xref: /freebsd-13-stable/sys/net/rtsock.c (revision 1a584160e3f4e2c7db60b8a1ed7a4d0be2017216)
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_dl.h>
60 #include <net/if_llatbl.h>
61 #include <net/if_types.h>
62 #include <net/netisr.h>
63 #include <net/raw_cb.h>
64 #include <net/route.h>
65 #include <net/route/route_ctl.h>
66 #include <net/route/route_var.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/if_ether.h>
71 #include <netinet/ip_carp.h>
72 #ifdef INET6
73 #include <netinet6/in6_var.h>
74 #include <netinet6/ip6_var.h>
75 #include <netinet6/scope6_var.h>
76 #endif
77 #include <net/route/nhop.h>
78 
79 #define	DEBUG_MOD_NAME	rtsock
80 #define	DEBUG_MAX_LEVEL	LOG_DEBUG
81 #include <net/route/route_debug.h>
82 _DECLARE_DEBUG(LOG_INFO);
83 
84 #ifdef COMPAT_FREEBSD32
85 #include <sys/mount.h>
86 #include <compat/freebsd32/freebsd32.h>
87 
88 struct if_msghdr32 {
89 	uint16_t ifm_msglen;
90 	uint8_t	ifm_version;
91 	uint8_t	ifm_type;
92 	int32_t	ifm_addrs;
93 	int32_t	ifm_flags;
94 	uint16_t ifm_index;
95 	uint16_t _ifm_spare1;
96 	struct	if_data ifm_data;
97 };
98 
99 struct if_msghdrl32 {
100 	uint16_t ifm_msglen;
101 	uint8_t	ifm_version;
102 	uint8_t	ifm_type;
103 	int32_t	ifm_addrs;
104 	int32_t	ifm_flags;
105 	uint16_t ifm_index;
106 	uint16_t _ifm_spare1;
107 	uint16_t ifm_len;
108 	uint16_t ifm_data_off;
109 	uint32_t _ifm_spare2;
110 	struct	if_data ifm_data;
111 };
112 
113 struct ifa_msghdrl32 {
114 	uint16_t ifam_msglen;
115 	uint8_t	ifam_version;
116 	uint8_t	ifam_type;
117 	int32_t	ifam_addrs;
118 	int32_t	ifam_flags;
119 	uint16_t ifam_index;
120 	uint16_t _ifam_spare1;
121 	uint16_t ifam_len;
122 	uint16_t ifam_data_off;
123 	int32_t	ifam_metric;
124 	struct	if_data ifam_data;
125 };
126 
127 #define SA_SIZE32(sa)						\
128     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
129 	sizeof(int)		:				\
130 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
131 
132 #endif /* COMPAT_FREEBSD32 */
133 
134 struct linear_buffer {
135 	char		*base;	/* Base allocated memory pointer */
136 	uint32_t	offset;	/* Currently used offset */
137 	uint32_t	size;	/* Total buffer size */
138 };
139 #define	SCRATCH_BUFFER_SIZE	1024
140 
141 #define	RTS_PID_LOG(_l, _fmt, ...)	RT_LOG_##_l(_l, "PID %d: " _fmt, curproc ? curproc->p_pid : 0, ## __VA_ARGS__)
142 
143 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
144 
145 /* NB: these are not modified */
146 static struct	sockaddr route_src = { 2, PF_ROUTE, };
147 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
148 
149 /* These are external hooks for CARP. */
150 int	(*carp_get_vhid_p)(struct ifaddr *);
151 
152 /*
153  * Used by rtsock/raw_input 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 typedef struct {
159 	int	ip_count;	/* attached w/ AF_INET */
160 	int	ip6_count;	/* attached w/ AF_INET6 */
161 	int	any_count;	/* total attached */
162 } route_cb_t;
163 VNET_DEFINE_STATIC(route_cb_t, route_cb);
164 #define	V_route_cb VNET(route_cb)
165 
166 struct mtx rtsock_mtx;
167 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
168 
169 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
170 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
171 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
172 
173 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
174 
175 struct walkarg {
176 	int	family;
177 	int	w_tmemsize;
178 	int	w_op, w_arg;
179 	caddr_t	w_tmem;
180 	struct sysctl_req *w_req;
181 	struct sockaddr *dst;
182 	struct sockaddr *mask;
183 };
184 
185 static void	rts_input(struct mbuf *m);
186 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
187 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
188 			struct walkarg *w, int *plen);
189 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
190 			struct rt_addrinfo *rtinfo);
191 static int	cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb);
192 static int	sysctl_dumpentry(struct rtentry *rt, void *vw);
193 static int	sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
194 			uint32_t weight, struct walkarg *w);
195 static int	sysctl_iflist(int af, struct walkarg *w);
196 static int	sysctl_ifmalist(int af, struct walkarg *w);
197 static int	route_output(struct mbuf *m, struct socket *so, ...);
198 static void	rt_getmetrics(const struct rtentry *rt,
199 			const struct nhop_object *nh, struct rt_metrics *out);
200 static void	rt_dispatch(struct mbuf *, sa_family_t);
201 static void	rt_ifannouncemsg(struct ifnet *ifp, int what);
202 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
203 			struct rt_msghdr *rtm, struct rib_cmd_info *rc);
204 static int	update_rtm_from_rc(struct rt_addrinfo *info,
205 			struct rt_msghdr **prtm, int alloc_len,
206 			struct rib_cmd_info *rc, struct nhop_object *nh);
207 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
208 			struct mbuf *m, sa_family_t saf, u_int fibnum,
209 			int rtm_errno);
210 static void	rtsock_notify_event(uint32_t fibnum, const struct rib_cmd_info *rc);
211 static void	rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask);
212 
213 static struct netisr_handler rtsock_nh = {
214 	.nh_name = "rtsock",
215 	.nh_handler = rts_input,
216 	.nh_proto = NETISR_ROUTE,
217 	.nh_policy = NETISR_POLICY_SOURCE,
218 };
219 
220 static int
sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)221 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
222 {
223 	int error, qlimit;
224 
225 	netisr_getqlimit(&rtsock_nh, &qlimit);
226 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
227         if (error || !req->newptr)
228                 return (error);
229 	if (qlimit < 1)
230 		return (EINVAL);
231 	return (netisr_setqlimit(&rtsock_nh, qlimit));
232 }
233 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
234     CTLTYPE_INT | CTLFLAG_RWTUN | CTLFLAG_NOFETCH | CTLFLAG_MPSAFE,
235     0, 0, sysctl_route_netisr_maxqlen, "I",
236     "maximum routing socket dispatch queue length");
237 
238 static void
vnet_rts_init(void)239 vnet_rts_init(void)
240 {
241 	int tmp;
242 
243 	if (IS_DEFAULT_VNET(curvnet)) {
244 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
245 			rtsock_nh.nh_qlimit = tmp;
246 		netisr_register(&rtsock_nh);
247 	}
248 #ifdef VIMAGE
249 	 else
250 		netisr_register_vnet(&rtsock_nh);
251 #endif
252 }
253 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
254     vnet_rts_init, NULL);
255 
256 #ifdef VIMAGE
257 static void
vnet_rts_uninit(void)258 vnet_rts_uninit(void)
259 {
260 
261 	netisr_unregister_vnet(&rtsock_nh);
262 }
263 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
264     vnet_rts_uninit, NULL);
265 #endif
266 
267 static void
report_route_event(const struct rib_cmd_info * rc,void * _cbdata)268 report_route_event(const struct rib_cmd_info *rc, void *_cbdata)
269 {
270 	uint32_t fibnum = (uint32_t)(uintptr_t)_cbdata;
271 	struct nhop_object *nh;
272 
273 	nh = rc->rc_cmd == RTM_DELETE ? rc->rc_nh_old : rc->rc_nh_new;
274 	rt_routemsg(rc->rc_cmd, rc->rc_rt, nh, fibnum);
275 }
276 
277 static void
rts_handle_route_event(uint32_t fibnum,const struct rib_cmd_info * rc)278 rts_handle_route_event(uint32_t fibnum, const struct rib_cmd_info *rc)
279 {
280 #ifdef ROUTE_MPATH
281 	if ((rc->rc_nh_new && NH_IS_NHGRP(rc->rc_nh_new)) ||
282 	    (rc->rc_nh_old && NH_IS_NHGRP(rc->rc_nh_old))) {
283 		rib_decompose_notification(rc, report_route_event,
284 		    (void *)(uintptr_t)fibnum);
285 	} else
286 #endif
287 		report_route_event(rc, (void *)(uintptr_t)fibnum);
288 }
289 static struct rtbridge rtsbridge = {
290 	.route_f = rts_handle_route_event,
291 	.ifmsg_f = rtsock_ifmsg,
292 };
293 static struct rtbridge *rtsbridge_orig_p;
294 
295 static void
rtsock_notify_event(uint32_t fibnum,const struct rib_cmd_info * rc)296 rtsock_notify_event(uint32_t fibnum, const struct rib_cmd_info *rc)
297 {
298 	netlink_callback_p->route_f(fibnum, rc);
299 }
300 
301 static void
rtsock_init(void)302 rtsock_init(void)
303 {
304 	rtsbridge_orig_p = rtsock_callback_p;
305 	rtsock_callback_p = &rtsbridge;
306 }
307 SYSINIT(rtsock_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rtsock_init, NULL);
308 
309 static void
rts_handle_ifnet_arrival(void * arg __unused,struct ifnet * ifp)310 rts_handle_ifnet_arrival(void *arg __unused, struct ifnet *ifp)
311 {
312 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
313 }
314 EVENTHANDLER_DEFINE(ifnet_arrival_event, rts_handle_ifnet_arrival, NULL, 0);
315 
316 static void
rts_handle_ifnet_departure(void * arg __unused,struct ifnet * ifp)317 rts_handle_ifnet_departure(void *arg __unused, struct ifnet *ifp)
318 {
319 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
320 }
321 EVENTHANDLER_DEFINE(ifnet_departure_event, rts_handle_ifnet_departure, NULL, 0);
322 
323 static int
raw_input_rts_cb(struct mbuf * m,struct sockproto * proto,struct sockaddr * src,struct rawcb * rp)324 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
325     struct rawcb *rp)
326 {
327 	int fibnum;
328 
329 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
330 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
331 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
332 
333 	/* No filtering requested. */
334 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
335 		return (0);
336 
337 	/* Check if it is a rts and the fib matches the one of the socket. */
338 	fibnum = M_GETFIB(m);
339 	if (proto->sp_family != PF_ROUTE ||
340 	    rp->rcb_socket == NULL ||
341 	    rp->rcb_socket->so_fibnum == fibnum)
342 		return (0);
343 
344 	/* Filtering requested and no match, the socket shall be skipped. */
345 	return (1);
346 }
347 
348 static void
rts_input(struct mbuf * m)349 rts_input(struct mbuf *m)
350 {
351 	struct sockproto route_proto;
352 	unsigned short *family;
353 	struct m_tag *tag;
354 
355 	route_proto.sp_family = PF_ROUTE;
356 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
357 	if (tag != NULL) {
358 		family = (unsigned short *)(tag + 1);
359 		route_proto.sp_protocol = *family;
360 		m_tag_delete(m, tag);
361 	} else
362 		route_proto.sp_protocol = 0;
363 
364 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
365 }
366 
367 /*
368  * It really doesn't make any sense at all for this code to share much
369  * with raw_usrreq.c, since its functionality is so restricted.  XXX
370  */
371 static void
rts_abort(struct socket * so)372 rts_abort(struct socket *so)
373 {
374 
375 	raw_usrreqs.pru_abort(so);
376 }
377 
378 static void
rts_close(struct socket * so)379 rts_close(struct socket *so)
380 {
381 
382 	raw_usrreqs.pru_close(so);
383 }
384 
385 /* pru_accept is EOPNOTSUPP */
386 
387 static int
rts_attach(struct socket * so,int proto,struct thread * td)388 rts_attach(struct socket *so, int proto, struct thread *td)
389 {
390 	struct rawcb *rp;
391 	int error;
392 
393 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
394 
395 	/* XXX */
396 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
397 
398 	so->so_pcb = (caddr_t)rp;
399 	so->so_fibnum = td->td_proc->p_fibnum;
400 	error = raw_attach(so, proto);
401 	rp = sotorawcb(so);
402 	if (error) {
403 		so->so_pcb = NULL;
404 		free(rp, M_PCB);
405 		return error;
406 	}
407 	RTSOCK_LOCK();
408 	switch(rp->rcb_proto.sp_protocol) {
409 	case AF_INET:
410 		V_route_cb.ip_count++;
411 		break;
412 	case AF_INET6:
413 		V_route_cb.ip6_count++;
414 		break;
415 	}
416 	V_route_cb.any_count++;
417 	RTSOCK_UNLOCK();
418 	soisconnected(so);
419 	so->so_options |= SO_USELOOPBACK;
420 	return 0;
421 }
422 
423 static int
rts_bind(struct socket * so,struct sockaddr * nam,struct thread * td)424 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
425 {
426 
427 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
428 }
429 
430 static int
rts_connect(struct socket * so,struct sockaddr * nam,struct thread * td)431 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
432 {
433 
434 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
435 }
436 
437 /* pru_connect2 is EOPNOTSUPP */
438 /* pru_control is EOPNOTSUPP */
439 
440 static void
rts_detach(struct socket * so)441 rts_detach(struct socket *so)
442 {
443 	struct rawcb *rp = sotorawcb(so);
444 
445 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
446 
447 	RTSOCK_LOCK();
448 	switch(rp->rcb_proto.sp_protocol) {
449 	case AF_INET:
450 		V_route_cb.ip_count--;
451 		break;
452 	case AF_INET6:
453 		V_route_cb.ip6_count--;
454 		break;
455 	}
456 	V_route_cb.any_count--;
457 	RTSOCK_UNLOCK();
458 	raw_usrreqs.pru_detach(so);
459 }
460 
461 static int
rts_disconnect(struct socket * so)462 rts_disconnect(struct socket *so)
463 {
464 
465 	return (raw_usrreqs.pru_disconnect(so));
466 }
467 
468 /* pru_listen is EOPNOTSUPP */
469 
470 static int
rts_peeraddr(struct socket * so,struct sockaddr ** nam)471 rts_peeraddr(struct socket *so, struct sockaddr **nam)
472 {
473 
474 	return (raw_usrreqs.pru_peeraddr(so, nam));
475 }
476 
477 /* pru_rcvd is EOPNOTSUPP */
478 /* pru_rcvoob is EOPNOTSUPP */
479 
480 static int
rts_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)481 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
482 	 struct mbuf *control, struct thread *td)
483 {
484 
485 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
486 }
487 
488 /* pru_sense is null */
489 
490 static int
rts_shutdown(struct socket * so)491 rts_shutdown(struct socket *so)
492 {
493 
494 	return (raw_usrreqs.pru_shutdown(so));
495 }
496 
497 static int
rts_sockaddr(struct socket * so,struct sockaddr ** nam)498 rts_sockaddr(struct socket *so, struct sockaddr **nam)
499 {
500 
501 	return (raw_usrreqs.pru_sockaddr(so, nam));
502 }
503 
504 static struct pr_usrreqs route_usrreqs = {
505 	.pru_abort =		rts_abort,
506 	.pru_attach =		rts_attach,
507 	.pru_bind =		rts_bind,
508 	.pru_connect =		rts_connect,
509 	.pru_detach =		rts_detach,
510 	.pru_disconnect =	rts_disconnect,
511 	.pru_peeraddr =		rts_peeraddr,
512 	.pru_send =		rts_send,
513 	.pru_shutdown =		rts_shutdown,
514 	.pru_sockaddr =		rts_sockaddr,
515 	.pru_close =		rts_close,
516 };
517 
518 #ifndef _SOCKADDR_UNION_DEFINED
519 #define	_SOCKADDR_UNION_DEFINED
520 /*
521  * The union of all possible address formats we handle.
522  */
523 union sockaddr_union {
524 	struct sockaddr		sa;
525 	struct sockaddr_in	sin;
526 	struct sockaddr_in6	sin6;
527 };
528 #endif /* _SOCKADDR_UNION_DEFINED */
529 
530 static int
rtm_get_jailed(struct rt_addrinfo * info,struct ifnet * ifp,struct nhop_object * nh,union sockaddr_union * saun,struct ucred * cred)531 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
532     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
533 {
534 #if defined(INET) || defined(INET6)
535 	struct epoch_tracker et;
536 #endif
537 
538 	/* First, see if the returned address is part of the jail. */
539 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
540 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
541 		return (0);
542 	}
543 
544 	switch (info->rti_info[RTAX_DST]->sa_family) {
545 #ifdef INET
546 	case AF_INET:
547 	{
548 		struct in_addr ia;
549 		struct ifaddr *ifa;
550 		int found;
551 
552 		found = 0;
553 		/*
554 		 * Try to find an address on the given outgoing interface
555 		 * that belongs to the jail.
556 		 */
557 		NET_EPOCH_ENTER(et);
558 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
559 			struct sockaddr *sa;
560 			sa = ifa->ifa_addr;
561 			if (sa->sa_family != AF_INET)
562 				continue;
563 			ia = ((struct sockaddr_in *)sa)->sin_addr;
564 			if (prison_check_ip4(cred, &ia) == 0) {
565 				found = 1;
566 				break;
567 			}
568 		}
569 		NET_EPOCH_EXIT(et);
570 		if (!found) {
571 			/*
572 			 * As a last resort return the 'default' jail address.
573 			 */
574 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
575 			    sin_addr;
576 			if (prison_get_ip4(cred, &ia) != 0)
577 				return (ESRCH);
578 		}
579 		bzero(&saun->sin, sizeof(struct sockaddr_in));
580 		saun->sin.sin_len = sizeof(struct sockaddr_in);
581 		saun->sin.sin_family = AF_INET;
582 		saun->sin.sin_addr.s_addr = ia.s_addr;
583 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
584 		break;
585 	}
586 #endif
587 #ifdef INET6
588 	case AF_INET6:
589 	{
590 		struct in6_addr ia6;
591 		struct ifaddr *ifa;
592 		int found;
593 
594 		found = 0;
595 		/*
596 		 * Try to find an address on the given outgoing interface
597 		 * that belongs to the jail.
598 		 */
599 		NET_EPOCH_ENTER(et);
600 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
601 			struct sockaddr *sa;
602 			sa = ifa->ifa_addr;
603 			if (sa->sa_family != AF_INET6)
604 				continue;
605 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
606 			    &ia6, sizeof(struct in6_addr));
607 			if (prison_check_ip6(cred, &ia6) == 0) {
608 				found = 1;
609 				break;
610 			}
611 		}
612 		NET_EPOCH_EXIT(et);
613 		if (!found) {
614 			/*
615 			 * As a last resort return the 'default' jail address.
616 			 */
617 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
618 			    sin6_addr;
619 			if (prison_get_ip6(cred, &ia6) != 0)
620 				return (ESRCH);
621 		}
622 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
623 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
624 		saun->sin6.sin6_family = AF_INET6;
625 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
626 		if (sa6_recoverscope(&saun->sin6) != 0)
627 			return (ESRCH);
628 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
629 		break;
630 	}
631 #endif
632 	default:
633 		return (ESRCH);
634 	}
635 	return (0);
636 }
637 
638 static int
fill_blackholeinfo(struct rt_addrinfo * info,union sockaddr_union * saun)639 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
640 {
641 	struct ifaddr *ifa;
642 	sa_family_t saf;
643 
644 	if (V_loif == NULL) {
645 		RTS_PID_LOG(LOG_INFO, "Unable to add blackhole/reject nhop without loopback");
646 		return (ENOTSUP);
647 	}
648 	info->rti_ifp = V_loif;
649 
650 	saf = info->rti_info[RTAX_DST]->sa_family;
651 
652 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
653 		if (ifa->ifa_addr->sa_family == saf) {
654 			info->rti_ifa = ifa;
655 			break;
656 		}
657 	}
658 	if (info->rti_ifa == NULL) {
659 		RTS_PID_LOG(LOG_INFO, "Unable to find ifa for blackhole/reject nhop");
660 		return (ENOTSUP);
661 	}
662 
663 	bzero(saun, sizeof(union sockaddr_union));
664 	switch (saf) {
665 #ifdef INET
666 	case AF_INET:
667 		saun->sin.sin_family = AF_INET;
668 		saun->sin.sin_len = sizeof(struct sockaddr_in);
669 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
670 		break;
671 #endif
672 #ifdef INET6
673 	case AF_INET6:
674 		saun->sin6.sin6_family = AF_INET6;
675 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
676 		saun->sin6.sin6_addr = in6addr_loopback;
677 		break;
678 #endif
679 	default:
680 		RTS_PID_LOG(LOG_INFO, "unsupported family: %d", saf);
681 		return (ENOTSUP);
682 	}
683 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
684 	info->rti_flags |= RTF_GATEWAY;
685 
686 	return (0);
687 }
688 
689 /*
690  * Fills in @info based on userland-provided @rtm message.
691  *
692  * Returns 0 on success.
693  */
694 static int
fill_addrinfo(struct rt_msghdr * rtm,int len,struct linear_buffer * lb,u_int fibnum,struct rt_addrinfo * info)695 fill_addrinfo(struct rt_msghdr *rtm, int len, struct linear_buffer *lb, u_int fibnum,
696     struct rt_addrinfo *info)
697 {
698 	int error;
699 	sa_family_t saf;
700 
701 	rtm->rtm_pid = curproc->p_pid;
702 	info->rti_addrs = rtm->rtm_addrs;
703 
704 	info->rti_mflags = rtm->rtm_inits;
705 	info->rti_rmx = &rtm->rtm_rmx;
706 
707 	/*
708 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
709 	 * link-local address because rtrequest requires addresses with
710 	 * embedded scope id.
711 	 */
712 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
713 		return (EINVAL);
714 
715 	info->rti_flags = rtm->rtm_flags;
716 	error = cleanup_xaddrs(info, lb);
717 	if (error != 0)
718 		return (error);
719 	saf = info->rti_info[RTAX_DST]->sa_family;
720 	/*
721 	 * Verify that the caller has the appropriate privilege; RTM_GET
722 	 * is the only operation the non-superuser is allowed.
723 	 */
724 	if (rtm->rtm_type != RTM_GET) {
725 		error = priv_check(curthread, PRIV_NET_ROUTE);
726 		if (error != 0)
727 			return (error);
728 	}
729 
730 	/*
731 	 * The given gateway address may be an interface address.
732 	 * For example, issuing a "route change" command on a route
733 	 * entry that was created from a tunnel, and the gateway
734 	 * address given is the local end point. In this case the
735 	 * RTF_GATEWAY flag must be cleared or the destination will
736 	 * not be reachable even though there is no error message.
737 	 */
738 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
739 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
740 		struct rt_addrinfo ginfo;
741 		struct sockaddr *gdst;
742 		struct sockaddr_storage ss;
743 
744 		bzero(&ginfo, sizeof(ginfo));
745 		bzero(&ss, sizeof(ss));
746 		ss.ss_len = sizeof(ss);
747 
748 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
749 		gdst = info->rti_info[RTAX_GATEWAY];
750 
751 		/*
752 		 * A host route through the loopback interface is
753 		 * installed for each interface address. In pre 8.0
754 		 * releases the interface address of a PPP link type
755 		 * is not reachable locally. This behavior is fixed as
756 		 * part of the new L2/L3 redesign and rewrite work. The
757 		 * signature of this interface address route is the
758 		 * AF_LINK sa_family type of the gateway, and the
759 		 * rt_ifp has the IFF_LOOPBACK flag set.
760 		 */
761 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
762 			if (ss.ss_family == AF_LINK &&
763 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
764 				info->rti_flags &= ~RTF_GATEWAY;
765 				info->rti_flags |= RTF_GWFLAG_COMPAT;
766 			}
767 			rib_free_info(&ginfo);
768 		}
769 	}
770 
771 	return (0);
772 }
773 
774 static struct nhop_object *
select_nhop(struct nhop_object * nh,const struct sockaddr * gw)775 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
776 {
777 	if (!NH_IS_NHGRP(nh))
778 		return (nh);
779 #ifdef ROUTE_MPATH
780 	const struct weightened_nhop *wn;
781 	uint32_t num_nhops;
782 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
783 	if (gw == NULL)
784 		return (wn[0].nh);
785 	for (int i = 0; i < num_nhops; i++) {
786 		if (match_nhop_gw(wn[i].nh, gw))
787 			return (wn[i].nh);
788 	}
789 #endif
790 	return (NULL);
791 }
792 
793 /*
794  * Handles RTM_GET message from routing socket, returning matching rt.
795  *
796  * Returns:
797  * 0 on success, with locked and referenced matching rt in @rt_nrt
798  * errno of failure
799  */
800 static int
handle_rtm_get(struct rt_addrinfo * info,u_int fibnum,struct rt_msghdr * rtm,struct rib_cmd_info * rc)801 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
802     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
803 {
804 	RIB_RLOCK_TRACKER;
805 	struct rib_head *rnh;
806 	struct nhop_object *nh;
807 	sa_family_t saf;
808 
809 	saf = info->rti_info[RTAX_DST]->sa_family;
810 
811 	rnh = rt_tables_get_rnh(fibnum, saf);
812 	if (rnh == NULL)
813 		return (EAFNOSUPPORT);
814 
815 	RIB_RLOCK(rnh);
816 
817 	/*
818 	 * By (implicit) convention host route (one without netmask)
819 	 * means longest-prefix-match request and the route with netmask
820 	 * means exact-match lookup.
821 	 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128
822 	 * prefixes, use original data to check for the netmask presence.
823 	 */
824 	if ((rtm->rtm_addrs & RTA_NETMASK) == 0) {
825 		/*
826 		 * Provide longest prefix match for
827 		 * address lookup (no mask).
828 		 * 'route -n get addr'
829 		 */
830 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
831 		    info->rti_info[RTAX_DST], &rnh->head);
832 	} else
833 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
834 		    info->rti_info[RTAX_DST],
835 		    info->rti_info[RTAX_NETMASK], &rnh->head);
836 
837 	if (rc->rc_rt == NULL) {
838 		RIB_RUNLOCK(rnh);
839 		return (ESRCH);
840 	}
841 
842 	nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
843 	if (nh == NULL) {
844 		RIB_RUNLOCK(rnh);
845 		return (ESRCH);
846 	}
847 	/*
848 	 * If performing proxied L2 entry insertion, and
849 	 * the actual PPP host entry is found, perform
850 	 * another search to retrieve the prefix route of
851 	 * the local end point of the PPP link.
852 	 * TODO: move this logic to userland.
853 	 */
854 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
855 		struct sockaddr_storage laddr;
856 
857 		if (nh->nh_ifp != NULL &&
858 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
859 			struct ifaddr *ifa;
860 
861 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
862 					RT_ALL_FIBS);
863 			if (ifa != NULL)
864 				rt_maskedcopy(ifa->ifa_addr,
865 					      (struct sockaddr *)&laddr,
866 					      ifa->ifa_netmask);
867 		} else
868 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
869 				      (struct sockaddr *)&laddr,
870 				      nh->nh_ifa->ifa_netmask);
871 		/*
872 		 * refactor rt and no lock operation necessary
873 		 */
874 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(
875 		    (struct sockaddr *)&laddr, &rnh->head);
876 		if (rc->rc_rt == NULL) {
877 			RIB_RUNLOCK(rnh);
878 			return (ESRCH);
879 		}
880 		nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
881 		if (nh == NULL) {
882 			RIB_RUNLOCK(rnh);
883 			return (ESRCH);
884 		}
885 	}
886 	rc->rc_nh_new = nh;
887 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
888 	RIB_RUNLOCK(rnh);
889 
890 	return (0);
891 }
892 
893 static void
init_sockaddrs_family(int family,struct sockaddr * dst,struct sockaddr * mask)894 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
895 {
896 #ifdef INET
897 	if (family == AF_INET) {
898 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
899 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
900 
901 		bzero(dst4, sizeof(struct sockaddr_in));
902 		bzero(mask4, sizeof(struct sockaddr_in));
903 
904 		dst4->sin_family = AF_INET;
905 		dst4->sin_len = sizeof(struct sockaddr_in);
906 		mask4->sin_family = AF_INET;
907 		mask4->sin_len = sizeof(struct sockaddr_in);
908 	}
909 #endif
910 #ifdef INET6
911 	if (family == AF_INET6) {
912 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
913 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
914 
915 		bzero(dst6, sizeof(struct sockaddr_in6));
916 		bzero(mask6, sizeof(struct sockaddr_in6));
917 
918 		dst6->sin6_family = AF_INET6;
919 		dst6->sin6_len = sizeof(struct sockaddr_in6);
920 		mask6->sin6_family = AF_INET6;
921 		mask6->sin6_len = sizeof(struct sockaddr_in6);
922 	}
923 #endif
924 }
925 
926 static void
export_rtaddrs(const struct rtentry * rt,struct sockaddr * dst,struct sockaddr * mask)927 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
928     struct sockaddr *mask)
929 {
930 #ifdef INET
931 	if (dst->sa_family == AF_INET) {
932 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
933 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
934 		uint32_t scopeid = 0;
935 		rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
936 		    &scopeid);
937 		return;
938 	}
939 #endif
940 #ifdef INET6
941 	if (dst->sa_family == AF_INET6) {
942 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
943 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
944 		uint32_t scopeid = 0;
945 		rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
946 		    &mask6->sin6_addr, &scopeid);
947 		dst6->sin6_scope_id = scopeid;
948 		return;
949 	}
950 #endif
951 }
952 
953 static int
update_rtm_from_info(struct rt_addrinfo * info,struct rt_msghdr ** prtm,int alloc_len)954 update_rtm_from_info(struct rt_addrinfo *info, struct rt_msghdr **prtm,
955     int alloc_len)
956 {
957 	struct rt_msghdr *rtm, *orig_rtm = NULL;
958 	struct walkarg w;
959 	int len;
960 
961 	rtm = *prtm;
962 	/* Check if we need to realloc storage */
963 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
964 	if (len > alloc_len) {
965 		struct rt_msghdr *tmp_rtm;
966 
967 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
968 		if (tmp_rtm == NULL)
969 			return (ENOBUFS);
970 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
971 		orig_rtm = rtm;
972 		rtm = tmp_rtm;
973 		alloc_len = len;
974 
975 		/*
976 		 * Delay freeing original rtm as info contains
977 		 * data referencing it.
978 		 */
979 	}
980 
981 	w.w_tmem = (caddr_t)rtm;
982 	w.w_tmemsize = alloc_len;
983 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
984 	rtm->rtm_addrs = info->rti_addrs;
985 
986 	if (orig_rtm != NULL)
987 		free(orig_rtm, M_TEMP);
988 	*prtm = rtm;
989 	return (0);
990 }
991 
992 
993 /*
994  * Update sockaddrs, flags, etc in @prtm based on @rc data.
995  * rtm can be reallocated.
996  *
997  * Returns 0 on success, along with pointer to (potentially reallocated)
998  *  rtm.
999  *
1000  */
1001 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)1002 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
1003     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
1004 {
1005 	union sockaddr_union saun;
1006 	struct rt_msghdr *rtm;
1007 	struct ifnet *ifp;
1008 	int error;
1009 
1010 	rtm = *prtm;
1011 	union sockaddr_union sa_dst, sa_mask;
1012 	int family = info->rti_info[RTAX_DST]->sa_family;
1013 	init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
1014 	export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
1015 
1016 	info->rti_info[RTAX_DST] = &sa_dst.sa;
1017 	info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
1018 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1019 	info->rti_info[RTAX_GENMASK] = 0;
1020 	ifp = nh->nh_ifp;
1021 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
1022 		if (ifp) {
1023 			info->rti_info[RTAX_IFP] =
1024 			    ifp->if_addr->ifa_addr;
1025 			error = rtm_get_jailed(info, ifp, nh,
1026 			    &saun, curthread->td_ucred);
1027 			if (error != 0)
1028 				return (error);
1029 			if (ifp->if_flags & IFF_POINTOPOINT)
1030 				info->rti_info[RTAX_BRD] =
1031 				    nh->nh_ifa->ifa_dstaddr;
1032 			rtm->rtm_index = ifp->if_index;
1033 		} else {
1034 			info->rti_info[RTAX_IFP] = NULL;
1035 			info->rti_info[RTAX_IFA] = NULL;
1036 		}
1037 	} else if (ifp != NULL)
1038 		rtm->rtm_index = ifp->if_index;
1039 
1040 	if ((error = update_rtm_from_info(info, prtm, alloc_len)) != 0)
1041 		return (error);
1042 
1043 	rtm = *prtm;
1044 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
1045 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
1046 		rtm->rtm_flags = RTF_GATEWAY |
1047 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
1048 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
1049 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
1050 
1051 	return (0);
1052 }
1053 
1054 #ifdef ROUTE_MPATH
1055 static void
save_del_notification(const struct rib_cmd_info * rc,void * _cbdata)1056 save_del_notification(const struct rib_cmd_info *rc, void *_cbdata)
1057 {
1058 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1059 
1060 	if (rc->rc_cmd == RTM_DELETE)
1061 		*rc_new = *rc;
1062 }
1063 
1064 static void
save_add_notification(const struct rib_cmd_info * rc,void * _cbdata)1065 save_add_notification(const struct rib_cmd_info *rc, void *_cbdata)
1066 {
1067 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1068 
1069 	if (rc->rc_cmd == RTM_ADD)
1070 		*rc_new = *rc;
1071 }
1072 #endif
1073 
1074 static struct sockaddr *
alloc_sockaddr_aligned(struct linear_buffer * lb,int len)1075 alloc_sockaddr_aligned(struct linear_buffer *lb, int len)
1076 {
1077 	len = roundup2(len, sizeof(uint64_t));
1078 	if (lb->offset + len > lb->size)
1079 		return (NULL);
1080 	struct sockaddr *sa = (struct sockaddr *)(lb->base + lb->offset);
1081 	lb->offset += len;
1082 	return (sa);
1083 }
1084 
1085 /*ARGSUSED*/
1086 static int
route_output(struct mbuf * m,struct socket * so,...)1087 route_output(struct mbuf *m, struct socket *so, ...)
1088 {
1089 	struct rt_msghdr *rtm = NULL;
1090 	struct rtentry *rt = NULL;
1091 	struct rt_addrinfo info;
1092 	struct epoch_tracker et;
1093 #ifdef INET6
1094 	struct sockaddr_storage ss;
1095 	struct sockaddr_in6 *sin6;
1096 	int i, rti_need_deembed = 0;
1097 #endif
1098 	int alloc_len = 0, len, error = 0, fibnum;
1099 	sa_family_t saf = AF_UNSPEC;
1100 	struct rib_cmd_info rc;
1101 	struct nhop_object *nh;
1102 
1103 	fibnum = so->so_fibnum;
1104 #define senderr(e) { error = e; goto flush;}
1105 	if (m == NULL || ((m->m_len < sizeof(long)) &&
1106 		       (m = m_pullup(m, sizeof(long))) == NULL))
1107 		return (ENOBUFS);
1108 	if ((m->m_flags & M_PKTHDR) == 0)
1109 		panic("route_output");
1110 	NET_EPOCH_ENTER(et);
1111 	len = m->m_pkthdr.len;
1112 	if (len < sizeof(*rtm) ||
1113 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1114 		senderr(EINVAL);
1115 
1116 	/*
1117 	 * Most of current messages are in range 200-240 bytes,
1118 	 * minimize possible re-allocation on reply using larger size
1119 	 * buffer aligned on 1k boundaty.
1120 	 */
1121 	alloc_len = roundup2(len, 1024);
1122 	int total_len = alloc_len + SCRATCH_BUFFER_SIZE;
1123 	if ((rtm = malloc(total_len, M_TEMP, M_NOWAIT)) == NULL)
1124 		senderr(ENOBUFS);
1125 
1126 	m_copydata(m, 0, len, (caddr_t)rtm);
1127 	bzero(&info, sizeof(info));
1128 	nh = NULL;
1129 	struct linear_buffer lb = {
1130 		.base = (char *)rtm + alloc_len,
1131 		.size = SCRATCH_BUFFER_SIZE,
1132 	};
1133 
1134 	if (rtm->rtm_version != RTM_VERSION) {
1135 		/* Do not touch message since format is unknown */
1136 		free(rtm, M_TEMP);
1137 		rtm = NULL;
1138 		senderr(EPROTONOSUPPORT);
1139 	}
1140 
1141 	/*
1142 	 * Starting from here, it is possible
1143 	 * to alter original message and insert
1144 	 * caller PID and error value.
1145 	 */
1146 
1147 	if ((error = fill_addrinfo(rtm, len, &lb, fibnum, &info)) != 0) {
1148 		senderr(error);
1149 	}
1150 	/* fill_addringo() embeds scope into IPv6 addresses */
1151 #ifdef INET6
1152 	rti_need_deembed = 1;
1153 #endif
1154 
1155 	saf = info.rti_info[RTAX_DST]->sa_family;
1156 
1157 	/* support for new ARP code */
1158 	if (rtm->rtm_flags & RTF_LLDATA) {
1159 		error = lla_rt_output(rtm, &info);
1160 		goto flush;
1161 	}
1162 
1163 	union sockaddr_union gw_saun;
1164 	int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1165 	if (blackhole_flags != 0) {
1166 		if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1167 			error = fill_blackholeinfo(&info, &gw_saun);
1168 		else {
1169 			RTS_PID_LOG(LOG_DEBUG, "both BLACKHOLE and REJECT flags specifiied");
1170 			error = EINVAL;
1171 		}
1172 		if (error != 0)
1173 			senderr(error);
1174 	}
1175 
1176 	switch (rtm->rtm_type) {
1177 	case RTM_ADD:
1178 	case RTM_CHANGE:
1179 		if (rtm->rtm_type == RTM_ADD) {
1180 			if (info.rti_info[RTAX_GATEWAY] == NULL) {
1181 				RTS_PID_LOG(LOG_DEBUG, "RTM_ADD w/o gateway");
1182 				senderr(EINVAL);
1183 			}
1184 		}
1185 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1186 		if (error == 0) {
1187 			rtsock_notify_event(fibnum, &rc);
1188 #ifdef ROUTE_MPATH
1189 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
1190 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1191 				struct rib_cmd_info rc_simple = {};
1192 				rib_decompose_notification(&rc,
1193 				    save_add_notification, (void *)&rc_simple);
1194 				rc = rc_simple;
1195 			}
1196 #endif
1197 			/* nh MAY be empty if RTM_CHANGE request is no-op */
1198 			nh = rc.rc_nh_new;
1199 			if (nh != NULL) {
1200 				rtm->rtm_index = nh->nh_ifp->if_index;
1201 				rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1202 			}
1203 		}
1204 		break;
1205 
1206 	case RTM_DELETE:
1207 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1208 		if (error == 0) {
1209 			rtsock_notify_event(fibnum, &rc);
1210 #ifdef ROUTE_MPATH
1211 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1212 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1213 				struct rib_cmd_info rc_simple = {};
1214 				rib_decompose_notification(&rc,
1215 				    save_del_notification, (void *)&rc_simple);
1216 				rc = rc_simple;
1217 			}
1218 #endif
1219 			nh = rc.rc_nh_old;
1220 		}
1221 		break;
1222 
1223 	case RTM_GET:
1224 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1225 		if (error != 0)
1226 			senderr(error);
1227 		nh = rc.rc_nh_new;
1228 
1229 		if (!rt_is_exportable(rc.rc_rt, curthread->td_ucred))
1230 			senderr(ESRCH);
1231 		break;
1232 
1233 	default:
1234 		senderr(EOPNOTSUPP);
1235 	}
1236 
1237 	if (error == 0 && nh != NULL) {
1238 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1239 		/*
1240 		 * Note that some sockaddr pointers may have changed to
1241 		 * point to memory outsize @rtm. Some may be pointing
1242 		 * to the on-stack variables.
1243 		 * Given that, any pointer in @info CANNOT BE USED.
1244 		 */
1245 
1246 		/*
1247 		 * scopeid deembedding has been performed while
1248 		 * writing updated rtm in rtsock_msg_buffer().
1249 		 * With that in mind, skip deembedding procedure below.
1250 		 */
1251 #ifdef INET6
1252 		rti_need_deembed = 0;
1253 #endif
1254 	}
1255 
1256 flush:
1257 	NET_EPOCH_EXIT(et);
1258 	rt = NULL;
1259 
1260 #ifdef INET6
1261 	if (rtm != NULL) {
1262 		if (rti_need_deembed) {
1263 			/* sin6_scope_id is recovered before sending rtm. */
1264 			sin6 = (struct sockaddr_in6 *)&ss;
1265 			for (i = 0; i < RTAX_MAX; i++) {
1266 				if (info.rti_info[i] == NULL)
1267 					continue;
1268 				if (info.rti_info[i]->sa_family != AF_INET6)
1269 					continue;
1270 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1271 				if (sa6_recoverscope(sin6) == 0)
1272 					bcopy(sin6, info.rti_info[i],
1273 						    sizeof(*sin6));
1274 			}
1275 			if (update_rtm_from_info(&info, &rtm, alloc_len) != 0) {
1276 				if (error != 0)
1277 					error = ENOBUFS;
1278 			}
1279 		}
1280 	}
1281 #endif
1282 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1283 
1284 	return (error);
1285 }
1286 
1287 /*
1288  * Sends the prepared reply message in @rtm to all rtsock clients.
1289  * Frees @m and @rtm.
1290  *
1291  */
1292 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)1293 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1294     sa_family_t saf, u_int fibnum, int rtm_errno)
1295 {
1296 	struct rawcb *rp = NULL;
1297 
1298 	/*
1299 	 * Check to see if we don't want our own messages.
1300 	 */
1301 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1302 		if (V_route_cb.any_count <= 1) {
1303 			if (rtm != NULL)
1304 				free(rtm, M_TEMP);
1305 			m_freem(m);
1306 			return;
1307 		}
1308 		/* There is another listener, so construct message */
1309 		rp = sotorawcb(so);
1310 	}
1311 
1312 	if (rtm != NULL) {
1313 		if (rtm_errno!= 0)
1314 			rtm->rtm_errno = rtm_errno;
1315 		else
1316 			rtm->rtm_flags |= RTF_DONE;
1317 
1318 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1319 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1320 			m_freem(m);
1321 			m = NULL;
1322 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1323 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1324 
1325 		free(rtm, M_TEMP);
1326 	}
1327 	if (m != NULL) {
1328 		M_SETFIB(m, fibnum);
1329 		m->m_flags |= RTS_FILTER_FIB;
1330 		if (rp) {
1331 			/*
1332 			 * XXX insure we don't get a copy by
1333 			 * invalidating our protocol
1334 			 */
1335 			unsigned short family = rp->rcb_proto.sp_family;
1336 			rp->rcb_proto.sp_family = 0;
1337 			rt_dispatch(m, saf);
1338 			rp->rcb_proto.sp_family = family;
1339 		} else
1340 			rt_dispatch(m, saf);
1341 	}
1342 }
1343 
1344 static void
rt_getmetrics(const struct rtentry * rt,const struct nhop_object * nh,struct rt_metrics * out)1345 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1346     struct rt_metrics *out)
1347 {
1348 
1349 	bzero(out, sizeof(*out));
1350 	out->rmx_mtu = nh->nh_mtu;
1351 	out->rmx_weight = rt->rt_weight;
1352 	out->rmx_nhidx = nhop_get_idx(nh);
1353 	/* Kernel -> userland timebase conversion. */
1354 	out->rmx_expire = nhop_get_expire(nh) ?
1355 	    nhop_get_expire(nh) - time_uptime + time_second : 0;
1356 }
1357 
1358 /*
1359  * Extract the addresses of the passed sockaddrs.
1360  * Do a little sanity checking so as to avoid bad memory references.
1361  * This data is derived straight from userland.
1362  */
1363 static int
rt_xaddrs(caddr_t cp,caddr_t cplim,struct rt_addrinfo * rtinfo)1364 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1365 {
1366 	struct sockaddr *sa;
1367 	int i;
1368 
1369 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1370 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1371 			continue;
1372 		sa = (struct sockaddr *)cp;
1373 		/*
1374 		 * It won't fit.
1375 		 */
1376 		if (cp + sa->sa_len > cplim) {
1377 			RTS_PID_LOG(LOG_DEBUG, "sa_len too big for sa type %d", i);
1378 			return (EINVAL);
1379 		}
1380 		/*
1381 		 * there are no more.. quit now
1382 		 * If there are more bits, they are in error.
1383 		 * I've seen this. route(1) can evidently generate these.
1384 		 * This causes kernel to core dump.
1385 		 * for compatibility, If we see this, point to a safe address.
1386 		 */
1387 		if (sa->sa_len == 0) {
1388 			rtinfo->rti_info[i] = &sa_zero;
1389 			return (0); /* should be EINVAL but for compat */
1390 		}
1391 		/* accept it */
1392 #ifdef INET6
1393 		if (sa->sa_family == AF_INET6)
1394 			sa6_embedscope((struct sockaddr_in6 *)sa,
1395 			    V_ip6_use_defzone);
1396 #endif
1397 		rtinfo->rti_info[i] = sa;
1398 		cp += SA_SIZE(sa);
1399 	}
1400 	return (0);
1401 }
1402 
1403 #ifdef INET
1404 static inline void
fill_sockaddr_inet(struct sockaddr_in * sin,struct in_addr addr)1405 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1406 {
1407 
1408 	const struct sockaddr_in nsin = {
1409 		.sin_family = AF_INET,
1410 		.sin_len = sizeof(struct sockaddr_in),
1411 		.sin_addr = addr,
1412 	};
1413 	*sin = nsin;
1414 }
1415 #endif
1416 
1417 #ifdef INET6
1418 static inline void
fill_sockaddr_inet6(struct sockaddr_in6 * sin6,const struct in6_addr * addr6,uint32_t scopeid)1419 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1420     uint32_t scopeid)
1421 {
1422 
1423 	const struct sockaddr_in6 nsin6 = {
1424 		.sin6_family = AF_INET6,
1425 		.sin6_len = sizeof(struct sockaddr_in6),
1426 		.sin6_addr = *addr6,
1427 		.sin6_scope_id = scopeid,
1428 	};
1429 	*sin6 = nsin6;
1430 }
1431 #endif
1432 
1433 /*
1434  * Checks if gateway is suitable for lltable operations.
1435  * Lltable code requires AF_LINK gateway with ifindex
1436  *  and mac address specified.
1437  * Returns 0 on success.
1438  */
1439 static int
cleanup_xaddrs_lladdr(struct rt_addrinfo * info)1440 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1441 {
1442 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1443 
1444 	if (sdl->sdl_family != AF_LINK)
1445 		return (EINVAL);
1446 
1447 	if (sdl->sdl_index == 0) {
1448 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gateway w/o ifindex");
1449 		return (EINVAL);
1450 	}
1451 
1452 	if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len) {
1453 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gw: sdl_nlen/sdl_alen too large");
1454 		return (EINVAL);
1455 	}
1456 
1457 	return (0);
1458 }
1459 
1460 static int
cleanup_xaddrs_gateway(struct rt_addrinfo * info,struct linear_buffer * lb)1461 cleanup_xaddrs_gateway(struct rt_addrinfo *info, struct linear_buffer *lb)
1462 {
1463 	struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1464 	struct sockaddr *sa;
1465 
1466 	if (info->rti_flags & RTF_LLDATA)
1467 		return (cleanup_xaddrs_lladdr(info));
1468 
1469 	switch (gw->sa_family) {
1470 #ifdef INET
1471 	case AF_INET:
1472 		{
1473 			struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw;
1474 
1475 			/* Ensure reads do not go beyoud SA boundary */
1476 			if (SA_SIZE(gw) < offsetof(struct sockaddr_in, sin_zero)) {
1477 				RTS_PID_LOG(LOG_DEBUG, "gateway sin_len too small: %d",
1478 				    gw->sa_len);
1479 				return (EINVAL);
1480 			}
1481 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_in));
1482 			if (sa == NULL)
1483 				return (ENOBUFS);
1484 			fill_sockaddr_inet((struct sockaddr_in *)sa, gw_sin->sin_addr);
1485 			info->rti_info[RTAX_GATEWAY] = sa;
1486 		}
1487 		break;
1488 #endif
1489 #ifdef INET6
1490 	case AF_INET6:
1491 		{
1492 			struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1493 			if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1494 				RTS_PID_LOG(LOG_DEBUG, "gateway sin6_len too small: %d",
1495 				    gw->sa_len);
1496 				return (EINVAL);
1497 			}
1498 			fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1499 			break;
1500 		}
1501 #endif
1502 	case AF_LINK:
1503 		{
1504 			struct sockaddr_dl *gw_sdl;
1505 
1506 			size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data);
1507 			gw_sdl = (struct sockaddr_dl *)gw;
1508 			if (gw_sdl->sdl_len < sdl_min_len) {
1509 				RTS_PID_LOG(LOG_DEBUG, "gateway sdl_len too small: %d",
1510 				    gw_sdl->sdl_len);
1511 				return (EINVAL);
1512 			}
1513 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_dl_short));
1514 			if (sa == NULL)
1515 				return (ENOBUFS);
1516 
1517 			const struct sockaddr_dl_short sdl = {
1518 				.sdl_family = AF_LINK,
1519 				.sdl_len = sizeof(struct sockaddr_dl_short),
1520 				.sdl_index = gw_sdl->sdl_index,
1521 			};
1522 			*((struct sockaddr_dl_short *)sa) = sdl;
1523 			info->rti_info[RTAX_GATEWAY] = sa;
1524 			break;
1525 		}
1526 	}
1527 
1528 	return (0);
1529 }
1530 
1531 static void
remove_netmask(struct rt_addrinfo * info)1532 remove_netmask(struct rt_addrinfo *info)
1533 {
1534 	info->rti_info[RTAX_NETMASK] = NULL;
1535 	info->rti_flags |= RTF_HOST;
1536 	info->rti_addrs &= ~RTA_NETMASK;
1537 }
1538 
1539 #ifdef INET
1540 static int
cleanup_xaddrs_inet(struct rt_addrinfo * info,struct linear_buffer * lb)1541 cleanup_xaddrs_inet(struct rt_addrinfo *info, struct linear_buffer *lb)
1542 {
1543 	struct sockaddr_in *dst_sa, *mask_sa;
1544 	const int sa_len = sizeof(struct sockaddr_in);
1545 	struct in_addr dst, mask;
1546 
1547 	/* Check & fixup dst/netmask combination first */
1548 	dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1549 	mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1550 
1551 	/* Ensure reads do not go beyound the buffer size */
1552 	if (SA_SIZE(dst_sa) < offsetof(struct sockaddr_in, sin_zero)) {
1553 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin_len too small: %d",
1554 		    dst_sa->sin_len);
1555 		return (EINVAL);
1556 	}
1557 
1558 	if ((mask_sa != NULL) && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1559 		/*
1560 		 * Some older routing software encode mask length into the
1561 		 * sin_len, thus resulting in "truncated" sockaddr.
1562 		 */
1563 		int len = mask_sa->sin_len - offsetof(struct sockaddr_in, sin_addr);
1564 		if (len >= 0) {
1565 			mask.s_addr = 0;
1566 			if (len > sizeof(struct in_addr))
1567 				len = sizeof(struct in_addr);
1568 			memcpy(&mask, &mask_sa->sin_addr, len);
1569 		} else {
1570 			RTS_PID_LOG(LOG_DEBUG, "prefix mask sin_len too small: %d",
1571 			    mask_sa->sin_len);
1572 			return (EINVAL);
1573 		}
1574 	} else
1575 		mask.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST;
1576 
1577 	dst.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr));
1578 
1579 	/* Construct new "clean" dst/mask sockaddresses */
1580 	if ((dst_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1581 		return (ENOBUFS);
1582 	fill_sockaddr_inet(dst_sa, dst);
1583 	info->rti_info[RTAX_DST] = (struct sockaddr *)dst_sa;
1584 
1585 	if (mask.s_addr != INADDR_BROADCAST) {
1586 		if ((mask_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1587 			return (ENOBUFS);
1588 		fill_sockaddr_inet(mask_sa, mask);
1589 		info->rti_info[RTAX_NETMASK] = (struct sockaddr *)mask_sa;
1590 		info->rti_flags &= ~RTF_HOST;
1591 	} else
1592 		remove_netmask(info);
1593 
1594 	/* Check gateway */
1595 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1596 		return (cleanup_xaddrs_gateway(info, lb));
1597 
1598 	return (0);
1599 }
1600 #endif
1601 
1602 #ifdef INET6
1603 static int
cleanup_xaddrs_inet6(struct rt_addrinfo * info,struct linear_buffer * lb)1604 cleanup_xaddrs_inet6(struct rt_addrinfo *info, struct linear_buffer *lb)
1605 {
1606 	struct sockaddr *sa;
1607 	struct sockaddr_in6 *dst_sa, *mask_sa;
1608 	struct in6_addr mask, *dst;
1609 	const int sa_len = sizeof(struct sockaddr_in6);
1610 
1611 	/* Check & fixup dst/netmask combination first */
1612 	dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1613 	mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1614 
1615 	if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1616 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin6_len too small: %d",
1617 		    dst_sa->sin6_len);
1618 		return (EINVAL);
1619 	}
1620 
1621 	if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1622 		/*
1623 		 * Some older routing software encode mask length into the
1624 		 * sin6_len, thus resulting in "truncated" sockaddr.
1625 		 */
1626 		int len = mask_sa->sin6_len - offsetof(struct sockaddr_in6, sin6_addr);
1627 		if (len >= 0) {
1628 			bzero(&mask, sizeof(mask));
1629 			if (len > sizeof(struct in6_addr))
1630 				len = sizeof(struct in6_addr);
1631 			memcpy(&mask, &mask_sa->sin6_addr, len);
1632 		} else {
1633 			RTS_PID_LOG(LOG_DEBUG, "rtsock: prefix mask sin6_len too small: %d",
1634 			    mask_sa->sin6_len);
1635 			return (EINVAL);
1636 		}
1637 	} else
1638 		mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1639 
1640 	dst = &dst_sa->sin6_addr;
1641 	IN6_MASK_ADDR(dst, &mask);
1642 
1643 	if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1644 		return (ENOBUFS);
1645 	fill_sockaddr_inet6((struct sockaddr_in6 *)sa, dst, 0);
1646 	info->rti_info[RTAX_DST] = sa;
1647 
1648 	if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128)) {
1649 		if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1650 			return (ENOBUFS);
1651 		fill_sockaddr_inet6((struct sockaddr_in6 *)sa, &mask, 0);
1652 		info->rti_info[RTAX_NETMASK] = sa;
1653 		info->rti_flags &= ~RTF_HOST;
1654 	} else
1655 		remove_netmask(info);
1656 
1657 	/* Check gateway */
1658 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1659 		return (cleanup_xaddrs_gateway(info, lb));
1660 
1661 	return (0);
1662 }
1663 #endif
1664 
1665 static int
cleanup_xaddrs(struct rt_addrinfo * info,struct linear_buffer * lb)1666 cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb)
1667 {
1668 	int error = EAFNOSUPPORT;
1669 
1670 	if (info->rti_info[RTAX_DST] == NULL) {
1671 		RTS_PID_LOG(LOG_DEBUG, "prefix dst is not set");
1672 		return (EINVAL);
1673 	}
1674 
1675 	if (info->rti_flags & RTF_LLDATA) {
1676 		/*
1677 		 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1678 		 * prefix along with the actual address in RTA_DST.
1679 		 * Remove netmask to avoid unnecessary address masking.
1680 		 */
1681 		remove_netmask(info);
1682 	}
1683 
1684 	switch (info->rti_info[RTAX_DST]->sa_family) {
1685 #ifdef INET
1686 	case AF_INET:
1687 		error = cleanup_xaddrs_inet(info, lb);
1688 		break;
1689 #endif
1690 #ifdef INET6
1691 	case AF_INET6:
1692 		error = cleanup_xaddrs_inet6(info, lb);
1693 		break;
1694 #endif
1695 	}
1696 
1697 	return (error);
1698 }
1699 
1700 /*
1701  * Fill in @dmask with valid netmask leaving original @smask
1702  * intact. Mostly used with radix netmasks.
1703  */
1704 struct sockaddr *
rtsock_fix_netmask(const struct sockaddr * dst,const struct sockaddr * smask,struct sockaddr_storage * dmask)1705 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1706     struct sockaddr_storage *dmask)
1707 {
1708 	if (dst == NULL || smask == NULL)
1709 		return (NULL);
1710 
1711 	memset(dmask, 0, dst->sa_len);
1712 	memcpy(dmask, smask, smask->sa_len);
1713 	dmask->ss_len = dst->sa_len;
1714 	dmask->ss_family = dst->sa_family;
1715 
1716 	return ((struct sockaddr *)dmask);
1717 }
1718 
1719 /*
1720  * Writes information related to @rtinfo object to newly-allocated mbuf.
1721  * Assumes MCLBYTES is enough to construct any message.
1722  * Used for OS notifications of vaious events (if/ifa announces,etc)
1723  *
1724  * Returns allocated mbuf or NULL on failure.
1725  */
1726 static struct mbuf *
rtsock_msg_mbuf(int type,struct rt_addrinfo * rtinfo)1727 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1728 {
1729 	struct sockaddr_storage ss;
1730 	struct rt_msghdr *rtm;
1731 	struct mbuf *m;
1732 	int i;
1733 	struct sockaddr *sa;
1734 #ifdef INET6
1735 	struct sockaddr_in6 *sin6;
1736 #endif
1737 	int len, dlen;
1738 
1739 	switch (type) {
1740 	case RTM_DELADDR:
1741 	case RTM_NEWADDR:
1742 		len = sizeof(struct ifa_msghdr);
1743 		break;
1744 
1745 	case RTM_DELMADDR:
1746 	case RTM_NEWMADDR:
1747 		len = sizeof(struct ifma_msghdr);
1748 		break;
1749 
1750 	case RTM_IFINFO:
1751 		len = sizeof(struct if_msghdr);
1752 		break;
1753 
1754 	case RTM_IFANNOUNCE:
1755 	case RTM_IEEE80211:
1756 		len = sizeof(struct if_announcemsghdr);
1757 		break;
1758 
1759 	default:
1760 		len = sizeof(struct rt_msghdr);
1761 	}
1762 
1763 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1764 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1765 	if (len > MHLEN)
1766 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1767 	else
1768 		m = m_gethdr(M_NOWAIT, MT_DATA);
1769 	if (m == NULL)
1770 		return (m);
1771 
1772 	m->m_pkthdr.len = m->m_len = len;
1773 	rtm = mtod(m, struct rt_msghdr *);
1774 	bzero((caddr_t)rtm, len);
1775 	for (i = 0; i < RTAX_MAX; i++) {
1776 		if ((sa = rtinfo->rti_info[i]) == NULL)
1777 			continue;
1778 		rtinfo->rti_addrs |= (1 << i);
1779 
1780 		dlen = SA_SIZE(sa);
1781 		KASSERT(dlen <= sizeof(ss),
1782 		    ("%s: sockaddr size overflow", __func__));
1783 		bzero(&ss, sizeof(ss));
1784 		bcopy(sa, &ss, sa->sa_len);
1785 		sa = (struct sockaddr *)&ss;
1786 #ifdef INET6
1787 		if (sa->sa_family == AF_INET6) {
1788 			sin6 = (struct sockaddr_in6 *)sa;
1789 			(void)sa6_recoverscope(sin6);
1790 		}
1791 #endif
1792 		m_copyback(m, len, dlen, (caddr_t)sa);
1793 		len += dlen;
1794 	}
1795 	if (m->m_pkthdr.len != len) {
1796 		m_freem(m);
1797 		return (NULL);
1798 	}
1799 	rtm->rtm_msglen = len;
1800 	rtm->rtm_version = RTM_VERSION;
1801 	rtm->rtm_type = type;
1802 	return (m);
1803 }
1804 
1805 /*
1806  * Writes information related to @rtinfo object to preallocated buffer.
1807  * Stores needed size in @plen. If @w is NULL, calculates size without
1808  * writing.
1809  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1810  *
1811  * Returns 0 on success.
1812  *
1813  */
1814 static int
rtsock_msg_buffer(int type,struct rt_addrinfo * rtinfo,struct walkarg * w,int * plen)1815 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1816 {
1817 	struct sockaddr_storage ss;
1818 	int len, buflen = 0, dlen, i;
1819 	caddr_t cp = NULL;
1820 	struct rt_msghdr *rtm = NULL;
1821 #ifdef INET6
1822 	struct sockaddr_in6 *sin6;
1823 #endif
1824 #ifdef COMPAT_FREEBSD32
1825 	bool compat32 = false;
1826 #endif
1827 
1828 	switch (type) {
1829 	case RTM_DELADDR:
1830 	case RTM_NEWADDR:
1831 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1832 #ifdef COMPAT_FREEBSD32
1833 			if (w->w_req->flags & SCTL_MASK32) {
1834 				len = sizeof(struct ifa_msghdrl32);
1835 				compat32 = true;
1836 			} else
1837 #endif
1838 				len = sizeof(struct ifa_msghdrl);
1839 		} else
1840 			len = sizeof(struct ifa_msghdr);
1841 		break;
1842 
1843 	case RTM_IFINFO:
1844 #ifdef COMPAT_FREEBSD32
1845 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1846 			if (w->w_op == NET_RT_IFLISTL)
1847 				len = sizeof(struct if_msghdrl32);
1848 			else
1849 				len = sizeof(struct if_msghdr32);
1850 			compat32 = true;
1851 			break;
1852 		}
1853 #endif
1854 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1855 			len = sizeof(struct if_msghdrl);
1856 		else
1857 			len = sizeof(struct if_msghdr);
1858 		break;
1859 
1860 	case RTM_NEWMADDR:
1861 		len = sizeof(struct ifma_msghdr);
1862 		break;
1863 
1864 	default:
1865 		len = sizeof(struct rt_msghdr);
1866 	}
1867 
1868 	if (w != NULL) {
1869 		rtm = (struct rt_msghdr *)w->w_tmem;
1870 		buflen = w->w_tmemsize - len;
1871 		cp = (caddr_t)w->w_tmem + len;
1872 	}
1873 
1874 	rtinfo->rti_addrs = 0;
1875 	for (i = 0; i < RTAX_MAX; i++) {
1876 		struct sockaddr *sa;
1877 
1878 		if ((sa = rtinfo->rti_info[i]) == NULL)
1879 			continue;
1880 		rtinfo->rti_addrs |= (1 << i);
1881 #ifdef COMPAT_FREEBSD32
1882 		if (compat32)
1883 			dlen = SA_SIZE32(sa);
1884 		else
1885 #endif
1886 			dlen = SA_SIZE(sa);
1887 		if (cp != NULL && buflen >= dlen) {
1888 			KASSERT(dlen <= sizeof(ss),
1889 			    ("%s: sockaddr size overflow", __func__));
1890 			bzero(&ss, sizeof(ss));
1891 			bcopy(sa, &ss, sa->sa_len);
1892 			sa = (struct sockaddr *)&ss;
1893 #ifdef INET6
1894 			if (sa->sa_family == AF_INET6) {
1895 				sin6 = (struct sockaddr_in6 *)sa;
1896 				(void)sa6_recoverscope(sin6);
1897 			}
1898 #endif
1899 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1900 			cp += dlen;
1901 			buflen -= dlen;
1902 		} else if (cp != NULL) {
1903 			/*
1904 			 * Buffer too small. Count needed size
1905 			 * and return with error.
1906 			 */
1907 			cp = NULL;
1908 		}
1909 
1910 		len += dlen;
1911 	}
1912 
1913 	if (cp != NULL) {
1914 		dlen = ALIGN(len) - len;
1915 		if (buflen < dlen)
1916 			cp = NULL;
1917 		else {
1918 			bzero(cp, dlen);
1919 			cp += dlen;
1920 			buflen -= dlen;
1921 		}
1922 	}
1923 	len = ALIGN(len);
1924 
1925 	if (cp != NULL) {
1926 		/* fill header iff buffer is large enough */
1927 		rtm->rtm_version = RTM_VERSION;
1928 		rtm->rtm_type = type;
1929 		rtm->rtm_msglen = len;
1930 	}
1931 
1932 	*plen = len;
1933 
1934 	if (w != NULL && cp == NULL)
1935 		return (ENOBUFS);
1936 
1937 	return (0);
1938 }
1939 
1940 /*
1941  * This routine is called to generate a message from the routing
1942  * socket indicating that a redirect has occurred, a routing lookup
1943  * has failed, or that a protocol has detected timeouts to a particular
1944  * destination.
1945  */
1946 void
rt_missmsg_fib(int type,struct rt_addrinfo * rtinfo,int flags,int error,int fibnum)1947 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1948     int fibnum)
1949 {
1950 	struct rt_msghdr *rtm;
1951 	struct mbuf *m;
1952 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1953 
1954 	if (V_route_cb.any_count == 0)
1955 		return;
1956 	m = rtsock_msg_mbuf(type, rtinfo);
1957 	if (m == NULL)
1958 		return;
1959 
1960 	if (fibnum != RT_ALL_FIBS) {
1961 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1962 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1963 		M_SETFIB(m, fibnum);
1964 		m->m_flags |= RTS_FILTER_FIB;
1965 	}
1966 
1967 	rtm = mtod(m, struct rt_msghdr *);
1968 	rtm->rtm_flags = RTF_DONE | flags;
1969 	rtm->rtm_errno = error;
1970 	rtm->rtm_addrs = rtinfo->rti_addrs;
1971 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1972 }
1973 
1974 void
rt_missmsg(int type,struct rt_addrinfo * rtinfo,int flags,int error)1975 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1976 {
1977 
1978 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1979 }
1980 
1981 /*
1982  * This routine is called to generate a message from the routing
1983  * socket indicating that the status of a network interface has changed.
1984  */
1985 static void
rtsock_ifmsg(struct ifnet * ifp,int if_flags_mask __unused)1986 rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask __unused)
1987 {
1988 	struct if_msghdr *ifm;
1989 	struct mbuf *m;
1990 	struct rt_addrinfo info;
1991 
1992 	if (V_route_cb.any_count == 0)
1993 		return;
1994 	bzero((caddr_t)&info, sizeof(info));
1995 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1996 	if (m == NULL)
1997 		return;
1998 	ifm = mtod(m, struct if_msghdr *);
1999 	ifm->ifm_index = ifp->if_index;
2000 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2001 	if_data_copy(ifp, &ifm->ifm_data);
2002 	ifm->ifm_addrs = 0;
2003 	rt_dispatch(m, AF_UNSPEC);
2004 }
2005 
2006 /*
2007  * Announce interface address arrival/withdraw.
2008  * Please do not call directly, use rt_addrmsg().
2009  * Assume input data to be valid.
2010  * Returns 0 on success.
2011  */
2012 int
rtsock_addrmsg(int cmd,struct ifaddr * ifa,int fibnum)2013 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
2014 {
2015 	struct rt_addrinfo info;
2016 	struct sockaddr *sa;
2017 	int ncmd;
2018 	struct mbuf *m;
2019 	struct ifa_msghdr *ifam;
2020 	struct ifnet *ifp = ifa->ifa_ifp;
2021 	struct sockaddr_storage ss;
2022 
2023 	if (V_route_cb.any_count == 0)
2024 		return (0);
2025 
2026 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
2027 
2028 	bzero((caddr_t)&info, sizeof(info));
2029 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
2030 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2031 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2032 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
2033 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2034 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
2035 		return (ENOBUFS);
2036 	ifam = mtod(m, struct ifa_msghdr *);
2037 	ifam->ifam_index = ifp->if_index;
2038 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2039 	ifam->ifam_flags = ifa->ifa_flags;
2040 	ifam->ifam_addrs = info.rti_addrs;
2041 
2042 	if (fibnum != RT_ALL_FIBS) {
2043 		M_SETFIB(m, fibnum);
2044 		m->m_flags |= RTS_FILTER_FIB;
2045 	}
2046 
2047 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2048 
2049 	return (0);
2050 }
2051 
2052 /*
2053  * Announce route addition/removal to rtsock based on @rt data.
2054  * Callers are advives to use rt_routemsg() instead of using this
2055  *  function directly.
2056  * Assume @rt data is consistent.
2057  *
2058  * Returns 0 on success.
2059  */
2060 int
rtsock_routemsg(int cmd,struct rtentry * rt,struct nhop_object * nh,int fibnum)2061 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
2062     int fibnum)
2063 {
2064 	union sockaddr_union dst, mask;
2065 	struct rt_addrinfo info;
2066 
2067 	if (V_route_cb.any_count == 0)
2068 		return (0);
2069 
2070 	int family = rt_get_family(rt);
2071 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
2072 	export_rtaddrs(rt, &dst.sa, &mask.sa);
2073 
2074 	bzero((caddr_t)&info, sizeof(info));
2075 	info.rti_info[RTAX_DST] = &dst.sa;
2076 	info.rti_info[RTAX_NETMASK] = &mask.sa;
2077 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2078 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
2079 	info.rti_ifp = nh->nh_ifp;
2080 
2081 	return (rtsock_routemsg_info(cmd, &info, fibnum));
2082 }
2083 
2084 int
rtsock_routemsg_info(int cmd,struct rt_addrinfo * info,int fibnum)2085 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
2086 {
2087 	struct rt_msghdr *rtm;
2088 	struct sockaddr *sa;
2089 	struct mbuf *m;
2090 
2091 	if (V_route_cb.any_count == 0)
2092 		return (0);
2093 
2094 	if (info->rti_flags & RTF_HOST)
2095 		info->rti_info[RTAX_NETMASK] = NULL;
2096 
2097 	m = rtsock_msg_mbuf(cmd, info);
2098 	if (m == NULL)
2099 		return (ENOBUFS);
2100 
2101 	if (fibnum != RT_ALL_FIBS) {
2102 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
2103 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
2104 		M_SETFIB(m, fibnum);
2105 		m->m_flags |= RTS_FILTER_FIB;
2106 	}
2107 
2108 	rtm = mtod(m, struct rt_msghdr *);
2109 	rtm->rtm_addrs = info->rti_addrs;
2110 	if (info->rti_ifp != NULL)
2111 		rtm->rtm_index = info->rti_ifp->if_index;
2112 	/* Add RTF_DONE to indicate command 'completion' required by API */
2113 	info->rti_flags |= RTF_DONE;
2114 	/* Reported routes has to be up */
2115 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
2116 		info->rti_flags |= RTF_UP;
2117 	rtm->rtm_flags = info->rti_flags;
2118 
2119 	sa = info->rti_info[RTAX_DST];
2120 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2121 
2122 	return (0);
2123 }
2124 
2125 /*
2126  * This is the analogue to the rt_newaddrmsg which performs the same
2127  * function but for multicast group memberhips.  This is easier since
2128  * there is no route state to worry about.
2129  */
2130 void
rt_newmaddrmsg(int cmd,struct ifmultiaddr * ifma)2131 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
2132 {
2133 	struct rt_addrinfo info;
2134 	struct mbuf *m = NULL;
2135 	struct ifnet *ifp = ifma->ifma_ifp;
2136 	struct ifma_msghdr *ifmam;
2137 
2138 	if (V_route_cb.any_count == 0)
2139 		return;
2140 
2141 	bzero((caddr_t)&info, sizeof(info));
2142 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2143 	if (ifp && ifp->if_addr)
2144 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2145 	else
2146 		info.rti_info[RTAX_IFP] = NULL;
2147 	/*
2148 	 * If a link-layer address is present, present it as a ``gateway''
2149 	 * (similarly to how ARP entries, e.g., are presented).
2150 	 */
2151 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
2152 	m = rtsock_msg_mbuf(cmd, &info);
2153 	if (m == NULL)
2154 		return;
2155 	ifmam = mtod(m, struct ifma_msghdr *);
2156 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
2157 	    __func__));
2158 	ifmam->ifmam_index = ifp->if_index;
2159 	ifmam->ifmam_addrs = info.rti_addrs;
2160 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
2161 }
2162 
2163 static struct mbuf *
rt_makeifannouncemsg(struct ifnet * ifp,int type,int what,struct rt_addrinfo * info)2164 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
2165 	struct rt_addrinfo *info)
2166 {
2167 	struct if_announcemsghdr *ifan;
2168 	struct mbuf *m;
2169 
2170 	if (V_route_cb.any_count == 0)
2171 		return NULL;
2172 	bzero((caddr_t)info, sizeof(*info));
2173 	m = rtsock_msg_mbuf(type, info);
2174 	if (m != NULL) {
2175 		ifan = mtod(m, struct if_announcemsghdr *);
2176 		ifan->ifan_index = ifp->if_index;
2177 		strlcpy(ifan->ifan_name, ifp->if_xname,
2178 			sizeof(ifan->ifan_name));
2179 		ifan->ifan_what = what;
2180 	}
2181 	return m;
2182 }
2183 
2184 /*
2185  * This is called to generate routing socket messages indicating
2186  * IEEE80211 wireless events.
2187  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2188  */
2189 void
rt_ieee80211msg(struct ifnet * ifp,int what,void * data,size_t data_len)2190 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2191 {
2192 	struct mbuf *m;
2193 	struct rt_addrinfo info;
2194 
2195 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2196 	if (m != NULL) {
2197 		/*
2198 		 * Append the ieee80211 data.  Try to stick it in the
2199 		 * mbuf containing the ifannounce msg; otherwise allocate
2200 		 * a new mbuf and append.
2201 		 *
2202 		 * NB: we assume m is a single mbuf.
2203 		 */
2204 		if (data_len > M_TRAILINGSPACE(m)) {
2205 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2206 			if (n == NULL) {
2207 				m_freem(m);
2208 				return;
2209 			}
2210 			bcopy(data, mtod(n, void *), data_len);
2211 			n->m_len = data_len;
2212 			m->m_next = n;
2213 		} else if (data_len > 0) {
2214 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2215 			m->m_len += data_len;
2216 		}
2217 		if (m->m_flags & M_PKTHDR)
2218 			m->m_pkthdr.len += data_len;
2219 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2220 		rt_dispatch(m, AF_UNSPEC);
2221 	}
2222 }
2223 
2224 /*
2225  * This is called to generate routing socket messages indicating
2226  * network interface arrival and departure.
2227  */
2228 static void
rt_ifannouncemsg(struct ifnet * ifp,int what)2229 rt_ifannouncemsg(struct ifnet *ifp, int what)
2230 {
2231 	struct mbuf *m;
2232 	struct rt_addrinfo info;
2233 
2234 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2235 	if (m != NULL)
2236 		rt_dispatch(m, AF_UNSPEC);
2237 }
2238 
2239 static void
rt_dispatch(struct mbuf * m,sa_family_t saf)2240 rt_dispatch(struct mbuf *m, sa_family_t saf)
2241 {
2242 	struct m_tag *tag;
2243 
2244 	/*
2245 	 * Preserve the family from the sockaddr, if any, in an m_tag for
2246 	 * use when injecting the mbuf into the routing socket buffer from
2247 	 * the netisr.
2248 	 */
2249 	if (saf != AF_UNSPEC) {
2250 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
2251 		    M_NOWAIT);
2252 		if (tag == NULL) {
2253 			m_freem(m);
2254 			return;
2255 		}
2256 		*(unsigned short *)(tag + 1) = saf;
2257 		m_tag_prepend(m, tag);
2258 	}
2259 #ifdef VIMAGE
2260 	if (V_loif)
2261 		m->m_pkthdr.rcvif = V_loif;
2262 	else {
2263 		m_freem(m);
2264 		return;
2265 	}
2266 #endif
2267 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
2268 }
2269 
2270 /*
2271  * This is used in dumping the kernel table via sysctl().
2272  */
2273 static int
sysctl_dumpentry(struct rtentry * rt,void * vw)2274 sysctl_dumpentry(struct rtentry *rt, void *vw)
2275 {
2276 	struct walkarg *w = vw;
2277 	struct nhop_object *nh;
2278 
2279 	NET_EPOCH_ASSERT();
2280 
2281 	if (!rt_is_exportable(rt, w->w_req->td->td_ucred))
2282 		return (0);
2283 
2284 	export_rtaddrs(rt, w->dst, w->mask);
2285 	nh = rt_get_raw_nhop(rt);
2286 #ifdef ROUTE_MPATH
2287 	if (NH_IS_NHGRP(nh)) {
2288 		const struct weightened_nhop *wn;
2289 		uint32_t num_nhops;
2290 		int error;
2291 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2292 		for (int i = 0; i < num_nhops; i++) {
2293 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2294 			if (error != 0)
2295 				return (error);
2296 		}
2297 	} else
2298 #endif
2299 		sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2300 
2301 	return (0);
2302 }
2303 
2304 
2305 static int
sysctl_dumpnhop(struct rtentry * rt,struct nhop_object * nh,uint32_t weight,struct walkarg * w)2306 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2307     struct walkarg *w)
2308 {
2309 	struct rt_addrinfo info;
2310 	int error = 0, size;
2311 	uint32_t rtflags;
2312 
2313 	rtflags = nhop_get_rtflags(nh);
2314 
2315 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2316 		return (0);
2317 
2318 	bzero((caddr_t)&info, sizeof(info));
2319 	info.rti_info[RTAX_DST] = w->dst;
2320 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2321 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2322 	info.rti_info[RTAX_GENMASK] = 0;
2323 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2324 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2325 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2326 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2327 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2328 	}
2329 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2330 		return (error);
2331 	if (w->w_req && w->w_tmem) {
2332 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2333 
2334 		bzero(&rtm->rtm_index,
2335 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2336 
2337 		/*
2338 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2339 		 * and RTF_UP (if entry is linked, which is always true here).
2340 		 * Given that, use nhop rtflags & add RTF_UP.
2341 		 */
2342 		rtm->rtm_flags = rtflags | RTF_UP;
2343 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2344 			rtm->rtm_flags = RTF_GATEWAY |
2345 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2346 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2347 		rtm->rtm_rmx.rmx_weight = weight;
2348 		rtm->rtm_index = nh->nh_ifp->if_index;
2349 		rtm->rtm_addrs = info.rti_addrs;
2350 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2351 		return (error);
2352 	}
2353 	return (error);
2354 }
2355 
2356 static int
sysctl_iflist_ifml(struct ifnet * ifp,const struct if_data * src_ifd,struct rt_addrinfo * info,struct walkarg * w,int len)2357 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2358     struct rt_addrinfo *info, struct walkarg *w, int len)
2359 {
2360 	struct if_msghdrl *ifm;
2361 	struct if_data *ifd;
2362 
2363 	ifm = (struct if_msghdrl *)w->w_tmem;
2364 
2365 #ifdef COMPAT_FREEBSD32
2366 	if (w->w_req->flags & SCTL_MASK32) {
2367 		struct if_msghdrl32 *ifm32;
2368 
2369 		ifm32 = (struct if_msghdrl32 *)ifm;
2370 		ifm32->ifm_addrs = info->rti_addrs;
2371 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2372 		ifm32->ifm_index = ifp->if_index;
2373 		ifm32->_ifm_spare1 = 0;
2374 		ifm32->ifm_len = sizeof(*ifm32);
2375 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2376 		ifm32->_ifm_spare2 = 0;
2377 		ifd = &ifm32->ifm_data;
2378 	} else
2379 #endif
2380 	{
2381 		ifm->ifm_addrs = info->rti_addrs;
2382 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2383 		ifm->ifm_index = ifp->if_index;
2384 		ifm->_ifm_spare1 = 0;
2385 		ifm->ifm_len = sizeof(*ifm);
2386 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2387 		ifm->_ifm_spare2 = 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_ifm(struct ifnet * ifp,const struct if_data * src_ifd,struct rt_addrinfo * info,struct walkarg * w,int len)2397 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2398     struct rt_addrinfo *info, struct walkarg *w, int len)
2399 {
2400 	struct if_msghdr *ifm;
2401 	struct if_data *ifd;
2402 
2403 	ifm = (struct if_msghdr *)w->w_tmem;
2404 
2405 #ifdef COMPAT_FREEBSD32
2406 	if (w->w_req->flags & SCTL_MASK32) {
2407 		struct if_msghdr32 *ifm32;
2408 
2409 		ifm32 = (struct if_msghdr32 *)ifm;
2410 		ifm32->ifm_addrs = info->rti_addrs;
2411 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2412 		ifm32->ifm_index = ifp->if_index;
2413 		ifm32->_ifm_spare1 = 0;
2414 		ifd = &ifm32->ifm_data;
2415 	} else
2416 #endif
2417 	{
2418 		ifm->ifm_addrs = info->rti_addrs;
2419 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2420 		ifm->ifm_index = ifp->if_index;
2421 		ifm->_ifm_spare1 = 0;
2422 		ifd = &ifm->ifm_data;
2423 	}
2424 
2425 	memcpy(ifd, src_ifd, sizeof(*ifd));
2426 
2427 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2428 }
2429 
2430 static int
sysctl_iflist_ifaml(struct ifaddr * ifa,struct rt_addrinfo * info,struct walkarg * w,int len)2431 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2432     struct walkarg *w, int len)
2433 {
2434 	struct ifa_msghdrl *ifam;
2435 	struct if_data *ifd;
2436 
2437 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2438 
2439 #ifdef COMPAT_FREEBSD32
2440 	if (w->w_req->flags & SCTL_MASK32) {
2441 		struct ifa_msghdrl32 *ifam32;
2442 
2443 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2444 		ifam32->ifam_addrs = info->rti_addrs;
2445 		ifam32->ifam_flags = ifa->ifa_flags;
2446 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2447 		ifam32->_ifam_spare1 = 0;
2448 		ifam32->ifam_len = sizeof(*ifam32);
2449 		ifam32->ifam_data_off =
2450 		    offsetof(struct ifa_msghdrl32, ifam_data);
2451 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2452 		ifd = &ifam32->ifam_data;
2453 	} else
2454 #endif
2455 	{
2456 		ifam->ifam_addrs = info->rti_addrs;
2457 		ifam->ifam_flags = ifa->ifa_flags;
2458 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2459 		ifam->_ifam_spare1 = 0;
2460 		ifam->ifam_len = sizeof(*ifam);
2461 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2462 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2463 		ifd = &ifam->ifam_data;
2464 	}
2465 
2466 	bzero(ifd, sizeof(*ifd));
2467 	ifd->ifi_datalen = sizeof(struct if_data);
2468 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2469 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2470 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2471 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2472 
2473 	/* Fixup if_data carp(4) vhid. */
2474 	if (carp_get_vhid_p != NULL)
2475 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2476 
2477 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2478 }
2479 
2480 static int
sysctl_iflist_ifam(struct ifaddr * ifa,struct rt_addrinfo * info,struct walkarg * w,int len)2481 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2482     struct walkarg *w, int len)
2483 {
2484 	struct ifa_msghdr *ifam;
2485 
2486 	ifam = (struct ifa_msghdr *)w->w_tmem;
2487 	ifam->ifam_addrs = info->rti_addrs;
2488 	ifam->ifam_flags = ifa->ifa_flags;
2489 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2490 	ifam->_ifam_spare1 = 0;
2491 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2492 
2493 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2494 }
2495 
2496 static int
sysctl_iflist(int af,struct walkarg * w)2497 sysctl_iflist(int af, struct walkarg *w)
2498 {
2499 	struct ifnet *ifp;
2500 	struct ifaddr *ifa;
2501 	struct if_data ifd;
2502 	struct rt_addrinfo info;
2503 	int len, error = 0;
2504 	struct sockaddr_storage ss;
2505 
2506 	bzero((caddr_t)&info, sizeof(info));
2507 	bzero(&ifd, sizeof(ifd));
2508 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2509 		if (w->w_arg && w->w_arg != ifp->if_index)
2510 			continue;
2511 		if_data_copy(ifp, &ifd);
2512 		ifa = ifp->if_addr;
2513 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2514 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2515 		if (error != 0)
2516 			goto done;
2517 		info.rti_info[RTAX_IFP] = NULL;
2518 		if (w->w_req && w->w_tmem) {
2519 			if (w->w_op == NET_RT_IFLISTL)
2520 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2521 				    len);
2522 			else
2523 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2524 				    len);
2525 			if (error)
2526 				goto done;
2527 		}
2528 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2529 			if (af && af != ifa->ifa_addr->sa_family)
2530 				continue;
2531 			if (prison_if(w->w_req->td->td_ucred,
2532 			    ifa->ifa_addr) != 0)
2533 				continue;
2534 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2535 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2536 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2537 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2538 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2539 			if (error != 0)
2540 				goto done;
2541 			if (w->w_req && w->w_tmem) {
2542 				if (w->w_op == NET_RT_IFLISTL)
2543 					error = sysctl_iflist_ifaml(ifa, &info,
2544 					    w, len);
2545 				else
2546 					error = sysctl_iflist_ifam(ifa, &info,
2547 					    w, len);
2548 				if (error)
2549 					goto done;
2550 			}
2551 		}
2552 		info.rti_info[RTAX_IFA] = NULL;
2553 		info.rti_info[RTAX_NETMASK] = NULL;
2554 		info.rti_info[RTAX_BRD] = NULL;
2555 	}
2556 done:
2557 	return (error);
2558 }
2559 
2560 static int
sysctl_ifmalist(int af,struct walkarg * w)2561 sysctl_ifmalist(int af, struct walkarg *w)
2562 {
2563 	struct rt_addrinfo info;
2564 	struct ifaddr *ifa;
2565 	struct ifmultiaddr *ifma;
2566 	struct ifnet *ifp;
2567 	int error, len;
2568 
2569 	NET_EPOCH_ASSERT();
2570 
2571 	error = 0;
2572 	bzero((caddr_t)&info, sizeof(info));
2573 
2574 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2575 		if (w->w_arg && w->w_arg != ifp->if_index)
2576 			continue;
2577 		ifa = ifp->if_addr;
2578 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2579 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2580 			if (af && af != ifma->ifma_addr->sa_family)
2581 				continue;
2582 			if (prison_if(w->w_req->td->td_ucred,
2583 			    ifma->ifma_addr) != 0)
2584 				continue;
2585 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2586 			info.rti_info[RTAX_GATEWAY] =
2587 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2588 			    ifma->ifma_lladdr : NULL;
2589 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2590 			if (error != 0)
2591 				break;
2592 			if (w->w_req && w->w_tmem) {
2593 				struct ifma_msghdr *ifmam;
2594 
2595 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2596 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2597 				ifmam->ifmam_flags = 0;
2598 				ifmam->ifmam_addrs = info.rti_addrs;
2599 				ifmam->_ifmam_spare1 = 0;
2600 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2601 				if (error != 0)
2602 					break;
2603 			}
2604 		}
2605 		if (error != 0)
2606 			break;
2607 	}
2608 	return (error);
2609 }
2610 
2611 static void
rtable_sysctl_dump(uint32_t fibnum,int family,struct walkarg * w)2612 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2613 {
2614 	union sockaddr_union sa_dst, sa_mask;
2615 
2616 	w->family = family;
2617 	w->dst = (struct sockaddr *)&sa_dst;
2618 	w->mask = (struct sockaddr *)&sa_mask;
2619 
2620 	init_sockaddrs_family(family, w->dst, w->mask);
2621 
2622 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2623 }
2624 
2625 static int
sysctl_rtsock(SYSCTL_HANDLER_ARGS)2626 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2627 {
2628 	struct epoch_tracker et;
2629 	int	*name = (int *)arg1;
2630 	u_int	namelen = arg2;
2631 	struct rib_head *rnh = NULL; /* silence compiler. */
2632 	int	i, lim, error = EINVAL;
2633 	int	fib = 0;
2634 	u_char	af;
2635 	struct	walkarg w;
2636 
2637 	if (namelen < 3)
2638 		return (EINVAL);
2639 
2640 	name++;
2641 	namelen--;
2642 	if (req->newptr)
2643 		return (EPERM);
2644 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2645 		if (namelen == 3)
2646 			fib = req->td->td_proc->p_fibnum;
2647 		else if (namelen == 4)
2648 			fib = (name[3] == RT_ALL_FIBS) ?
2649 			    req->td->td_proc->p_fibnum : name[3];
2650 		else
2651 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2652 		if (fib < 0 || fib >= rt_numfibs)
2653 			return (EINVAL);
2654 	} else if (namelen != 3)
2655 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2656 	af = name[0];
2657 	if (af > AF_MAX)
2658 		return (EINVAL);
2659 	bzero(&w, sizeof(w));
2660 	w.w_op = name[1];
2661 	w.w_arg = name[2];
2662 	w.w_req = req;
2663 
2664 	error = sysctl_wire_old_buffer(req, 0);
2665 	if (error)
2666 		return (error);
2667 
2668 	/*
2669 	 * Allocate reply buffer in advance.
2670 	 * All rtsock messages has maximum length of u_short.
2671 	 */
2672 	w.w_tmemsize = 65536;
2673 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2674 
2675 	NET_EPOCH_ENTER(et);
2676 	switch (w.w_op) {
2677 	case NET_RT_DUMP:
2678 	case NET_RT_FLAGS:
2679 		if (af == 0) {			/* dump all tables */
2680 			i = 1;
2681 			lim = AF_MAX;
2682 		} else				/* dump only one table */
2683 			i = lim = af;
2684 
2685 		/*
2686 		 * take care of llinfo entries, the caller must
2687 		 * specify an AF
2688 		 */
2689 		if (w.w_op == NET_RT_FLAGS &&
2690 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2691 			if (af != 0)
2692 				error = lltable_sysctl_dumparp(af, w.w_req);
2693 			else
2694 				error = EINVAL;
2695 			break;
2696 		}
2697 		/*
2698 		 * take care of routing entries
2699 		 */
2700 		for (error = 0; error == 0 && i <= lim; i++) {
2701 			rnh = rt_tables_get_rnh(fib, i);
2702 			if (rnh != NULL) {
2703 				rtable_sysctl_dump(fib, i, &w);
2704 			} else if (af != 0)
2705 				error = EAFNOSUPPORT;
2706 		}
2707 		break;
2708 	case NET_RT_NHOP:
2709 	case NET_RT_NHGRP:
2710 		/* Allow dumping one specific af/fib at a time */
2711 		if (namelen < 4) {
2712 			error = EINVAL;
2713 			break;
2714 		}
2715 		fib = name[3];
2716 		if (fib < 0 || fib > rt_numfibs) {
2717 			error = EINVAL;
2718 			break;
2719 		}
2720 		rnh = rt_tables_get_rnh(fib, af);
2721 		if (rnh == NULL) {
2722 			error = EAFNOSUPPORT;
2723 			break;
2724 		}
2725 		if (w.w_op == NET_RT_NHOP)
2726 			error = nhops_dump_sysctl(rnh, w.w_req);
2727 		else
2728 #ifdef ROUTE_MPATH
2729 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2730 #else
2731 			error = ENOTSUP;
2732 #endif
2733 		break;
2734 	case NET_RT_IFLIST:
2735 	case NET_RT_IFLISTL:
2736 		error = sysctl_iflist(af, &w);
2737 		break;
2738 
2739 	case NET_RT_IFMALIST:
2740 		error = sysctl_ifmalist(af, &w);
2741 		break;
2742 	}
2743 	NET_EPOCH_EXIT(et);
2744 
2745 	free(w.w_tmem, M_TEMP);
2746 	return (error);
2747 }
2748 
2749 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2750     sysctl_rtsock, "Return route tables and interface/address lists");
2751 
2752 /*
2753  * Definitions of protocols supported in the ROUTE domain.
2754  */
2755 
2756 static struct domain routedomain;		/* or at least forward */
2757 
2758 static struct protosw routesw[] = {
2759 {
2760 	.pr_type =		SOCK_RAW,
2761 	.pr_domain =		&routedomain,
2762 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2763 	.pr_output =		route_output,
2764 	.pr_ctlinput =		raw_ctlinput,
2765 	.pr_init =		raw_init,
2766 	.pr_usrreqs =		&route_usrreqs
2767 }
2768 };
2769 
2770 static struct domain routedomain = {
2771 	.dom_family =		PF_ROUTE,
2772 	.dom_name =		"route",
2773 	.dom_protosw =		routesw,
2774 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
2775 };
2776 
2777 VNET_DOMAIN_SET(route);
2778