xref: /freebsd/sys/net/rtsock.c (revision caccbaef8e263b1d769e7bcac1c4617bdc12d484)
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
sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)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
vnet_rts_init(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, NULL);
265 
266 #ifdef VIMAGE
267 static void
vnet_rts_uninit(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, NULL);
275 #endif
276 
277 static void
report_route_event(const struct rib_cmd_info * rc,void * _cbdata)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
rts_handle_route_event(uint32_t fibnum,const struct rib_cmd_info * rc)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
rtsock_notify_event(uint32_t fibnum,const struct rib_cmd_info * rc)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
rtsock_init(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
rts_handle_ifnet_arrival(void * arg __unused,struct ifnet * ifp)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
rts_handle_ifnet_departure(void * arg __unused,struct ifnet * ifp)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
rts_append_data(struct socket * so,struct mbuf * m)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
rts_input(struct mbuf * m)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
rts_close(struct socket * so)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
rts_attach(struct socket * so,int proto,struct thread * td)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 int
rts_ctloutput(struct socket * so,struct sockopt * sopt)427 rts_ctloutput(struct socket *so, struct sockopt *sopt)
428 {
429 	int error, optval;
430 
431 	error = ENOPROTOOPT;
432 	if (sopt->sopt_dir == SOPT_SET) {
433 		switch (sopt->sopt_level) {
434 		case SOL_SOCKET:
435 			switch (sopt->sopt_name) {
436 			case SO_SETFIB:
437 				error = sooptcopyin(sopt, &optval,
438 				    sizeof(optval), sizeof(optval));
439 				if (error != 0)
440 					break;
441 				error = sosetfib(so, optval);
442 				break;
443 			}
444 			break;
445 		}
446 	}
447 	return (error);
448 }
449 
450 static void
rts_detach(struct socket * so)451 rts_detach(struct socket *so)
452 {
453 	struct rcb *rcb = so->so_pcb;
454 
455 	RTSOCK_LOCK();
456 	LIST_REMOVE(rcb, list);
457 	switch(rcb->rcb_family) {
458 	case AF_INET:
459 		V_route_cb.ip_count--;
460 		break;
461 	case AF_INET6:
462 		V_route_cb.ip6_count--;
463 		break;
464 	}
465 	V_route_cb.any_count--;
466 	RTSOCK_UNLOCK();
467 	free(rcb, M_PCB);
468 	so->so_pcb = NULL;
469 }
470 
471 static int
rts_disconnect(struct socket * so)472 rts_disconnect(struct socket *so)
473 {
474 
475 	return (ENOTCONN);
476 }
477 
478 static int
rts_shutdown(struct socket * so,enum shutdown_how how)479 rts_shutdown(struct socket *so, enum shutdown_how how)
480 {
481 	/*
482 	 * Note: route socket marks itself as connected through its lifetime.
483 	 */
484 	switch (how) {
485 	case SHUT_RD:
486 		sorflush(so);
487 		break;
488 	case SHUT_RDWR:
489 		sorflush(so);
490 		/* FALLTHROUGH */
491 	case SHUT_WR:
492 		socantsendmore(so);
493 	}
494 
495 	return (0);
496 }
497 
498 #ifndef _SOCKADDR_UNION_DEFINED
499 #define	_SOCKADDR_UNION_DEFINED
500 /*
501  * The union of all possible address formats we handle.
502  */
503 union sockaddr_union {
504 	struct sockaddr		sa;
505 	struct sockaddr_in	sin;
506 	struct sockaddr_in6	sin6;
507 };
508 #endif /* _SOCKADDR_UNION_DEFINED */
509 
510 static int
rtm_get_jailed(struct rt_addrinfo * info,struct ifnet * ifp,struct nhop_object * nh,union sockaddr_union * saun,struct ucred * cred)511 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
512     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
513 {
514 #if defined(INET) || defined(INET6)
515 	struct epoch_tracker et;
516 #endif
517 
518 	/* First, see if the returned address is part of the jail. */
519 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
520 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
521 		return (0);
522 	}
523 
524 	switch (info->rti_info[RTAX_DST]->sa_family) {
525 #ifdef INET
526 	case AF_INET:
527 	{
528 		struct in_addr ia;
529 		struct ifaddr *ifa;
530 		int found;
531 
532 		found = 0;
533 		/*
534 		 * Try to find an address on the given outgoing interface
535 		 * that belongs to the jail.
536 		 */
537 		NET_EPOCH_ENTER(et);
538 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
539 			struct sockaddr *sa;
540 			sa = ifa->ifa_addr;
541 			if (sa->sa_family != AF_INET)
542 				continue;
543 			ia = ((struct sockaddr_in *)sa)->sin_addr;
544 			if (prison_check_ip4(cred, &ia) == 0) {
545 				found = 1;
546 				break;
547 			}
548 		}
549 		NET_EPOCH_EXIT(et);
550 		if (!found) {
551 			/*
552 			 * As a last resort return the 'default' jail address.
553 			 */
554 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
555 			    sin_addr;
556 			if (prison_get_ip4(cred, &ia) != 0)
557 				return (ESRCH);
558 		}
559 		bzero(&saun->sin, sizeof(struct sockaddr_in));
560 		saun->sin.sin_len = sizeof(struct sockaddr_in);
561 		saun->sin.sin_family = AF_INET;
562 		saun->sin.sin_addr.s_addr = ia.s_addr;
563 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
564 		break;
565 	}
566 #endif
567 #ifdef INET6
568 	case AF_INET6:
569 	{
570 		struct in6_addr ia6;
571 		struct ifaddr *ifa;
572 		int found;
573 
574 		found = 0;
575 		/*
576 		 * Try to find an address on the given outgoing interface
577 		 * that belongs to the jail.
578 		 */
579 		NET_EPOCH_ENTER(et);
580 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
581 			struct sockaddr *sa;
582 			sa = ifa->ifa_addr;
583 			if (sa->sa_family != AF_INET6)
584 				continue;
585 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
586 			    &ia6, sizeof(struct in6_addr));
587 			if (prison_check_ip6(cred, &ia6) == 0) {
588 				found = 1;
589 				break;
590 			}
591 		}
592 		NET_EPOCH_EXIT(et);
593 		if (!found) {
594 			/*
595 			 * As a last resort return the 'default' jail address.
596 			 */
597 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
598 			    sin6_addr;
599 			if (prison_get_ip6(cred, &ia6) != 0)
600 				return (ESRCH);
601 		}
602 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
603 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
604 		saun->sin6.sin6_family = AF_INET6;
605 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
606 		if (sa6_recoverscope(&saun->sin6) != 0)
607 			return (ESRCH);
608 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
609 		break;
610 	}
611 #endif
612 	default:
613 		return (ESRCH);
614 	}
615 	return (0);
616 }
617 
618 static int
fill_blackholeinfo(struct rt_addrinfo * info,union sockaddr_union * saun)619 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
620 {
621 	struct ifaddr *ifa;
622 	sa_family_t saf;
623 
624 	if (V_loif == NULL) {
625 		RTS_PID_LOG(LOG_INFO, "Unable to add blackhole/reject nhop without loopback");
626 		return (ENOTSUP);
627 	}
628 	info->rti_ifp = V_loif;
629 
630 	saf = info->rti_info[RTAX_DST]->sa_family;
631 
632 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
633 		if (ifa->ifa_addr->sa_family == saf) {
634 			info->rti_ifa = ifa;
635 			break;
636 		}
637 	}
638 	if (info->rti_ifa == NULL) {
639 		RTS_PID_LOG(LOG_INFO, "Unable to find ifa for blackhole/reject nhop");
640 		return (ENOTSUP);
641 	}
642 
643 	bzero(saun, sizeof(union sockaddr_union));
644 	switch (saf) {
645 #ifdef INET
646 	case AF_INET:
647 		saun->sin.sin_family = AF_INET;
648 		saun->sin.sin_len = sizeof(struct sockaddr_in);
649 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
650 		break;
651 #endif
652 #ifdef INET6
653 	case AF_INET6:
654 		saun->sin6.sin6_family = AF_INET6;
655 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
656 		saun->sin6.sin6_addr = in6addr_loopback;
657 		break;
658 #endif
659 	default:
660 		RTS_PID_LOG(LOG_INFO, "unsupported family: %d", saf);
661 		return (ENOTSUP);
662 	}
663 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
664 	info->rti_flags |= RTF_GATEWAY;
665 
666 	return (0);
667 }
668 
669 /*
670  * Fills in @info based on userland-provided @rtm message.
671  *
672  * Returns 0 on success.
673  */
674 static int
fill_addrinfo(struct rt_msghdr * rtm,int len,struct linear_buffer * lb,u_int fibnum,struct rt_addrinfo * info)675 fill_addrinfo(struct rt_msghdr *rtm, int len, struct linear_buffer *lb, u_int fibnum,
676     struct rt_addrinfo *info)
677 {
678 	int error;
679 
680 	rtm->rtm_pid = curproc->p_pid;
681 	info->rti_addrs = rtm->rtm_addrs;
682 
683 	info->rti_mflags = rtm->rtm_inits;
684 	info->rti_rmx = &rtm->rtm_rmx;
685 
686 	/*
687 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
688 	 * link-local address because rtrequest requires addresses with
689 	 * embedded scope id.
690 	 */
691 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
692 		return (EINVAL);
693 
694 	info->rti_flags = rtm->rtm_flags;
695 	error = cleanup_xaddrs(info, lb);
696 	if (error != 0)
697 		return (error);
698 	/*
699 	 * Verify that the caller has the appropriate privilege; RTM_GET
700 	 * is the only operation the non-superuser is allowed.
701 	 */
702 	if (rtm->rtm_type != RTM_GET) {
703 		error = priv_check(curthread, PRIV_NET_ROUTE);
704 		if (error != 0)
705 			return (error);
706 	}
707 
708 	/*
709 	 * The given gateway address may be an interface address.
710 	 * For example, issuing a "route change" command on a route
711 	 * entry that was created from a tunnel, and the gateway
712 	 * address given is the local end point. In this case the
713 	 * RTF_GATEWAY flag must be cleared or the destination will
714 	 * not be reachable even though there is no error message.
715 	 */
716 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
717 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
718 		struct nhop_object *nh;
719 
720 		/*
721 		 * A host route through the loopback interface is
722 		 * installed for each interface address. In pre 8.0
723 		 * releases the interface address of a PPP link type
724 		 * is not reachable locally. This behavior is fixed as
725 		 * part of the new L2/L3 redesign and rewrite work. The
726 		 * signature of this interface address route is the
727 		 * AF_LINK sa_family type of the gateway, and the
728 		 * rt_ifp has the IFF_LOOPBACK flag set.
729 		 */
730 		nh = rib_lookup(fibnum, info->rti_info[RTAX_GATEWAY], NHR_NONE, 0);
731 		if (nh != NULL && nh->gw_sa.sa_family == AF_LINK &&
732 		    nh->nh_ifp->if_flags & IFF_LOOPBACK) {
733 				info->rti_flags &= ~RTF_GATEWAY;
734 				info->rti_flags |= RTF_GWFLAG_COMPAT;
735 		}
736 	}
737 
738 	return (0);
739 }
740 
741 static struct nhop_object *
select_nhop(struct nhop_object * nh,const struct sockaddr * gw)742 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
743 {
744 	if (!NH_IS_NHGRP(nh))
745 		return (nh);
746 #ifdef ROUTE_MPATH
747 	const struct weightened_nhop *wn;
748 	uint32_t num_nhops;
749 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
750 	if (gw == NULL)
751 		return (wn[0].nh);
752 	for (int i = 0; i < num_nhops; i++) {
753 		if (match_nhop_gw(wn[i].nh, gw))
754 			return (wn[i].nh);
755 	}
756 #endif
757 	return (NULL);
758 }
759 
760 /*
761  * Handles RTM_GET message from routing socket, returning matching rt.
762  *
763  * Returns:
764  * 0 on success, with locked and referenced matching rt in @rt_nrt
765  * errno of failure
766  */
767 static int
handle_rtm_get(struct rt_addrinfo * info,u_int fibnum,struct rt_msghdr * rtm,struct rib_cmd_info * rc)768 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
769     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
770 {
771 	RIB_RLOCK_TRACKER;
772 	struct rib_head *rnh;
773 	struct nhop_object *nh;
774 	sa_family_t saf;
775 
776 	saf = info->rti_info[RTAX_DST]->sa_family;
777 
778 	rnh = rt_tables_get_rnh(fibnum, saf);
779 	if (rnh == NULL)
780 		return (EAFNOSUPPORT);
781 
782 	RIB_RLOCK(rnh);
783 
784 	/*
785 	 * By (implicit) convention host route (one without netmask)
786 	 * means longest-prefix-match request and the route with netmask
787 	 * means exact-match lookup.
788 	 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128
789 	 * prefixes, use original data to check for the netmask presence.
790 	 */
791 	if ((rtm->rtm_addrs & RTA_NETMASK) == 0) {
792 		/*
793 		 * Provide longest prefix match for
794 		 * address lookup (no mask).
795 		 * 'route -n get addr'
796 		 */
797 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
798 		    info->rti_info[RTAX_DST], &rnh->head);
799 	} else
800 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
801 		    info->rti_info[RTAX_DST],
802 		    info->rti_info[RTAX_NETMASK], &rnh->head);
803 
804 	if (rc->rc_rt == NULL) {
805 		RIB_RUNLOCK(rnh);
806 		return (ESRCH);
807 	}
808 
809 	nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
810 	if (nh == NULL) {
811 		RIB_RUNLOCK(rnh);
812 		return (ESRCH);
813 	}
814 	/*
815 	 * If performing proxied L2 entry insertion, and
816 	 * the actual PPP host entry is found, perform
817 	 * another search to retrieve the prefix route of
818 	 * the local end point of the PPP link.
819 	 * TODO: move this logic to userland.
820 	 */
821 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
822 		struct sockaddr_storage laddr;
823 
824 		if (nh->nh_ifp != NULL &&
825 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
826 			struct ifaddr *ifa;
827 
828 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
829 					RT_ALL_FIBS);
830 			if (ifa != NULL)
831 				rt_maskedcopy(ifa->ifa_addr,
832 					      (struct sockaddr *)&laddr,
833 					      ifa->ifa_netmask);
834 		} else
835 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
836 				      (struct sockaddr *)&laddr,
837 				      nh->nh_ifa->ifa_netmask);
838 		/*
839 		 * refactor rt and no lock operation necessary
840 		 */
841 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(
842 		    (struct sockaddr *)&laddr, &rnh->head);
843 		if (rc->rc_rt == NULL) {
844 			RIB_RUNLOCK(rnh);
845 			return (ESRCH);
846 		}
847 		nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
848 		if (nh == NULL) {
849 			RIB_RUNLOCK(rnh);
850 			return (ESRCH);
851 		}
852 	}
853 	rc->rc_nh_new = nh;
854 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
855 	RIB_RUNLOCK(rnh);
856 
857 	return (0);
858 }
859 
860 static void
init_sockaddrs_family(int family,struct sockaddr * dst,struct sockaddr * mask)861 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
862 {
863 #ifdef INET
864 	if (family == AF_INET) {
865 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
866 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
867 
868 		bzero(dst4, sizeof(struct sockaddr_in));
869 		bzero(mask4, sizeof(struct sockaddr_in));
870 
871 		dst4->sin_family = AF_INET;
872 		dst4->sin_len = sizeof(struct sockaddr_in);
873 		mask4->sin_family = AF_INET;
874 		mask4->sin_len = sizeof(struct sockaddr_in);
875 	}
876 #endif
877 #ifdef INET6
878 	if (family == AF_INET6) {
879 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
880 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
881 
882 		bzero(dst6, sizeof(struct sockaddr_in6));
883 		bzero(mask6, sizeof(struct sockaddr_in6));
884 
885 		dst6->sin6_family = AF_INET6;
886 		dst6->sin6_len = sizeof(struct sockaddr_in6);
887 		mask6->sin6_family = AF_INET6;
888 		mask6->sin6_len = sizeof(struct sockaddr_in6);
889 	}
890 #endif
891 }
892 
893 static void
export_rtaddrs(const struct rtentry * rt,struct sockaddr * dst,struct sockaddr * mask)894 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
895     struct sockaddr *mask)
896 {
897 #ifdef INET
898 	if (dst->sa_family == AF_INET) {
899 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
900 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
901 		uint32_t scopeid = 0;
902 		rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
903 		    &scopeid);
904 		return;
905 	}
906 #endif
907 #ifdef INET6
908 	if (dst->sa_family == AF_INET6) {
909 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
910 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
911 		uint32_t scopeid = 0;
912 		rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
913 		    &mask6->sin6_addr, &scopeid);
914 		dst6->sin6_scope_id = scopeid;
915 		return;
916 	}
917 #endif
918 }
919 
920 static int
update_rtm_from_info(struct rt_addrinfo * info,struct rt_msghdr ** prtm,int alloc_len)921 update_rtm_from_info(struct rt_addrinfo *info, struct rt_msghdr **prtm,
922     int alloc_len)
923 {
924 	struct rt_msghdr *rtm, *orig_rtm = NULL;
925 	struct walkarg w;
926 	int len;
927 
928 	rtm = *prtm;
929 	/* Check if we need to realloc storage */
930 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
931 	if (len > alloc_len) {
932 		struct rt_msghdr *tmp_rtm;
933 
934 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
935 		if (tmp_rtm == NULL)
936 			return (ENOBUFS);
937 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
938 		orig_rtm = rtm;
939 		rtm = tmp_rtm;
940 		alloc_len = len;
941 
942 		/*
943 		 * Delay freeing original rtm as info contains
944 		 * data referencing it.
945 		 */
946 	}
947 
948 	w = (struct walkarg ){
949 		.w_tmem = (caddr_t)rtm,
950 		.w_tmemsize = alloc_len,
951 	};
952 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
953 	rtm->rtm_addrs = info->rti_addrs;
954 
955 	if (orig_rtm != NULL)
956 		free(orig_rtm, M_TEMP);
957 	*prtm = rtm;
958 	return (0);
959 }
960 
961 
962 /*
963  * Update sockaddrs, flags, etc in @prtm based on @rc data.
964  * rtm can be reallocated.
965  *
966  * Returns 0 on success, along with pointer to (potentially reallocated)
967  *  rtm.
968  *
969  */
970 static int
update_rtm_from_rc(struct rt_addrinfo * info,struct rt_msghdr ** prtm,int alloc_len,struct rib_cmd_info * rc,struct nhop_object * nh)971 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
972     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
973 {
974 	union sockaddr_union saun;
975 	struct rt_msghdr *rtm;
976 	struct ifnet *ifp;
977 	int error;
978 
979 	rtm = *prtm;
980 	union sockaddr_union sa_dst, sa_mask;
981 	int family = info->rti_info[RTAX_DST]->sa_family;
982 	init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
983 	export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
984 
985 	info->rti_info[RTAX_DST] = &sa_dst.sa;
986 	info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
987 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
988 	info->rti_info[RTAX_GENMASK] = 0;
989 	ifp = nh->nh_ifp;
990 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
991 		if (ifp) {
992 			info->rti_info[RTAX_IFP] =
993 			    ifp->if_addr->ifa_addr;
994 			error = rtm_get_jailed(info, ifp, nh,
995 			    &saun, curthread->td_ucred);
996 			if (error != 0)
997 				return (error);
998 			if (ifp->if_flags & IFF_POINTOPOINT)
999 				info->rti_info[RTAX_BRD] =
1000 				    nh->nh_ifa->ifa_dstaddr;
1001 			rtm->rtm_index = ifp->if_index;
1002 		} else {
1003 			info->rti_info[RTAX_IFP] = NULL;
1004 			info->rti_info[RTAX_IFA] = NULL;
1005 		}
1006 	} else if (ifp != NULL)
1007 		rtm->rtm_index = ifp->if_index;
1008 
1009 	if ((error = update_rtm_from_info(info, prtm, alloc_len)) != 0)
1010 		return (error);
1011 
1012 	rtm = *prtm;
1013 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
1014 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
1015 		rtm->rtm_flags = RTF_GATEWAY |
1016 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
1017 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
1018 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
1019 
1020 	return (0);
1021 }
1022 
1023 #ifdef ROUTE_MPATH
1024 static void
save_del_notification(const struct rib_cmd_info * rc,void * _cbdata)1025 save_del_notification(const struct rib_cmd_info *rc, void *_cbdata)
1026 {
1027 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1028 
1029 	if (rc->rc_cmd == RTM_DELETE)
1030 		*rc_new = *rc;
1031 }
1032 
1033 static void
save_add_notification(const struct rib_cmd_info * rc,void * _cbdata)1034 save_add_notification(const struct rib_cmd_info *rc, void *_cbdata)
1035 {
1036 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1037 
1038 	if (rc->rc_cmd == RTM_ADD)
1039 		*rc_new = *rc;
1040 }
1041 #endif
1042 
1043 #if defined(INET6) || defined(INET)
1044 static struct sockaddr *
alloc_sockaddr_aligned(struct linear_buffer * lb,int len)1045 alloc_sockaddr_aligned(struct linear_buffer *lb, int len)
1046 {
1047 	len = roundup2(len, sizeof(uint64_t));
1048 	if (lb->offset + len > lb->size)
1049 		return (NULL);
1050 	struct sockaddr *sa = (struct sockaddr *)(lb->base + lb->offset);
1051 	lb->offset += len;
1052 	return (sa);
1053 }
1054 #endif
1055 
1056 static int
rts_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)1057 rts_send(struct socket *so, int flags, struct mbuf *m,
1058     struct sockaddr *nam, struct mbuf *control, struct thread *td)
1059 {
1060 	struct rt_msghdr *rtm = NULL;
1061 	struct rt_addrinfo info;
1062 	struct epoch_tracker et;
1063 #ifdef INET6
1064 	struct sockaddr_storage ss;
1065 	struct sockaddr_in6 *sin6;
1066 	int i, rti_need_deembed = 0;
1067 #endif
1068 	int alloc_len = 0, len, error = 0, fibnum;
1069 	sa_family_t saf = AF_UNSPEC;
1070 	struct rib_cmd_info rc;
1071 	struct nhop_object *nh;
1072 
1073 	if ((flags & PRUS_OOB) || control != NULL) {
1074 		m_freem(m);
1075 		if (control != NULL)
1076 			m_freem(control);
1077 		return (EOPNOTSUPP);
1078 	}
1079 
1080 	fibnum = so->so_fibnum;
1081 #define senderr(e) { error = e; goto flush;}
1082 	if (m == NULL || ((m->m_len < sizeof(long)) &&
1083 		       (m = m_pullup(m, sizeof(long))) == NULL))
1084 		return (ENOBUFS);
1085 	if ((m->m_flags & M_PKTHDR) == 0)
1086 		panic("route_output");
1087 	NET_EPOCH_ENTER(et);
1088 	len = m->m_pkthdr.len;
1089 	if (len < sizeof(*rtm) ||
1090 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1091 		senderr(EINVAL);
1092 
1093 	/*
1094 	 * Most of current messages are in range 200-240 bytes,
1095 	 * minimize possible re-allocation on reply using larger size
1096 	 * buffer aligned on 1k boundaty.
1097 	 */
1098 	alloc_len = roundup2(len, 1024);
1099 	int total_len = alloc_len + SCRATCH_BUFFER_SIZE;
1100 	if ((rtm = malloc(total_len, M_TEMP, M_NOWAIT)) == NULL)
1101 		senderr(ENOBUFS);
1102 
1103 	m_copydata(m, 0, len, (caddr_t)rtm);
1104 	bzero(&info, sizeof(info));
1105 	nh = NULL;
1106 	struct linear_buffer lb = {
1107 		.base = (char *)rtm + alloc_len,
1108 		.size = SCRATCH_BUFFER_SIZE,
1109 	};
1110 
1111 	if (rtm->rtm_version != RTM_VERSION) {
1112 		/* Do not touch message since format is unknown */
1113 		free(rtm, M_TEMP);
1114 		rtm = NULL;
1115 		senderr(EPROTONOSUPPORT);
1116 	}
1117 
1118 	/*
1119 	 * Starting from here, it is possible
1120 	 * to alter original message and insert
1121 	 * caller PID and error value.
1122 	 */
1123 
1124 	if ((error = fill_addrinfo(rtm, len, &lb, fibnum, &info)) != 0) {
1125 		senderr(error);
1126 	}
1127 	/* fill_addringo() embeds scope into IPv6 addresses */
1128 #ifdef INET6
1129 	rti_need_deembed = 1;
1130 #endif
1131 
1132 	saf = info.rti_info[RTAX_DST]->sa_family;
1133 
1134 	/* support for new ARP code */
1135 	if (rtm->rtm_flags & RTF_LLDATA) {
1136 		error = lla_rt_output(rtm, &info);
1137 		goto flush;
1138 	}
1139 
1140 	union sockaddr_union gw_saun;
1141 	int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1142 	if (blackhole_flags != 0) {
1143 		if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1144 			error = fill_blackholeinfo(&info, &gw_saun);
1145 		else {
1146 			RTS_PID_LOG(LOG_DEBUG, "both BLACKHOLE and REJECT flags specifiied");
1147 			error = EINVAL;
1148 		}
1149 		if (error != 0)
1150 			senderr(error);
1151 	}
1152 
1153 	switch (rtm->rtm_type) {
1154 	case RTM_ADD:
1155 	case RTM_CHANGE:
1156 		if (rtm->rtm_type == RTM_ADD) {
1157 			if (info.rti_info[RTAX_GATEWAY] == NULL) {
1158 				RTS_PID_LOG(LOG_DEBUG, "RTM_ADD w/o gateway");
1159 				senderr(EINVAL);
1160 			}
1161 		}
1162 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1163 		if (error == 0) {
1164 			rtsock_notify_event(fibnum, &rc);
1165 #ifdef ROUTE_MPATH
1166 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
1167 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1168 				struct rib_cmd_info rc_simple = {};
1169 				rib_decompose_notification(&rc,
1170 				    save_add_notification, (void *)&rc_simple);
1171 				rc = rc_simple;
1172 			}
1173 #endif
1174 			/* nh MAY be empty if RTM_CHANGE request is no-op */
1175 			nh = rc.rc_nh_new;
1176 			if (nh != NULL) {
1177 				rtm->rtm_index = nh->nh_ifp->if_index;
1178 				rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1179 			}
1180 		}
1181 		break;
1182 
1183 	case RTM_DELETE:
1184 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1185 		if (error == 0) {
1186 			rtsock_notify_event(fibnum, &rc);
1187 #ifdef ROUTE_MPATH
1188 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1189 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1190 				struct rib_cmd_info rc_simple = {};
1191 				rib_decompose_notification(&rc,
1192 				    save_del_notification, (void *)&rc_simple);
1193 				rc = rc_simple;
1194 			}
1195 #endif
1196 			nh = rc.rc_nh_old;
1197 		}
1198 		break;
1199 
1200 	case RTM_GET:
1201 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1202 		if (error != 0)
1203 			senderr(error);
1204 		nh = rc.rc_nh_new;
1205 
1206 		if (!rt_is_exportable(rc.rc_rt, curthread->td_ucred))
1207 			senderr(ESRCH);
1208 		break;
1209 
1210 	default:
1211 		senderr(EOPNOTSUPP);
1212 	}
1213 
1214 	if (error == 0 && nh != NULL) {
1215 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1216 		/*
1217 		 * Note that some sockaddr pointers may have changed to
1218 		 * point to memory outsize @rtm. Some may be pointing
1219 		 * to the on-stack variables.
1220 		 * Given that, any pointer in @info CANNOT BE USED.
1221 		 */
1222 
1223 		/*
1224 		 * scopeid deembedding has been performed while
1225 		 * writing updated rtm in rtsock_msg_buffer().
1226 		 * With that in mind, skip deembedding procedure below.
1227 		 */
1228 #ifdef INET6
1229 		rti_need_deembed = 0;
1230 #endif
1231 	}
1232 
1233 flush:
1234 	NET_EPOCH_EXIT(et);
1235 
1236 #ifdef INET6
1237 	if (rtm != NULL) {
1238 		if (rti_need_deembed) {
1239 			/* sin6_scope_id is recovered before sending rtm. */
1240 			sin6 = (struct sockaddr_in6 *)&ss;
1241 			for (i = 0; i < RTAX_MAX; i++) {
1242 				if (info.rti_info[i] == NULL)
1243 					continue;
1244 				if (info.rti_info[i]->sa_family != AF_INET6)
1245 					continue;
1246 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1247 				if (sa6_recoverscope(sin6) == 0)
1248 					bcopy(sin6, info.rti_info[i],
1249 						    sizeof(*sin6));
1250 			}
1251 			if (update_rtm_from_info(&info, &rtm, alloc_len) != 0) {
1252 				if (error != 0)
1253 					error = ENOBUFS;
1254 			}
1255 		}
1256 	}
1257 #endif
1258 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1259 
1260 	return (error);
1261 }
1262 
1263 /*
1264  * Sends the prepared reply message in @rtm to all rtsock clients.
1265  * Frees @m and @rtm.
1266  *
1267  */
1268 static void
send_rtm_reply(struct socket * so,struct rt_msghdr * rtm,struct mbuf * m,sa_family_t saf,u_int fibnum,int rtm_errno)1269 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1270     sa_family_t saf, u_int fibnum, int rtm_errno)
1271 {
1272 	struct rcb *rcb = NULL;
1273 
1274 	/*
1275 	 * Check to see if we don't want our own messages.
1276 	 */
1277 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1278 		if (V_route_cb.any_count <= 1) {
1279 			if (rtm != NULL)
1280 				free(rtm, M_TEMP);
1281 			m_freem(m);
1282 			return;
1283 		}
1284 		/* There is another listener, so construct message */
1285 		rcb = so->so_pcb;
1286 	}
1287 
1288 	if (rtm != NULL) {
1289 		if (rtm_errno!= 0)
1290 			rtm->rtm_errno = rtm_errno;
1291 		else
1292 			rtm->rtm_flags |= RTF_DONE;
1293 
1294 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1295 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1296 			m_freem(m);
1297 			m = NULL;
1298 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1299 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1300 
1301 		free(rtm, M_TEMP);
1302 	}
1303 	if (m != NULL) {
1304 		M_SETFIB(m, fibnum);
1305 		m->m_flags |= RTS_FILTER_FIB;
1306 		if (rcb) {
1307 			/*
1308 			 * XXX insure we don't get a copy by
1309 			 * invalidating our protocol
1310 			 */
1311 			sa_family_t family = rcb->rcb_family;
1312 			rcb->rcb_family = AF_UNSPEC;
1313 			rt_dispatch(m, saf);
1314 			rcb->rcb_family = family;
1315 		} else
1316 			rt_dispatch(m, saf);
1317 	}
1318 }
1319 
1320 static void
rt_getmetrics(const struct rtentry * rt,const struct nhop_object * nh,struct rt_metrics * out)1321 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1322     struct rt_metrics *out)
1323 {
1324 
1325 	bzero(out, sizeof(*out));
1326 	out->rmx_mtu = nh->nh_mtu;
1327 	out->rmx_weight = rt->rt_weight;
1328 	out->rmx_nhidx = nhop_get_idx(nh);
1329 	/* Kernel -> userland timebase conversion. */
1330 	out->rmx_expire = nhop_get_expire(nh) ?
1331 	    nhop_get_expire(nh) - time_uptime + time_second : 0;
1332 }
1333 
1334 /*
1335  * Extract the addresses of the passed sockaddrs.
1336  * Do a little sanity checking so as to avoid bad memory references.
1337  * This data is derived straight from userland.
1338  */
1339 static int
rt_xaddrs(caddr_t cp,caddr_t cplim,struct rt_addrinfo * rtinfo)1340 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1341 {
1342 	struct sockaddr *sa;
1343 	int i;
1344 
1345 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1346 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1347 			continue;
1348 		sa = (struct sockaddr *)cp;
1349 		/*
1350 		 * It won't fit.
1351 		 */
1352 		if (cp + sa->sa_len > cplim) {
1353 			RTS_PID_LOG(LOG_DEBUG, "sa_len too big for sa type %d", i);
1354 			return (EINVAL);
1355 		}
1356 		/*
1357 		 * there are no more.. quit now
1358 		 * If there are more bits, they are in error.
1359 		 * I've seen this. route(1) can evidently generate these.
1360 		 * This causes kernel to core dump.
1361 		 * for compatibility, If we see this, point to a safe address.
1362 		 */
1363 		if (sa->sa_len == 0) {
1364 			rtinfo->rti_info[i] = &sa_zero;
1365 			return (0); /* should be EINVAL but for compat */
1366 		}
1367 		/* accept it */
1368 #ifdef INET6
1369 		if (sa->sa_family == AF_INET6)
1370 			sa6_embedscope((struct sockaddr_in6 *)sa,
1371 			    V_ip6_use_defzone);
1372 #endif
1373 		rtinfo->rti_info[i] = sa;
1374 		cp += SA_SIZE(sa);
1375 	}
1376 	return (0);
1377 }
1378 
1379 #ifdef INET
1380 static inline void
fill_sockaddr_inet(struct sockaddr_in * sin,struct in_addr addr)1381 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1382 {
1383 
1384 	const struct sockaddr_in nsin = {
1385 		.sin_family = AF_INET,
1386 		.sin_len = sizeof(struct sockaddr_in),
1387 		.sin_addr = addr,
1388 	};
1389 	*sin = nsin;
1390 }
1391 #endif
1392 
1393 #ifdef INET6
1394 static inline void
fill_sockaddr_inet6(struct sockaddr_in6 * sin6,const struct in6_addr * addr6,uint32_t scopeid)1395 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1396     uint32_t scopeid)
1397 {
1398 
1399 	const struct sockaddr_in6 nsin6 = {
1400 		.sin6_family = AF_INET6,
1401 		.sin6_len = sizeof(struct sockaddr_in6),
1402 		.sin6_addr = *addr6,
1403 		.sin6_scope_id = scopeid,
1404 	};
1405 	*sin6 = nsin6;
1406 }
1407 #endif
1408 
1409 #if defined(INET6) || defined(INET)
1410 /*
1411  * Checks if gateway is suitable for lltable operations.
1412  * Lltable code requires AF_LINK gateway with ifindex
1413  *  and mac address specified.
1414  * Returns 0 on success.
1415  */
1416 static int
cleanup_xaddrs_lladdr(struct rt_addrinfo * info)1417 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1418 {
1419 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1420 
1421 	if (sdl->sdl_family != AF_LINK)
1422 		return (EINVAL);
1423 
1424 	if (sdl->sdl_index == 0) {
1425 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gateway w/o ifindex");
1426 		return (EINVAL);
1427 	}
1428 
1429 	if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len) {
1430 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gw: sdl_nlen/sdl_alen too large");
1431 		return (EINVAL);
1432 	}
1433 
1434 	return (0);
1435 }
1436 
1437 static int
cleanup_xaddrs_gateway(struct rt_addrinfo * info,struct linear_buffer * lb)1438 cleanup_xaddrs_gateway(struct rt_addrinfo *info, struct linear_buffer *lb)
1439 {
1440 	struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1441 	struct sockaddr *sa;
1442 
1443 	if (info->rti_flags & RTF_LLDATA)
1444 		return (cleanup_xaddrs_lladdr(info));
1445 
1446 	switch (gw->sa_family) {
1447 #ifdef INET
1448 	case AF_INET:
1449 		{
1450 			struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw;
1451 
1452 			/* Ensure reads do not go beyoud SA boundary */
1453 			if (SA_SIZE(gw) < offsetof(struct sockaddr_in, sin_zero)) {
1454 				RTS_PID_LOG(LOG_DEBUG, "gateway sin_len too small: %d",
1455 				    gw->sa_len);
1456 				return (EINVAL);
1457 			}
1458 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_in));
1459 			if (sa == NULL)
1460 				return (ENOBUFS);
1461 			fill_sockaddr_inet((struct sockaddr_in *)sa, gw_sin->sin_addr);
1462 			info->rti_info[RTAX_GATEWAY] = sa;
1463 		}
1464 		break;
1465 #endif
1466 #ifdef INET6
1467 	case AF_INET6:
1468 		{
1469 			struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1470 			if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1471 				RTS_PID_LOG(LOG_DEBUG, "gateway sin6_len too small: %d",
1472 				    gw->sa_len);
1473 				return (EINVAL);
1474 			}
1475 			fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1476 			break;
1477 		}
1478 #endif
1479 	case AF_LINK:
1480 		{
1481 			struct sockaddr_dl *gw_sdl;
1482 
1483 			size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data);
1484 			gw_sdl = (struct sockaddr_dl *)gw;
1485 			if (gw_sdl->sdl_len < sdl_min_len) {
1486 				RTS_PID_LOG(LOG_DEBUG, "gateway sdl_len too small: %d",
1487 				    gw_sdl->sdl_len);
1488 				return (EINVAL);
1489 			}
1490 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_dl_short));
1491 			if (sa == NULL)
1492 				return (ENOBUFS);
1493 
1494 			const struct sockaddr_dl_short sdl = {
1495 				.sdl_family = AF_LINK,
1496 				.sdl_len = sizeof(struct sockaddr_dl_short),
1497 				.sdl_index = gw_sdl->sdl_index,
1498 			};
1499 			*((struct sockaddr_dl_short *)sa) = sdl;
1500 			info->rti_info[RTAX_GATEWAY] = sa;
1501 			break;
1502 		}
1503 	}
1504 
1505 	return (0);
1506 }
1507 #endif
1508 
1509 static void
remove_netmask(struct rt_addrinfo * info)1510 remove_netmask(struct rt_addrinfo *info)
1511 {
1512 	info->rti_info[RTAX_NETMASK] = NULL;
1513 	info->rti_flags |= RTF_HOST;
1514 	info->rti_addrs &= ~RTA_NETMASK;
1515 }
1516 
1517 #ifdef INET
1518 static int
cleanup_xaddrs_inet(struct rt_addrinfo * info,struct linear_buffer * lb)1519 cleanup_xaddrs_inet(struct rt_addrinfo *info, struct linear_buffer *lb)
1520 {
1521 	struct sockaddr_in *dst_sa, *mask_sa;
1522 	const int sa_len = sizeof(struct sockaddr_in);
1523 	struct in_addr dst, mask;
1524 
1525 	/* Check & fixup dst/netmask combination first */
1526 	dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1527 	mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1528 
1529 	/* Ensure reads do not go beyound the buffer size */
1530 	if (SA_SIZE(dst_sa) < offsetof(struct sockaddr_in, sin_zero)) {
1531 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin_len too small: %d",
1532 		    dst_sa->sin_len);
1533 		return (EINVAL);
1534 	}
1535 
1536 	if ((mask_sa != NULL) && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1537 		/*
1538 		 * Some older routing software encode mask length into the
1539 		 * sin_len, thus resulting in "truncated" sockaddr.
1540 		 */
1541 		int len = mask_sa->sin_len - offsetof(struct sockaddr_in, sin_addr);
1542 		if (len >= 0) {
1543 			mask.s_addr = 0;
1544 			if (len > sizeof(struct in_addr))
1545 				len = sizeof(struct in_addr);
1546 			memcpy(&mask, &mask_sa->sin_addr, len);
1547 		} else {
1548 			RTS_PID_LOG(LOG_DEBUG, "prefix mask sin_len too small: %d",
1549 			    mask_sa->sin_len);
1550 			return (EINVAL);
1551 		}
1552 	} else
1553 		mask.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST;
1554 
1555 	dst.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr));
1556 
1557 	/* Construct new "clean" dst/mask sockaddresses */
1558 	if ((dst_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1559 		return (ENOBUFS);
1560 	fill_sockaddr_inet(dst_sa, dst);
1561 	info->rti_info[RTAX_DST] = (struct sockaddr *)dst_sa;
1562 
1563 	if (mask.s_addr != INADDR_BROADCAST) {
1564 		if ((mask_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1565 			return (ENOBUFS);
1566 		fill_sockaddr_inet(mask_sa, mask);
1567 		info->rti_info[RTAX_NETMASK] = (struct sockaddr *)mask_sa;
1568 		info->rti_flags &= ~RTF_HOST;
1569 	} else
1570 		remove_netmask(info);
1571 
1572 	/* Check gateway */
1573 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1574 		return (cleanup_xaddrs_gateway(info, lb));
1575 
1576 	return (0);
1577 }
1578 #endif
1579 
1580 #ifdef INET6
1581 static int
cleanup_xaddrs_inet6(struct rt_addrinfo * info,struct linear_buffer * lb)1582 cleanup_xaddrs_inet6(struct rt_addrinfo *info, struct linear_buffer *lb)
1583 {
1584 	struct sockaddr *sa;
1585 	struct sockaddr_in6 *dst_sa, *mask_sa;
1586 	struct in6_addr mask, *dst;
1587 	const int sa_len = sizeof(struct sockaddr_in6);
1588 
1589 	/* Check & fixup dst/netmask combination first */
1590 	dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1591 	mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1592 
1593 	if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1594 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin6_len too small: %d",
1595 		    dst_sa->sin6_len);
1596 		return (EINVAL);
1597 	}
1598 
1599 	if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1600 		/*
1601 		 * Some older routing software encode mask length into the
1602 		 * sin6_len, thus resulting in "truncated" sockaddr.
1603 		 */
1604 		int len = mask_sa->sin6_len - offsetof(struct sockaddr_in6, sin6_addr);
1605 		if (len >= 0) {
1606 			bzero(&mask, sizeof(mask));
1607 			if (len > sizeof(struct in6_addr))
1608 				len = sizeof(struct in6_addr);
1609 			memcpy(&mask, &mask_sa->sin6_addr, len);
1610 		} else {
1611 			RTS_PID_LOG(LOG_DEBUG, "rtsock: prefix mask sin6_len too small: %d",
1612 			    mask_sa->sin6_len);
1613 			return (EINVAL);
1614 		}
1615 	} else
1616 		mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1617 
1618 	dst = &dst_sa->sin6_addr;
1619 	IN6_MASK_ADDR(dst, &mask);
1620 
1621 	if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1622 		return (ENOBUFS);
1623 	fill_sockaddr_inet6((struct sockaddr_in6 *)sa, dst, 0);
1624 	info->rti_info[RTAX_DST] = sa;
1625 
1626 	if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128)) {
1627 		if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1628 			return (ENOBUFS);
1629 		fill_sockaddr_inet6((struct sockaddr_in6 *)sa, &mask, 0);
1630 		info->rti_info[RTAX_NETMASK] = sa;
1631 		info->rti_flags &= ~RTF_HOST;
1632 	} else
1633 		remove_netmask(info);
1634 
1635 	/* Check gateway */
1636 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1637 		return (cleanup_xaddrs_gateway(info, lb));
1638 
1639 	return (0);
1640 }
1641 #endif
1642 
1643 static int
cleanup_xaddrs(struct rt_addrinfo * info,struct linear_buffer * lb)1644 cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb)
1645 {
1646 	int error = EAFNOSUPPORT;
1647 
1648 	if (info->rti_info[RTAX_DST] == NULL) {
1649 		RTS_PID_LOG(LOG_DEBUG, "prefix dst is not set");
1650 		return (EINVAL);
1651 	}
1652 
1653 	if (info->rti_flags & RTF_LLDATA) {
1654 		/*
1655 		 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1656 		 * prefix along with the actual address in RTA_DST.
1657 		 * Remove netmask to avoid unnecessary address masking.
1658 		 */
1659 		remove_netmask(info);
1660 	}
1661 
1662 	switch (info->rti_info[RTAX_DST]->sa_family) {
1663 #ifdef INET
1664 	case AF_INET:
1665 		error = cleanup_xaddrs_inet(info, lb);
1666 		break;
1667 #endif
1668 #ifdef INET6
1669 	case AF_INET6:
1670 		error = cleanup_xaddrs_inet6(info, lb);
1671 		break;
1672 #endif
1673 	}
1674 
1675 	return (error);
1676 }
1677 
1678 /*
1679  * Fill in @dmask with valid netmask leaving original @smask
1680  * intact. Mostly used with radix netmasks.
1681  */
1682 struct sockaddr *
rtsock_fix_netmask(const struct sockaddr * dst,const struct sockaddr * smask,struct sockaddr_storage * dmask)1683 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1684     struct sockaddr_storage *dmask)
1685 {
1686 	if (dst == NULL || smask == NULL)
1687 		return (NULL);
1688 
1689 	memset(dmask, 0, dst->sa_len);
1690 	memcpy(dmask, smask, smask->sa_len);
1691 	dmask->ss_len = dst->sa_len;
1692 	dmask->ss_family = dst->sa_family;
1693 
1694 	return ((struct sockaddr *)dmask);
1695 }
1696 
1697 /*
1698  * Writes information related to @rtinfo object to newly-allocated mbuf.
1699  * Assumes MCLBYTES is enough to construct any message.
1700  * Used for OS notifications of vaious events (if/ifa announces,etc)
1701  *
1702  * Returns allocated mbuf or NULL on failure.
1703  */
1704 static struct mbuf *
rtsock_msg_mbuf(int type,struct rt_addrinfo * rtinfo)1705 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1706 {
1707 	struct sockaddr_storage ss;
1708 	struct rt_msghdr *rtm;
1709 	struct mbuf *m;
1710 	int i;
1711 	struct sockaddr *sa;
1712 #ifdef INET6
1713 	struct sockaddr_in6 *sin6;
1714 #endif
1715 	int len, dlen;
1716 
1717 	switch (type) {
1718 	case RTM_DELADDR:
1719 	case RTM_NEWADDR:
1720 		len = sizeof(struct ifa_msghdr);
1721 		break;
1722 
1723 	case RTM_DELMADDR:
1724 	case RTM_NEWMADDR:
1725 		len = sizeof(struct ifma_msghdr);
1726 		break;
1727 
1728 	case RTM_IFINFO:
1729 		len = sizeof(struct if_msghdr);
1730 		break;
1731 
1732 	case RTM_IFANNOUNCE:
1733 	case RTM_IEEE80211:
1734 		len = sizeof(struct if_announcemsghdr);
1735 		break;
1736 
1737 	default:
1738 		len = sizeof(struct rt_msghdr);
1739 	}
1740 
1741 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1742 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1743 	if (len > MHLEN)
1744 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1745 	else
1746 		m = m_gethdr(M_NOWAIT, MT_DATA);
1747 	if (m == NULL)
1748 		return (m);
1749 
1750 	m->m_pkthdr.len = m->m_len = len;
1751 	rtm = mtod(m, struct rt_msghdr *);
1752 	bzero((caddr_t)rtm, len);
1753 	for (i = 0; i < RTAX_MAX; i++) {
1754 		if ((sa = rtinfo->rti_info[i]) == NULL)
1755 			continue;
1756 		rtinfo->rti_addrs |= (1 << i);
1757 
1758 		dlen = SA_SIZE(sa);
1759 		KASSERT(dlen <= sizeof(ss),
1760 		    ("%s: sockaddr size overflow", __func__));
1761 		bzero(&ss, sizeof(ss));
1762 		bcopy(sa, &ss, sa->sa_len);
1763 		sa = (struct sockaddr *)&ss;
1764 #ifdef INET6
1765 		if (sa->sa_family == AF_INET6) {
1766 			sin6 = (struct sockaddr_in6 *)sa;
1767 			(void)sa6_recoverscope(sin6);
1768 		}
1769 #endif
1770 		m_copyback(m, len, dlen, (caddr_t)sa);
1771 		len += dlen;
1772 	}
1773 	if (m->m_pkthdr.len != len) {
1774 		m_freem(m);
1775 		return (NULL);
1776 	}
1777 	rtm->rtm_msglen = len;
1778 	rtm->rtm_version = RTM_VERSION;
1779 	rtm->rtm_type = type;
1780 	return (m);
1781 }
1782 
1783 /*
1784  * Writes information related to @rtinfo object to preallocated buffer.
1785  * Stores needed size in @plen. If @w is NULL, calculates size without
1786  * writing.
1787  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1788  *
1789  * Returns 0 on success.
1790  *
1791  */
1792 static int
rtsock_msg_buffer(int type,struct rt_addrinfo * rtinfo,struct walkarg * w,int * plen)1793 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1794 {
1795 	struct sockaddr_storage ss;
1796 	int len, buflen = 0, dlen, i;
1797 	caddr_t cp = NULL;
1798 	struct rt_msghdr *rtm = NULL;
1799 #ifdef INET6
1800 	struct sockaddr_in6 *sin6;
1801 #endif
1802 #ifdef COMPAT_FREEBSD32
1803 	bool compat32;
1804 
1805 	compat32 = w != NULL && w->w_req != NULL &&
1806 	    (w->w_req->flags & SCTL_MASK32);
1807 #endif
1808 
1809 	switch (type) {
1810 	case RTM_DELADDR:
1811 	case RTM_NEWADDR:
1812 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1813 #ifdef COMPAT_FREEBSD32
1814 			if (compat32)
1815 				len = sizeof(struct ifa_msghdrl32);
1816 			else
1817 #endif
1818 				len = sizeof(struct ifa_msghdrl);
1819 		} else
1820 			len = sizeof(struct ifa_msghdr);
1821 		break;
1822 
1823 	case RTM_IFINFO:
1824 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1825 #ifdef COMPAT_FREEBSD32
1826 			if (compat32)
1827 				len = sizeof(struct if_msghdrl32);
1828 			else
1829 #endif
1830 				len = sizeof(struct if_msghdrl);
1831 		} else {
1832 #ifdef COMPAT_FREEBSD32
1833 			if (compat32)
1834 				len = sizeof(struct if_msghdr32);
1835 			else
1836 #endif
1837 				len = sizeof(struct if_msghdr);
1838 		}
1839 		break;
1840 
1841 	case RTM_NEWMADDR:
1842 		len = sizeof(struct ifma_msghdr);
1843 		break;
1844 
1845 	default:
1846 		len = sizeof(struct rt_msghdr);
1847 	}
1848 
1849 	if (w != NULL) {
1850 		rtm = (struct rt_msghdr *)w->w_tmem;
1851 		buflen = w->w_tmemsize - len;
1852 		cp = (caddr_t)w->w_tmem + len;
1853 	}
1854 
1855 	rtinfo->rti_addrs = 0;
1856 	for (i = 0; i < RTAX_MAX; i++) {
1857 		struct sockaddr *sa;
1858 
1859 		if ((sa = rtinfo->rti_info[i]) == NULL)
1860 			continue;
1861 		rtinfo->rti_addrs |= (1 << i);
1862 #ifdef COMPAT_FREEBSD32
1863 		if (compat32)
1864 			dlen = SA_SIZE32(sa);
1865 		else
1866 #endif
1867 			dlen = SA_SIZE(sa);
1868 		if (cp != NULL && buflen >= dlen) {
1869 			KASSERT(dlen <= sizeof(ss),
1870 			    ("%s: sockaddr size overflow", __func__));
1871 			bzero(&ss, sizeof(ss));
1872 			bcopy(sa, &ss, sa->sa_len);
1873 			sa = (struct sockaddr *)&ss;
1874 #ifdef INET6
1875 			if (sa->sa_family == AF_INET6) {
1876 				sin6 = (struct sockaddr_in6 *)sa;
1877 				(void)sa6_recoverscope(sin6);
1878 			}
1879 #endif
1880 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1881 			cp += dlen;
1882 			buflen -= dlen;
1883 		} else if (cp != NULL) {
1884 			/*
1885 			 * Buffer too small. Count needed size
1886 			 * and return with error.
1887 			 */
1888 			cp = NULL;
1889 		}
1890 
1891 		len += dlen;
1892 	}
1893 
1894 	if (cp != NULL) {
1895 		dlen = ALIGN(len) - len;
1896 		if (buflen < dlen)
1897 			cp = NULL;
1898 		else {
1899 			bzero(cp, dlen);
1900 			cp += dlen;
1901 			buflen -= dlen;
1902 		}
1903 	}
1904 	len = ALIGN(len);
1905 
1906 	if (cp != NULL) {
1907 		/* fill header iff buffer is large enough */
1908 		rtm->rtm_version = RTM_VERSION;
1909 		rtm->rtm_type = type;
1910 		rtm->rtm_msglen = len;
1911 	}
1912 
1913 	*plen = len;
1914 
1915 	if (w != NULL && cp == NULL)
1916 		return (ENOBUFS);
1917 
1918 	return (0);
1919 }
1920 
1921 /*
1922  * This routine is called to generate a message from the routing
1923  * socket indicating that a redirect has occurred, a routing lookup
1924  * has failed, or that a protocol has detected timeouts to a particular
1925  * destination.
1926  */
1927 void
rt_missmsg_fib(int type,struct rt_addrinfo * rtinfo,int flags,int error,int fibnum)1928 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1929     int fibnum)
1930 {
1931 	struct rt_msghdr *rtm;
1932 	struct mbuf *m;
1933 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1934 
1935 	if (V_route_cb.any_count == 0)
1936 		return;
1937 	m = rtsock_msg_mbuf(type, rtinfo);
1938 	if (m == NULL)
1939 		return;
1940 
1941 	if (fibnum != RT_ALL_FIBS) {
1942 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1943 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1944 		M_SETFIB(m, fibnum);
1945 		m->m_flags |= RTS_FILTER_FIB;
1946 	}
1947 
1948 	rtm = mtod(m, struct rt_msghdr *);
1949 	rtm->rtm_flags = RTF_DONE | flags;
1950 	rtm->rtm_errno = error;
1951 	rtm->rtm_addrs = rtinfo->rti_addrs;
1952 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1953 }
1954 
1955 void
rt_missmsg(int type,struct rt_addrinfo * rtinfo,int flags,int error)1956 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1957 {
1958 
1959 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1960 }
1961 
1962 /*
1963  * This routine is called to generate a message from the routing
1964  * socket indicating that the status of a network interface has changed.
1965  */
1966 static void
rtsock_ifmsg(struct ifnet * ifp,int if_flags_mask __unused)1967 rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask __unused)
1968 {
1969 	struct if_msghdr *ifm;
1970 	struct mbuf *m;
1971 	struct rt_addrinfo info;
1972 
1973 	if (V_route_cb.any_count == 0)
1974 		return;
1975 	bzero((caddr_t)&info, sizeof(info));
1976 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1977 	if (m == NULL)
1978 		return;
1979 	ifm = mtod(m, struct if_msghdr *);
1980 	ifm->ifm_index = ifp->if_index;
1981 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1982 	if_data_copy(ifp, &ifm->ifm_data);
1983 	ifm->ifm_addrs = 0;
1984 	rt_dispatch(m, AF_UNSPEC);
1985 }
1986 
1987 /*
1988  * Announce interface address arrival/withdraw.
1989  * Please do not call directly, use rt_addrmsg().
1990  * Assume input data to be valid.
1991  * Returns 0 on success.
1992  */
1993 int
rtsock_addrmsg(int cmd,struct ifaddr * ifa,int fibnum)1994 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1995 {
1996 	struct rt_addrinfo info;
1997 	struct sockaddr *sa;
1998 	int ncmd;
1999 	struct mbuf *m;
2000 	struct ifa_msghdr *ifam;
2001 	struct ifnet *ifp = ifa->ifa_ifp;
2002 	struct sockaddr_storage ss;
2003 
2004 	if (V_route_cb.any_count == 0)
2005 		return (0);
2006 
2007 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
2008 
2009 	bzero((caddr_t)&info, sizeof(info));
2010 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
2011 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2012 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2013 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
2014 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2015 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
2016 		return (ENOBUFS);
2017 	ifam = mtod(m, struct ifa_msghdr *);
2018 	ifam->ifam_index = ifp->if_index;
2019 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2020 	ifam->ifam_flags = ifa->ifa_flags;
2021 	ifam->ifam_addrs = info.rti_addrs;
2022 
2023 	if (fibnum != RT_ALL_FIBS) {
2024 		M_SETFIB(m, fibnum);
2025 		m->m_flags |= RTS_FILTER_FIB;
2026 	}
2027 
2028 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2029 
2030 	return (0);
2031 }
2032 
2033 /*
2034  * Announce route addition/removal to rtsock based on @rt data.
2035  * Callers are advives to use rt_routemsg() instead of using this
2036  *  function directly.
2037  * Assume @rt data is consistent.
2038  *
2039  * Returns 0 on success.
2040  */
2041 int
rtsock_routemsg(int cmd,struct rtentry * rt,struct nhop_object * nh,int fibnum)2042 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
2043     int fibnum)
2044 {
2045 	union sockaddr_union dst, mask;
2046 	struct rt_addrinfo info;
2047 
2048 	if (V_route_cb.any_count == 0)
2049 		return (0);
2050 
2051 	int family = rt_get_family(rt);
2052 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
2053 	export_rtaddrs(rt, &dst.sa, &mask.sa);
2054 
2055 	bzero((caddr_t)&info, sizeof(info));
2056 	info.rti_info[RTAX_DST] = &dst.sa;
2057 	info.rti_info[RTAX_NETMASK] = &mask.sa;
2058 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2059 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
2060 	info.rti_ifp = nh->nh_ifp;
2061 
2062 	return (rtsock_routemsg_info(cmd, &info, fibnum));
2063 }
2064 
2065 int
rtsock_routemsg_info(int cmd,struct rt_addrinfo * info,int fibnum)2066 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
2067 {
2068 	struct rt_msghdr *rtm;
2069 	struct sockaddr *sa;
2070 	struct mbuf *m;
2071 
2072 	if (V_route_cb.any_count == 0)
2073 		return (0);
2074 
2075 	if (info->rti_flags & RTF_HOST)
2076 		info->rti_info[RTAX_NETMASK] = NULL;
2077 
2078 	m = rtsock_msg_mbuf(cmd, info);
2079 	if (m == NULL)
2080 		return (ENOBUFS);
2081 
2082 	if (fibnum != RT_ALL_FIBS) {
2083 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
2084 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
2085 		M_SETFIB(m, fibnum);
2086 		m->m_flags |= RTS_FILTER_FIB;
2087 	}
2088 
2089 	rtm = mtod(m, struct rt_msghdr *);
2090 	rtm->rtm_addrs = info->rti_addrs;
2091 	if (info->rti_ifp != NULL)
2092 		rtm->rtm_index = info->rti_ifp->if_index;
2093 	/* Add RTF_DONE to indicate command 'completion' required by API */
2094 	info->rti_flags |= RTF_DONE;
2095 	/* Reported routes has to be up */
2096 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
2097 		info->rti_flags |= RTF_UP;
2098 	rtm->rtm_flags = info->rti_flags;
2099 
2100 	sa = info->rti_info[RTAX_DST];
2101 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2102 
2103 	return (0);
2104 }
2105 
2106 /*
2107  * This is the analogue to the rt_newaddrmsg which performs the same
2108  * function but for multicast group memberhips.  This is easier since
2109  * there is no route state to worry about.
2110  */
2111 void
rt_newmaddrmsg(int cmd,struct ifmultiaddr * ifma)2112 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
2113 {
2114 	struct rt_addrinfo info;
2115 	struct mbuf *m = NULL;
2116 	struct ifnet *ifp = ifma->ifma_ifp;
2117 	struct ifma_msghdr *ifmam;
2118 
2119 	if (V_route_cb.any_count == 0)
2120 		return;
2121 
2122 	bzero((caddr_t)&info, sizeof(info));
2123 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2124 	if (ifp && ifp->if_addr)
2125 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2126 	else
2127 		info.rti_info[RTAX_IFP] = NULL;
2128 	/*
2129 	 * If a link-layer address is present, present it as a ``gateway''
2130 	 * (similarly to how ARP entries, e.g., are presented).
2131 	 */
2132 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
2133 	m = rtsock_msg_mbuf(cmd, &info);
2134 	if (m == NULL)
2135 		return;
2136 	ifmam = mtod(m, struct ifma_msghdr *);
2137 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
2138 	    __func__));
2139 	ifmam->ifmam_index = ifp->if_index;
2140 	ifmam->ifmam_addrs = info.rti_addrs;
2141 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
2142 }
2143 
2144 static struct mbuf *
rt_makeifannouncemsg(struct ifnet * ifp,int type,int what,struct rt_addrinfo * info)2145 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
2146 	struct rt_addrinfo *info)
2147 {
2148 	struct if_announcemsghdr *ifan;
2149 	struct mbuf *m;
2150 
2151 	if (V_route_cb.any_count == 0)
2152 		return NULL;
2153 	bzero((caddr_t)info, sizeof(*info));
2154 	m = rtsock_msg_mbuf(type, info);
2155 	if (m != NULL) {
2156 		ifan = mtod(m, struct if_announcemsghdr *);
2157 		ifan->ifan_index = ifp->if_index;
2158 		strlcpy(ifan->ifan_name, ifp->if_xname,
2159 			sizeof(ifan->ifan_name));
2160 		ifan->ifan_what = what;
2161 	}
2162 	return m;
2163 }
2164 
2165 /*
2166  * This is called to generate routing socket messages indicating
2167  * IEEE80211 wireless events.
2168  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2169  */
2170 void
rt_ieee80211msg(struct ifnet * ifp,int what,void * data,size_t data_len)2171 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2172 {
2173 	struct mbuf *m;
2174 	struct rt_addrinfo info;
2175 
2176 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2177 	if (m != NULL) {
2178 		/*
2179 		 * Append the ieee80211 data.  Try to stick it in the
2180 		 * mbuf containing the ifannounce msg; otherwise allocate
2181 		 * a new mbuf and append.
2182 		 *
2183 		 * NB: we assume m is a single mbuf.
2184 		 */
2185 		if (data_len > M_TRAILINGSPACE(m)) {
2186 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2187 			if (n == NULL) {
2188 				m_freem(m);
2189 				return;
2190 			}
2191 			bcopy(data, mtod(n, void *), data_len);
2192 			n->m_len = data_len;
2193 			m->m_next = n;
2194 		} else if (data_len > 0) {
2195 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2196 			m->m_len += data_len;
2197 		}
2198 		if (m->m_flags & M_PKTHDR)
2199 			m->m_pkthdr.len += data_len;
2200 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2201 		rt_dispatch(m, AF_UNSPEC);
2202 	}
2203 }
2204 
2205 /*
2206  * This is called to generate routing socket messages indicating
2207  * network interface arrival and departure.
2208  */
2209 static void
rt_ifannouncemsg(struct ifnet * ifp,int what)2210 rt_ifannouncemsg(struct ifnet *ifp, int what)
2211 {
2212 	struct mbuf *m;
2213 	struct rt_addrinfo info;
2214 
2215 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2216 	if (m != NULL)
2217 		rt_dispatch(m, AF_UNSPEC);
2218 }
2219 
2220 static void
rt_dispatch(struct mbuf * m,sa_family_t saf)2221 rt_dispatch(struct mbuf *m, sa_family_t saf)
2222 {
2223 
2224 	M_ASSERTPKTHDR(m);
2225 
2226 	m->m_rtsock_family = saf;
2227 	if (V_loif)
2228 		m->m_pkthdr.rcvif = V_loif;
2229 	else {
2230 		m_freem(m);
2231 		return;
2232 	}
2233 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
2234 }
2235 
2236 /*
2237  * This is used in dumping the kernel table via sysctl().
2238  */
2239 static int
sysctl_dumpentry(struct rtentry * rt,void * vw)2240 sysctl_dumpentry(struct rtentry *rt, void *vw)
2241 {
2242 	struct walkarg *w = vw;
2243 	struct nhop_object *nh;
2244 
2245 	NET_EPOCH_ASSERT();
2246 
2247 	if (!rt_is_exportable(rt, w->w_req->td->td_ucred))
2248 		return (0);
2249 
2250 	export_rtaddrs(rt, w->dst, w->mask);
2251 	nh = rt_get_raw_nhop(rt);
2252 #ifdef ROUTE_MPATH
2253 	if (NH_IS_NHGRP(nh)) {
2254 		const struct weightened_nhop *wn;
2255 		uint32_t num_nhops;
2256 		int error;
2257 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2258 		for (int i = 0; i < num_nhops; i++) {
2259 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2260 			if (error != 0)
2261 				return (error);
2262 		}
2263 	} else
2264 #endif
2265 		sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2266 
2267 	return (0);
2268 }
2269 
2270 
2271 static int
sysctl_dumpnhop(struct rtentry * rt,struct nhop_object * nh,uint32_t weight,struct walkarg * w)2272 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2273     struct walkarg *w)
2274 {
2275 	struct rt_addrinfo info;
2276 	int error = 0, size;
2277 	uint32_t rtflags;
2278 
2279 	rtflags = nhop_get_rtflags(nh);
2280 
2281 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2282 		return (0);
2283 
2284 	bzero((caddr_t)&info, sizeof(info));
2285 	info.rti_info[RTAX_DST] = w->dst;
2286 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2287 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2288 	info.rti_info[RTAX_GENMASK] = 0;
2289 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2290 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2291 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2292 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2293 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2294 	}
2295 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2296 		return (error);
2297 	if (w->w_req && w->w_tmem) {
2298 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2299 
2300 		bzero(&rtm->rtm_index,
2301 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2302 
2303 		/*
2304 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2305 		 * and RTF_UP (if entry is linked, which is always true here).
2306 		 * Given that, use nhop rtflags & add RTF_UP.
2307 		 */
2308 		rtm->rtm_flags = rtflags | RTF_UP;
2309 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2310 			rtm->rtm_flags = RTF_GATEWAY |
2311 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2312 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2313 		rtm->rtm_rmx.rmx_weight = weight;
2314 		rtm->rtm_index = nh->nh_ifp->if_index;
2315 		rtm->rtm_addrs = info.rti_addrs;
2316 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2317 		return (error);
2318 	}
2319 	return (error);
2320 }
2321 
2322 static int
sysctl_iflist_ifml(struct ifnet * ifp,const struct if_data * src_ifd,struct rt_addrinfo * info,struct walkarg * w,int len)2323 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2324     struct rt_addrinfo *info, struct walkarg *w, int len)
2325 {
2326 	struct if_msghdrl *ifm;
2327 	struct if_data *ifd;
2328 
2329 	ifm = (struct if_msghdrl *)w->w_tmem;
2330 
2331 #ifdef COMPAT_FREEBSD32
2332 	if (w->w_req->flags & SCTL_MASK32) {
2333 		struct if_msghdrl32 *ifm32;
2334 
2335 		ifm32 = (struct if_msghdrl32 *)ifm;
2336 		ifm32->ifm_addrs = info->rti_addrs;
2337 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2338 		ifm32->ifm_index = ifp->if_index;
2339 		ifm32->_ifm_spare1 = 0;
2340 		ifm32->ifm_len = sizeof(*ifm32);
2341 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2342 		ifm32->_ifm_spare2 = 0;
2343 		ifd = &ifm32->ifm_data;
2344 	} else
2345 #endif
2346 	{
2347 		ifm->ifm_addrs = info->rti_addrs;
2348 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2349 		ifm->ifm_index = ifp->if_index;
2350 		ifm->_ifm_spare1 = 0;
2351 		ifm->ifm_len = sizeof(*ifm);
2352 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2353 		ifm->_ifm_spare2 = 0;
2354 		ifd = &ifm->ifm_data;
2355 	}
2356 
2357 	memcpy(ifd, src_ifd, sizeof(*ifd));
2358 
2359 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2360 }
2361 
2362 static int
sysctl_iflist_ifm(struct ifnet * ifp,const struct if_data * src_ifd,struct rt_addrinfo * info,struct walkarg * w,int len)2363 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2364     struct rt_addrinfo *info, struct walkarg *w, int len)
2365 {
2366 	struct if_msghdr *ifm;
2367 	struct if_data *ifd;
2368 
2369 	ifm = (struct if_msghdr *)w->w_tmem;
2370 
2371 #ifdef COMPAT_FREEBSD32
2372 	if (w->w_req->flags & SCTL_MASK32) {
2373 		struct if_msghdr32 *ifm32;
2374 
2375 		ifm32 = (struct if_msghdr32 *)ifm;
2376 		ifm32->ifm_addrs = info->rti_addrs;
2377 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2378 		ifm32->ifm_index = ifp->if_index;
2379 		ifm32->_ifm_spare1 = 0;
2380 		ifd = &ifm32->ifm_data;
2381 	} else
2382 #endif
2383 	{
2384 		ifm->ifm_addrs = info->rti_addrs;
2385 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2386 		ifm->ifm_index = ifp->if_index;
2387 		ifm->_ifm_spare1 = 0;
2388 		ifd = &ifm->ifm_data;
2389 	}
2390 
2391 	memcpy(ifd, src_ifd, sizeof(*ifd));
2392 
2393 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2394 }
2395 
2396 static int
sysctl_iflist_ifaml(struct ifaddr * ifa,struct rt_addrinfo * info,struct walkarg * w,int len)2397 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2398     struct walkarg *w, int len)
2399 {
2400 	struct ifa_msghdrl *ifam;
2401 	struct if_data *ifd;
2402 
2403 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2404 
2405 #ifdef COMPAT_FREEBSD32
2406 	if (w->w_req->flags & SCTL_MASK32) {
2407 		struct ifa_msghdrl32 *ifam32;
2408 
2409 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2410 		ifam32->ifam_addrs = info->rti_addrs;
2411 		ifam32->ifam_flags = ifa->ifa_flags;
2412 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2413 		ifam32->_ifam_spare1 = 0;
2414 		ifam32->ifam_len = sizeof(*ifam32);
2415 		ifam32->ifam_data_off =
2416 		    offsetof(struct ifa_msghdrl32, ifam_data);
2417 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2418 		ifd = &ifam32->ifam_data;
2419 	} else
2420 #endif
2421 	{
2422 		ifam->ifam_addrs = info->rti_addrs;
2423 		ifam->ifam_flags = ifa->ifa_flags;
2424 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2425 		ifam->_ifam_spare1 = 0;
2426 		ifam->ifam_len = sizeof(*ifam);
2427 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2428 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2429 		ifd = &ifam->ifam_data;
2430 	}
2431 
2432 	bzero(ifd, sizeof(*ifd));
2433 	ifd->ifi_datalen = sizeof(struct if_data);
2434 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2435 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2436 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2437 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2438 
2439 	/* Fixup if_data carp(4) vhid. */
2440 	if (carp_get_vhid_p != NULL)
2441 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2442 
2443 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2444 }
2445 
2446 static int
sysctl_iflist_ifam(struct ifaddr * ifa,struct rt_addrinfo * info,struct walkarg * w,int len)2447 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2448     struct walkarg *w, int len)
2449 {
2450 	struct ifa_msghdr *ifam;
2451 
2452 	ifam = (struct ifa_msghdr *)w->w_tmem;
2453 	ifam->ifam_addrs = info->rti_addrs;
2454 	ifam->ifam_flags = ifa->ifa_flags;
2455 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2456 	ifam->_ifam_spare1 = 0;
2457 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2458 
2459 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2460 }
2461 
2462 static int
sysctl_iflist(int af,struct walkarg * w)2463 sysctl_iflist(int af, struct walkarg *w)
2464 {
2465 	struct ifnet *ifp;
2466 	struct ifaddr *ifa;
2467 	struct if_data ifd;
2468 	struct rt_addrinfo info;
2469 	int len, error = 0;
2470 	struct sockaddr_storage ss;
2471 
2472 	bzero((caddr_t)&info, sizeof(info));
2473 	bzero(&ifd, sizeof(ifd));
2474 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2475 		if (w->w_arg && w->w_arg != ifp->if_index)
2476 			continue;
2477 		if_data_copy(ifp, &ifd);
2478 		ifa = ifp->if_addr;
2479 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2480 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2481 		if (error != 0)
2482 			goto done;
2483 		info.rti_info[RTAX_IFP] = NULL;
2484 		if (w->w_req && w->w_tmem) {
2485 			if (w->w_op == NET_RT_IFLISTL)
2486 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2487 				    len);
2488 			else
2489 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2490 				    len);
2491 			if (error)
2492 				goto done;
2493 		}
2494 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2495 			if (af && af != ifa->ifa_addr->sa_family)
2496 				continue;
2497 			if (prison_if(w->w_req->td->td_ucred,
2498 			    ifa->ifa_addr) != 0)
2499 				continue;
2500 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2501 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2502 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2503 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2504 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2505 			if (error != 0)
2506 				goto done;
2507 			if (w->w_req && w->w_tmem) {
2508 				if (w->w_op == NET_RT_IFLISTL)
2509 					error = sysctl_iflist_ifaml(ifa, &info,
2510 					    w, len);
2511 				else
2512 					error = sysctl_iflist_ifam(ifa, &info,
2513 					    w, len);
2514 				if (error)
2515 					goto done;
2516 			}
2517 		}
2518 		info.rti_info[RTAX_IFA] = NULL;
2519 		info.rti_info[RTAX_NETMASK] = NULL;
2520 		info.rti_info[RTAX_BRD] = NULL;
2521 	}
2522 done:
2523 	return (error);
2524 }
2525 
2526 static int
sysctl_ifmalist(int af,struct walkarg * w)2527 sysctl_ifmalist(int af, struct walkarg *w)
2528 {
2529 	struct rt_addrinfo info;
2530 	struct ifaddr *ifa;
2531 	struct ifmultiaddr *ifma;
2532 	struct ifnet *ifp;
2533 	int error, len;
2534 
2535 	NET_EPOCH_ASSERT();
2536 
2537 	error = 0;
2538 	bzero((caddr_t)&info, sizeof(info));
2539 
2540 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2541 		if (w->w_arg && w->w_arg != ifp->if_index)
2542 			continue;
2543 		ifa = ifp->if_addr;
2544 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2545 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2546 			if (af && af != ifma->ifma_addr->sa_family)
2547 				continue;
2548 			if (prison_if(w->w_req->td->td_ucred,
2549 			    ifma->ifma_addr) != 0)
2550 				continue;
2551 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2552 			info.rti_info[RTAX_GATEWAY] =
2553 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2554 			    ifma->ifma_lladdr : NULL;
2555 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2556 			if (error != 0)
2557 				break;
2558 			if (w->w_req && w->w_tmem) {
2559 				struct ifma_msghdr *ifmam;
2560 
2561 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2562 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2563 				ifmam->ifmam_flags = 0;
2564 				ifmam->ifmam_addrs = info.rti_addrs;
2565 				ifmam->_ifmam_spare1 = 0;
2566 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2567 				if (error != 0)
2568 					break;
2569 			}
2570 		}
2571 		if (error != 0)
2572 			break;
2573 	}
2574 	return (error);
2575 }
2576 
2577 static void
rtable_sysctl_dump(uint32_t fibnum,int family,struct walkarg * w)2578 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2579 {
2580 	union sockaddr_union sa_dst, sa_mask;
2581 
2582 	w->family = family;
2583 	w->dst = (struct sockaddr *)&sa_dst;
2584 	w->mask = (struct sockaddr *)&sa_mask;
2585 
2586 	init_sockaddrs_family(family, w->dst, w->mask);
2587 
2588 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2589 }
2590 
2591 static int
sysctl_rtsock(SYSCTL_HANDLER_ARGS)2592 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2593 {
2594 	struct epoch_tracker et;
2595 	int	*name = (int *)arg1;
2596 	u_int	namelen = arg2;
2597 	struct rib_head *rnh = NULL; /* silence compiler. */
2598 	int	i, lim, error = EINVAL;
2599 	int	fib = 0;
2600 	u_char	af;
2601 	struct	walkarg w;
2602 
2603 	if (namelen < 3)
2604 		return (EINVAL);
2605 
2606 	name++;
2607 	namelen--;
2608 	if (req->newptr)
2609 		return (EPERM);
2610 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2611 		if (namelen == 3)
2612 			fib = req->td->td_proc->p_fibnum;
2613 		else if (namelen == 4)
2614 			fib = (name[3] == RT_ALL_FIBS) ?
2615 			    req->td->td_proc->p_fibnum : name[3];
2616 		else
2617 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2618 		if (fib < 0 || fib >= rt_numfibs)
2619 			return (EINVAL);
2620 	} else if (namelen != 3)
2621 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2622 	af = name[0];
2623 	if (af > AF_MAX)
2624 		return (EINVAL);
2625 	bzero(&w, sizeof(w));
2626 	w.w_op = name[1];
2627 	w.w_arg = name[2];
2628 	w.w_req = req;
2629 
2630 	error = sysctl_wire_old_buffer(req, 0);
2631 	if (error)
2632 		return (error);
2633 
2634 	/*
2635 	 * Allocate reply buffer in advance.
2636 	 * All rtsock messages has maximum length of u_short.
2637 	 */
2638 	w.w_tmemsize = 65536;
2639 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2640 
2641 	NET_EPOCH_ENTER(et);
2642 	switch (w.w_op) {
2643 	case NET_RT_DUMP:
2644 	case NET_RT_FLAGS:
2645 		if (af == 0) {			/* dump all tables */
2646 			i = 1;
2647 			lim = AF_MAX;
2648 		} else				/* dump only one table */
2649 			i = lim = af;
2650 
2651 		/*
2652 		 * take care of llinfo entries, the caller must
2653 		 * specify an AF
2654 		 */
2655 		if (w.w_op == NET_RT_FLAGS &&
2656 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2657 			if (af != 0)
2658 				error = lltable_sysctl_dumparp(af, w.w_req);
2659 			else
2660 				error = EINVAL;
2661 			break;
2662 		}
2663 		/*
2664 		 * take care of routing entries
2665 		 */
2666 		for (error = 0; error == 0 && i <= lim; i++) {
2667 			rnh = rt_tables_get_rnh(fib, i);
2668 			if (rnh != NULL) {
2669 				rtable_sysctl_dump(fib, i, &w);
2670 			} else if (af != 0)
2671 				error = EAFNOSUPPORT;
2672 		}
2673 		break;
2674 	case NET_RT_NHOP:
2675 	case NET_RT_NHGRP:
2676 		/* Allow dumping one specific af/fib at a time */
2677 		if (namelen < 4) {
2678 			error = EINVAL;
2679 			break;
2680 		}
2681 		fib = name[3];
2682 		if (fib < 0 || fib > rt_numfibs) {
2683 			error = EINVAL;
2684 			break;
2685 		}
2686 		rnh = rt_tables_get_rnh(fib, af);
2687 		if (rnh == NULL) {
2688 			error = EAFNOSUPPORT;
2689 			break;
2690 		}
2691 		if (w.w_op == NET_RT_NHOP)
2692 			error = nhops_dump_sysctl(rnh, w.w_req);
2693 		else
2694 #ifdef ROUTE_MPATH
2695 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2696 #else
2697 			error = ENOTSUP;
2698 #endif
2699 		break;
2700 	case NET_RT_IFLIST:
2701 	case NET_RT_IFLISTL:
2702 		error = sysctl_iflist(af, &w);
2703 		break;
2704 
2705 	case NET_RT_IFMALIST:
2706 		error = sysctl_ifmalist(af, &w);
2707 		break;
2708 	}
2709 	NET_EPOCH_EXIT(et);
2710 
2711 	free(w.w_tmem, M_TEMP);
2712 	return (error);
2713 }
2714 
2715 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2716     sysctl_rtsock, "Return route tables and interface/address lists");
2717 
2718 /*
2719  * Definitions of protocols supported in the ROUTE domain.
2720  */
2721 
2722 static struct domain routedomain;		/* or at least forward */
2723 
2724 static struct protosw routesw = {
2725 	.pr_type =		SOCK_RAW,
2726 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2727 	.pr_abort =		rts_close,
2728 	.pr_attach =		rts_attach,
2729 	.pr_ctloutput =		rts_ctloutput,
2730 	.pr_detach =		rts_detach,
2731 	.pr_send =		rts_send,
2732 	.pr_shutdown =		rts_shutdown,
2733 	.pr_disconnect =	rts_disconnect,
2734 	.pr_close =		rts_close,
2735 };
2736 
2737 static struct domain routedomain = {
2738 	.dom_family =		PF_ROUTE,
2739 	.dom_name =		"route",
2740 	.dom_nprotosw =		1,
2741 	.dom_protosw =		{ &routesw },
2742 };
2743 
2744 DOMAIN_SET(route);
2745