xref: /freebsd-14.2/sys/net/rtsock.c (revision 3985e96a)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1988, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
32  */
33 #include "opt_ddb.h"
34 #include "opt_route.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/jail.h>
40 #include <sys/kernel.h>
41 #include <sys/eventhandler.h>
42 #include <sys/domain.h>
43 #include <sys/lock.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/priv.h>
47 #include <sys/proc.h>
48 #include <sys/protosw.h>
49 #include <sys/rmlock.h>
50 #include <sys/rwlock.h>
51 #include <sys/signalvar.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/sysctl.h>
55 #include <sys/systm.h>
56 
57 #include <net/if.h>
58 #include <net/if_var.h>
59 #include <net/if_private.h>
60 #include <net/if_dl.h>
61 #include <net/if_llatbl.h>
62 #include <net/if_types.h>
63 #include <net/netisr.h>
64 #include <net/route.h>
65 #include <net/route/route_ctl.h>
66 #include <net/route/route_var.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/if_ether.h>
71 #include <netinet/ip_carp.h>
72 #ifdef INET6
73 #include <netinet6/in6_var.h>
74 #include <netinet6/ip6_var.h>
75 #include <netinet6/scope6_var.h>
76 #endif
77 #include <net/route/nhop.h>
78 
79 #define	DEBUG_MOD_NAME	rtsock
80 #define	DEBUG_MAX_LEVEL	LOG_DEBUG
81 #include <net/route/route_debug.h>
82 _DECLARE_DEBUG(LOG_INFO);
83 
84 #ifdef COMPAT_FREEBSD32
85 #include <sys/mount.h>
86 #include <compat/freebsd32/freebsd32.h>
87 
88 struct if_msghdr32 {
89 	uint16_t ifm_msglen;
90 	uint8_t	ifm_version;
91 	uint8_t	ifm_type;
92 	int32_t	ifm_addrs;
93 	int32_t	ifm_flags;
94 	uint16_t ifm_index;
95 	uint16_t _ifm_spare1;
96 	struct	if_data ifm_data;
97 };
98 
99 struct if_msghdrl32 {
100 	uint16_t ifm_msglen;
101 	uint8_t	ifm_version;
102 	uint8_t	ifm_type;
103 	int32_t	ifm_addrs;
104 	int32_t	ifm_flags;
105 	uint16_t ifm_index;
106 	uint16_t _ifm_spare1;
107 	uint16_t ifm_len;
108 	uint16_t ifm_data_off;
109 	uint32_t _ifm_spare2;
110 	struct	if_data ifm_data;
111 };
112 
113 struct ifa_msghdrl32 {
114 	uint16_t ifam_msglen;
115 	uint8_t	ifam_version;
116 	uint8_t	ifam_type;
117 	int32_t	ifam_addrs;
118 	int32_t	ifam_flags;
119 	uint16_t ifam_index;
120 	uint16_t _ifam_spare1;
121 	uint16_t ifam_len;
122 	uint16_t ifam_data_off;
123 	int32_t	ifam_metric;
124 	struct	if_data ifam_data;
125 };
126 
127 #define SA_SIZE32(sa)						\
128     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
129 	sizeof(int)		:				\
130 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
131 
132 #endif /* COMPAT_FREEBSD32 */
133 
134 struct linear_buffer {
135 	char		*base;	/* Base allocated memory pointer */
136 	uint32_t	offset;	/* Currently used offset */
137 	uint32_t	size;	/* Total buffer size */
138 };
139 #define	SCRATCH_BUFFER_SIZE	1024
140 
141 #define	RTS_PID_LOG(_l, _fmt, ...)					\
142 	RT_LOG_##_l(_l, "PID %d: " _fmt, curproc ? curproc->p_pid : 0,	\
143 	    ## __VA_ARGS__)
144 
145 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
146 
147 /* NB: these are not modified */
148 static struct	sockaddr route_src = { 2, PF_ROUTE, };
149 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
150 
151 /* These are external hooks for CARP. */
152 int	(*carp_get_vhid_p)(struct ifaddr *);
153 
154 /*
155  * Used by rtsock callback code to decide whether to filter the update
156  * notification to a socket bound to a particular FIB.
157  */
158 #define	RTS_FILTER_FIB	M_PROTO8
159 /*
160  * Used to store address family of the notification.
161  */
162 #define	m_rtsock_family	m_pkthdr.PH_loc.eight[0]
163 
164 struct rcb {
165 	LIST_ENTRY(rcb) list;
166 	struct socket	*rcb_socket;
167 	sa_family_t	rcb_family;
168 };
169 
170 typedef struct {
171 	LIST_HEAD(, rcb)	cblist;
172 	int	ip_count;	/* attached w/ AF_INET */
173 	int	ip6_count;	/* attached w/ AF_INET6 */
174 	int	any_count;	/* total attached */
175 } route_cb_t;
176 VNET_DEFINE_STATIC(route_cb_t, route_cb);
177 #define	V_route_cb VNET(route_cb)
178 
179 struct mtx rtsock_mtx;
180 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
181 
182 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
183 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
184 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
185 
186 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
187 
188 struct walkarg {
189 	int	family;
190 	int	w_tmemsize;
191 	int	w_op, w_arg;
192 	caddr_t	w_tmem;
193 	struct sysctl_req *w_req;
194 	struct sockaddr *dst;
195 	struct sockaddr *mask;
196 };
197 
198 static void	rts_input(struct mbuf *m);
199 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
200 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
201 			struct walkarg *w, int *plen);
202 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
203 			struct rt_addrinfo *rtinfo);
204 static int	cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb);
205 static int	sysctl_dumpentry(struct rtentry *rt, void *vw);
206 static int	sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
207 			uint32_t weight, struct walkarg *w);
208 static int	sysctl_iflist(int af, struct walkarg *w);
209 static int	sysctl_ifmalist(int af, struct walkarg *w);
210 static void	rt_getmetrics(const struct rtentry *rt,
211 			const struct nhop_object *nh, struct rt_metrics *out);
212 static void	rt_dispatch(struct mbuf *, sa_family_t);
213 static void	rt_ifannouncemsg(struct ifnet *ifp, int what);
214 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
215 			struct rt_msghdr *rtm, struct rib_cmd_info *rc);
216 static int	update_rtm_from_rc(struct rt_addrinfo *info,
217 			struct rt_msghdr **prtm, int alloc_len,
218 			struct rib_cmd_info *rc, struct nhop_object *nh);
219 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
220 			struct mbuf *m, sa_family_t saf, u_int fibnum,
221 			int rtm_errno);
222 static void	rtsock_notify_event(uint32_t fibnum, const struct rib_cmd_info *rc);
223 static void	rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask);
224 
225 static struct netisr_handler rtsock_nh = {
226 	.nh_name = "rtsock",
227 	.nh_handler = rts_input,
228 	.nh_proto = NETISR_ROUTE,
229 	.nh_policy = NETISR_POLICY_SOURCE,
230 };
231 
232 static int
sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)233 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
234 {
235 	int error, qlimit;
236 
237 	netisr_getqlimit(&rtsock_nh, &qlimit);
238 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
239         if (error || !req->newptr)
240                 return (error);
241 	if (qlimit < 1)
242 		return (EINVAL);
243 	return (netisr_setqlimit(&rtsock_nh, qlimit));
244 }
245 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
246     CTLTYPE_INT | CTLFLAG_RWTUN | CTLFLAG_NOFETCH | CTLFLAG_MPSAFE,
247     0, 0, sysctl_route_netisr_maxqlen, "I",
248     "maximum routing socket dispatch queue length");
249 
250 static void
vnet_rts_init(void)251 vnet_rts_init(void)
252 {
253 	int tmp;
254 
255 	if (IS_DEFAULT_VNET(curvnet)) {
256 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
257 			rtsock_nh.nh_qlimit = tmp;
258 		netisr_register(&rtsock_nh);
259 	}
260 #ifdef VIMAGE
261 	 else
262 		netisr_register_vnet(&rtsock_nh);
263 #endif
264 }
265 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
266     vnet_rts_init, 0);
267 
268 #ifdef VIMAGE
269 static void
vnet_rts_uninit(void)270 vnet_rts_uninit(void)
271 {
272 
273 	netisr_unregister_vnet(&rtsock_nh);
274 }
275 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
276     vnet_rts_uninit, 0);
277 #endif
278 
279 static void
report_route_event(const struct rib_cmd_info * rc,void * _cbdata)280 report_route_event(const struct rib_cmd_info *rc, void *_cbdata)
281 {
282 	uint32_t fibnum = (uint32_t)(uintptr_t)_cbdata;
283 	struct nhop_object *nh;
284 
285 	nh = rc->rc_cmd == RTM_DELETE ? rc->rc_nh_old : rc->rc_nh_new;
286 	rt_routemsg(rc->rc_cmd, rc->rc_rt, nh, fibnum);
287 }
288 
289 static void
rts_handle_route_event(uint32_t fibnum,const struct rib_cmd_info * rc)290 rts_handle_route_event(uint32_t fibnum, const struct rib_cmd_info *rc)
291 {
292 #ifdef ROUTE_MPATH
293 	if ((rc->rc_nh_new && NH_IS_NHGRP(rc->rc_nh_new)) ||
294 	    (rc->rc_nh_old && NH_IS_NHGRP(rc->rc_nh_old))) {
295 		rib_decompose_notification(rc, report_route_event,
296 		    (void *)(uintptr_t)fibnum);
297 	} else
298 #endif
299 		report_route_event(rc, (void *)(uintptr_t)fibnum);
300 }
301 static struct rtbridge rtsbridge = {
302 	.route_f = rts_handle_route_event,
303 	.ifmsg_f = rtsock_ifmsg,
304 };
305 static struct rtbridge *rtsbridge_orig_p;
306 
307 static void
rtsock_notify_event(uint32_t fibnum,const struct rib_cmd_info * rc)308 rtsock_notify_event(uint32_t fibnum, const struct rib_cmd_info *rc)
309 {
310 	netlink_callback_p->route_f(fibnum, rc);
311 }
312 
313 static void
rtsock_init(void)314 rtsock_init(void)
315 {
316 	rtsbridge_orig_p = rtsock_callback_p;
317 	rtsock_callback_p = &rtsbridge;
318 }
319 SYSINIT(rtsock_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rtsock_init, NULL);
320 
321 static void
rts_handle_ifnet_arrival(void * arg __unused,struct ifnet * ifp)322 rts_handle_ifnet_arrival(void *arg __unused, struct ifnet *ifp)
323 {
324 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
325 }
326 EVENTHANDLER_DEFINE(ifnet_arrival_event, rts_handle_ifnet_arrival, NULL, 0);
327 
328 static void
rts_handle_ifnet_departure(void * arg __unused,struct ifnet * ifp)329 rts_handle_ifnet_departure(void *arg __unused, struct ifnet *ifp)
330 {
331 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
332 }
333 EVENTHANDLER_DEFINE(ifnet_departure_event, rts_handle_ifnet_departure, NULL, 0);
334 
335 static void
rts_append_data(struct socket * so,struct mbuf * m)336 rts_append_data(struct socket *so, struct mbuf *m)
337 {
338 
339 	if (sbappendaddr(&so->so_rcv, &route_src, m, NULL) == 0) {
340 		soroverflow(so);
341 		m_freem(m);
342 	} else
343 		sorwakeup(so);
344 }
345 
346 static void
rts_input(struct mbuf * m)347 rts_input(struct mbuf *m)
348 {
349 	struct rcb *rcb;
350 	struct socket *last;
351 
352 	last = NULL;
353 	RTSOCK_LOCK();
354 	LIST_FOREACH(rcb, &V_route_cb.cblist, list) {
355 		if (rcb->rcb_family != AF_UNSPEC &&
356 		    rcb->rcb_family != m->m_rtsock_family)
357 			continue;
358 		if ((m->m_flags & RTS_FILTER_FIB) &&
359 		    M_GETFIB(m) != rcb->rcb_socket->so_fibnum)
360 			continue;
361 		if (last != NULL) {
362 			struct mbuf *n;
363 
364 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
365 			if (n != NULL)
366 				rts_append_data(last, n);
367 		}
368 		last = rcb->rcb_socket;
369 	}
370 	if (last != NULL)
371 		rts_append_data(last, m);
372 	else
373 		m_freem(m);
374 	RTSOCK_UNLOCK();
375 }
376 
377 static void
rts_close(struct socket * so)378 rts_close(struct socket *so)
379 {
380 
381 	soisdisconnected(so);
382 }
383 
384 static SYSCTL_NODE(_net, OID_AUTO, rtsock, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
385     "Routing socket infrastructure");
386 static u_long rts_sendspace = 8192;
387 SYSCTL_ULONG(_net_rtsock, OID_AUTO, sendspace, CTLFLAG_RW, &rts_sendspace, 0,
388     "Default routing socket send space");
389 static u_long rts_recvspace = 8192;
390 SYSCTL_ULONG(_net_rtsock, OID_AUTO, recvspace, CTLFLAG_RW, &rts_recvspace, 0,
391     "Default routing socket receive space");
392 
393 static int
rts_attach(struct socket * so,int proto,struct thread * td)394 rts_attach(struct socket *so, int proto, struct thread *td)
395 {
396 	struct rcb *rcb;
397 	int error;
398 
399 	error = soreserve(so, rts_sendspace, rts_recvspace);
400 	if (error)
401 		return (error);
402 
403 	rcb = malloc(sizeof(*rcb), M_PCB, M_WAITOK);
404 	rcb->rcb_socket = so;
405 	rcb->rcb_family = proto;
406 
407 	so->so_pcb = rcb;
408 	so->so_fibnum = td->td_proc->p_fibnum;
409 	so->so_options |= SO_USELOOPBACK;
410 
411 	RTSOCK_LOCK();
412 	LIST_INSERT_HEAD(&V_route_cb.cblist, rcb, list);
413 	switch (proto) {
414 	case AF_INET:
415 		V_route_cb.ip_count++;
416 		break;
417 	case AF_INET6:
418 		V_route_cb.ip6_count++;
419 		break;
420 	}
421 	V_route_cb.any_count++;
422 	RTSOCK_UNLOCK();
423 	soisconnected(so);
424 
425 	return (0);
426 }
427 
428 static void
rts_detach(struct socket * so)429 rts_detach(struct socket *so)
430 {
431 	struct rcb *rcb = so->so_pcb;
432 
433 	RTSOCK_LOCK();
434 	LIST_REMOVE(rcb, list);
435 	switch(rcb->rcb_family) {
436 	case AF_INET:
437 		V_route_cb.ip_count--;
438 		break;
439 	case AF_INET6:
440 		V_route_cb.ip6_count--;
441 		break;
442 	}
443 	V_route_cb.any_count--;
444 	RTSOCK_UNLOCK();
445 	free(rcb, M_PCB);
446 	so->so_pcb = NULL;
447 }
448 
449 static int
rts_disconnect(struct socket * so)450 rts_disconnect(struct socket *so)
451 {
452 
453 	return (ENOTCONN);
454 }
455 
456 static int
rts_shutdown(struct socket * so)457 rts_shutdown(struct socket *so)
458 {
459 
460 	socantsendmore(so);
461 	return (0);
462 }
463 
464 #ifndef _SOCKADDR_UNION_DEFINED
465 #define	_SOCKADDR_UNION_DEFINED
466 /*
467  * The union of all possible address formats we handle.
468  */
469 union sockaddr_union {
470 	struct sockaddr		sa;
471 	struct sockaddr_in	sin;
472 	struct sockaddr_in6	sin6;
473 };
474 #endif /* _SOCKADDR_UNION_DEFINED */
475 
476 static int
rtm_get_jailed(struct rt_addrinfo * info,struct ifnet * ifp,struct nhop_object * nh,union sockaddr_union * saun,struct ucred * cred)477 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
478     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
479 {
480 #if defined(INET) || defined(INET6)
481 	struct epoch_tracker et;
482 #endif
483 
484 	/* First, see if the returned address is part of the jail. */
485 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
486 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
487 		return (0);
488 	}
489 
490 	switch (info->rti_info[RTAX_DST]->sa_family) {
491 #ifdef INET
492 	case AF_INET:
493 	{
494 		struct in_addr ia;
495 		struct ifaddr *ifa;
496 		int found;
497 
498 		found = 0;
499 		/*
500 		 * Try to find an address on the given outgoing interface
501 		 * that belongs to the jail.
502 		 */
503 		NET_EPOCH_ENTER(et);
504 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
505 			struct sockaddr *sa;
506 			sa = ifa->ifa_addr;
507 			if (sa->sa_family != AF_INET)
508 				continue;
509 			ia = ((struct sockaddr_in *)sa)->sin_addr;
510 			if (prison_check_ip4(cred, &ia) == 0) {
511 				found = 1;
512 				break;
513 			}
514 		}
515 		NET_EPOCH_EXIT(et);
516 		if (!found) {
517 			/*
518 			 * As a last resort return the 'default' jail address.
519 			 */
520 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
521 			    sin_addr;
522 			if (prison_get_ip4(cred, &ia) != 0)
523 				return (ESRCH);
524 		}
525 		bzero(&saun->sin, sizeof(struct sockaddr_in));
526 		saun->sin.sin_len = sizeof(struct sockaddr_in);
527 		saun->sin.sin_family = AF_INET;
528 		saun->sin.sin_addr.s_addr = ia.s_addr;
529 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
530 		break;
531 	}
532 #endif
533 #ifdef INET6
534 	case AF_INET6:
535 	{
536 		struct in6_addr ia6;
537 		struct ifaddr *ifa;
538 		int found;
539 
540 		found = 0;
541 		/*
542 		 * Try to find an address on the given outgoing interface
543 		 * that belongs to the jail.
544 		 */
545 		NET_EPOCH_ENTER(et);
546 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
547 			struct sockaddr *sa;
548 			sa = ifa->ifa_addr;
549 			if (sa->sa_family != AF_INET6)
550 				continue;
551 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
552 			    &ia6, sizeof(struct in6_addr));
553 			if (prison_check_ip6(cred, &ia6) == 0) {
554 				found = 1;
555 				break;
556 			}
557 		}
558 		NET_EPOCH_EXIT(et);
559 		if (!found) {
560 			/*
561 			 * As a last resort return the 'default' jail address.
562 			 */
563 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
564 			    sin6_addr;
565 			if (prison_get_ip6(cred, &ia6) != 0)
566 				return (ESRCH);
567 		}
568 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
569 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
570 		saun->sin6.sin6_family = AF_INET6;
571 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
572 		if (sa6_recoverscope(&saun->sin6) != 0)
573 			return (ESRCH);
574 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
575 		break;
576 	}
577 #endif
578 	default:
579 		return (ESRCH);
580 	}
581 	return (0);
582 }
583 
584 static int
fill_blackholeinfo(struct rt_addrinfo * info,union sockaddr_union * saun)585 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
586 {
587 	struct ifaddr *ifa;
588 	sa_family_t saf;
589 
590 	if (V_loif == NULL) {
591 		RTS_PID_LOG(LOG_INFO, "Unable to add blackhole/reject nhop without loopback");
592 		return (ENOTSUP);
593 	}
594 	info->rti_ifp = V_loif;
595 
596 	saf = info->rti_info[RTAX_DST]->sa_family;
597 
598 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
599 		if (ifa->ifa_addr->sa_family == saf) {
600 			info->rti_ifa = ifa;
601 			break;
602 		}
603 	}
604 	if (info->rti_ifa == NULL) {
605 		RTS_PID_LOG(LOG_INFO, "Unable to find ifa for blackhole/reject nhop");
606 		return (ENOTSUP);
607 	}
608 
609 	bzero(saun, sizeof(union sockaddr_union));
610 	switch (saf) {
611 #ifdef INET
612 	case AF_INET:
613 		saun->sin.sin_family = AF_INET;
614 		saun->sin.sin_len = sizeof(struct sockaddr_in);
615 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
616 		break;
617 #endif
618 #ifdef INET6
619 	case AF_INET6:
620 		saun->sin6.sin6_family = AF_INET6;
621 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
622 		saun->sin6.sin6_addr = in6addr_loopback;
623 		break;
624 #endif
625 	default:
626 		RTS_PID_LOG(LOG_INFO, "unsupported family: %d", saf);
627 		return (ENOTSUP);
628 	}
629 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
630 	info->rti_flags |= RTF_GATEWAY;
631 
632 	return (0);
633 }
634 
635 /*
636  * Fills in @info based on userland-provided @rtm message.
637  *
638  * Returns 0 on success.
639  */
640 static int
fill_addrinfo(struct rt_msghdr * rtm,int len,struct linear_buffer * lb,u_int fibnum,struct rt_addrinfo * info)641 fill_addrinfo(struct rt_msghdr *rtm, int len, struct linear_buffer *lb, u_int fibnum,
642     struct rt_addrinfo *info)
643 {
644 	int error;
645 
646 	rtm->rtm_pid = curproc->p_pid;
647 	info->rti_addrs = rtm->rtm_addrs;
648 
649 	info->rti_mflags = rtm->rtm_inits;
650 	info->rti_rmx = &rtm->rtm_rmx;
651 
652 	/*
653 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
654 	 * link-local address because rtrequest requires addresses with
655 	 * embedded scope id.
656 	 */
657 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
658 		return (EINVAL);
659 
660 	info->rti_flags = rtm->rtm_flags;
661 	error = cleanup_xaddrs(info, lb);
662 	if (error != 0)
663 		return (error);
664 	/*
665 	 * Verify that the caller has the appropriate privilege; RTM_GET
666 	 * is the only operation the non-superuser is allowed.
667 	 */
668 	if (rtm->rtm_type != RTM_GET) {
669 		error = priv_check(curthread, PRIV_NET_ROUTE);
670 		if (error != 0)
671 			return (error);
672 	}
673 
674 	/*
675 	 * The given gateway address may be an interface address.
676 	 * For example, issuing a "route change" command on a route
677 	 * entry that was created from a tunnel, and the gateway
678 	 * address given is the local end point. In this case the
679 	 * RTF_GATEWAY flag must be cleared or the destination will
680 	 * not be reachable even though there is no error message.
681 	 */
682 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
683 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
684 		struct nhop_object *nh;
685 
686 		/*
687 		 * A host route through the loopback interface is
688 		 * installed for each interface address. In pre 8.0
689 		 * releases the interface address of a PPP link type
690 		 * is not reachable locally. This behavior is fixed as
691 		 * part of the new L2/L3 redesign and rewrite work. The
692 		 * signature of this interface address route is the
693 		 * AF_LINK sa_family type of the gateway, and the
694 		 * rt_ifp has the IFF_LOOPBACK flag set.
695 		 */
696 		nh = rib_lookup(fibnum, info->rti_info[RTAX_GATEWAY], NHR_NONE, 0);
697 		if (nh != NULL && nh->gw_sa.sa_family == AF_LINK &&
698 		    nh->nh_ifp->if_flags & IFF_LOOPBACK) {
699 				info->rti_flags &= ~RTF_GATEWAY;
700 				info->rti_flags |= RTF_GWFLAG_COMPAT;
701 		}
702 	}
703 
704 	return (0);
705 }
706 
707 static struct nhop_object *
select_nhop(struct nhop_object * nh,const struct sockaddr * gw)708 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
709 {
710 	if (!NH_IS_NHGRP(nh))
711 		return (nh);
712 #ifdef ROUTE_MPATH
713 	const struct weightened_nhop *wn;
714 	uint32_t num_nhops;
715 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
716 	if (gw == NULL)
717 		return (wn[0].nh);
718 	for (int i = 0; i < num_nhops; i++) {
719 		if (match_nhop_gw(wn[i].nh, gw))
720 			return (wn[i].nh);
721 	}
722 #endif
723 	return (NULL);
724 }
725 
726 /*
727  * Handles RTM_GET message from routing socket, returning matching rt.
728  *
729  * Returns:
730  * 0 on success, with locked and referenced matching rt in @rt_nrt
731  * errno of failure
732  */
733 static int
handle_rtm_get(struct rt_addrinfo * info,u_int fibnum,struct rt_msghdr * rtm,struct rib_cmd_info * rc)734 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
735     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
736 {
737 	RIB_RLOCK_TRACKER;
738 	struct rib_head *rnh;
739 	struct nhop_object *nh;
740 	sa_family_t saf;
741 
742 	saf = info->rti_info[RTAX_DST]->sa_family;
743 
744 	rnh = rt_tables_get_rnh(fibnum, saf);
745 	if (rnh == NULL)
746 		return (EAFNOSUPPORT);
747 
748 	RIB_RLOCK(rnh);
749 
750 	/*
751 	 * By (implicit) convention host route (one without netmask)
752 	 * means longest-prefix-match request and the route with netmask
753 	 * means exact-match lookup.
754 	 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128
755 	 * prefixes, use original data to check for the netmask presence.
756 	 */
757 	if ((rtm->rtm_addrs & RTA_NETMASK) == 0) {
758 		/*
759 		 * Provide longest prefix match for
760 		 * address lookup (no mask).
761 		 * 'route -n get addr'
762 		 */
763 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
764 		    info->rti_info[RTAX_DST], &rnh->head);
765 	} else
766 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
767 		    info->rti_info[RTAX_DST],
768 		    info->rti_info[RTAX_NETMASK], &rnh->head);
769 
770 	if (rc->rc_rt == NULL) {
771 		RIB_RUNLOCK(rnh);
772 		return (ESRCH);
773 	}
774 
775 	nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
776 	if (nh == NULL) {
777 		RIB_RUNLOCK(rnh);
778 		return (ESRCH);
779 	}
780 	/*
781 	 * If performing proxied L2 entry insertion, and
782 	 * the actual PPP host entry is found, perform
783 	 * another search to retrieve the prefix route of
784 	 * the local end point of the PPP link.
785 	 * TODO: move this logic to userland.
786 	 */
787 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
788 		struct sockaddr_storage laddr;
789 
790 		if (nh->nh_ifp != NULL &&
791 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
792 			struct ifaddr *ifa;
793 
794 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
795 					RT_ALL_FIBS);
796 			if (ifa != NULL)
797 				rt_maskedcopy(ifa->ifa_addr,
798 					      (struct sockaddr *)&laddr,
799 					      ifa->ifa_netmask);
800 		} else
801 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
802 				      (struct sockaddr *)&laddr,
803 				      nh->nh_ifa->ifa_netmask);
804 		/*
805 		 * refactor rt and no lock operation necessary
806 		 */
807 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(
808 		    (struct sockaddr *)&laddr, &rnh->head);
809 		if (rc->rc_rt == NULL) {
810 			RIB_RUNLOCK(rnh);
811 			return (ESRCH);
812 		}
813 		nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
814 		if (nh == NULL) {
815 			RIB_RUNLOCK(rnh);
816 			return (ESRCH);
817 		}
818 	}
819 	rc->rc_nh_new = nh;
820 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
821 	RIB_RUNLOCK(rnh);
822 
823 	return (0);
824 }
825 
826 static void
init_sockaddrs_family(int family,struct sockaddr * dst,struct sockaddr * mask)827 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
828 {
829 #ifdef INET
830 	if (family == AF_INET) {
831 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
832 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
833 
834 		bzero(dst4, sizeof(struct sockaddr_in));
835 		bzero(mask4, sizeof(struct sockaddr_in));
836 
837 		dst4->sin_family = AF_INET;
838 		dst4->sin_len = sizeof(struct sockaddr_in);
839 		mask4->sin_family = AF_INET;
840 		mask4->sin_len = sizeof(struct sockaddr_in);
841 	}
842 #endif
843 #ifdef INET6
844 	if (family == AF_INET6) {
845 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
846 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
847 
848 		bzero(dst6, sizeof(struct sockaddr_in6));
849 		bzero(mask6, sizeof(struct sockaddr_in6));
850 
851 		dst6->sin6_family = AF_INET6;
852 		dst6->sin6_len = sizeof(struct sockaddr_in6);
853 		mask6->sin6_family = AF_INET6;
854 		mask6->sin6_len = sizeof(struct sockaddr_in6);
855 	}
856 #endif
857 }
858 
859 static void
export_rtaddrs(const struct rtentry * rt,struct sockaddr * dst,struct sockaddr * mask)860 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
861     struct sockaddr *mask)
862 {
863 #ifdef INET
864 	if (dst->sa_family == AF_INET) {
865 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
866 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
867 		uint32_t scopeid = 0;
868 		rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
869 		    &scopeid);
870 		return;
871 	}
872 #endif
873 #ifdef INET6
874 	if (dst->sa_family == AF_INET6) {
875 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
876 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
877 		uint32_t scopeid = 0;
878 		rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
879 		    &mask6->sin6_addr, &scopeid);
880 		dst6->sin6_scope_id = scopeid;
881 		return;
882 	}
883 #endif
884 }
885 
886 static int
update_rtm_from_info(struct rt_addrinfo * info,struct rt_msghdr ** prtm,int alloc_len)887 update_rtm_from_info(struct rt_addrinfo *info, struct rt_msghdr **prtm,
888     int alloc_len)
889 {
890 	struct rt_msghdr *rtm, *orig_rtm = NULL;
891 	struct walkarg w;
892 	int len;
893 
894 	rtm = *prtm;
895 	/* Check if we need to realloc storage */
896 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
897 	if (len > alloc_len) {
898 		struct rt_msghdr *tmp_rtm;
899 
900 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
901 		if (tmp_rtm == NULL)
902 			return (ENOBUFS);
903 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
904 		orig_rtm = rtm;
905 		rtm = tmp_rtm;
906 		alloc_len = len;
907 
908 		/*
909 		 * Delay freeing original rtm as info contains
910 		 * data referencing it.
911 		 */
912 	}
913 
914 	w.w_tmem = (caddr_t)rtm;
915 	w.w_tmemsize = alloc_len;
916 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
917 	rtm->rtm_addrs = info->rti_addrs;
918 
919 	if (orig_rtm != NULL)
920 		free(orig_rtm, M_TEMP);
921 	*prtm = rtm;
922 	return (0);
923 }
924 
925 
926 /*
927  * Update sockaddrs, flags, etc in @prtm based on @rc data.
928  * rtm can be reallocated.
929  *
930  * Returns 0 on success, along with pointer to (potentially reallocated)
931  *  rtm.
932  *
933  */
934 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)935 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
936     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
937 {
938 	union sockaddr_union saun;
939 	struct rt_msghdr *rtm;
940 	struct ifnet *ifp;
941 	int error;
942 
943 	rtm = *prtm;
944 	union sockaddr_union sa_dst, sa_mask;
945 	int family = info->rti_info[RTAX_DST]->sa_family;
946 	init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
947 	export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
948 
949 	info->rti_info[RTAX_DST] = &sa_dst.sa;
950 	info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
951 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
952 	info->rti_info[RTAX_GENMASK] = 0;
953 	ifp = nh->nh_ifp;
954 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
955 		if (ifp) {
956 			info->rti_info[RTAX_IFP] =
957 			    ifp->if_addr->ifa_addr;
958 			error = rtm_get_jailed(info, ifp, nh,
959 			    &saun, curthread->td_ucred);
960 			if (error != 0)
961 				return (error);
962 			if (ifp->if_flags & IFF_POINTOPOINT)
963 				info->rti_info[RTAX_BRD] =
964 				    nh->nh_ifa->ifa_dstaddr;
965 			rtm->rtm_index = ifp->if_index;
966 		} else {
967 			info->rti_info[RTAX_IFP] = NULL;
968 			info->rti_info[RTAX_IFA] = NULL;
969 		}
970 	} else if (ifp != NULL)
971 		rtm->rtm_index = ifp->if_index;
972 
973 	if ((error = update_rtm_from_info(info, prtm, alloc_len)) != 0)
974 		return (error);
975 
976 	rtm = *prtm;
977 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
978 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
979 		rtm->rtm_flags = RTF_GATEWAY |
980 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
981 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
982 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
983 
984 	return (0);
985 }
986 
987 #ifdef ROUTE_MPATH
988 static void
save_del_notification(const struct rib_cmd_info * rc,void * _cbdata)989 save_del_notification(const struct rib_cmd_info *rc, void *_cbdata)
990 {
991 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
992 
993 	if (rc->rc_cmd == RTM_DELETE)
994 		*rc_new = *rc;
995 }
996 
997 static void
save_add_notification(const struct rib_cmd_info * rc,void * _cbdata)998 save_add_notification(const struct rib_cmd_info *rc, void *_cbdata)
999 {
1000 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1001 
1002 	if (rc->rc_cmd == RTM_ADD)
1003 		*rc_new = *rc;
1004 }
1005 #endif
1006 
1007 #if defined(INET6) || defined(INET)
1008 static struct sockaddr *
alloc_sockaddr_aligned(struct linear_buffer * lb,int len)1009 alloc_sockaddr_aligned(struct linear_buffer *lb, int len)
1010 {
1011 	len = roundup2(len, sizeof(uint64_t));
1012 	if (lb->offset + len > lb->size)
1013 		return (NULL);
1014 	struct sockaddr *sa = (struct sockaddr *)(lb->base + lb->offset);
1015 	lb->offset += len;
1016 	return (sa);
1017 }
1018 #endif
1019 
1020 static int
rts_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)1021 rts_send(struct socket *so, int flags, struct mbuf *m,
1022     struct sockaddr *nam, struct mbuf *control, struct thread *td)
1023 {
1024 	struct rt_msghdr *rtm = NULL;
1025 	struct rt_addrinfo info;
1026 	struct epoch_tracker et;
1027 #ifdef INET6
1028 	struct sockaddr_storage ss;
1029 	struct sockaddr_in6 *sin6;
1030 	int i, rti_need_deembed = 0;
1031 #endif
1032 	int alloc_len = 0, len, error = 0, fibnum;
1033 	sa_family_t saf = AF_UNSPEC;
1034 	struct rib_cmd_info rc;
1035 	struct nhop_object *nh;
1036 
1037 	if ((flags & PRUS_OOB) || control != NULL) {
1038 		m_freem(m);
1039 		if (control != NULL)
1040 			m_freem(control);
1041 		return (EOPNOTSUPP);
1042 	}
1043 
1044 	fibnum = so->so_fibnum;
1045 #define senderr(e) { error = e; goto flush;}
1046 	if (m == NULL || ((m->m_len < sizeof(long)) &&
1047 		       (m = m_pullup(m, sizeof(long))) == NULL))
1048 		return (ENOBUFS);
1049 	if ((m->m_flags & M_PKTHDR) == 0)
1050 		panic("route_output");
1051 	NET_EPOCH_ENTER(et);
1052 	len = m->m_pkthdr.len;
1053 	if (len < sizeof(*rtm) ||
1054 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1055 		senderr(EINVAL);
1056 
1057 	/*
1058 	 * Most of current messages are in range 200-240 bytes,
1059 	 * minimize possible re-allocation on reply using larger size
1060 	 * buffer aligned on 1k boundaty.
1061 	 */
1062 	alloc_len = roundup2(len, 1024);
1063 	int total_len = alloc_len + SCRATCH_BUFFER_SIZE;
1064 	if ((rtm = malloc(total_len, M_TEMP, M_NOWAIT)) == NULL)
1065 		senderr(ENOBUFS);
1066 
1067 	m_copydata(m, 0, len, (caddr_t)rtm);
1068 	bzero(&info, sizeof(info));
1069 	nh = NULL;
1070 	struct linear_buffer lb = {
1071 		.base = (char *)rtm + alloc_len,
1072 		.size = SCRATCH_BUFFER_SIZE,
1073 	};
1074 
1075 	if (rtm->rtm_version != RTM_VERSION) {
1076 		/* Do not touch message since format is unknown */
1077 		free(rtm, M_TEMP);
1078 		rtm = NULL;
1079 		senderr(EPROTONOSUPPORT);
1080 	}
1081 
1082 	/*
1083 	 * Starting from here, it is possible
1084 	 * to alter original message and insert
1085 	 * caller PID and error value.
1086 	 */
1087 
1088 	if ((error = fill_addrinfo(rtm, len, &lb, fibnum, &info)) != 0) {
1089 		senderr(error);
1090 	}
1091 	/* fill_addringo() embeds scope into IPv6 addresses */
1092 #ifdef INET6
1093 	rti_need_deembed = 1;
1094 #endif
1095 
1096 	saf = info.rti_info[RTAX_DST]->sa_family;
1097 
1098 	/* support for new ARP code */
1099 	if (rtm->rtm_flags & RTF_LLDATA) {
1100 		error = lla_rt_output(rtm, &info);
1101 		goto flush;
1102 	}
1103 
1104 	union sockaddr_union gw_saun;
1105 	int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1106 	if (blackhole_flags != 0) {
1107 		if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1108 			error = fill_blackholeinfo(&info, &gw_saun);
1109 		else {
1110 			RTS_PID_LOG(LOG_DEBUG, "both BLACKHOLE and REJECT flags specifiied");
1111 			error = EINVAL;
1112 		}
1113 		if (error != 0)
1114 			senderr(error);
1115 	}
1116 
1117 	switch (rtm->rtm_type) {
1118 	case RTM_ADD:
1119 	case RTM_CHANGE:
1120 		if (rtm->rtm_type == RTM_ADD) {
1121 			if (info.rti_info[RTAX_GATEWAY] == NULL) {
1122 				RTS_PID_LOG(LOG_DEBUG, "RTM_ADD w/o gateway");
1123 				senderr(EINVAL);
1124 			}
1125 		}
1126 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1127 		if (error == 0) {
1128 			rtsock_notify_event(fibnum, &rc);
1129 #ifdef ROUTE_MPATH
1130 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
1131 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1132 				struct rib_cmd_info rc_simple = {};
1133 				rib_decompose_notification(&rc,
1134 				    save_add_notification, (void *)&rc_simple);
1135 				rc = rc_simple;
1136 			}
1137 #endif
1138 			/* nh MAY be empty if RTM_CHANGE request is no-op */
1139 			nh = rc.rc_nh_new;
1140 			if (nh != NULL) {
1141 				rtm->rtm_index = nh->nh_ifp->if_index;
1142 				rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1143 			}
1144 		}
1145 		break;
1146 
1147 	case RTM_DELETE:
1148 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1149 		if (error == 0) {
1150 			rtsock_notify_event(fibnum, &rc);
1151 #ifdef ROUTE_MPATH
1152 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1153 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1154 				struct rib_cmd_info rc_simple = {};
1155 				rib_decompose_notification(&rc,
1156 				    save_del_notification, (void *)&rc_simple);
1157 				rc = rc_simple;
1158 			}
1159 #endif
1160 			nh = rc.rc_nh_old;
1161 		}
1162 		break;
1163 
1164 	case RTM_GET:
1165 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1166 		if (error != 0)
1167 			senderr(error);
1168 		nh = rc.rc_nh_new;
1169 
1170 		if (!rt_is_exportable(rc.rc_rt, curthread->td_ucred))
1171 			senderr(ESRCH);
1172 		break;
1173 
1174 	default:
1175 		senderr(EOPNOTSUPP);
1176 	}
1177 
1178 	if (error == 0 && nh != NULL) {
1179 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1180 		/*
1181 		 * Note that some sockaddr pointers may have changed to
1182 		 * point to memory outsize @rtm. Some may be pointing
1183 		 * to the on-stack variables.
1184 		 * Given that, any pointer in @info CANNOT BE USED.
1185 		 */
1186 
1187 		/*
1188 		 * scopeid deembedding has been performed while
1189 		 * writing updated rtm in rtsock_msg_buffer().
1190 		 * With that in mind, skip deembedding procedure below.
1191 		 */
1192 #ifdef INET6
1193 		rti_need_deembed = 0;
1194 #endif
1195 	}
1196 
1197 flush:
1198 	NET_EPOCH_EXIT(et);
1199 
1200 #ifdef INET6
1201 	if (rtm != NULL) {
1202 		if (rti_need_deembed) {
1203 			/* sin6_scope_id is recovered before sending rtm. */
1204 			sin6 = (struct sockaddr_in6 *)&ss;
1205 			for (i = 0; i < RTAX_MAX; i++) {
1206 				if (info.rti_info[i] == NULL)
1207 					continue;
1208 				if (info.rti_info[i]->sa_family != AF_INET6)
1209 					continue;
1210 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1211 				if (sa6_recoverscope(sin6) == 0)
1212 					bcopy(sin6, info.rti_info[i],
1213 						    sizeof(*sin6));
1214 			}
1215 			if (update_rtm_from_info(&info, &rtm, alloc_len) != 0) {
1216 				if (error != 0)
1217 					error = ENOBUFS;
1218 			}
1219 		}
1220 	}
1221 #endif
1222 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1223 
1224 	return (error);
1225 }
1226 
1227 /*
1228  * Sends the prepared reply message in @rtm to all rtsock clients.
1229  * Frees @m and @rtm.
1230  *
1231  */
1232 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)1233 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1234     sa_family_t saf, u_int fibnum, int rtm_errno)
1235 {
1236 	struct rcb *rcb = NULL;
1237 
1238 	/*
1239 	 * Check to see if we don't want our own messages.
1240 	 */
1241 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1242 		if (V_route_cb.any_count <= 1) {
1243 			if (rtm != NULL)
1244 				free(rtm, M_TEMP);
1245 			m_freem(m);
1246 			return;
1247 		}
1248 		/* There is another listener, so construct message */
1249 		rcb = so->so_pcb;
1250 	}
1251 
1252 	if (rtm != NULL) {
1253 		if (rtm_errno!= 0)
1254 			rtm->rtm_errno = rtm_errno;
1255 		else
1256 			rtm->rtm_flags |= RTF_DONE;
1257 
1258 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1259 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1260 			m_freem(m);
1261 			m = NULL;
1262 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1263 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1264 
1265 		free(rtm, M_TEMP);
1266 	}
1267 	if (m != NULL) {
1268 		M_SETFIB(m, fibnum);
1269 		m->m_flags |= RTS_FILTER_FIB;
1270 		if (rcb) {
1271 			/*
1272 			 * XXX insure we don't get a copy by
1273 			 * invalidating our protocol
1274 			 */
1275 			sa_family_t family = rcb->rcb_family;
1276 			rcb->rcb_family = AF_UNSPEC;
1277 			rt_dispatch(m, saf);
1278 			rcb->rcb_family = family;
1279 		} else
1280 			rt_dispatch(m, saf);
1281 	}
1282 }
1283 
1284 static void
rt_getmetrics(const struct rtentry * rt,const struct nhop_object * nh,struct rt_metrics * out)1285 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1286     struct rt_metrics *out)
1287 {
1288 
1289 	bzero(out, sizeof(*out));
1290 	out->rmx_mtu = nh->nh_mtu;
1291 	out->rmx_weight = rt->rt_weight;
1292 	out->rmx_nhidx = nhop_get_idx(nh);
1293 	/* Kernel -> userland timebase conversion. */
1294 	out->rmx_expire = nhop_get_expire(nh) ?
1295 	    nhop_get_expire(nh) - time_uptime + time_second : 0;
1296 }
1297 
1298 /*
1299  * Extract the addresses of the passed sockaddrs.
1300  * Do a little sanity checking so as to avoid bad memory references.
1301  * This data is derived straight from userland.
1302  */
1303 static int
rt_xaddrs(caddr_t cp,caddr_t cplim,struct rt_addrinfo * rtinfo)1304 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1305 {
1306 	struct sockaddr *sa;
1307 	int i;
1308 
1309 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1310 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1311 			continue;
1312 		sa = (struct sockaddr *)cp;
1313 		/*
1314 		 * It won't fit.
1315 		 */
1316 		if (cp + sa->sa_len > cplim) {
1317 			RTS_PID_LOG(LOG_DEBUG, "sa_len too big for sa type %d", i);
1318 			return (EINVAL);
1319 		}
1320 		/*
1321 		 * there are no more.. quit now
1322 		 * If there are more bits, they are in error.
1323 		 * I've seen this. route(1) can evidently generate these.
1324 		 * This causes kernel to core dump.
1325 		 * for compatibility, If we see this, point to a safe address.
1326 		 */
1327 		if (sa->sa_len == 0) {
1328 			rtinfo->rti_info[i] = &sa_zero;
1329 			return (0); /* should be EINVAL but for compat */
1330 		}
1331 		/* accept it */
1332 #ifdef INET6
1333 		if (sa->sa_family == AF_INET6)
1334 			sa6_embedscope((struct sockaddr_in6 *)sa,
1335 			    V_ip6_use_defzone);
1336 #endif
1337 		rtinfo->rti_info[i] = sa;
1338 		cp += SA_SIZE(sa);
1339 	}
1340 	return (0);
1341 }
1342 
1343 #ifdef INET
1344 static inline void
fill_sockaddr_inet(struct sockaddr_in * sin,struct in_addr addr)1345 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1346 {
1347 
1348 	const struct sockaddr_in nsin = {
1349 		.sin_family = AF_INET,
1350 		.sin_len = sizeof(struct sockaddr_in),
1351 		.sin_addr = addr,
1352 	};
1353 	*sin = nsin;
1354 }
1355 #endif
1356 
1357 #ifdef INET6
1358 static inline void
fill_sockaddr_inet6(struct sockaddr_in6 * sin6,const struct in6_addr * addr6,uint32_t scopeid)1359 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1360     uint32_t scopeid)
1361 {
1362 
1363 	const struct sockaddr_in6 nsin6 = {
1364 		.sin6_family = AF_INET6,
1365 		.sin6_len = sizeof(struct sockaddr_in6),
1366 		.sin6_addr = *addr6,
1367 		.sin6_scope_id = scopeid,
1368 	};
1369 	*sin6 = nsin6;
1370 }
1371 #endif
1372 
1373 #if defined(INET6) || defined(INET)
1374 /*
1375  * Checks if gateway is suitable for lltable operations.
1376  * Lltable code requires AF_LINK gateway with ifindex
1377  *  and mac address specified.
1378  * Returns 0 on success.
1379  */
1380 static int
cleanup_xaddrs_lladdr(struct rt_addrinfo * info)1381 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1382 {
1383 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1384 
1385 	if (sdl->sdl_family != AF_LINK)
1386 		return (EINVAL);
1387 
1388 	if (sdl->sdl_index == 0) {
1389 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gateway w/o ifindex");
1390 		return (EINVAL);
1391 	}
1392 
1393 	if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len) {
1394 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gw: sdl_nlen/sdl_alen too large");
1395 		return (EINVAL);
1396 	}
1397 
1398 	return (0);
1399 }
1400 
1401 static int
cleanup_xaddrs_gateway(struct rt_addrinfo * info,struct linear_buffer * lb)1402 cleanup_xaddrs_gateway(struct rt_addrinfo *info, struct linear_buffer *lb)
1403 {
1404 	struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1405 	struct sockaddr *sa;
1406 
1407 	if (info->rti_flags & RTF_LLDATA)
1408 		return (cleanup_xaddrs_lladdr(info));
1409 
1410 	switch (gw->sa_family) {
1411 #ifdef INET
1412 	case AF_INET:
1413 		{
1414 			struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw;
1415 
1416 			/* Ensure reads do not go beyoud SA boundary */
1417 			if (SA_SIZE(gw) < offsetof(struct sockaddr_in, sin_zero)) {
1418 				RTS_PID_LOG(LOG_DEBUG, "gateway sin_len too small: %d",
1419 				    gw->sa_len);
1420 				return (EINVAL);
1421 			}
1422 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_in));
1423 			if (sa == NULL)
1424 				return (ENOBUFS);
1425 			fill_sockaddr_inet((struct sockaddr_in *)sa, gw_sin->sin_addr);
1426 			info->rti_info[RTAX_GATEWAY] = sa;
1427 		}
1428 		break;
1429 #endif
1430 #ifdef INET6
1431 	case AF_INET6:
1432 		{
1433 			struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1434 			if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1435 				RTS_PID_LOG(LOG_DEBUG, "gateway sin6_len too small: %d",
1436 				    gw->sa_len);
1437 				return (EINVAL);
1438 			}
1439 			fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1440 			break;
1441 		}
1442 #endif
1443 	case AF_LINK:
1444 		{
1445 			struct sockaddr_dl *gw_sdl;
1446 
1447 			size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data);
1448 			gw_sdl = (struct sockaddr_dl *)gw;
1449 			if (gw_sdl->sdl_len < sdl_min_len) {
1450 				RTS_PID_LOG(LOG_DEBUG, "gateway sdl_len too small: %d",
1451 				    gw_sdl->sdl_len);
1452 				return (EINVAL);
1453 			}
1454 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_dl_short));
1455 			if (sa == NULL)
1456 				return (ENOBUFS);
1457 
1458 			const struct sockaddr_dl_short sdl = {
1459 				.sdl_family = AF_LINK,
1460 				.sdl_len = sizeof(struct sockaddr_dl_short),
1461 				.sdl_index = gw_sdl->sdl_index,
1462 			};
1463 			*((struct sockaddr_dl_short *)sa) = sdl;
1464 			info->rti_info[RTAX_GATEWAY] = sa;
1465 			break;
1466 		}
1467 	}
1468 
1469 	return (0);
1470 }
1471 #endif
1472 
1473 static void
remove_netmask(struct rt_addrinfo * info)1474 remove_netmask(struct rt_addrinfo *info)
1475 {
1476 	info->rti_info[RTAX_NETMASK] = NULL;
1477 	info->rti_flags |= RTF_HOST;
1478 	info->rti_addrs &= ~RTA_NETMASK;
1479 }
1480 
1481 #ifdef INET
1482 static int
cleanup_xaddrs_inet(struct rt_addrinfo * info,struct linear_buffer * lb)1483 cleanup_xaddrs_inet(struct rt_addrinfo *info, struct linear_buffer *lb)
1484 {
1485 	struct sockaddr_in *dst_sa, *mask_sa;
1486 	const int sa_len = sizeof(struct sockaddr_in);
1487 	struct in_addr dst, mask;
1488 
1489 	/* Check & fixup dst/netmask combination first */
1490 	dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1491 	mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1492 
1493 	/* Ensure reads do not go beyound the buffer size */
1494 	if (SA_SIZE(dst_sa) < offsetof(struct sockaddr_in, sin_zero)) {
1495 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin_len too small: %d",
1496 		    dst_sa->sin_len);
1497 		return (EINVAL);
1498 	}
1499 
1500 	if ((mask_sa != NULL) && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1501 		/*
1502 		 * Some older routing software encode mask length into the
1503 		 * sin_len, thus resulting in "truncated" sockaddr.
1504 		 */
1505 		int len = mask_sa->sin_len - offsetof(struct sockaddr_in, sin_addr);
1506 		if (len >= 0) {
1507 			mask.s_addr = 0;
1508 			if (len > sizeof(struct in_addr))
1509 				len = sizeof(struct in_addr);
1510 			memcpy(&mask, &mask_sa->sin_addr, len);
1511 		} else {
1512 			RTS_PID_LOG(LOG_DEBUG, "prefix mask sin_len too small: %d",
1513 			    mask_sa->sin_len);
1514 			return (EINVAL);
1515 		}
1516 	} else
1517 		mask.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST;
1518 
1519 	dst.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr));
1520 
1521 	/* Construct new "clean" dst/mask sockaddresses */
1522 	if ((dst_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1523 		return (ENOBUFS);
1524 	fill_sockaddr_inet(dst_sa, dst);
1525 	info->rti_info[RTAX_DST] = (struct sockaddr *)dst_sa;
1526 
1527 	if (mask.s_addr != INADDR_BROADCAST) {
1528 		if ((mask_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1529 			return (ENOBUFS);
1530 		fill_sockaddr_inet(mask_sa, mask);
1531 		info->rti_info[RTAX_NETMASK] = (struct sockaddr *)mask_sa;
1532 		info->rti_flags &= ~RTF_HOST;
1533 	} else
1534 		remove_netmask(info);
1535 
1536 	/* Check gateway */
1537 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1538 		return (cleanup_xaddrs_gateway(info, lb));
1539 
1540 	return (0);
1541 }
1542 #endif
1543 
1544 #ifdef INET6
1545 static int
cleanup_xaddrs_inet6(struct rt_addrinfo * info,struct linear_buffer * lb)1546 cleanup_xaddrs_inet6(struct rt_addrinfo *info, struct linear_buffer *lb)
1547 {
1548 	struct sockaddr *sa;
1549 	struct sockaddr_in6 *dst_sa, *mask_sa;
1550 	struct in6_addr mask, *dst;
1551 	const int sa_len = sizeof(struct sockaddr_in6);
1552 
1553 	/* Check & fixup dst/netmask combination first */
1554 	dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1555 	mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1556 
1557 	if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1558 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin6_len too small: %d",
1559 		    dst_sa->sin6_len);
1560 		return (EINVAL);
1561 	}
1562 
1563 	if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1564 		/*
1565 		 * Some older routing software encode mask length into the
1566 		 * sin6_len, thus resulting in "truncated" sockaddr.
1567 		 */
1568 		int len = mask_sa->sin6_len - offsetof(struct sockaddr_in6, sin6_addr);
1569 		if (len >= 0) {
1570 			bzero(&mask, sizeof(mask));
1571 			if (len > sizeof(struct in6_addr))
1572 				len = sizeof(struct in6_addr);
1573 			memcpy(&mask, &mask_sa->sin6_addr, len);
1574 		} else {
1575 			RTS_PID_LOG(LOG_DEBUG, "rtsock: prefix mask sin6_len too small: %d",
1576 			    mask_sa->sin6_len);
1577 			return (EINVAL);
1578 		}
1579 	} else
1580 		mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1581 
1582 	dst = &dst_sa->sin6_addr;
1583 	IN6_MASK_ADDR(dst, &mask);
1584 
1585 	if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1586 		return (ENOBUFS);
1587 	fill_sockaddr_inet6((struct sockaddr_in6 *)sa, dst, 0);
1588 	info->rti_info[RTAX_DST] = sa;
1589 
1590 	if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128)) {
1591 		if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1592 			return (ENOBUFS);
1593 		fill_sockaddr_inet6((struct sockaddr_in6 *)sa, &mask, 0);
1594 		info->rti_info[RTAX_NETMASK] = sa;
1595 		info->rti_flags &= ~RTF_HOST;
1596 	} else
1597 		remove_netmask(info);
1598 
1599 	/* Check gateway */
1600 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1601 		return (cleanup_xaddrs_gateway(info, lb));
1602 
1603 	return (0);
1604 }
1605 #endif
1606 
1607 static int
cleanup_xaddrs(struct rt_addrinfo * info,struct linear_buffer * lb)1608 cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb)
1609 {
1610 	int error = EAFNOSUPPORT;
1611 
1612 	if (info->rti_info[RTAX_DST] == NULL) {
1613 		RTS_PID_LOG(LOG_DEBUG, "prefix dst is not set");
1614 		return (EINVAL);
1615 	}
1616 
1617 	if (info->rti_flags & RTF_LLDATA) {
1618 		/*
1619 		 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1620 		 * prefix along with the actual address in RTA_DST.
1621 		 * Remove netmask to avoid unnecessary address masking.
1622 		 */
1623 		remove_netmask(info);
1624 	}
1625 
1626 	switch (info->rti_info[RTAX_DST]->sa_family) {
1627 #ifdef INET
1628 	case AF_INET:
1629 		error = cleanup_xaddrs_inet(info, lb);
1630 		break;
1631 #endif
1632 #ifdef INET6
1633 	case AF_INET6:
1634 		error = cleanup_xaddrs_inet6(info, lb);
1635 		break;
1636 #endif
1637 	}
1638 
1639 	return (error);
1640 }
1641 
1642 /*
1643  * Fill in @dmask with valid netmask leaving original @smask
1644  * intact. Mostly used with radix netmasks.
1645  */
1646 struct sockaddr *
rtsock_fix_netmask(const struct sockaddr * dst,const struct sockaddr * smask,struct sockaddr_storage * dmask)1647 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1648     struct sockaddr_storage *dmask)
1649 {
1650 	if (dst == NULL || smask == NULL)
1651 		return (NULL);
1652 
1653 	memset(dmask, 0, dst->sa_len);
1654 	memcpy(dmask, smask, smask->sa_len);
1655 	dmask->ss_len = dst->sa_len;
1656 	dmask->ss_family = dst->sa_family;
1657 
1658 	return ((struct sockaddr *)dmask);
1659 }
1660 
1661 /*
1662  * Writes information related to @rtinfo object to newly-allocated mbuf.
1663  * Assumes MCLBYTES is enough to construct any message.
1664  * Used for OS notifications of vaious events (if/ifa announces,etc)
1665  *
1666  * Returns allocated mbuf or NULL on failure.
1667  */
1668 static struct mbuf *
rtsock_msg_mbuf(int type,struct rt_addrinfo * rtinfo)1669 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1670 {
1671 	struct sockaddr_storage ss;
1672 	struct rt_msghdr *rtm;
1673 	struct mbuf *m;
1674 	int i;
1675 	struct sockaddr *sa;
1676 #ifdef INET6
1677 	struct sockaddr_in6 *sin6;
1678 #endif
1679 	int len, dlen;
1680 
1681 	switch (type) {
1682 	case RTM_DELADDR:
1683 	case RTM_NEWADDR:
1684 		len = sizeof(struct ifa_msghdr);
1685 		break;
1686 
1687 	case RTM_DELMADDR:
1688 	case RTM_NEWMADDR:
1689 		len = sizeof(struct ifma_msghdr);
1690 		break;
1691 
1692 	case RTM_IFINFO:
1693 		len = sizeof(struct if_msghdr);
1694 		break;
1695 
1696 	case RTM_IFANNOUNCE:
1697 	case RTM_IEEE80211:
1698 		len = sizeof(struct if_announcemsghdr);
1699 		break;
1700 
1701 	default:
1702 		len = sizeof(struct rt_msghdr);
1703 	}
1704 
1705 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1706 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1707 	if (len > MHLEN)
1708 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1709 	else
1710 		m = m_gethdr(M_NOWAIT, MT_DATA);
1711 	if (m == NULL)
1712 		return (m);
1713 
1714 	m->m_pkthdr.len = m->m_len = len;
1715 	rtm = mtod(m, struct rt_msghdr *);
1716 	bzero((caddr_t)rtm, len);
1717 	for (i = 0; i < RTAX_MAX; i++) {
1718 		if ((sa = rtinfo->rti_info[i]) == NULL)
1719 			continue;
1720 		rtinfo->rti_addrs |= (1 << i);
1721 
1722 		dlen = SA_SIZE(sa);
1723 		KASSERT(dlen <= sizeof(ss),
1724 		    ("%s: sockaddr size overflow", __func__));
1725 		bzero(&ss, sizeof(ss));
1726 		bcopy(sa, &ss, sa->sa_len);
1727 		sa = (struct sockaddr *)&ss;
1728 #ifdef INET6
1729 		if (sa->sa_family == AF_INET6) {
1730 			sin6 = (struct sockaddr_in6 *)sa;
1731 			(void)sa6_recoverscope(sin6);
1732 		}
1733 #endif
1734 		m_copyback(m, len, dlen, (caddr_t)sa);
1735 		len += dlen;
1736 	}
1737 	if (m->m_pkthdr.len != len) {
1738 		m_freem(m);
1739 		return (NULL);
1740 	}
1741 	rtm->rtm_msglen = len;
1742 	rtm->rtm_version = RTM_VERSION;
1743 	rtm->rtm_type = type;
1744 	return (m);
1745 }
1746 
1747 /*
1748  * Writes information related to @rtinfo object to preallocated buffer.
1749  * Stores needed size in @plen. If @w is NULL, calculates size without
1750  * writing.
1751  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1752  *
1753  * Returns 0 on success.
1754  *
1755  */
1756 static int
rtsock_msg_buffer(int type,struct rt_addrinfo * rtinfo,struct walkarg * w,int * plen)1757 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1758 {
1759 	struct sockaddr_storage ss;
1760 	int len, buflen = 0, dlen, i;
1761 	caddr_t cp = NULL;
1762 	struct rt_msghdr *rtm = NULL;
1763 #ifdef INET6
1764 	struct sockaddr_in6 *sin6;
1765 #endif
1766 #ifdef COMPAT_FREEBSD32
1767 	bool compat32 = false;
1768 #endif
1769 
1770 	switch (type) {
1771 	case RTM_DELADDR:
1772 	case RTM_NEWADDR:
1773 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1774 #ifdef COMPAT_FREEBSD32
1775 			if (w->w_req->flags & SCTL_MASK32) {
1776 				len = sizeof(struct ifa_msghdrl32);
1777 				compat32 = true;
1778 			} else
1779 #endif
1780 				len = sizeof(struct ifa_msghdrl);
1781 		} else
1782 			len = sizeof(struct ifa_msghdr);
1783 		break;
1784 
1785 	case RTM_IFINFO:
1786 #ifdef COMPAT_FREEBSD32
1787 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1788 			if (w->w_op == NET_RT_IFLISTL)
1789 				len = sizeof(struct if_msghdrl32);
1790 			else
1791 				len = sizeof(struct if_msghdr32);
1792 			compat32 = true;
1793 			break;
1794 		}
1795 #endif
1796 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1797 			len = sizeof(struct if_msghdrl);
1798 		else
1799 			len = sizeof(struct if_msghdr);
1800 		break;
1801 
1802 	case RTM_NEWMADDR:
1803 		len = sizeof(struct ifma_msghdr);
1804 		break;
1805 
1806 	default:
1807 		len = sizeof(struct rt_msghdr);
1808 	}
1809 
1810 	if (w != NULL) {
1811 		rtm = (struct rt_msghdr *)w->w_tmem;
1812 		buflen = w->w_tmemsize - len;
1813 		cp = (caddr_t)w->w_tmem + len;
1814 	}
1815 
1816 	rtinfo->rti_addrs = 0;
1817 	for (i = 0; i < RTAX_MAX; i++) {
1818 		struct sockaddr *sa;
1819 
1820 		if ((sa = rtinfo->rti_info[i]) == NULL)
1821 			continue;
1822 		rtinfo->rti_addrs |= (1 << i);
1823 #ifdef COMPAT_FREEBSD32
1824 		if (compat32)
1825 			dlen = SA_SIZE32(sa);
1826 		else
1827 #endif
1828 			dlen = SA_SIZE(sa);
1829 		if (cp != NULL && buflen >= dlen) {
1830 			KASSERT(dlen <= sizeof(ss),
1831 			    ("%s: sockaddr size overflow", __func__));
1832 			bzero(&ss, sizeof(ss));
1833 			bcopy(sa, &ss, sa->sa_len);
1834 			sa = (struct sockaddr *)&ss;
1835 #ifdef INET6
1836 			if (sa->sa_family == AF_INET6) {
1837 				sin6 = (struct sockaddr_in6 *)sa;
1838 				(void)sa6_recoverscope(sin6);
1839 			}
1840 #endif
1841 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1842 			cp += dlen;
1843 			buflen -= dlen;
1844 		} else if (cp != NULL) {
1845 			/*
1846 			 * Buffer too small. Count needed size
1847 			 * and return with error.
1848 			 */
1849 			cp = NULL;
1850 		}
1851 
1852 		len += dlen;
1853 	}
1854 
1855 	if (cp != NULL) {
1856 		dlen = ALIGN(len) - len;
1857 		if (buflen < dlen)
1858 			cp = NULL;
1859 		else {
1860 			bzero(cp, dlen);
1861 			cp += dlen;
1862 			buflen -= dlen;
1863 		}
1864 	}
1865 	len = ALIGN(len);
1866 
1867 	if (cp != NULL) {
1868 		/* fill header iff buffer is large enough */
1869 		rtm->rtm_version = RTM_VERSION;
1870 		rtm->rtm_type = type;
1871 		rtm->rtm_msglen = len;
1872 	}
1873 
1874 	*plen = len;
1875 
1876 	if (w != NULL && cp == NULL)
1877 		return (ENOBUFS);
1878 
1879 	return (0);
1880 }
1881 
1882 /*
1883  * This routine is called to generate a message from the routing
1884  * socket indicating that a redirect has occurred, a routing lookup
1885  * has failed, or that a protocol has detected timeouts to a particular
1886  * destination.
1887  */
1888 void
rt_missmsg_fib(int type,struct rt_addrinfo * rtinfo,int flags,int error,int fibnum)1889 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1890     int fibnum)
1891 {
1892 	struct rt_msghdr *rtm;
1893 	struct mbuf *m;
1894 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1895 
1896 	if (V_route_cb.any_count == 0)
1897 		return;
1898 	m = rtsock_msg_mbuf(type, rtinfo);
1899 	if (m == NULL)
1900 		return;
1901 
1902 	if (fibnum != RT_ALL_FIBS) {
1903 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1904 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1905 		M_SETFIB(m, fibnum);
1906 		m->m_flags |= RTS_FILTER_FIB;
1907 	}
1908 
1909 	rtm = mtod(m, struct rt_msghdr *);
1910 	rtm->rtm_flags = RTF_DONE | flags;
1911 	rtm->rtm_errno = error;
1912 	rtm->rtm_addrs = rtinfo->rti_addrs;
1913 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1914 }
1915 
1916 void
rt_missmsg(int type,struct rt_addrinfo * rtinfo,int flags,int error)1917 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1918 {
1919 
1920 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1921 }
1922 
1923 /*
1924  * This routine is called to generate a message from the routing
1925  * socket indicating that the status of a network interface has changed.
1926  */
1927 static void
rtsock_ifmsg(struct ifnet * ifp,int if_flags_mask __unused)1928 rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask __unused)
1929 {
1930 	struct if_msghdr *ifm;
1931 	struct mbuf *m;
1932 	struct rt_addrinfo info;
1933 
1934 	if (V_route_cb.any_count == 0)
1935 		return;
1936 	bzero((caddr_t)&info, sizeof(info));
1937 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1938 	if (m == NULL)
1939 		return;
1940 	ifm = mtod(m, struct if_msghdr *);
1941 	ifm->ifm_index = ifp->if_index;
1942 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1943 	if_data_copy(ifp, &ifm->ifm_data);
1944 	ifm->ifm_addrs = 0;
1945 	rt_dispatch(m, AF_UNSPEC);
1946 }
1947 
1948 /*
1949  * Announce interface address arrival/withdraw.
1950  * Please do not call directly, use rt_addrmsg().
1951  * Assume input data to be valid.
1952  * Returns 0 on success.
1953  */
1954 int
rtsock_addrmsg(int cmd,struct ifaddr * ifa,int fibnum)1955 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1956 {
1957 	struct rt_addrinfo info;
1958 	struct sockaddr *sa;
1959 	int ncmd;
1960 	struct mbuf *m;
1961 	struct ifa_msghdr *ifam;
1962 	struct ifnet *ifp = ifa->ifa_ifp;
1963 	struct sockaddr_storage ss;
1964 
1965 	if (V_route_cb.any_count == 0)
1966 		return (0);
1967 
1968 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1969 
1970 	bzero((caddr_t)&info, sizeof(info));
1971 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1972 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1973 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1974 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1975 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1976 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1977 		return (ENOBUFS);
1978 	ifam = mtod(m, struct ifa_msghdr *);
1979 	ifam->ifam_index = ifp->if_index;
1980 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1981 	ifam->ifam_flags = ifa->ifa_flags;
1982 	ifam->ifam_addrs = info.rti_addrs;
1983 
1984 	if (fibnum != RT_ALL_FIBS) {
1985 		M_SETFIB(m, fibnum);
1986 		m->m_flags |= RTS_FILTER_FIB;
1987 	}
1988 
1989 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1990 
1991 	return (0);
1992 }
1993 
1994 /*
1995  * Announce route addition/removal to rtsock based on @rt data.
1996  * Callers are advives to use rt_routemsg() instead of using this
1997  *  function directly.
1998  * Assume @rt data is consistent.
1999  *
2000  * Returns 0 on success.
2001  */
2002 int
rtsock_routemsg(int cmd,struct rtentry * rt,struct nhop_object * nh,int fibnum)2003 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
2004     int fibnum)
2005 {
2006 	union sockaddr_union dst, mask;
2007 	struct rt_addrinfo info;
2008 
2009 	if (V_route_cb.any_count == 0)
2010 		return (0);
2011 
2012 	int family = rt_get_family(rt);
2013 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
2014 	export_rtaddrs(rt, &dst.sa, &mask.sa);
2015 
2016 	bzero((caddr_t)&info, sizeof(info));
2017 	info.rti_info[RTAX_DST] = &dst.sa;
2018 	info.rti_info[RTAX_NETMASK] = &mask.sa;
2019 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2020 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
2021 	info.rti_ifp = nh->nh_ifp;
2022 
2023 	return (rtsock_routemsg_info(cmd, &info, fibnum));
2024 }
2025 
2026 int
rtsock_routemsg_info(int cmd,struct rt_addrinfo * info,int fibnum)2027 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
2028 {
2029 	struct rt_msghdr *rtm;
2030 	struct sockaddr *sa;
2031 	struct mbuf *m;
2032 
2033 	if (V_route_cb.any_count == 0)
2034 		return (0);
2035 
2036 	if (info->rti_flags & RTF_HOST)
2037 		info->rti_info[RTAX_NETMASK] = NULL;
2038 
2039 	m = rtsock_msg_mbuf(cmd, info);
2040 	if (m == NULL)
2041 		return (ENOBUFS);
2042 
2043 	if (fibnum != RT_ALL_FIBS) {
2044 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
2045 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
2046 		M_SETFIB(m, fibnum);
2047 		m->m_flags |= RTS_FILTER_FIB;
2048 	}
2049 
2050 	rtm = mtod(m, struct rt_msghdr *);
2051 	rtm->rtm_addrs = info->rti_addrs;
2052 	if (info->rti_ifp != NULL)
2053 		rtm->rtm_index = info->rti_ifp->if_index;
2054 	/* Add RTF_DONE to indicate command 'completion' required by API */
2055 	info->rti_flags |= RTF_DONE;
2056 	/* Reported routes has to be up */
2057 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
2058 		info->rti_flags |= RTF_UP;
2059 	rtm->rtm_flags = info->rti_flags;
2060 
2061 	sa = info->rti_info[RTAX_DST];
2062 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2063 
2064 	return (0);
2065 }
2066 
2067 /*
2068  * This is the analogue to the rt_newaddrmsg which performs the same
2069  * function but for multicast group memberhips.  This is easier since
2070  * there is no route state to worry about.
2071  */
2072 void
rt_newmaddrmsg(int cmd,struct ifmultiaddr * ifma)2073 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
2074 {
2075 	struct rt_addrinfo info;
2076 	struct mbuf *m = NULL;
2077 	struct ifnet *ifp = ifma->ifma_ifp;
2078 	struct ifma_msghdr *ifmam;
2079 
2080 	if (V_route_cb.any_count == 0)
2081 		return;
2082 
2083 	bzero((caddr_t)&info, sizeof(info));
2084 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2085 	if (ifp && ifp->if_addr)
2086 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2087 	else
2088 		info.rti_info[RTAX_IFP] = NULL;
2089 	/*
2090 	 * If a link-layer address is present, present it as a ``gateway''
2091 	 * (similarly to how ARP entries, e.g., are presented).
2092 	 */
2093 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
2094 	m = rtsock_msg_mbuf(cmd, &info);
2095 	if (m == NULL)
2096 		return;
2097 	ifmam = mtod(m, struct ifma_msghdr *);
2098 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
2099 	    __func__));
2100 	ifmam->ifmam_index = ifp->if_index;
2101 	ifmam->ifmam_addrs = info.rti_addrs;
2102 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
2103 }
2104 
2105 static struct mbuf *
rt_makeifannouncemsg(struct ifnet * ifp,int type,int what,struct rt_addrinfo * info)2106 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
2107 	struct rt_addrinfo *info)
2108 {
2109 	struct if_announcemsghdr *ifan;
2110 	struct mbuf *m;
2111 
2112 	if (V_route_cb.any_count == 0)
2113 		return NULL;
2114 	bzero((caddr_t)info, sizeof(*info));
2115 	m = rtsock_msg_mbuf(type, info);
2116 	if (m != NULL) {
2117 		ifan = mtod(m, struct if_announcemsghdr *);
2118 		ifan->ifan_index = ifp->if_index;
2119 		strlcpy(ifan->ifan_name, ifp->if_xname,
2120 			sizeof(ifan->ifan_name));
2121 		ifan->ifan_what = what;
2122 	}
2123 	return m;
2124 }
2125 
2126 /*
2127  * This is called to generate routing socket messages indicating
2128  * IEEE80211 wireless events.
2129  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2130  */
2131 void
rt_ieee80211msg(struct ifnet * ifp,int what,void * data,size_t data_len)2132 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2133 {
2134 	struct mbuf *m;
2135 	struct rt_addrinfo info;
2136 
2137 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2138 	if (m != NULL) {
2139 		/*
2140 		 * Append the ieee80211 data.  Try to stick it in the
2141 		 * mbuf containing the ifannounce msg; otherwise allocate
2142 		 * a new mbuf and append.
2143 		 *
2144 		 * NB: we assume m is a single mbuf.
2145 		 */
2146 		if (data_len > M_TRAILINGSPACE(m)) {
2147 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2148 			if (n == NULL) {
2149 				m_freem(m);
2150 				return;
2151 			}
2152 			bcopy(data, mtod(n, void *), data_len);
2153 			n->m_len = data_len;
2154 			m->m_next = n;
2155 		} else if (data_len > 0) {
2156 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2157 			m->m_len += data_len;
2158 		}
2159 		if (m->m_flags & M_PKTHDR)
2160 			m->m_pkthdr.len += data_len;
2161 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2162 		rt_dispatch(m, AF_UNSPEC);
2163 	}
2164 }
2165 
2166 /*
2167  * This is called to generate routing socket messages indicating
2168  * network interface arrival and departure.
2169  */
2170 static void
rt_ifannouncemsg(struct ifnet * ifp,int what)2171 rt_ifannouncemsg(struct ifnet *ifp, int what)
2172 {
2173 	struct mbuf *m;
2174 	struct rt_addrinfo info;
2175 
2176 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2177 	if (m != NULL)
2178 		rt_dispatch(m, AF_UNSPEC);
2179 }
2180 
2181 static void
rt_dispatch(struct mbuf * m,sa_family_t saf)2182 rt_dispatch(struct mbuf *m, sa_family_t saf)
2183 {
2184 
2185 	M_ASSERTPKTHDR(m);
2186 
2187 	m->m_rtsock_family = saf;
2188 	if (V_loif)
2189 		m->m_pkthdr.rcvif = V_loif;
2190 	else {
2191 		m_freem(m);
2192 		return;
2193 	}
2194 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
2195 }
2196 
2197 /*
2198  * This is used in dumping the kernel table via sysctl().
2199  */
2200 static int
sysctl_dumpentry(struct rtentry * rt,void * vw)2201 sysctl_dumpentry(struct rtentry *rt, void *vw)
2202 {
2203 	struct walkarg *w = vw;
2204 	struct nhop_object *nh;
2205 
2206 	NET_EPOCH_ASSERT();
2207 
2208 	if (!rt_is_exportable(rt, w->w_req->td->td_ucred))
2209 		return (0);
2210 
2211 	export_rtaddrs(rt, w->dst, w->mask);
2212 	nh = rt_get_raw_nhop(rt);
2213 #ifdef ROUTE_MPATH
2214 	if (NH_IS_NHGRP(nh)) {
2215 		const struct weightened_nhop *wn;
2216 		uint32_t num_nhops;
2217 		int error;
2218 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2219 		for (int i = 0; i < num_nhops; i++) {
2220 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2221 			if (error != 0)
2222 				return (error);
2223 		}
2224 	} else
2225 #endif
2226 		sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2227 
2228 	return (0);
2229 }
2230 
2231 
2232 static int
sysctl_dumpnhop(struct rtentry * rt,struct nhop_object * nh,uint32_t weight,struct walkarg * w)2233 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2234     struct walkarg *w)
2235 {
2236 	struct rt_addrinfo info;
2237 	int error = 0, size;
2238 	uint32_t rtflags;
2239 
2240 	rtflags = nhop_get_rtflags(nh);
2241 
2242 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2243 		return (0);
2244 
2245 	bzero((caddr_t)&info, sizeof(info));
2246 	info.rti_info[RTAX_DST] = w->dst;
2247 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2248 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2249 	info.rti_info[RTAX_GENMASK] = 0;
2250 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2251 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2252 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2253 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2254 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2255 	}
2256 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2257 		return (error);
2258 	if (w->w_req && w->w_tmem) {
2259 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2260 
2261 		bzero(&rtm->rtm_index,
2262 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2263 
2264 		/*
2265 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2266 		 * and RTF_UP (if entry is linked, which is always true here).
2267 		 * Given that, use nhop rtflags & add RTF_UP.
2268 		 */
2269 		rtm->rtm_flags = rtflags | RTF_UP;
2270 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2271 			rtm->rtm_flags = RTF_GATEWAY |
2272 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2273 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2274 		rtm->rtm_rmx.rmx_weight = weight;
2275 		rtm->rtm_index = nh->nh_ifp->if_index;
2276 		rtm->rtm_addrs = info.rti_addrs;
2277 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2278 		return (error);
2279 	}
2280 	return (error);
2281 }
2282 
2283 static int
sysctl_iflist_ifml(struct ifnet * ifp,const struct if_data * src_ifd,struct rt_addrinfo * info,struct walkarg * w,int len)2284 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2285     struct rt_addrinfo *info, struct walkarg *w, int len)
2286 {
2287 	struct if_msghdrl *ifm;
2288 	struct if_data *ifd;
2289 
2290 	ifm = (struct if_msghdrl *)w->w_tmem;
2291 
2292 #ifdef COMPAT_FREEBSD32
2293 	if (w->w_req->flags & SCTL_MASK32) {
2294 		struct if_msghdrl32 *ifm32;
2295 
2296 		ifm32 = (struct if_msghdrl32 *)ifm;
2297 		ifm32->ifm_addrs = info->rti_addrs;
2298 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2299 		ifm32->ifm_index = ifp->if_index;
2300 		ifm32->_ifm_spare1 = 0;
2301 		ifm32->ifm_len = sizeof(*ifm32);
2302 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2303 		ifm32->_ifm_spare2 = 0;
2304 		ifd = &ifm32->ifm_data;
2305 	} else
2306 #endif
2307 	{
2308 		ifm->ifm_addrs = info->rti_addrs;
2309 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2310 		ifm->ifm_index = ifp->if_index;
2311 		ifm->_ifm_spare1 = 0;
2312 		ifm->ifm_len = sizeof(*ifm);
2313 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2314 		ifm->_ifm_spare2 = 0;
2315 		ifd = &ifm->ifm_data;
2316 	}
2317 
2318 	memcpy(ifd, src_ifd, sizeof(*ifd));
2319 
2320 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2321 }
2322 
2323 static int
sysctl_iflist_ifm(struct ifnet * ifp,const struct if_data * src_ifd,struct rt_addrinfo * info,struct walkarg * w,int len)2324 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2325     struct rt_addrinfo *info, struct walkarg *w, int len)
2326 {
2327 	struct if_msghdr *ifm;
2328 	struct if_data *ifd;
2329 
2330 	ifm = (struct if_msghdr *)w->w_tmem;
2331 
2332 #ifdef COMPAT_FREEBSD32
2333 	if (w->w_req->flags & SCTL_MASK32) {
2334 		struct if_msghdr32 *ifm32;
2335 
2336 		ifm32 = (struct if_msghdr32 *)ifm;
2337 		ifm32->ifm_addrs = info->rti_addrs;
2338 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2339 		ifm32->ifm_index = ifp->if_index;
2340 		ifm32->_ifm_spare1 = 0;
2341 		ifd = &ifm32->ifm_data;
2342 	} else
2343 #endif
2344 	{
2345 		ifm->ifm_addrs = info->rti_addrs;
2346 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2347 		ifm->ifm_index = ifp->if_index;
2348 		ifm->_ifm_spare1 = 0;
2349 		ifd = &ifm->ifm_data;
2350 	}
2351 
2352 	memcpy(ifd, src_ifd, sizeof(*ifd));
2353 
2354 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2355 }
2356 
2357 static int
sysctl_iflist_ifaml(struct ifaddr * ifa,struct rt_addrinfo * info,struct walkarg * w,int len)2358 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2359     struct walkarg *w, int len)
2360 {
2361 	struct ifa_msghdrl *ifam;
2362 	struct if_data *ifd;
2363 
2364 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2365 
2366 #ifdef COMPAT_FREEBSD32
2367 	if (w->w_req->flags & SCTL_MASK32) {
2368 		struct ifa_msghdrl32 *ifam32;
2369 
2370 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2371 		ifam32->ifam_addrs = info->rti_addrs;
2372 		ifam32->ifam_flags = ifa->ifa_flags;
2373 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2374 		ifam32->_ifam_spare1 = 0;
2375 		ifam32->ifam_len = sizeof(*ifam32);
2376 		ifam32->ifam_data_off =
2377 		    offsetof(struct ifa_msghdrl32, ifam_data);
2378 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2379 		ifd = &ifam32->ifam_data;
2380 	} else
2381 #endif
2382 	{
2383 		ifam->ifam_addrs = info->rti_addrs;
2384 		ifam->ifam_flags = ifa->ifa_flags;
2385 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2386 		ifam->_ifam_spare1 = 0;
2387 		ifam->ifam_len = sizeof(*ifam);
2388 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2389 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2390 		ifd = &ifam->ifam_data;
2391 	}
2392 
2393 	bzero(ifd, sizeof(*ifd));
2394 	ifd->ifi_datalen = sizeof(struct if_data);
2395 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2396 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2397 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2398 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2399 
2400 	/* Fixup if_data carp(4) vhid. */
2401 	if (carp_get_vhid_p != NULL)
2402 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2403 
2404 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2405 }
2406 
2407 static int
sysctl_iflist_ifam(struct ifaddr * ifa,struct rt_addrinfo * info,struct walkarg * w,int len)2408 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2409     struct walkarg *w, int len)
2410 {
2411 	struct ifa_msghdr *ifam;
2412 
2413 	ifam = (struct ifa_msghdr *)w->w_tmem;
2414 	ifam->ifam_addrs = info->rti_addrs;
2415 	ifam->ifam_flags = ifa->ifa_flags;
2416 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2417 	ifam->_ifam_spare1 = 0;
2418 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2419 
2420 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2421 }
2422 
2423 static int
sysctl_iflist(int af,struct walkarg * w)2424 sysctl_iflist(int af, struct walkarg *w)
2425 {
2426 	struct ifnet *ifp;
2427 	struct ifaddr *ifa;
2428 	struct if_data ifd;
2429 	struct rt_addrinfo info;
2430 	int len, error = 0;
2431 	struct sockaddr_storage ss;
2432 
2433 	bzero((caddr_t)&info, sizeof(info));
2434 	bzero(&ifd, sizeof(ifd));
2435 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2436 		if (w->w_arg && w->w_arg != ifp->if_index)
2437 			continue;
2438 		if_data_copy(ifp, &ifd);
2439 		ifa = ifp->if_addr;
2440 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2441 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2442 		if (error != 0)
2443 			goto done;
2444 		info.rti_info[RTAX_IFP] = NULL;
2445 		if (w->w_req && w->w_tmem) {
2446 			if (w->w_op == NET_RT_IFLISTL)
2447 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2448 				    len);
2449 			else
2450 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2451 				    len);
2452 			if (error)
2453 				goto done;
2454 		}
2455 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2456 			if (af && af != ifa->ifa_addr->sa_family)
2457 				continue;
2458 			if (prison_if(w->w_req->td->td_ucred,
2459 			    ifa->ifa_addr) != 0)
2460 				continue;
2461 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2462 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2463 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2464 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2465 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2466 			if (error != 0)
2467 				goto done;
2468 			if (w->w_req && w->w_tmem) {
2469 				if (w->w_op == NET_RT_IFLISTL)
2470 					error = sysctl_iflist_ifaml(ifa, &info,
2471 					    w, len);
2472 				else
2473 					error = sysctl_iflist_ifam(ifa, &info,
2474 					    w, len);
2475 				if (error)
2476 					goto done;
2477 			}
2478 		}
2479 		info.rti_info[RTAX_IFA] = NULL;
2480 		info.rti_info[RTAX_NETMASK] = NULL;
2481 		info.rti_info[RTAX_BRD] = NULL;
2482 	}
2483 done:
2484 	return (error);
2485 }
2486 
2487 static int
sysctl_ifmalist(int af,struct walkarg * w)2488 sysctl_ifmalist(int af, struct walkarg *w)
2489 {
2490 	struct rt_addrinfo info;
2491 	struct ifaddr *ifa;
2492 	struct ifmultiaddr *ifma;
2493 	struct ifnet *ifp;
2494 	int error, len;
2495 
2496 	NET_EPOCH_ASSERT();
2497 
2498 	error = 0;
2499 	bzero((caddr_t)&info, sizeof(info));
2500 
2501 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2502 		if (w->w_arg && w->w_arg != ifp->if_index)
2503 			continue;
2504 		ifa = ifp->if_addr;
2505 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2506 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2507 			if (af && af != ifma->ifma_addr->sa_family)
2508 				continue;
2509 			if (prison_if(w->w_req->td->td_ucred,
2510 			    ifma->ifma_addr) != 0)
2511 				continue;
2512 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2513 			info.rti_info[RTAX_GATEWAY] =
2514 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2515 			    ifma->ifma_lladdr : NULL;
2516 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2517 			if (error != 0)
2518 				break;
2519 			if (w->w_req && w->w_tmem) {
2520 				struct ifma_msghdr *ifmam;
2521 
2522 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2523 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2524 				ifmam->ifmam_flags = 0;
2525 				ifmam->ifmam_addrs = info.rti_addrs;
2526 				ifmam->_ifmam_spare1 = 0;
2527 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2528 				if (error != 0)
2529 					break;
2530 			}
2531 		}
2532 		if (error != 0)
2533 			break;
2534 	}
2535 	return (error);
2536 }
2537 
2538 static void
rtable_sysctl_dump(uint32_t fibnum,int family,struct walkarg * w)2539 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2540 {
2541 	union sockaddr_union sa_dst, sa_mask;
2542 
2543 	w->family = family;
2544 	w->dst = (struct sockaddr *)&sa_dst;
2545 	w->mask = (struct sockaddr *)&sa_mask;
2546 
2547 	init_sockaddrs_family(family, w->dst, w->mask);
2548 
2549 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2550 }
2551 
2552 static int
sysctl_rtsock(SYSCTL_HANDLER_ARGS)2553 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2554 {
2555 	struct epoch_tracker et;
2556 	int	*name = (int *)arg1;
2557 	u_int	namelen = arg2;
2558 	struct rib_head *rnh = NULL; /* silence compiler. */
2559 	int	i, lim, error = EINVAL;
2560 	int	fib = 0;
2561 	u_char	af;
2562 	struct	walkarg w;
2563 
2564 	if (namelen < 3)
2565 		return (EINVAL);
2566 
2567 	name++;
2568 	namelen--;
2569 	if (req->newptr)
2570 		return (EPERM);
2571 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2572 		if (namelen == 3)
2573 			fib = req->td->td_proc->p_fibnum;
2574 		else if (namelen == 4)
2575 			fib = (name[3] == RT_ALL_FIBS) ?
2576 			    req->td->td_proc->p_fibnum : name[3];
2577 		else
2578 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2579 		if (fib < 0 || fib >= rt_numfibs)
2580 			return (EINVAL);
2581 	} else if (namelen != 3)
2582 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2583 	af = name[0];
2584 	if (af > AF_MAX)
2585 		return (EINVAL);
2586 	bzero(&w, sizeof(w));
2587 	w.w_op = name[1];
2588 	w.w_arg = name[2];
2589 	w.w_req = req;
2590 
2591 	error = sysctl_wire_old_buffer(req, 0);
2592 	if (error)
2593 		return (error);
2594 
2595 	/*
2596 	 * Allocate reply buffer in advance.
2597 	 * All rtsock messages has maximum length of u_short.
2598 	 */
2599 	w.w_tmemsize = 65536;
2600 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2601 
2602 	NET_EPOCH_ENTER(et);
2603 	switch (w.w_op) {
2604 	case NET_RT_DUMP:
2605 	case NET_RT_FLAGS:
2606 		if (af == 0) {			/* dump all tables */
2607 			i = 1;
2608 			lim = AF_MAX;
2609 		} else				/* dump only one table */
2610 			i = lim = af;
2611 
2612 		/*
2613 		 * take care of llinfo entries, the caller must
2614 		 * specify an AF
2615 		 */
2616 		if (w.w_op == NET_RT_FLAGS &&
2617 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2618 			if (af != 0)
2619 				error = lltable_sysctl_dumparp(af, w.w_req);
2620 			else
2621 				error = EINVAL;
2622 			break;
2623 		}
2624 		/*
2625 		 * take care of routing entries
2626 		 */
2627 		for (error = 0; error == 0 && i <= lim; i++) {
2628 			rnh = rt_tables_get_rnh(fib, i);
2629 			if (rnh != NULL) {
2630 				rtable_sysctl_dump(fib, i, &w);
2631 			} else if (af != 0)
2632 				error = EAFNOSUPPORT;
2633 		}
2634 		break;
2635 	case NET_RT_NHOP:
2636 	case NET_RT_NHGRP:
2637 		/* Allow dumping one specific af/fib at a time */
2638 		if (namelen < 4) {
2639 			error = EINVAL;
2640 			break;
2641 		}
2642 		fib = name[3];
2643 		if (fib < 0 || fib > rt_numfibs) {
2644 			error = EINVAL;
2645 			break;
2646 		}
2647 		rnh = rt_tables_get_rnh(fib, af);
2648 		if (rnh == NULL) {
2649 			error = EAFNOSUPPORT;
2650 			break;
2651 		}
2652 		if (w.w_op == NET_RT_NHOP)
2653 			error = nhops_dump_sysctl(rnh, w.w_req);
2654 		else
2655 #ifdef ROUTE_MPATH
2656 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2657 #else
2658 			error = ENOTSUP;
2659 #endif
2660 		break;
2661 	case NET_RT_IFLIST:
2662 	case NET_RT_IFLISTL:
2663 		error = sysctl_iflist(af, &w);
2664 		break;
2665 
2666 	case NET_RT_IFMALIST:
2667 		error = sysctl_ifmalist(af, &w);
2668 		break;
2669 	}
2670 	NET_EPOCH_EXIT(et);
2671 
2672 	free(w.w_tmem, M_TEMP);
2673 	return (error);
2674 }
2675 
2676 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2677     sysctl_rtsock, "Return route tables and interface/address lists");
2678 
2679 /*
2680  * Definitions of protocols supported in the ROUTE domain.
2681  */
2682 
2683 static struct domain routedomain;		/* or at least forward */
2684 
2685 static struct protosw routesw = {
2686 	.pr_type =		SOCK_RAW,
2687 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2688 	.pr_abort =		rts_close,
2689 	.pr_attach =		rts_attach,
2690 	.pr_detach =		rts_detach,
2691 	.pr_send =		rts_send,
2692 	.pr_shutdown =		rts_shutdown,
2693 	.pr_disconnect =	rts_disconnect,
2694 	.pr_close =		rts_close,
2695 };
2696 
2697 static struct domain routedomain = {
2698 	.dom_family =		PF_ROUTE,
2699 	.dom_name =		"route",
2700 	.dom_nprotosw =		1,
2701 	.dom_protosw =		{ &routesw },
2702 };
2703 
2704 DOMAIN_SET(route);
2705