1909 	rtm->rtm_errno = error;
1910 	rtm->rtm_addrs = rtinfo->rti_addrs;
1911 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1912 }
1913 
1914 void
1915 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1916 {
1917 
1918 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1919 }
1920 
1921 /*
1922  * This routine is called to generate a message from the routing
1923  * socket indicating that the status of a network interface has changed.
1924  */
1925 static void
1926 rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask __unused)
1927 {
1928 	struct if_msghdr *ifm;
1929 	struct mbuf *m;
1930 	struct rt_addrinfo info;
1931 
1932 	if (V_route_cb.any_count == 0)
1933 		return;
1934 	bzero((caddr_t)&info, sizeof(info));
1935 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1936 	if (m == NULL)
1937 		return;
1938 	ifm = mtod(m, struct if_msghdr *);
1939 	ifm->ifm_index = ifp->if_index;
1940 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1941 	if_data_copy(ifp, &ifm->ifm_data);
1942 	ifm->ifm_addrs = 0;
1943 	rt_dispatch(m, AF_UNSPEC);
1944 }
1945 
1946 /*
1947  * Announce interface address arrival/withdraw.
1948  * Please do not call directly, use rt_addrmsg().
1949  * Assume input data to be valid.
1950  * Returns 0 on success.
1951  */
1952 int
1953 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1954 {
1955 	struct rt_addrinfo info;
1956 	struct sockaddr *sa;
1957 	int ncmd;
1958 	struct mbuf *m;
1959 	struct ifa_msghdr *ifam;
1960 	struct ifnet *ifp = ifa->ifa_ifp;
1961 	struct sockaddr_storage ss;
1962 
1963 	if (V_route_cb.any_count == 0)
1964 		return (0);
1965 
1966 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1967 
1968 	bzero((caddr_t)&info, sizeof(info));
1969 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1970 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1971 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1972 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1973 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1974 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1975 		return (ENOBUFS);
1976 	ifam = mtod(m, struct ifa_msghdr *);
1977 	ifam->ifam_index = ifp->if_index;
1978 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1979 	ifam->ifam_flags = ifa->ifa_flags;
1980 	ifam->ifam_addrs = info.rti_addrs;
1981 
1982 	if (fibnum != RT_ALL_FIBS) {
1983 		M_SETFIB(m, fibnum);
1984 		m->m_flags |= RTS_FILTER_FIB;
1985 	}
1986 
1987 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1988 
1989 	return (0);
1990 }
1991 
1992 /*
1993  * Announce route addition/removal to rtsock based on @rt data.
1994  * Callers are advives to use rt_routemsg() instead of using this
1995  *  function directly.
1996  * Assume @rt data is consistent.
1997  *
1998  * Returns 0 on success.
1999  */
2000 int
2001 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
2002     int fibnum)
2003 {
2004 	union sockaddr_union dst, mask;
2005 	struct rt_addrinfo info;
2006 
2007 	if (V_route_cb.any_count == 0)
2008 		return (0);
2009 
2010 	int family = rt_get_family(rt);
2011 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
2012 	export_rtaddrs(rt, &dst.sa, &mask.sa);
2013 
2014 	bzero((caddr_t)&info, sizeof(info));
2015 	info.rti_info[RTAX_DST] = &dst.sa;
2016 	info.rti_info[RTAX_NETMASK] = &mask.sa;
2017 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2018 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
2019 	info.rti_ifp = nh->nh_ifp;
2020 
2021 	return (rtsock_routemsg_info(cmd, &info, fibnum));
2022 }
2023 
2024 int
2025 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
2026 {
2027 	struct rt_msghdr *rtm;
2028 	struct sockaddr *sa;
2029 	struct mbuf *m;
2030 
2031 	if (V_route_cb.any_count == 0)
2032 		return (0);
2033 
2034 	if (info->rti_flags & RTF_HOST)
2035 		info->rti_info[RTAX_NETMASK] = NULL;
2036 
2037 	m = rtsock_msg_mbuf(cmd, info);
2038 	if (m == NULL)
2039 		return (ENOBUFS);
2040 
2041 	if (fibnum != RT_ALL_FIBS) {
2042 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
2043 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
2044 		M_SETFIB(m, fibnum);
2045 		m->m_flags |= RTS_FILTER_FIB;
2046 	}
2047 
2048 	rtm = mtod(m, struct rt_msghdr *);
2049 	rtm->rtm_addrs = info->rti_addrs;
2050 	if (info->rti_ifp != NULL)
2051 		rtm->rtm_index = info->rti_ifp->if_index;
2052 	/* Add RTF_DONE to indicate command 'completion' required by API */
2053 	info->rti_flags |= RTF_DONE;
2054 	/* Reported routes has to be up */
2055 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
2056 		info->rti_flags |= RTF_UP;
2057 	rtm->rtm_flags = info->rti_flags;
2058 
2059 	sa = info->rti_info[RTAX_DST];
2060 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2061 
2062 	return (0);
2063 }
2064 
2065 /*
2066  * This is the analogue to the rt_newaddrmsg which performs the same
2067  * function but for multicast group memberhips.  This is easier since
2068  * there is no route state to worry about.
2069  */
2070 void
2071 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
2072 {
2073 	struct rt_addrinfo info;
2074 	struct mbuf *m = NULL;
2075 	struct ifnet *ifp = ifma->ifma_ifp;
2076 	struct ifma_msghdr *ifmam;
2077 
2078 	if (V_route_cb.any_count == 0)
2079 		return;
2080 
2081 	bzero((caddr_t)&info, sizeof(info));
2082 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2083 	if (ifp && ifp->if_addr)
2084 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2085 	else
2086 		info.rti_info[RTAX_IFP] = NULL;
2087 	/*
2088 	 * If a link-layer address is present, present it as a ``gateway''
2089 	 * (similarly to how ARP entries, e.g., are presented).
2090 	 */
2091 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
2092 	m = rtsock_msg_mbuf(cmd, &info);
2093 	if (m == NULL)
2094 		return;
2095 	ifmam = mtod(m, struct ifma_msghdr *);
2096 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
2097 	    __func__));
2098 	ifmam->ifmam_index = ifp->if_index;
2099 	ifmam->ifmam_addrs = info.rti_addrs;
2100 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
2101 }
2102 
2103 static struct mbuf *
2104 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
2105 	struct rt_addrinfo *info)
2106 {
2107 	struct if_announcemsghdr *ifan;
2108 	struct mbuf *m;
2109 
2110 	if (V_route_cb.any_count == 0)
2111 		return NULL;
2112 	bzero((caddr_t)info, sizeof(*info));
2113 	m = rtsock_msg_mbuf(type, info);
2114 	if (m != NULL) {
2115 		ifan = mtod(m, struct if_announcemsghdr *);
2116 		ifan->ifan_index = ifp->if_index;
2117 		strlcpy(ifan->ifan_name, ifp->if_xname,
2118 			sizeof(ifan->ifan_name));
2119 		ifan->ifan_what = what;
2120 	}
2121 	return m;
2122 }
2123 
2124 /*
2125  * This is called to generate routing socket messages indicating
2126  * IEEE80211 wireless events.
2127  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2128  */
2129 void
2130 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2131 {
2132 	struct mbuf *m;
2133 	struct rt_addrinfo info;
2134 
2135 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2136 	if (m != NULL) {
2137 		/*
2138 		 * Append the ieee80211 data.  Try to stick it in the
2139 		 * mbuf containing the ifannounce msg; otherwise allocate
2140 		 * a new mbuf and append.
2141 		 *
2142 		 * NB: we assume m is a single mbuf.
2143 		 */
2144 		if (data_len > M_TRAILINGSPACE(m)) {
2145 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2146 			if (n == NULL) {
2147 				m_freem(m);
2148 				return;
2149 			}
2150 			bcopy(data, mtod(n, void *), data_len);
2151 			n->m_len = data_len;
2152 			m->m_next = n;
2153 		} else if (data_len > 0) {
2154 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2155 			m->m_len += data_len;
2156 		}
2157 		if (m->m_flags & M_PKTHDR)
2158 			m->m_pkthdr.len += data_len;
2159 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2160 		rt_dispatch(m, AF_UNSPEC);
2161 	}
2162 }
2163 
2164 /*
2165  * This is called to generate routing socket messages indicating
2166  * network interface arrival and departure.
2167  */
2168 static void
2169 rt_ifannouncemsg(struct ifnet *ifp, int what)
2170 {
2171 	struct mbuf *m;
2172 	struct rt_addrinfo info;
2173 
2174 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2175 	if (m != NULL)
2176 		rt_dispatch(m, AF_UNSPEC);
2177 }
2178 
2179 static void
2180 rt_dispatch(struct mbuf *m, sa_family_t saf)
2181 {
2182 
2183 	M_ASSERTPKTHDR(m);
2184 
2185 	m->m_rtsock_family = saf;
2186 	if (V_loif)
2187 		m->m_pkthdr.rcvif = V_loif;
2188 	else {
2189 		m_freem(m);
2190 		return;
2191 	}
2192 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
2193 }
2194 
2195 /*
2196  * This is used in dumping the kernel table via sysctl().
2197  */
2198 static int
2199 sysctl_dumpentry(struct rtentry *rt, void *vw)
2200 {
2201 	struct walkarg *w = vw;
2202 	struct nhop_object *nh;
2203 
2204 	NET_EPOCH_ASSERT();
2205 
2206 	if (!rt_is_exportable(rt, w->w_req->td->td_ucred))
2207 		return (0);
2208 
2209 	export_rtaddrs(rt, w->dst, w->mask);
2210 	nh = rt_get_raw_nhop(rt);
2211 #ifdef ROUTE_MPATH
2212 	if (NH_IS_NHGRP(nh)) {
2213 		const struct weightened_nhop *wn;
2214 		uint32_t num_nhops;
2215 		int error;
2216 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2217 		for (int i = 0; i < num_nhops; i++) {
2218 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2219 			if (error != 0)
2220 				return (error);
2221 		}
2222 	} else
2223 #endif
2224 		sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2225 
2226 	return (0);
2227 }
2228 
2229 
2230 static int
2231 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2232     struct walkarg *w)
2233 {
2234 	struct rt_addrinfo info;
2235 	int error = 0, size;
2236 	uint32_t rtflags;
2237 
2238 	rtflags = nhop_get_rtflags(nh);
2239 
2240 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2241 		return (0);
2242 
2243 	bzero((caddr_t)&info, sizeof(info));
2244 	info.rti_info[RTAX_DST] = w->dst;
2245 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2246 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2247 	info.rti_info[RTAX_GENMASK] = 0;
2248 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2249 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2250 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2251 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2252 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2253 	}
2254 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2255 		return (error);
2256 	if (w->w_req && w->w_tmem) {
2257 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2258 
2259 		bzero(&rtm->rtm_index,
2260 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2261 
2262 		/*
2263 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2264 		 * and RTF_UP (if entry is linked, which is always true here).
2265 		 * Given that, use nhop rtflags & add RTF_UP.
2266 		 */
2267 		rtm->rtm_flags = rtflags | RTF_UP;
2268 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2269 			rtm->rtm_flags = RTF_GATEWAY |
2270 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2271 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2272 		rtm->rtm_rmx.rmx_weight = weight;
2273 		rtm->rtm_index = nh->nh_ifp->if_index;
2274 		rtm->rtm_addrs = info.rti_addrs;
2275 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2276 		return (error);
2277 	}
2278 	return (error);
2279 }
2280 
2281 static int
2282 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2283     struct rt_addrinfo *info, struct walkarg *w, int len)
2284 {
2285 	struct if_msghdrl *ifm;
2286 	struct if_data *ifd;
2287 
2288 	ifm = (struct if_msghdrl *)w->w_tmem;
2289 
2290 #ifdef COMPAT_FREEBSD32
2291 	if (w->w_req->flags & SCTL_MASK32) {
2292 		struct if_msghdrl32 *ifm32;
2293 
2294 		ifm32 = (struct if_msghdrl32 *)ifm;
2295 		ifm32->ifm_addrs = info->rti_addrs;
2296 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2297 		ifm32->ifm_index = ifp->if_index;
2298 		ifm32->_ifm_spare1 = 0;
2299 		ifm32->ifm_len = sizeof(*ifm32);
2300 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2301 		ifm32->_ifm_spare2 = 0;
2302 		ifd = &ifm32->ifm_data;
2303 	} else
2304 #endif
2305 	{
2306 		ifm->ifm_addrs = info->rti_addrs;
2307 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2308 		ifm->ifm_index = ifp->if_index;
2309 		ifm->_ifm_spare1 = 0;
2310 		ifm->ifm_len = sizeof(*ifm);
2311 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2312 		ifm->_ifm_spare2 = 0;
2313 		ifd = &ifm->ifm_data;
2314 	}
2315 
2316 	memcpy(ifd, src_ifd, sizeof(*ifd));
2317 
2318 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2319 }
2320 
2321 static int
2322 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2323     struct rt_addrinfo *info, struct walkarg *w, int len)
2324 {
2325 	struct if_msghdr *ifm;
2326 	struct if_data *ifd;
2327 
2328 	ifm = (struct if_msghdr *)w->w_tmem;
2329 
2330 #ifdef COMPAT_FREEBSD32
2331 	if (w->w_req->flags & SCTL_MASK32) {
2332 		struct if_msghdr32 *ifm32;
2333 
2334 		ifm32 = (struct if_msghdr32 *)ifm;
2335 		ifm32->ifm_addrs = info->rti_addrs;
2336 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2337 		ifm32->ifm_index = ifp->if_index;
2338 		ifm32->_ifm_spare1 = 0;
2339 		ifd = &ifm32->ifm_data;
2340 	} else
2341 #endif
2342 	{
2343 		ifm->ifm_addrs = info->rti_addrs;
2344 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2345 		ifm->ifm_index = ifp->if_index;
2346 		ifm->_ifm_spare1 = 0;
2347 		ifd = &ifm->ifm_data;
2348 	}
2349 
2350 	memcpy(ifd, src_ifd, sizeof(*ifd));
2351 
2352 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2353 }
2354 
2355 static int
2356 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2357     struct walkarg *w, int len)
2358 {
2359 	struct ifa_msghdrl *ifam;
2360 	struct if_data *ifd;
2361 
2362 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2363 
2364 #ifdef COMPAT_FREEBSD32
2365 	if (w->w_req->flags & SCTL_MASK32) {
2366 		struct ifa_msghdrl32 *ifam32;
2367 
2368 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2369 		ifam32->ifam_addrs = info->rti_addrs;
2370 		ifam32->ifam_flags = ifa->ifa_flags;
2371 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2372 		ifam32->_ifam_spare1 = 0;
2373 		ifam32->ifam_len = sizeof(*ifam32);
2374 		ifam32->ifam_data_off =
2375 		    offsetof(struct ifa_msghdrl32, ifam_data);
2376 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2377 		ifd = &ifam32->ifam_data;
2378 	} else
2379 #endif
2380 	{
2381 		ifam->ifam_addrs = info->rti_addrs;
2382 		ifam->ifam_flags = ifa->ifa_flags;
2383 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2384 		ifam->_ifam_spare1 = 0;
2385 		ifam->ifam_len = sizeof(*ifam);
2386 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2387 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2388 		ifd = &ifam->ifam_data;
2389 	}
2390 
2391 	bzero(ifd, sizeof(*ifd));
2392 	ifd->ifi_datalen = sizeof(struct if_data);
2393 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2394 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2395 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2396 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2397 
2398 	/* Fixup if_data carp(4) vhid. */
2399 	if (carp_get_vhid_p != NULL)
2400 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2401 
2402 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2403 }
2404 
2405 static int
2406 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2407     struct walkarg *w, int len)
2408 {
2409 	struct ifa_msghdr *ifam;
2410 
2411 	ifam = (struct ifa_msghdr *)w->w_tmem;
2412 	ifam->ifam_addrs = info->rti_addrs;
2413 	ifam->ifam_flags = ifa->ifa_flags;
2414 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2415 	ifam->_ifam_spare1 = 0;
2416 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2417 
2418 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2419 }
2420 
2421 static int
2422 sysctl_iflist(int af, struct walkarg *w)
2423 {
2424 	struct ifnet *ifp;
2425 	struct ifaddr *ifa;
2426 	struct if_data ifd;
2427 	struct rt_addrinfo info;
2428 	int len, error = 0;
2429 	struct sockaddr_storage ss;
2430 
2431 	bzero((caddr_t)&info, sizeof(info));
2432 	bzero(&ifd, sizeof(ifd));
2433 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2434 		if (w->w_arg && w->w_arg != ifp->if_index)
2435 			continue;
2436 		if_data_copy(ifp, &ifd);
2437 		ifa = ifp->if_addr;
2438 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2439 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2440 		if (error != 0)
2441 			goto done;
2442 		info.rti_info[RTAX_IFP] = NULL;
2443 		if (w->w_req && w->w_tmem) {
2444 			if (w->w_op == NET_RT_IFLISTL)
2445 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2446 				    len);
2447 			else
2448 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2449 				    len);
2450 			if (error)
2451 				goto done;
2452 		}
2453 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2454 			if (af && af != ifa->ifa_addr->sa_family)
2455 				continue;
2456 			if (prison_if(w->w_req->td->td_ucred,
2457 			    ifa->ifa_addr) != 0)
2458 				continue;
2459 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2460 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2461 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2462 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2463 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2464 			if (error != 0)
2465 				goto done;
2466 			if (w->w_req && w->w_tmem) {
2467 				if (w->w_op == NET_RT_IFLISTL)
2468 					error = sysctl_iflist_ifaml(ifa, &info,
2469 					    w, len);
2470 				else
2471 					error = sysctl_iflist_ifam(ifa, &info,
2472 					    w, len);
2473 				if (error)
2474 					goto done;
2475 			}
2476 		}
2477 		info.rti_info[RTAX_IFA] = NULL;
2478 		info.rti_info[RTAX_NETMASK] = NULL;
2479 		info.rti_info[RTAX_BRD] = NULL;
2480 	}
2481 done:
2482 	return (error);
2483 }
2484 
2485 static int
2486 sysctl_ifmalist(int af, struct walkarg *w)
2487 {
2488 	struct rt_addrinfo info;
2489 	struct ifaddr *ifa;
2490 	struct ifmultiaddr *ifma;
2491 	struct ifnet *ifp;
2492 	int error, len;
2493 
2494 	NET_EPOCH_ASSERT();
2495 
2496 	error = 0;
2497 	bzero((caddr_t)&info, sizeof(info));
2498 
2499 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2500 		if (w->w_arg && w->w_arg != ifp->if_index)
2501 			continue;
2502 		ifa = ifp->if_addr;
2503 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2504 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2505 			if (af && af != ifma->ifma_addr->sa_family)
2506 				continue;
2507 			if (prison_if(w->w_req->td->td_ucred,
2508 			    ifma->ifma_addr) != 0)
2509 				continue;
2510 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2511 			info.rti_info[RTAX_GATEWAY] =
2512 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2513 			    ifma->ifma_lladdr : NULL;
2514 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2515 			if (error != 0)
2516 				break;
2517 			if (w->w_req && w->w_tmem) {
2518 				struct ifma_msghdr *ifmam;
2519 
2520 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2521 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2522 				ifmam->ifmam_flags = 0;
2523 				ifmam->ifmam_addrs = info.rti_addrs;
2524 				ifmam->_ifmam_spare1 = 0;
2525 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2526 				if (error != 0)
2527 					break;
2528 			}
2529 		}
2530 		if (error != 0)
2531 			break;
2532 	}
2533 	return (error);
2534 }
2535 
2536 static void
2537 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2538 {
2539 	union sockaddr_union sa_dst, sa_mask;
2540 
2541 	w->family = family;
2542 	w->dst = (struct sockaddr *)&sa_dst;
2543 	w->mask = (struct sockaddr *)&sa_mask;
2544 
2545 	init_sockaddrs_family(family, w->dst, w->mask);
2546 
2547 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2548 }
2549 
2550 static int
2551 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2552 {
2553 	struct epoch_tracker et;
2554 	int	*name = (int *)arg1;
2555 	u_int	namelen = arg2;
2556 	struct rib_head *rnh = NULL; /* silence compiler. */
2557 	int	i, lim, error = EINVAL;
2558 	int	fib = 0;
2559 	u_char	af;
2560 	struct	walkarg w;
2561 
2562 	if (namelen < 3)
2563 		return (EINVAL);
2564 
2565 	name++;
2566 	namelen--;
2567 	if (req->newptr)
2568 		return (EPERM);
2569 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2570 		if (namelen == 3)
2571 			fib = req->td->td_proc->p_fibnum;
2572 		else if (namelen == 4)
2573 			fib = (name[3] == RT_ALL_FIBS) ?
2574 			    req->td->td_proc->p_fibnum : name[3];
2575 		else
2576 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2577 		if (fib < 0 || fib >= rt_numfibs)
2578 			return (EINVAL);
2579 	} else if (namelen != 3)
2580 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2581 	af = name[0];
2582 	if (af > AF_MAX)
2583 		return (EINVAL);
2584 	bzero(&w, sizeof(w));
2585 	w.w_op = name[1];
2586 	w.w_arg = name[2];
2587 	w.w_req = req;
2588 
2589 	error = sysctl_wire_old_buffer(req, 0);
2590 	if (error)
2591 		return (error);
2592 
2593 	/*
2594 	 * Allocate reply buffer in advance.
2595 	 * All rtsock messages has maximum length of u_short.
2596 	 */
2597 	w.w_tmemsize = 65536;
2598 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2599 
2600 	NET_EPOCH_ENTER(et);
2601 	switch (w.w_op) {
2602 	case NET_RT_DUMP:
2603 	case NET_RT_FLAGS:
2604 		if (af == 0) {			/* dump all tables */
2605 			i = 1;
2606 			lim = AF_MAX;
2607 		} else				/* dump only one table */
2608 			i = lim = af;
2609 
2610 		/*
2611 		 * take care of llinfo entries, the caller must
2612 		 * specify an AF
2613 		 */
2614 		if (w.w_op == NET_RT_FLAGS &&
2615 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2616 			if (af != 0)
2617 				error = lltable_sysctl_dumparp(af, w.w_req);
2618 			else
2619 				error = EINVAL;
2620 			break;
2621 		}
2622 		/*
2623 		 * take care of routing entries
2624 		 */
2625 		for (error = 0; error == 0 && i <= lim; i++) {
2626 			rnh = rt_tables_get_rnh(fib, i);
2627 			if (rnh != NULL) {
2628 				rtable_sysctl_dump(fib, i, &w);
2629 			} else if (af != 0)
2630 				error = EAFNOSUPPORT;
2631 		}
2632 		break;
2633 	case NET_RT_NHOP:
2634 	case NET_RT_NHGRP:
2635 		/* Allow dumping one specific af/fib at a time */
2636 		if (namelen < 4) {
2637 			error = EINVAL;
2638 			break;
2639 		}
2640 		fib = name[3];
2641 		if (fib < 0 || fib > rt_numfibs) {
2642 			error = EINVAL;
2643 			break;
2644 		}
2645 		rnh = rt_tables_get_rnh(fib, af);
2646 		if (rnh == NULL) {
2647 			error = EAFNOSUPPORT;
2648 			break;
2649 		}
2650 		if (w.w_op == NET_RT_NHOP)
2651 			error = nhops_dump_sysctl(rnh, w.w_req);
2652 		else
2653 #ifdef ROUTE_MPATH
2654 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2655 #else
2656 			error = ENOTSUP;
2657 #endif
2658 		break;
2659 	case NET_RT_IFLIST:
2660 	case NET_RT_IFLISTL:
2661 		error = sysctl_iflist(af, &w);
2662 		break;
2663 
2664 	case NET_RT_IFMALIST:
2665 		error = sysctl_ifmalist(af, &w);
2666 		break;
2667 	}
2668 	NET_EPOCH_EXIT(et);
2669 
2670 	free(w.w_tmem, M_TEMP);
2671 	return (error);
2672 }
2673 
2674 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2675     sysctl_rtsock, "Return route tables and interface/address lists");
2676 
2677 /*
2678  * Definitions of protocols supported in the ROUTE domain.
2679  */
2680 
2681 static struct domain routedomain;		/* or at least forward */
2682 
2683 static struct protosw routesw = {
2684 	.pr_type =		SOCK_RAW,
2685 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2686 	.pr_abort =		rts_close,
2687 	.pr_attach =		rts_attach,
2688 	.pr_detach =		rts_detach,
2689 	.pr_send =		rts_send,
2690 	.pr_shutdown =		rts_shutdown,
2691 	.pr_disconnect =	rts_disconnect,
2692 	.pr_close =		rts_close,
2693 };
2694 
2695 static struct domain routedomain = {
2696 	.dom_family =		PF_ROUTE,
2697 	.dom_name =		"route",
2698 	.dom_nprotosw =		1,
2699 	.dom_protosw =		{ &routesw },
2700 };
2701 
2702 DOMAIN_SET(route);
2703