xref: /freebsd/sys/net/rtsock.c (revision 780f28929782a104eefbc81f031836bf1febb6de)
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, ...)					\
140 	RT_LOG_##_l(_l, "PID %d: " _fmt, curproc ? curproc->p_pid : 0,	\
141 	    ## __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_RWTUN | CTLFLAG_NOFETCH | 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, enum shutdown_how how)
456 {
457 	/*
458 	 * Note: route socket marks itself as connected through its lifetime.
459 	 */
460 	switch (how) {
461 	case SHUT_RD:
462 		sorflush(so);
463 		break;
464 	case SHUT_RDWR:
465 		sorflush(so);
466 		/* FALLTHROUGH */
467 	case SHUT_WR:
468 		socantsendmore(so);
469 	}
470 
471 	return (0);
472 }
473 
474 #ifndef _SOCKADDR_UNION_DEFINED
475 #define	_SOCKADDR_UNION_DEFINED
476 /*
477  * The union of all possible address formats we handle.
478  */
479 union sockaddr_union {
480 	struct sockaddr		sa;
481 	struct sockaddr_in	sin;
482 	struct sockaddr_in6	sin6;
483 };
484 #endif /* _SOCKADDR_UNION_DEFINED */
485 
486 static int
487 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
488     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
489 {
490 #if defined(INET) || defined(INET6)
491 	struct epoch_tracker et;
492 #endif
493 
494 	/* First, see if the returned address is part of the jail. */
495 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
496 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
497 		return (0);
498 	}
499 
500 	switch (info->rti_info[RTAX_DST]->sa_family) {
501 #ifdef INET
502 	case AF_INET:
503 	{
504 		struct in_addr ia;
505 		struct ifaddr *ifa;
506 		int found;
507 
508 		found = 0;
509 		/*
510 		 * Try to find an address on the given outgoing interface
511 		 * that belongs to the jail.
512 		 */
513 		NET_EPOCH_ENTER(et);
514 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
515 			struct sockaddr *sa;
516 			sa = ifa->ifa_addr;
517 			if (sa->sa_family != AF_INET)
518 				continue;
519 			ia = ((struct sockaddr_in *)sa)->sin_addr;
520 			if (prison_check_ip4(cred, &ia) == 0) {
521 				found = 1;
522 				break;
523 			}
524 		}
525 		NET_EPOCH_EXIT(et);
526 		if (!found) {
527 			/*
528 			 * As a last resort return the 'default' jail address.
529 			 */
530 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
531 			    sin_addr;
532 			if (prison_get_ip4(cred, &ia) != 0)
533 				return (ESRCH);
534 		}
535 		bzero(&saun->sin, sizeof(struct sockaddr_in));
536 		saun->sin.sin_len = sizeof(struct sockaddr_in);
537 		saun->sin.sin_family = AF_INET;
538 		saun->sin.sin_addr.s_addr = ia.s_addr;
539 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
540 		break;
541 	}
542 #endif
543 #ifdef INET6
544 	case AF_INET6:
545 	{
546 		struct in6_addr ia6;
547 		struct ifaddr *ifa;
548 		int found;
549 
550 		found = 0;
551 		/*
552 		 * Try to find an address on the given outgoing interface
553 		 * that belongs to the jail.
554 		 */
555 		NET_EPOCH_ENTER(et);
556 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
557 			struct sockaddr *sa;
558 			sa = ifa->ifa_addr;
559 			if (sa->sa_family != AF_INET6)
560 				continue;
561 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
562 			    &ia6, sizeof(struct in6_addr));
563 			if (prison_check_ip6(cred, &ia6) == 0) {
564 				found = 1;
565 				break;
566 			}
567 		}
568 		NET_EPOCH_EXIT(et);
569 		if (!found) {
570 			/*
571 			 * As a last resort return the 'default' jail address.
572 			 */
573 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
574 			    sin6_addr;
575 			if (prison_get_ip6(cred, &ia6) != 0)
576 				return (ESRCH);
577 		}
578 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
579 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
580 		saun->sin6.sin6_family = AF_INET6;
581 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
582 		if (sa6_recoverscope(&saun->sin6) != 0)
583 			return (ESRCH);
584 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
585 		break;
586 	}
587 #endif
588 	default:
589 		return (ESRCH);
590 	}
591 	return (0);
592 }
593 
594 static int
595 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
596 {
597 	struct ifaddr *ifa;
598 	sa_family_t saf;
599 
600 	if (V_loif == NULL) {
601 		RTS_PID_LOG(LOG_INFO, "Unable to add blackhole/reject nhop without loopback");
602 		return (ENOTSUP);
603 	}
604 	info->rti_ifp = V_loif;
605 
606 	saf = info->rti_info[RTAX_DST]->sa_family;
607 
608 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
609 		if (ifa->ifa_addr->sa_family == saf) {
610 			info->rti_ifa = ifa;
611 			break;
612 		}
613 	}
614 	if (info->rti_ifa == NULL) {
615 		RTS_PID_LOG(LOG_INFO, "Unable to find ifa for blackhole/reject nhop");
616 		return (ENOTSUP);
617 	}
618 
619 	bzero(saun, sizeof(union sockaddr_union));
620 	switch (saf) {
621 #ifdef INET
622 	case AF_INET:
623 		saun->sin.sin_family = AF_INET;
624 		saun->sin.sin_len = sizeof(struct sockaddr_in);
625 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
626 		break;
627 #endif
628 #ifdef INET6
629 	case AF_INET6:
630 		saun->sin6.sin6_family = AF_INET6;
631 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
632 		saun->sin6.sin6_addr = in6addr_loopback;
633 		break;
634 #endif
635 	default:
636 		RTS_PID_LOG(LOG_INFO, "unsupported family: %d", saf);
637 		return (ENOTSUP);
638 	}
639 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
640 	info->rti_flags |= RTF_GATEWAY;
641 
642 	return (0);
643 }
644 
645 /*
646  * Fills in @info based on userland-provided @rtm message.
647  *
648  * Returns 0 on success.
649  */
650 static int
651 fill_addrinfo(struct rt_msghdr *rtm, int len, struct linear_buffer *lb, u_int fibnum,
652     struct rt_addrinfo *info)
653 {
654 	int error;
655 
656 	rtm->rtm_pid = curproc->p_pid;
657 	info->rti_addrs = rtm->rtm_addrs;
658 
659 	info->rti_mflags = rtm->rtm_inits;
660 	info->rti_rmx = &rtm->rtm_rmx;
661 
662 	/*
663 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
664 	 * link-local address because rtrequest requires addresses with
665 	 * embedded scope id.
666 	 */
667 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
668 		return (EINVAL);
669 
670 	info->rti_flags = rtm->rtm_flags;
671 	error = cleanup_xaddrs(info, lb);
672 	if (error != 0)
673 		return (error);
674 	/*
675 	 * Verify that the caller has the appropriate privilege; RTM_GET
676 	 * is the only operation the non-superuser is allowed.
677 	 */
678 	if (rtm->rtm_type != RTM_GET) {
679 		error = priv_check(curthread, PRIV_NET_ROUTE);
680 		if (error != 0)
681 			return (error);
682 	}
683 
684 	/*
685 	 * The given gateway address may be an interface address.
686 	 * For example, issuing a "route change" command on a route
687 	 * entry that was created from a tunnel, and the gateway
688 	 * address given is the local end point. In this case the
689 	 * RTF_GATEWAY flag must be cleared or the destination will
690 	 * not be reachable even though there is no error message.
691 	 */
692 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
693 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
694 		struct nhop_object *nh;
695 
696 		/*
697 		 * A host route through the loopback interface is
698 		 * installed for each interface address. In pre 8.0
699 		 * releases the interface address of a PPP link type
700 		 * is not reachable locally. This behavior is fixed as
701 		 * part of the new L2/L3 redesign and rewrite work. The
702 		 * signature of this interface address route is the
703 		 * AF_LINK sa_family type of the gateway, and the
704 		 * rt_ifp has the IFF_LOOPBACK flag set.
705 		 */
706 		nh = rib_lookup(fibnum, info->rti_info[RTAX_GATEWAY], NHR_NONE, 0);
707 		if (nh != NULL && nh->gw_sa.sa_family == AF_LINK &&
708 		    nh->nh_ifp->if_flags & IFF_LOOPBACK) {
709 				info->rti_flags &= ~RTF_GATEWAY;
710 				info->rti_flags |= RTF_GWFLAG_COMPAT;
711 		}
712 	}
713 
714 	return (0);
715 }
716 
717 static struct nhop_object *
718 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
719 {
720 	if (!NH_IS_NHGRP(nh))
721 		return (nh);
722 #ifdef ROUTE_MPATH
723 	const struct weightened_nhop *wn;
724 	uint32_t num_nhops;
725 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
726 	if (gw == NULL)
727 		return (wn[0].nh);
728 	for (int i = 0; i < num_nhops; i++) {
729 		if (match_nhop_gw(wn[i].nh, gw))
730 			return (wn[i].nh);
731 	}
732 #endif
733 	return (NULL);
734 }
735 
736 /*
737  * Handles RTM_GET message from routing socket, returning matching rt.
738  *
739  * Returns:
740  * 0 on success, with locked and referenced matching rt in @rt_nrt
741  * errno of failure
742  */
743 static int
744 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
745     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
746 {
747 	RIB_RLOCK_TRACKER;
748 	struct rib_head *rnh;
749 	struct nhop_object *nh;
750 	sa_family_t saf;
751 
752 	saf = info->rti_info[RTAX_DST]->sa_family;
753 
754 	rnh = rt_tables_get_rnh(fibnum, saf);
755 	if (rnh == NULL)
756 		return (EAFNOSUPPORT);
757 
758 	RIB_RLOCK(rnh);
759 
760 	/*
761 	 * By (implicit) convention host route (one without netmask)
762 	 * means longest-prefix-match request and the route with netmask
763 	 * means exact-match lookup.
764 	 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128
765 	 * prefixes, use original data to check for the netmask presence.
766 	 */
767 	if ((rtm->rtm_addrs & RTA_NETMASK) == 0) {
768 		/*
769 		 * Provide longest prefix match for
770 		 * address lookup (no mask).
771 		 * 'route -n get addr'
772 		 */
773 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
774 		    info->rti_info[RTAX_DST], &rnh->head);
775 	} else
776 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
777 		    info->rti_info[RTAX_DST],
778 		    info->rti_info[RTAX_NETMASK], &rnh->head);
779 
780 	if (rc->rc_rt == NULL) {
781 		RIB_RUNLOCK(rnh);
782 		return (ESRCH);
783 	}
784 
785 	nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
786 	if (nh == NULL) {
787 		RIB_RUNLOCK(rnh);
788 		return (ESRCH);
789 	}
790 	/*
791 	 * If performing proxied L2 entry insertion, and
792 	 * the actual PPP host entry is found, perform
793 	 * another search to retrieve the prefix route of
794 	 * the local end point of the PPP link.
795 	 * TODO: move this logic to userland.
796 	 */
797 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
798 		struct sockaddr_storage laddr;
799 
800 		if (nh->nh_ifp != NULL &&
801 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
802 			struct ifaddr *ifa;
803 
804 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
805 					RT_ALL_FIBS);
806 			if (ifa != NULL)
807 				rt_maskedcopy(ifa->ifa_addr,
808 					      (struct sockaddr *)&laddr,
809 					      ifa->ifa_netmask);
810 		} else
811 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
812 				      (struct sockaddr *)&laddr,
813 				      nh->nh_ifa->ifa_netmask);
814 		/*
815 		 * refactor rt and no lock operation necessary
816 		 */
817 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(
818 		    (struct sockaddr *)&laddr, &rnh->head);
819 		if (rc->rc_rt == NULL) {
820 			RIB_RUNLOCK(rnh);
821 			return (ESRCH);
822 		}
823 		nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
824 		if (nh == NULL) {
825 			RIB_RUNLOCK(rnh);
826 			return (ESRCH);
827 		}
828 	}
829 	rc->rc_nh_new = nh;
830 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
831 	RIB_RUNLOCK(rnh);
832 
833 	return (0);
834 }
835 
836 static void
837 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
838 {
839 #ifdef INET
840 	if (family == AF_INET) {
841 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
842 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
843 
844 		bzero(dst4, sizeof(struct sockaddr_in));
845 		bzero(mask4, sizeof(struct sockaddr_in));
846 
847 		dst4->sin_family = AF_INET;
848 		dst4->sin_len = sizeof(struct sockaddr_in);
849 		mask4->sin_family = AF_INET;
850 		mask4->sin_len = sizeof(struct sockaddr_in);
851 	}
852 #endif
853 #ifdef INET6
854 	if (family == AF_INET6) {
855 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
856 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
857 
858 		bzero(dst6, sizeof(struct sockaddr_in6));
859 		bzero(mask6, sizeof(struct sockaddr_in6));
860 
861 		dst6->sin6_family = AF_INET6;
862 		dst6->sin6_len = sizeof(struct sockaddr_in6);
863 		mask6->sin6_family = AF_INET6;
864 		mask6->sin6_len = sizeof(struct sockaddr_in6);
865 	}
866 #endif
867 }
868 
869 static void
870 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
871     struct sockaddr *mask)
872 {
873 #ifdef INET
874 	if (dst->sa_family == AF_INET) {
875 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
876 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
877 		uint32_t scopeid = 0;
878 		rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
879 		    &scopeid);
880 		return;
881 	}
882 #endif
883 #ifdef INET6
884 	if (dst->sa_family == AF_INET6) {
885 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
886 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
887 		uint32_t scopeid = 0;
888 		rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
889 		    &mask6->sin6_addr, &scopeid);
890 		dst6->sin6_scope_id = scopeid;
891 		return;
892 	}
893 #endif
894 }
895 
896 static int
897 update_rtm_from_info(struct rt_addrinfo *info, struct rt_msghdr **prtm,
898     int alloc_len)
899 {
900 	struct rt_msghdr *rtm, *orig_rtm = NULL;
901 	struct walkarg w;
902 	int len;
903 
904 	rtm = *prtm;
905 	/* Check if we need to realloc storage */
906 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
907 	if (len > alloc_len) {
908 		struct rt_msghdr *tmp_rtm;
909 
910 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
911 		if (tmp_rtm == NULL)
912 			return (ENOBUFS);
913 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
914 		orig_rtm = rtm;
915 		rtm = tmp_rtm;
916 		alloc_len = len;
917 
918 		/*
919 		 * Delay freeing original rtm as info contains
920 		 * data referencing it.
921 		 */
922 	}
923 
924 	w = (struct walkarg ){
925 		.w_tmem = (caddr_t)rtm,
926 		.w_tmemsize = alloc_len,
927 	};
928 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
929 	rtm->rtm_addrs = info->rti_addrs;
930 
931 	if (orig_rtm != NULL)
932 		free(orig_rtm, M_TEMP);
933 	*prtm = rtm;
934 	return (0);
935 }
936 
937 
938 /*
939  * Update sockaddrs, flags, etc in @prtm based on @rc data.
940  * rtm can be reallocated.
941  *
942  * Returns 0 on success, along with pointer to (potentially reallocated)
943  *  rtm.
944  *
945  */
946 static int
947 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
948     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
949 {
950 	union sockaddr_union saun;
951 	struct rt_msghdr *rtm;
952 	struct ifnet *ifp;
953 	int error;
954 
955 	rtm = *prtm;
956 	union sockaddr_union sa_dst, sa_mask;
957 	int family = info->rti_info[RTAX_DST]->sa_family;
958 	init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
959 	export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
960 
961 	info->rti_info[RTAX_DST] = &sa_dst.sa;
962 	info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
963 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
964 	info->rti_info[RTAX_GENMASK] = 0;
965 	ifp = nh->nh_ifp;
966 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
967 		if (ifp) {
968 			info->rti_info[RTAX_IFP] =
969 			    ifp->if_addr->ifa_addr;
970 			error = rtm_get_jailed(info, ifp, nh,
971 			    &saun, curthread->td_ucred);
972 			if (error != 0)
973 				return (error);
974 			if (ifp->if_flags & IFF_POINTOPOINT)
975 				info->rti_info[RTAX_BRD] =
976 				    nh->nh_ifa->ifa_dstaddr;
977 			rtm->rtm_index = ifp->if_index;
978 		} else {
979 			info->rti_info[RTAX_IFP] = NULL;
980 			info->rti_info[RTAX_IFA] = NULL;
981 		}
982 	} else if (ifp != NULL)
983 		rtm->rtm_index = ifp->if_index;
984 
985 	if ((error = update_rtm_from_info(info, prtm, alloc_len)) != 0)
986 		return (error);
987 
988 	rtm = *prtm;
989 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
990 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
991 		rtm->rtm_flags = RTF_GATEWAY |
992 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
993 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
994 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
995 
996 	return (0);
997 }
998 
999 #ifdef ROUTE_MPATH
1000 static void
1001 save_del_notification(const struct rib_cmd_info *rc, void *_cbdata)
1002 {
1003 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1004 
1005 	if (rc->rc_cmd == RTM_DELETE)
1006 		*rc_new = *rc;
1007 }
1008 
1009 static void
1010 save_add_notification(const struct rib_cmd_info *rc, void *_cbdata)
1011 {
1012 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1013 
1014 	if (rc->rc_cmd == RTM_ADD)
1015 		*rc_new = *rc;
1016 }
1017 #endif
1018 
1019 #if defined(INET6) || defined(INET)
1020 static struct sockaddr *
1021 alloc_sockaddr_aligned(struct linear_buffer *lb, int len)
1022 {
1023 	len = roundup2(len, sizeof(uint64_t));
1024 	if (lb->offset + len > lb->size)
1025 		return (NULL);
1026 	struct sockaddr *sa = (struct sockaddr *)(lb->base + lb->offset);
1027 	lb->offset += len;
1028 	return (sa);
1029 }
1030 #endif
1031 
1032 static int
1033 rts_send(struct socket *so, int flags, struct mbuf *m,
1034     struct sockaddr *nam, struct mbuf *control, struct thread *td)
1035 {
1036 	struct rt_msghdr *rtm = NULL;
1037 	struct rt_addrinfo info;
1038 	struct epoch_tracker et;
1039 #ifdef INET6
1040 	struct sockaddr_storage ss;
1041 	struct sockaddr_in6 *sin6;
1042 	int i, rti_need_deembed = 0;
1043 #endif
1044 	int alloc_len = 0, len, error = 0, fibnum;
1045 	sa_family_t saf = AF_UNSPEC;
1046 	struct rib_cmd_info rc;
1047 	struct nhop_object *nh;
1048 
1049 	if ((flags & PRUS_OOB) || control != NULL) {
1050 		m_freem(m);
1051 		if (control != NULL)
1052 			m_freem(control);
1053 		return (EOPNOTSUPP);
1054 	}
1055 
1056 	fibnum = so->so_fibnum;
1057 #define senderr(e) { error = e; goto flush;}
1058 	if (m == NULL || ((m->m_len < sizeof(long)) &&
1059 		       (m = m_pullup(m, sizeof(long))) == NULL))
1060 		return (ENOBUFS);
1061 	if ((m->m_flags & M_PKTHDR) == 0)
1062 		panic("route_output");
1063 	NET_EPOCH_ENTER(et);
1064 	len = m->m_pkthdr.len;
1065 	if (len < sizeof(*rtm) ||
1066 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1067 		senderr(EINVAL);
1068 
1069 	/*
1070 	 * Most of current messages are in range 200-240 bytes,
1071 	 * minimize possible re-allocation on reply using larger size
1072 	 * buffer aligned on 1k boundaty.
1073 	 */
1074 	alloc_len = roundup2(len, 1024);
1075 	int total_len = alloc_len + SCRATCH_BUFFER_SIZE;
1076 	if ((rtm = malloc(total_len, M_TEMP, M_NOWAIT)) == NULL)
1077 		senderr(ENOBUFS);
1078 
1079 	m_copydata(m, 0, len, (caddr_t)rtm);
1080 	bzero(&info, sizeof(info));
1081 	nh = NULL;
1082 	struct linear_buffer lb = {
1083 		.base = (char *)rtm + alloc_len,
1084 		.size = SCRATCH_BUFFER_SIZE,
1085 	};
1086 
1087 	if (rtm->rtm_version != RTM_VERSION) {
1088 		/* Do not touch message since format is unknown */
1089 		free(rtm, M_TEMP);
1090 		rtm = NULL;
1091 		senderr(EPROTONOSUPPORT);
1092 	}
1093 
1094 	/*
1095 	 * Starting from here, it is possible
1096 	 * to alter original message and insert
1097 	 * caller PID and error value.
1098 	 */
1099 
1100 	if ((error = fill_addrinfo(rtm, len, &lb, fibnum, &info)) != 0) {
1101 		senderr(error);
1102 	}
1103 	/* fill_addringo() embeds scope into IPv6 addresses */
1104 #ifdef INET6
1105 	rti_need_deembed = 1;
1106 #endif
1107 
1108 	saf = info.rti_info[RTAX_DST]->sa_family;
1109 
1110 	/* support for new ARP code */
1111 	if (rtm->rtm_flags & RTF_LLDATA) {
1112 		error = lla_rt_output(rtm, &info);
1113 		goto flush;
1114 	}
1115 
1116 	union sockaddr_union gw_saun;
1117 	int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1118 	if (blackhole_flags != 0) {
1119 		if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1120 			error = fill_blackholeinfo(&info, &gw_saun);
1121 		else {
1122 			RTS_PID_LOG(LOG_DEBUG, "both BLACKHOLE and REJECT flags specifiied");
1123 			error = EINVAL;
1124 		}
1125 		if (error != 0)
1126 			senderr(error);
1127 	}
1128 
1129 	switch (rtm->rtm_type) {
1130 	case RTM_ADD:
1131 	case RTM_CHANGE:
1132 		if (rtm->rtm_type == RTM_ADD) {
1133 			if (info.rti_info[RTAX_GATEWAY] == NULL) {
1134 				RTS_PID_LOG(LOG_DEBUG, "RTM_ADD w/o gateway");
1135 				senderr(EINVAL);
1136 			}
1137 		}
1138 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1139 		if (error == 0) {
1140 			rtsock_notify_event(fibnum, &rc);
1141 #ifdef ROUTE_MPATH
1142 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
1143 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1144 				struct rib_cmd_info rc_simple = {};
1145 				rib_decompose_notification(&rc,
1146 				    save_add_notification, (void *)&rc_simple);
1147 				rc = rc_simple;
1148 			}
1149 #endif
1150 			/* nh MAY be empty if RTM_CHANGE request is no-op */
1151 			nh = rc.rc_nh_new;
1152 			if (nh != NULL) {
1153 				rtm->rtm_index = nh->nh_ifp->if_index;
1154 				rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1155 			}
1156 		}
1157 		break;
1158 
1159 	case RTM_DELETE:
1160 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1161 		if (error == 0) {
1162 			rtsock_notify_event(fibnum, &rc);
1163 #ifdef ROUTE_MPATH
1164 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1165 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1166 				struct rib_cmd_info rc_simple = {};
1167 				rib_decompose_notification(&rc,
1168 				    save_del_notification, (void *)&rc_simple);
1169 				rc = rc_simple;
1170 			}
1171 #endif
1172 			nh = rc.rc_nh_old;
1173 		}
1174 		break;
1175 
1176 	case RTM_GET:
1177 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1178 		if (error != 0)
1179 			senderr(error);
1180 		nh = rc.rc_nh_new;
1181 
1182 		if (!rt_is_exportable(rc.rc_rt, curthread->td_ucred))
1183 			senderr(ESRCH);
1184 		break;
1185 
1186 	default:
1187 		senderr(EOPNOTSUPP);
1188 	}
1189 
1190 	if (error == 0 && nh != NULL) {
1191 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1192 		/*
1193 		 * Note that some sockaddr pointers may have changed to
1194 		 * point to memory outsize @rtm. Some may be pointing
1195 		 * to the on-stack variables.
1196 		 * Given that, any pointer in @info CANNOT BE USED.
1197 		 */
1198 
1199 		/*
1200 		 * scopeid deembedding has been performed while
1201 		 * writing updated rtm in rtsock_msg_buffer().
1202 		 * With that in mind, skip deembedding procedure below.
1203 		 */
1204 #ifdef INET6
1205 		rti_need_deembed = 0;
1206 #endif
1207 	}
1208 
1209 flush:
1210 	NET_EPOCH_EXIT(et);
1211 
1212 #ifdef INET6
1213 	if (rtm != NULL) {
1214 		if (rti_need_deembed) {
1215 			/* sin6_scope_id is recovered before sending rtm. */
1216 			sin6 = (struct sockaddr_in6 *)&ss;
1217 			for (i = 0; i < RTAX_MAX; i++) {
1218 				if (info.rti_info[i] == NULL)
1219 					continue;
1220 				if (info.rti_info[i]->sa_family != AF_INET6)
1221 					continue;
1222 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1223 				if (sa6_recoverscope(sin6) == 0)
1224 					bcopy(sin6, info.rti_info[i],
1225 						    sizeof(*sin6));
1226 			}
1227 			if (update_rtm_from_info(&info, &rtm, alloc_len) != 0) {
1228 				if (error != 0)
1229 					error = ENOBUFS;
1230 			}
1231 		}
1232 	}
1233 #endif
1234 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1235 
1236 	return (error);
1237 }
1238 
1239 /*
1240  * Sends the prepared reply message in @rtm to all rtsock clients.
1241  * Frees @m and @rtm.
1242  *
1243  */
1244 static void
1245 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1246     sa_family_t saf, u_int fibnum, int rtm_errno)
1247 {
1248 	struct rcb *rcb = NULL;
1249 
1250 	/*
1251 	 * Check to see if we don't want our own messages.
1252 	 */
1253 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1254 		if (V_route_cb.any_count <= 1) {
1255 			if (rtm != NULL)
1256 				free(rtm, M_TEMP);
1257 			m_freem(m);
1258 			return;
1259 		}
1260 		/* There is another listener, so construct message */
1261 		rcb = so->so_pcb;
1262 	}
1263 
1264 	if (rtm != NULL) {
1265 		if (rtm_errno!= 0)
1266 			rtm->rtm_errno = rtm_errno;
1267 		else
1268 			rtm->rtm_flags |= RTF_DONE;
1269 
1270 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1271 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1272 			m_freem(m);
1273 			m = NULL;
1274 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1275 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1276 
1277 		free(rtm, M_TEMP);
1278 	}
1279 	if (m != NULL) {
1280 		M_SETFIB(m, fibnum);
1281 		m->m_flags |= RTS_FILTER_FIB;
1282 		if (rcb) {
1283 			/*
1284 			 * XXX insure we don't get a copy by
1285 			 * invalidating our protocol
1286 			 */
1287 			sa_family_t family = rcb->rcb_family;
1288 			rcb->rcb_family = AF_UNSPEC;
1289 			rt_dispatch(m, saf);
1290 			rcb->rcb_family = family;
1291 		} else
1292 			rt_dispatch(m, saf);
1293 	}
1294 }
1295 
1296 static void
1297 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1298     struct rt_metrics *out)
1299 {
1300 
1301 	bzero(out, sizeof(*out));
1302 	out->rmx_mtu = nh->nh_mtu;
1303 	out->rmx_weight = rt->rt_weight;
1304 	out->rmx_nhidx = nhop_get_idx(nh);
1305 	/* Kernel -> userland timebase conversion. */
1306 	out->rmx_expire = nhop_get_expire(nh) ?
1307 	    nhop_get_expire(nh) - time_uptime + time_second : 0;
1308 }
1309 
1310 /*
1311  * Extract the addresses of the passed sockaddrs.
1312  * Do a little sanity checking so as to avoid bad memory references.
1313  * This data is derived straight from userland.
1314  */
1315 static int
1316 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1317 {
1318 	struct sockaddr *sa;
1319 	int i;
1320 
1321 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1322 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1323 			continue;
1324 		sa = (struct sockaddr *)cp;
1325 		/*
1326 		 * It won't fit.
1327 		 */
1328 		if (cp + sa->sa_len > cplim) {
1329 			RTS_PID_LOG(LOG_DEBUG, "sa_len too big for sa type %d", i);
1330 			return (EINVAL);
1331 		}
1332 		/*
1333 		 * there are no more.. quit now
1334 		 * If there are more bits, they are in error.
1335 		 * I've seen this. route(1) can evidently generate these.
1336 		 * This causes kernel to core dump.
1337 		 * for compatibility, If we see this, point to a safe address.
1338 		 */
1339 		if (sa->sa_len == 0) {
1340 			rtinfo->rti_info[i] = &sa_zero;
1341 			return (0); /* should be EINVAL but for compat */
1342 		}
1343 		/* accept it */
1344 #ifdef INET6
1345 		if (sa->sa_family == AF_INET6)
1346 			sa6_embedscope((struct sockaddr_in6 *)sa,
1347 			    V_ip6_use_defzone);
1348 #endif
1349 		rtinfo->rti_info[i] = sa;
1350 		cp += SA_SIZE(sa);
1351 	}
1352 	return (0);
1353 }
1354 
1355 #ifdef INET
1356 static inline void
1357 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1358 {
1359 
1360 	const struct sockaddr_in nsin = {
1361 		.sin_family = AF_INET,
1362 		.sin_len = sizeof(struct sockaddr_in),
1363 		.sin_addr = addr,
1364 	};
1365 	*sin = nsin;
1366 }
1367 #endif
1368 
1369 #ifdef INET6
1370 static inline void
1371 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1372     uint32_t scopeid)
1373 {
1374 
1375 	const struct sockaddr_in6 nsin6 = {
1376 		.sin6_family = AF_INET6,
1377 		.sin6_len = sizeof(struct sockaddr_in6),
1378 		.sin6_addr = *addr6,
1379 		.sin6_scope_id = scopeid,
1380 	};
1381 	*sin6 = nsin6;
1382 }
1383 #endif
1384 
1385 #if defined(INET6) || defined(INET)
1386 /*
1387  * Checks if gateway is suitable for lltable operations.
1388  * Lltable code requires AF_LINK gateway with ifindex
1389  *  and mac address specified.
1390  * Returns 0 on success.
1391  */
1392 static int
1393 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1394 {
1395 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1396 
1397 	if (sdl->sdl_family != AF_LINK)
1398 		return (EINVAL);
1399 
1400 	if (sdl->sdl_index == 0) {
1401 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gateway w/o ifindex");
1402 		return (EINVAL);
1403 	}
1404 
1405 	if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len) {
1406 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gw: sdl_nlen/sdl_alen too large");
1407 		return (EINVAL);
1408 	}
1409 
1410 	return (0);
1411 }
1412 
1413 static int
1414 cleanup_xaddrs_gateway(struct rt_addrinfo *info, struct linear_buffer *lb)
1415 {
1416 	struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1417 	struct sockaddr *sa;
1418 
1419 	if (info->rti_flags & RTF_LLDATA)
1420 		return (cleanup_xaddrs_lladdr(info));
1421 
1422 	switch (gw->sa_family) {
1423 #ifdef INET
1424 	case AF_INET:
1425 		{
1426 			struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw;
1427 
1428 			/* Ensure reads do not go beyoud SA boundary */
1429 			if (SA_SIZE(gw) < offsetof(struct sockaddr_in, sin_zero)) {
1430 				RTS_PID_LOG(LOG_DEBUG, "gateway sin_len too small: %d",
1431 				    gw->sa_len);
1432 				return (EINVAL);
1433 			}
1434 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_in));
1435 			if (sa == NULL)
1436 				return (ENOBUFS);
1437 			fill_sockaddr_inet((struct sockaddr_in *)sa, gw_sin->sin_addr);
1438 			info->rti_info[RTAX_GATEWAY] = sa;
1439 		}
1440 		break;
1441 #endif
1442 #ifdef INET6
1443 	case AF_INET6:
1444 		{
1445 			struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1446 			if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1447 				RTS_PID_LOG(LOG_DEBUG, "gateway sin6_len too small: %d",
1448 				    gw->sa_len);
1449 				return (EINVAL);
1450 			}
1451 			fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1452 			break;
1453 		}
1454 #endif
1455 	case AF_LINK:
1456 		{
1457 			struct sockaddr_dl *gw_sdl;
1458 
1459 			size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data);
1460 			gw_sdl = (struct sockaddr_dl *)gw;
1461 			if (gw_sdl->sdl_len < sdl_min_len) {
1462 				RTS_PID_LOG(LOG_DEBUG, "gateway sdl_len too small: %d",
1463 				    gw_sdl->sdl_len);
1464 				return (EINVAL);
1465 			}
1466 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_dl_short));
1467 			if (sa == NULL)
1468 				return (ENOBUFS);
1469 
1470 			const struct sockaddr_dl_short sdl = {
1471 				.sdl_family = AF_LINK,
1472 				.sdl_len = sizeof(struct sockaddr_dl_short),
1473 				.sdl_index = gw_sdl->sdl_index,
1474 			};
1475 			*((struct sockaddr_dl_short *)sa) = sdl;
1476 			info->rti_info[RTAX_GATEWAY] = sa;
1477 			break;
1478 		}
1479 	}
1480 
1481 	return (0);
1482 }
1483 #endif
1484 
1485 static void
1486 remove_netmask(struct rt_addrinfo *info)
1487 {
1488 	info->rti_info[RTAX_NETMASK] = NULL;
1489 	info->rti_flags |= RTF_HOST;
1490 	info->rti_addrs &= ~RTA_NETMASK;
1491 }
1492 
1493 #ifdef INET
1494 static int
1495 cleanup_xaddrs_inet(struct rt_addrinfo *info, struct linear_buffer *lb)
1496 {
1497 	struct sockaddr_in *dst_sa, *mask_sa;
1498 	const int sa_len = sizeof(struct sockaddr_in);
1499 	struct in_addr dst, mask;
1500 
1501 	/* Check & fixup dst/netmask combination first */
1502 	dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1503 	mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1504 
1505 	/* Ensure reads do not go beyound the buffer size */
1506 	if (SA_SIZE(dst_sa) < offsetof(struct sockaddr_in, sin_zero)) {
1507 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin_len too small: %d",
1508 		    dst_sa->sin_len);
1509 		return (EINVAL);
1510 	}
1511 
1512 	if ((mask_sa != NULL) && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1513 		/*
1514 		 * Some older routing software encode mask length into the
1515 		 * sin_len, thus resulting in "truncated" sockaddr.
1516 		 */
1517 		int len = mask_sa->sin_len - offsetof(struct sockaddr_in, sin_addr);
1518 		if (len >= 0) {
1519 			mask.s_addr = 0;
1520 			if (len > sizeof(struct in_addr))
1521 				len = sizeof(struct in_addr);
1522 			memcpy(&mask, &mask_sa->sin_addr, len);
1523 		} else {
1524 			RTS_PID_LOG(LOG_DEBUG, "prefix mask sin_len too small: %d",
1525 			    mask_sa->sin_len);
1526 			return (EINVAL);
1527 		}
1528 	} else
1529 		mask.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST;
1530 
1531 	dst.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr));
1532 
1533 	/* Construct new "clean" dst/mask sockaddresses */
1534 	if ((dst_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1535 		return (ENOBUFS);
1536 	fill_sockaddr_inet(dst_sa, dst);
1537 	info->rti_info[RTAX_DST] = (struct sockaddr *)dst_sa;
1538 
1539 	if (mask.s_addr != INADDR_BROADCAST) {
1540 		if ((mask_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1541 			return (ENOBUFS);
1542 		fill_sockaddr_inet(mask_sa, mask);
1543 		info->rti_info[RTAX_NETMASK] = (struct sockaddr *)mask_sa;
1544 		info->rti_flags &= ~RTF_HOST;
1545 	} else
1546 		remove_netmask(info);
1547 
1548 	/* Check gateway */
1549 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1550 		return (cleanup_xaddrs_gateway(info, lb));
1551 
1552 	return (0);
1553 }
1554 #endif
1555 
1556 #ifdef INET6
1557 static int
1558 cleanup_xaddrs_inet6(struct rt_addrinfo *info, struct linear_buffer *lb)
1559 {
1560 	struct sockaddr *sa;
1561 	struct sockaddr_in6 *dst_sa, *mask_sa;
1562 	struct in6_addr mask, *dst;
1563 	const int sa_len = sizeof(struct sockaddr_in6);
1564 
1565 	/* Check & fixup dst/netmask combination first */
1566 	dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1567 	mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1568 
1569 	if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1570 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin6_len too small: %d",
1571 		    dst_sa->sin6_len);
1572 		return (EINVAL);
1573 	}
1574 
1575 	if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1576 		/*
1577 		 * Some older routing software encode mask length into the
1578 		 * sin6_len, thus resulting in "truncated" sockaddr.
1579 		 */
1580 		int len = mask_sa->sin6_len - offsetof(struct sockaddr_in6, sin6_addr);
1581 		if (len >= 0) {
1582 			bzero(&mask, sizeof(mask));
1583 			if (len > sizeof(struct in6_addr))
1584 				len = sizeof(struct in6_addr);
1585 			memcpy(&mask, &mask_sa->sin6_addr, len);
1586 		} else {
1587 			RTS_PID_LOG(LOG_DEBUG, "rtsock: prefix mask sin6_len too small: %d",
1588 			    mask_sa->sin6_len);
1589 			return (EINVAL);
1590 		}
1591 	} else
1592 		mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1593 
1594 	dst = &dst_sa->sin6_addr;
1595 	IN6_MASK_ADDR(dst, &mask);
1596 
1597 	if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1598 		return (ENOBUFS);
1599 	fill_sockaddr_inet6((struct sockaddr_in6 *)sa, dst, 0);
1600 	info->rti_info[RTAX_DST] = sa;
1601 
1602 	if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128)) {
1603 		if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1604 			return (ENOBUFS);
1605 		fill_sockaddr_inet6((struct sockaddr_in6 *)sa, &mask, 0);
1606 		info->rti_info[RTAX_NETMASK] = sa;
1607 		info->rti_flags &= ~RTF_HOST;
1608 	} else
1609 		remove_netmask(info);
1610 
1611 	/* Check gateway */
1612 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1613 		return (cleanup_xaddrs_gateway(info, lb));
1614 
1615 	return (0);
1616 }
1617 #endif
1618 
1619 static int
1620 cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb)
1621 {
1622 	int error = EAFNOSUPPORT;
1623 
1624 	if (info->rti_info[RTAX_DST] == NULL) {
1625 		RTS_PID_LOG(LOG_DEBUG, "prefix dst is not set");
1626 		return (EINVAL);
1627 	}
1628 
1629 	if (info->rti_flags & RTF_LLDATA) {
1630 		/*
1631 		 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1632 		 * prefix along with the actual address in RTA_DST.
1633 		 * Remove netmask to avoid unnecessary address masking.
1634 		 */
1635 		remove_netmask(info);
1636 	}
1637 
1638 	switch (info->rti_info[RTAX_DST]->sa_family) {
1639 #ifdef INET
1640 	case AF_INET:
1641 		error = cleanup_xaddrs_inet(info, lb);
1642 		break;
1643 #endif
1644 #ifdef INET6
1645 	case AF_INET6:
1646 		error = cleanup_xaddrs_inet6(info, lb);
1647 		break;
1648 #endif
1649 	}
1650 
1651 	return (error);
1652 }
1653 
1654 /*
1655  * Fill in @dmask with valid netmask leaving original @smask
1656  * intact. Mostly used with radix netmasks.
1657  */
1658 struct sockaddr *
1659 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1660     struct sockaddr_storage *dmask)
1661 {
1662 	if (dst == NULL || smask == NULL)
1663 		return (NULL);
1664 
1665 	memset(dmask, 0, dst->sa_len);
1666 	memcpy(dmask, smask, smask->sa_len);
1667 	dmask->ss_len = dst->sa_len;
1668 	dmask->ss_family = dst->sa_family;
1669 
1670 	return ((struct sockaddr *)dmask);
1671 }
1672 
1673 /*
1674  * Writes information related to @rtinfo object to newly-allocated mbuf.
1675  * Assumes MCLBYTES is enough to construct any message.
1676  * Used for OS notifications of vaious events (if/ifa announces,etc)
1677  *
1678  * Returns allocated mbuf or NULL on failure.
1679  */
1680 static struct mbuf *
1681 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1682 {
1683 	struct sockaddr_storage ss;
1684 	struct rt_msghdr *rtm;
1685 	struct mbuf *m;
1686 	int i;
1687 	struct sockaddr *sa;
1688 #ifdef INET6
1689 	struct sockaddr_in6 *sin6;
1690 #endif
1691 	int len, dlen;
1692 
1693 	switch (type) {
1694 	case RTM_DELADDR:
1695 	case RTM_NEWADDR:
1696 		len = sizeof(struct ifa_msghdr);
1697 		break;
1698 
1699 	case RTM_DELMADDR:
1700 	case RTM_NEWMADDR:
1701 		len = sizeof(struct ifma_msghdr);
1702 		break;
1703 
1704 	case RTM_IFINFO:
1705 		len = sizeof(struct if_msghdr);
1706 		break;
1707 
1708 	case RTM_IFANNOUNCE:
1709 	case RTM_IEEE80211:
1710 		len = sizeof(struct if_announcemsghdr);
1711 		break;
1712 
1713 	default:
1714 		len = sizeof(struct rt_msghdr);
1715 	}
1716 
1717 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1718 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1719 	if (len > MHLEN)
1720 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1721 	else
1722 		m = m_gethdr(M_NOWAIT, MT_DATA);
1723 	if (m == NULL)
1724 		return (m);
1725 
1726 	m->m_pkthdr.len = m->m_len = len;
1727 	rtm = mtod(m, struct rt_msghdr *);
1728 	bzero((caddr_t)rtm, len);
1729 	for (i = 0; i < RTAX_MAX; i++) {
1730 		if ((sa = rtinfo->rti_info[i]) == NULL)
1731 			continue;
1732 		rtinfo->rti_addrs |= (1 << i);
1733 
1734 		dlen = SA_SIZE(sa);
1735 		KASSERT(dlen <= sizeof(ss),
1736 		    ("%s: sockaddr size overflow", __func__));
1737 		bzero(&ss, sizeof(ss));
1738 		bcopy(sa, &ss, sa->sa_len);
1739 		sa = (struct sockaddr *)&ss;
1740 #ifdef INET6
1741 		if (sa->sa_family == AF_INET6) {
1742 			sin6 = (struct sockaddr_in6 *)sa;
1743 			(void)sa6_recoverscope(sin6);
1744 		}
1745 #endif
1746 		m_copyback(m, len, dlen, (caddr_t)sa);
1747 		len += dlen;
1748 	}
1749 	if (m->m_pkthdr.len != len) {
1750 		m_freem(m);
1751 		return (NULL);
1752 	}
1753 	rtm->rtm_msglen = len;
1754 	rtm->rtm_version = RTM_VERSION;
1755 	rtm->rtm_type = type;
1756 	return (m);
1757 }
1758 
1759 /*
1760  * Writes information related to @rtinfo object to preallocated buffer.
1761  * Stores needed size in @plen. If @w is NULL, calculates size without
1762  * writing.
1763  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1764  *
1765  * Returns 0 on success.
1766  *
1767  */
1768 static int
1769 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1770 {
1771 	struct sockaddr_storage ss;
1772 	int len, buflen = 0, dlen, i;
1773 	caddr_t cp = NULL;
1774 	struct rt_msghdr *rtm = NULL;
1775 #ifdef INET6
1776 	struct sockaddr_in6 *sin6;
1777 #endif
1778 #ifdef COMPAT_FREEBSD32
1779 	bool compat32;
1780 
1781 	compat32 = w != NULL && w->w_req != NULL &&
1782 	    (w->w_req->flags & SCTL_MASK32);
1783 #endif
1784 
1785 	switch (type) {
1786 	case RTM_DELADDR:
1787 	case RTM_NEWADDR:
1788 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1789 #ifdef COMPAT_FREEBSD32
1790 			if (compat32)
1791 				len = sizeof(struct ifa_msghdrl32);
1792 			else
1793 #endif
1794 				len = sizeof(struct ifa_msghdrl);
1795 		} else
1796 			len = sizeof(struct ifa_msghdr);
1797 		break;
1798 
1799 	case RTM_IFINFO:
1800 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1801 #ifdef COMPAT_FREEBSD32
1802 			if (compat32)
1803 				len = sizeof(struct if_msghdrl32);
1804 			else
1805 #endif
1806 				len = sizeof(struct if_msghdrl);
1807 		} else {
1808 #ifdef COMPAT_FREEBSD32
1809 			if (compat32)
1810 				len = sizeof(struct if_msghdr32);
1811 			else
1812 #endif
1813 				len = sizeof(struct if_msghdr);
1814 		}
1815 		break;
1816 
1817 	case RTM_NEWMADDR:
1818 		len = sizeof(struct ifma_msghdr);
1819 		break;
1820 
1821 	default:
1822 		len = sizeof(struct rt_msghdr);
1823 	}
1824 
1825 	if (w != NULL) {
1826 		rtm = (struct rt_msghdr *)w->w_tmem;
1827 		buflen = w->w_tmemsize - len;
1828 		cp = (caddr_t)w->w_tmem + len;
1829 	}
1830 
1831 	rtinfo->rti_addrs = 0;
1832 	for (i = 0; i < RTAX_MAX; i++) {
1833 		struct sockaddr *sa;
1834 
1835 		if ((sa = rtinfo->rti_info[i]) == NULL)
1836 			continue;
1837 		rtinfo->rti_addrs |= (1 << i);
1838 #ifdef COMPAT_FREEBSD32
1839 		if (compat32)
1840 			dlen = SA_SIZE32(sa);
1841 		else
1842 #endif
1843 			dlen = SA_SIZE(sa);
1844 		if (cp != NULL && buflen >= dlen) {
1845 			KASSERT(dlen <= sizeof(ss),
1846 			    ("%s: sockaddr size overflow", __func__));
1847 			bzero(&ss, sizeof(ss));
1848 			bcopy(sa, &ss, sa->sa_len);
1849 			sa = (struct sockaddr *)&ss;
1850 #ifdef INET6
1851 			if (sa->sa_family == AF_INET6) {
1852 				sin6 = (struct sockaddr_in6 *)sa;
1853 				(void)sa6_recoverscope(sin6);
1854 			}
1855 #endif
1856 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1857 			cp += dlen;
1858 			buflen -= dlen;
1859 		} else if (cp != NULL) {
1860 			/*
1861 			 * Buffer too small. Count needed size
1862 			 * and return with error.
1863 			 */
1864 			cp = NULL;
1865 		}
1866 
1867 		len += dlen;
1868 	}
1869 
1870 	if (cp != NULL) {
1871 		dlen = ALIGN(len) - len;
1872 		if (buflen < dlen)
1873 			cp = NULL;
1874 		else {
1875 			bzero(cp, dlen);
1876 			cp += dlen;
1877 			buflen -= dlen;
1878 		}
1879 	}
1880 	len = ALIGN(len);
1881 
1882 	if (cp != NULL) {
1883 		/* fill header iff buffer is large enough */
1884 		rtm->rtm_version = RTM_VERSION;
1885 		rtm->rtm_type = type;
1886 		rtm->rtm_msglen = len;
1887 	}
1888 
1889 	*plen = len;
1890 
1891 	if (w != NULL && cp == NULL)
1892 		return (ENOBUFS);
1893 
1894 	return (0);
1895 }
1896 
1897 /*
1898  * This routine is called to generate a message from the routing
1899  * socket indicating that a redirect has occurred, a routing lookup
1900  * has failed, or that a protocol has detected timeouts to a particular
1901  * destination.
1902  */
1903 void
1904 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1905     int fibnum)
1906 {
1907 	struct rt_msghdr *rtm;
1908 	struct mbuf *m;
1909 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1910 
1911 	if (V_route_cb.any_count == 0)
1912 		return;
1913 	m = rtsock_msg_mbuf(type, rtinfo);
1914 	if (m == NULL)
1915 		return;
1916 
1917 	if (fibnum != RT_ALL_FIBS) {
1918 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1919 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1920 		M_SETFIB(m, fibnum);
1921 		m->m_flags |= RTS_FILTER_FIB;
1922 	}
1923 
1924 	rtm = mtod(m, struct rt_msghdr *);
1925 	rtm->rtm_flags = RTF_DONE | flags;
1926 	rtm->rtm_errno = error;
1927 	rtm->rtm_addrs = rtinfo->rti_addrs;
1928 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1929 }
1930 
1931 void
1932 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1933 {
1934 
1935 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1936 }
1937 
1938 /*
1939  * This routine is called to generate a message from the routing
1940  * socket indicating that the status of a network interface has changed.
1941  */
1942 static void
1943 rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask __unused)
1944 {
1945 	struct if_msghdr *ifm;
1946 	struct mbuf *m;
1947 	struct rt_addrinfo info;
1948 
1949 	if (V_route_cb.any_count == 0)
1950 		return;
1951 	bzero((caddr_t)&info, sizeof(info));
1952 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1953 	if (m == NULL)
1954 		return;
1955 	ifm = mtod(m, struct if_msghdr *);
1956 	ifm->ifm_index = ifp->if_index;
1957 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1958 	if_data_copy(ifp, &ifm->ifm_data);
1959 	ifm->ifm_addrs = 0;
1960 	rt_dispatch(m, AF_UNSPEC);
1961 }
1962 
1963 /*
1964  * Announce interface address arrival/withdraw.
1965  * Please do not call directly, use rt_addrmsg().
1966  * Assume input data to be valid.
1967  * Returns 0 on success.
1968  */
1969 int
1970 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1971 {
1972 	struct rt_addrinfo info;
1973 	struct sockaddr *sa;
1974 	int ncmd;
1975 	struct mbuf *m;
1976 	struct ifa_msghdr *ifam;
1977 	struct ifnet *ifp = ifa->ifa_ifp;
1978 	struct sockaddr_storage ss;
1979 
1980 	if (V_route_cb.any_count == 0)
1981 		return (0);
1982 
1983 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1984 
1985 	bzero((caddr_t)&info, sizeof(info));
1986 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1987 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1988 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1989 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1990 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1991 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1992 		return (ENOBUFS);
1993 	ifam = mtod(m, struct ifa_msghdr *);
1994 	ifam->ifam_index = ifp->if_index;
1995 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1996 	ifam->ifam_flags = ifa->ifa_flags;
1997 	ifam->ifam_addrs = info.rti_addrs;
1998 
1999 	if (fibnum != RT_ALL_FIBS) {
2000 		M_SETFIB(m, fibnum);
2001 		m->m_flags |= RTS_FILTER_FIB;
2002 	}
2003 
2004 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2005 
2006 	return (0);
2007 }
2008 
2009 /*
2010  * Announce route addition/removal to rtsock based on @rt data.
2011  * Callers are advives to use rt_routemsg() instead of using this
2012  *  function directly.
2013  * Assume @rt data is consistent.
2014  *
2015  * Returns 0 on success.
2016  */
2017 int
2018 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
2019     int fibnum)
2020 {
2021 	union sockaddr_union dst, mask;
2022 	struct rt_addrinfo info;
2023 
2024 	if (V_route_cb.any_count == 0)
2025 		return (0);
2026 
2027 	int family = rt_get_family(rt);
2028 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
2029 	export_rtaddrs(rt, &dst.sa, &mask.sa);
2030 
2031 	bzero((caddr_t)&info, sizeof(info));
2032 	info.rti_info[RTAX_DST] = &dst.sa;
2033 	info.rti_info[RTAX_NETMASK] = &mask.sa;
2034 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2035 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
2036 	info.rti_ifp = nh->nh_ifp;
2037 
2038 	return (rtsock_routemsg_info(cmd, &info, fibnum));
2039 }
2040 
2041 int
2042 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
2043 {
2044 	struct rt_msghdr *rtm;
2045 	struct sockaddr *sa;
2046 	struct mbuf *m;
2047 
2048 	if (V_route_cb.any_count == 0)
2049 		return (0);
2050 
2051 	if (info->rti_flags & RTF_HOST)
2052 		info->rti_info[RTAX_NETMASK] = NULL;
2053 
2054 	m = rtsock_msg_mbuf(cmd, info);
2055 	if (m == NULL)
2056 		return (ENOBUFS);
2057 
2058 	if (fibnum != RT_ALL_FIBS) {
2059 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
2060 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
2061 		M_SETFIB(m, fibnum);
2062 		m->m_flags |= RTS_FILTER_FIB;
2063 	}
2064 
2065 	rtm = mtod(m, struct rt_msghdr *);
2066 	rtm->rtm_addrs = info->rti_addrs;
2067 	if (info->rti_ifp != NULL)
2068 		rtm->rtm_index = info->rti_ifp->if_index;
2069 	/* Add RTF_DONE to indicate command 'completion' required by API */
2070 	info->rti_flags |= RTF_DONE;
2071 	/* Reported routes has to be up */
2072 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
2073 		info->rti_flags |= RTF_UP;
2074 	rtm->rtm_flags = info->rti_flags;
2075 
2076 	sa = info->rti_info[RTAX_DST];
2077 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2078 
2079 	return (0);
2080 }
2081 
2082 /*
2083  * This is the analogue to the rt_newaddrmsg which performs the same
2084  * function but for multicast group memberhips.  This is easier since
2085  * there is no route state to worry about.
2086  */
2087 void
2088 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
2089 {
2090 	struct rt_addrinfo info;
2091 	struct mbuf *m = NULL;
2092 	struct ifnet *ifp = ifma->ifma_ifp;
2093 	struct ifma_msghdr *ifmam;
2094 
2095 	if (V_route_cb.any_count == 0)
2096 		return;
2097 
2098 	bzero((caddr_t)&info, sizeof(info));
2099 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2100 	if (ifp && ifp->if_addr)
2101 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2102 	else
2103 		info.rti_info[RTAX_IFP] = NULL;
2104 	/*
2105 	 * If a link-layer address is present, present it as a ``gateway''
2106 	 * (similarly to how ARP entries, e.g., are presented).
2107 	 */
2108 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
2109 	m = rtsock_msg_mbuf(cmd, &info);
2110 	if (m == NULL)
2111 		return;
2112 	ifmam = mtod(m, struct ifma_msghdr *);
2113 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
2114 	    __func__));
2115 	ifmam->ifmam_index = ifp->if_index;
2116 	ifmam->ifmam_addrs = info.rti_addrs;
2117 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
2118 }
2119 
2120 static struct mbuf *
2121 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
2122 	struct rt_addrinfo *info)
2123 {
2124 	struct if_announcemsghdr *ifan;
2125 	struct mbuf *m;
2126 
2127 	if (V_route_cb.any_count == 0)
2128 		return NULL;
2129 	bzero((caddr_t)info, sizeof(*info));
2130 	m = rtsock_msg_mbuf(type, info);
2131 	if (m != NULL) {
2132 		ifan = mtod(m, struct if_announcemsghdr *);
2133 		ifan->ifan_index = ifp->if_index;
2134 		strlcpy(ifan->ifan_name, ifp->if_xname,
2135 			sizeof(ifan->ifan_name));
2136 		ifan->ifan_what = what;
2137 	}
2138 	return m;
2139 }
2140 
2141 /*
2142  * This is called to generate routing socket messages indicating
2143  * IEEE80211 wireless events.
2144  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2145  */
2146 void
2147 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2148 {
2149 	struct mbuf *m;
2150 	struct rt_addrinfo info;
2151 
2152 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2153 	if (m != NULL) {
2154 		/*
2155 		 * Append the ieee80211 data.  Try to stick it in the
2156 		 * mbuf containing the ifannounce msg; otherwise allocate
2157 		 * a new mbuf and append.
2158 		 *
2159 		 * NB: we assume m is a single mbuf.
2160 		 */
2161 		if (data_len > M_TRAILINGSPACE(m)) {
2162 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2163 			if (n == NULL) {
2164 				m_freem(m);
2165 				return;
2166 			}
2167 			bcopy(data, mtod(n, void *), data_len);
2168 			n->m_len = data_len;
2169 			m->m_next = n;
2170 		} else if (data_len > 0) {
2171 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2172 			m->m_len += data_len;
2173 		}
2174 		if (m->m_flags & M_PKTHDR)
2175 			m->m_pkthdr.len += data_len;
2176 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2177 		rt_dispatch(m, AF_UNSPEC);
2178 	}
2179 }
2180 
2181 /*
2182  * This is called to generate routing socket messages indicating
2183  * network interface arrival and departure.
2184  */
2185 static void
2186 rt_ifannouncemsg(struct ifnet *ifp, int what)
2187 {
2188 	struct mbuf *m;
2189 	struct rt_addrinfo info;
2190 
2191 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2192 	if (m != NULL)
2193 		rt_dispatch(m, AF_UNSPEC);
2194 }
2195 
2196 static void
2197 rt_dispatch(struct mbuf *m, sa_family_t saf)
2198 {
2199 
2200 	M_ASSERTPKTHDR(m);
2201 
2202 	m->m_rtsock_family = saf;
2203 	if (V_loif)
2204 		m->m_pkthdr.rcvif = V_loif;
2205 	else {
2206 		m_freem(m);
2207 		return;
2208 	}
2209 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
2210 }
2211 
2212 /*
2213  * This is used in dumping the kernel table via sysctl().
2214  */
2215 static int
2216 sysctl_dumpentry(struct rtentry *rt, void *vw)
2217 {
2218 	struct walkarg *w = vw;
2219 	struct nhop_object *nh;
2220 
2221 	NET_EPOCH_ASSERT();
2222 
2223 	if (!rt_is_exportable(rt, w->w_req->td->td_ucred))
2224 		return (0);
2225 
2226 	export_rtaddrs(rt, w->dst, w->mask);
2227 	nh = rt_get_raw_nhop(rt);
2228 #ifdef ROUTE_MPATH
2229 	if (NH_IS_NHGRP(nh)) {
2230 		const struct weightened_nhop *wn;
2231 		uint32_t num_nhops;
2232 		int error;
2233 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2234 		for (int i = 0; i < num_nhops; i++) {
2235 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2236 			if (error != 0)
2237 				return (error);
2238 		}
2239 	} else
2240 #endif
2241 		sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2242 
2243 	return (0);
2244 }
2245 
2246 
2247 static int
2248 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2249     struct walkarg *w)
2250 {
2251 	struct rt_addrinfo info;
2252 	int error = 0, size;
2253 	uint32_t rtflags;
2254 
2255 	rtflags = nhop_get_rtflags(nh);
2256 
2257 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2258 		return (0);
2259 
2260 	bzero((caddr_t)&info, sizeof(info));
2261 	info.rti_info[RTAX_DST] = w->dst;
2262 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2263 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2264 	info.rti_info[RTAX_GENMASK] = 0;
2265 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2266 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2267 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2268 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2269 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2270 	}
2271 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2272 		return (error);
2273 	if (w->w_req && w->w_tmem) {
2274 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2275 
2276 		bzero(&rtm->rtm_index,
2277 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2278 
2279 		/*
2280 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2281 		 * and RTF_UP (if entry is linked, which is always true here).
2282 		 * Given that, use nhop rtflags & add RTF_UP.
2283 		 */
2284 		rtm->rtm_flags = rtflags | RTF_UP;
2285 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2286 			rtm->rtm_flags = RTF_GATEWAY |
2287 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2288 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2289 		rtm->rtm_rmx.rmx_weight = weight;
2290 		rtm->rtm_index = nh->nh_ifp->if_index;
2291 		rtm->rtm_addrs = info.rti_addrs;
2292 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2293 		return (error);
2294 	}
2295 	return (error);
2296 }
2297 
2298 static int
2299 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2300     struct rt_addrinfo *info, struct walkarg *w, int len)
2301 {
2302 	struct if_msghdrl *ifm;
2303 	struct if_data *ifd;
2304 
2305 	ifm = (struct if_msghdrl *)w->w_tmem;
2306 
2307 #ifdef COMPAT_FREEBSD32
2308 	if (w->w_req->flags & SCTL_MASK32) {
2309 		struct if_msghdrl32 *ifm32;
2310 
2311 		ifm32 = (struct if_msghdrl32 *)ifm;
2312 		ifm32->ifm_addrs = info->rti_addrs;
2313 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2314 		ifm32->ifm_index = ifp->if_index;
2315 		ifm32->_ifm_spare1 = 0;
2316 		ifm32->ifm_len = sizeof(*ifm32);
2317 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2318 		ifm32->_ifm_spare2 = 0;
2319 		ifd = &ifm32->ifm_data;
2320 	} else
2321 #endif
2322 	{
2323 		ifm->ifm_addrs = info->rti_addrs;
2324 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2325 		ifm->ifm_index = ifp->if_index;
2326 		ifm->_ifm_spare1 = 0;
2327 		ifm->ifm_len = sizeof(*ifm);
2328 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2329 		ifm->_ifm_spare2 = 0;
2330 		ifd = &ifm->ifm_data;
2331 	}
2332 
2333 	memcpy(ifd, src_ifd, sizeof(*ifd));
2334 
2335 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2336 }
2337 
2338 static int
2339 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2340     struct rt_addrinfo *info, struct walkarg *w, int len)
2341 {
2342 	struct if_msghdr *ifm;
2343 	struct if_data *ifd;
2344 
2345 	ifm = (struct if_msghdr *)w->w_tmem;
2346 
2347 #ifdef COMPAT_FREEBSD32
2348 	if (w->w_req->flags & SCTL_MASK32) {
2349 		struct if_msghdr32 *ifm32;
2350 
2351 		ifm32 = (struct if_msghdr32 *)ifm;
2352 		ifm32->ifm_addrs = info->rti_addrs;
2353 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2354 		ifm32->ifm_index = ifp->if_index;
2355 		ifm32->_ifm_spare1 = 0;
2356 		ifd = &ifm32->ifm_data;
2357 	} else
2358 #endif
2359 	{
2360 		ifm->ifm_addrs = info->rti_addrs;
2361 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2362 		ifm->ifm_index = ifp->if_index;
2363 		ifm->_ifm_spare1 = 0;
2364 		ifd = &ifm->ifm_data;
2365 	}
2366 
2367 	memcpy(ifd, src_ifd, sizeof(*ifd));
2368 
2369 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2370 }
2371 
2372 static int
2373 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2374     struct walkarg *w, int len)
2375 {
2376 	struct ifa_msghdrl *ifam;
2377 	struct if_data *ifd;
2378 
2379 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2380 
2381 #ifdef COMPAT_FREEBSD32
2382 	if (w->w_req->flags & SCTL_MASK32) {
2383 		struct ifa_msghdrl32 *ifam32;
2384 
2385 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2386 		ifam32->ifam_addrs = info->rti_addrs;
2387 		ifam32->ifam_flags = ifa->ifa_flags;
2388 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2389 		ifam32->_ifam_spare1 = 0;
2390 		ifam32->ifam_len = sizeof(*ifam32);
2391 		ifam32->ifam_data_off =
2392 		    offsetof(struct ifa_msghdrl32, ifam_data);
2393 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2394 		ifd = &ifam32->ifam_data;
2395 	} else
2396 #endif
2397 	{
2398 		ifam->ifam_addrs = info->rti_addrs;
2399 		ifam->ifam_flags = ifa->ifa_flags;
2400 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2401 		ifam->_ifam_spare1 = 0;
2402 		ifam->ifam_len = sizeof(*ifam);
2403 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2404 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2405 		ifd = &ifam->ifam_data;
2406 	}
2407 
2408 	bzero(ifd, sizeof(*ifd));
2409 	ifd->ifi_datalen = sizeof(struct if_data);
2410 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2411 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2412 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2413 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2414 
2415 	/* Fixup if_data carp(4) vhid. */
2416 	if (carp_get_vhid_p != NULL)
2417 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2418 
2419 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2420 }
2421 
2422 static int
2423 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2424     struct walkarg *w, int len)
2425 {
2426 	struct ifa_msghdr *ifam;
2427 
2428 	ifam = (struct ifa_msghdr *)w->w_tmem;
2429 	ifam->ifam_addrs = info->rti_addrs;
2430 	ifam->ifam_flags = ifa->ifa_flags;
2431 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2432 	ifam->_ifam_spare1 = 0;
2433 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2434 
2435 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2436 }
2437 
2438 static int
2439 sysctl_iflist(int af, struct walkarg *w)
2440 {
2441 	struct ifnet *ifp;
2442 	struct ifaddr *ifa;
2443 	struct if_data ifd;
2444 	struct rt_addrinfo info;
2445 	int len, error = 0;
2446 	struct sockaddr_storage ss;
2447 
2448 	bzero((caddr_t)&info, sizeof(info));
2449 	bzero(&ifd, sizeof(ifd));
2450 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2451 		if (w->w_arg && w->w_arg != ifp->if_index)
2452 			continue;
2453 		if_data_copy(ifp, &ifd);
2454 		ifa = ifp->if_addr;
2455 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2456 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2457 		if (error != 0)
2458 			goto done;
2459 		info.rti_info[RTAX_IFP] = NULL;
2460 		if (w->w_req && w->w_tmem) {
2461 			if (w->w_op == NET_RT_IFLISTL)
2462 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2463 				    len);
2464 			else
2465 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2466 				    len);
2467 			if (error)
2468 				goto done;
2469 		}
2470 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2471 			if (af && af != ifa->ifa_addr->sa_family)
2472 				continue;
2473 			if (prison_if(w->w_req->td->td_ucred,
2474 			    ifa->ifa_addr) != 0)
2475 				continue;
2476 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2477 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2478 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2479 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2480 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2481 			if (error != 0)
2482 				goto done;
2483 			if (w->w_req && w->w_tmem) {
2484 				if (w->w_op == NET_RT_IFLISTL)
2485 					error = sysctl_iflist_ifaml(ifa, &info,
2486 					    w, len);
2487 				else
2488 					error = sysctl_iflist_ifam(ifa, &info,
2489 					    w, len);
2490 				if (error)
2491 					goto done;
2492 			}
2493 		}
2494 		info.rti_info[RTAX_IFA] = NULL;
2495 		info.rti_info[RTAX_NETMASK] = NULL;
2496 		info.rti_info[RTAX_BRD] = NULL;
2497 	}
2498 done:
2499 	return (error);
2500 }
2501 
2502 static int
2503 sysctl_ifmalist(int af, struct walkarg *w)
2504 {
2505 	struct rt_addrinfo info;
2506 	struct ifaddr *ifa;
2507 	struct ifmultiaddr *ifma;
2508 	struct ifnet *ifp;
2509 	int error, len;
2510 
2511 	NET_EPOCH_ASSERT();
2512 
2513 	error = 0;
2514 	bzero((caddr_t)&info, sizeof(info));
2515 
2516 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2517 		if (w->w_arg && w->w_arg != ifp->if_index)
2518 			continue;
2519 		ifa = ifp->if_addr;
2520 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2521 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2522 			if (af && af != ifma->ifma_addr->sa_family)
2523 				continue;
2524 			if (prison_if(w->w_req->td->td_ucred,
2525 			    ifma->ifma_addr) != 0)
2526 				continue;
2527 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2528 			info.rti_info[RTAX_GATEWAY] =
2529 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2530 			    ifma->ifma_lladdr : NULL;
2531 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2532 			if (error != 0)
2533 				break;
2534 			if (w->w_req && w->w_tmem) {
2535 				struct ifma_msghdr *ifmam;
2536 
2537 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2538 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2539 				ifmam->ifmam_flags = 0;
2540 				ifmam->ifmam_addrs = info.rti_addrs;
2541 				ifmam->_ifmam_spare1 = 0;
2542 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2543 				if (error != 0)
2544 					break;
2545 			}
2546 		}
2547 		if (error != 0)
2548 			break;
2549 	}
2550 	return (error);
2551 }
2552 
2553 static void
2554 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2555 {
2556 	union sockaddr_union sa_dst, sa_mask;
2557 
2558 	w->family = family;
2559 	w->dst = (struct sockaddr *)&sa_dst;
2560 	w->mask = (struct sockaddr *)&sa_mask;
2561 
2562 	init_sockaddrs_family(family, w->dst, w->mask);
2563 
2564 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2565 }
2566 
2567 static int
2568 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2569 {
2570 	struct epoch_tracker et;
2571 	int	*name = (int *)arg1;
2572 	u_int	namelen = arg2;
2573 	struct rib_head *rnh = NULL; /* silence compiler. */
2574 	int	i, lim, error = EINVAL;
2575 	int	fib = 0;
2576 	u_char	af;
2577 	struct	walkarg w;
2578 
2579 	if (namelen < 3)
2580 		return (EINVAL);
2581 
2582 	name++;
2583 	namelen--;
2584 	if (req->newptr)
2585 		return (EPERM);
2586 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2587 		if (namelen == 3)
2588 			fib = req->td->td_proc->p_fibnum;
2589 		else if (namelen == 4)
2590 			fib = (name[3] == RT_ALL_FIBS) ?
2591 			    req->td->td_proc->p_fibnum : name[3];
2592 		else
2593 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2594 		if (fib < 0 || fib >= rt_numfibs)
2595 			return (EINVAL);
2596 	} else if (namelen != 3)
2597 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2598 	af = name[0];
2599 	if (af > AF_MAX)
2600 		return (EINVAL);
2601 	bzero(&w, sizeof(w));
2602 	w.w_op = name[1];
2603 	w.w_arg = name[2];
2604 	w.w_req = req;
2605 
2606 	error = sysctl_wire_old_buffer(req, 0);
2607 	if (error)
2608 		return (error);
2609 
2610 	/*
2611 	 * Allocate reply buffer in advance.
2612 	 * All rtsock messages has maximum length of u_short.
2613 	 */
2614 	w.w_tmemsize = 65536;
2615 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2616 
2617 	NET_EPOCH_ENTER(et);
2618 	switch (w.w_op) {
2619 	case NET_RT_DUMP:
2620 	case NET_RT_FLAGS:
2621 		if (af == 0) {			/* dump all tables */
2622 			i = 1;
2623 			lim = AF_MAX;
2624 		} else				/* dump only one table */
2625 			i = lim = af;
2626 
2627 		/*
2628 		 * take care of llinfo entries, the caller must
2629 		 * specify an AF
2630 		 */
2631 		if (w.w_op == NET_RT_FLAGS &&
2632 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2633 			if (af != 0)
2634 				error = lltable_sysctl_dumparp(af, w.w_req);
2635 			else
2636 				error = EINVAL;
2637 			break;
2638 		}
2639 		/*
2640 		 * take care of routing entries
2641 		 */
2642 		for (error = 0; error == 0 && i <= lim; i++) {
2643 			rnh = rt_tables_get_rnh(fib, i);
2644 			if (rnh != NULL) {
2645 				rtable_sysctl_dump(fib, i, &w);
2646 			} else if (af != 0)
2647 				error = EAFNOSUPPORT;
2648 		}
2649 		break;
2650 	case NET_RT_NHOP:
2651 	case NET_RT_NHGRP:
2652 		/* Allow dumping one specific af/fib at a time */
2653 		if (namelen < 4) {
2654 			error = EINVAL;
2655 			break;
2656 		}
2657 		fib = name[3];
2658 		if (fib < 0 || fib > rt_numfibs) {
2659 			error = EINVAL;
2660 			break;
2661 		}
2662 		rnh = rt_tables_get_rnh(fib, af);
2663 		if (rnh == NULL) {
2664 			error = EAFNOSUPPORT;
2665 			break;
2666 		}
2667 		if (w.w_op == NET_RT_NHOP)
2668 			error = nhops_dump_sysctl(rnh, w.w_req);
2669 		else
2670 #ifdef ROUTE_MPATH
2671 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2672 #else
2673 			error = ENOTSUP;
2674 #endif
2675 		break;
2676 	case NET_RT_IFLIST:
2677 	case NET_RT_IFLISTL:
2678 		error = sysctl_iflist(af, &w);
2679 		break;
2680 
2681 	case NET_RT_IFMALIST:
2682 		error = sysctl_ifmalist(af, &w);
2683 		break;
2684 	}
2685 	NET_EPOCH_EXIT(et);
2686 
2687 	free(w.w_tmem, M_TEMP);
2688 	return (error);
2689 }
2690 
2691 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2692     sysctl_rtsock, "Return route tables and interface/address lists");
2693 
2694 /*
2695  * Definitions of protocols supported in the ROUTE domain.
2696  */
2697 
2698 static struct domain routedomain;		/* or at least forward */
2699 
2700 static struct protosw routesw = {
2701 	.pr_type =		SOCK_RAW,
2702 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2703 	.pr_abort =		rts_close,
2704 	.pr_attach =		rts_attach,
2705 	.pr_detach =		rts_detach,
2706 	.pr_send =		rts_send,
2707 	.pr_shutdown =		rts_shutdown,
2708 	.pr_disconnect =	rts_disconnect,
2709 	.pr_close =		rts_close,
2710 };
2711 
2712 static struct domain routedomain = {
2713 	.dom_family =		PF_ROUTE,
2714 	.dom_name =		"route",
2715 	.dom_nprotosw =		1,
2716 	.dom_protosw =		{ &routesw },
2717 };
2718 
2719 DOMAIN_SET(route);
2720