1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1988, 1991, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31 #include "opt_ddb.h"
32 #include "opt_route.h"
33 #include "opt_inet.h"
34 #include "opt_inet6.h"
35
36 #include <sys/param.h>
37 #include <sys/jail.h>
38 #include <sys/kernel.h>
39 #include <sys/eventhandler.h>
40 #include <sys/domain.h>
41 #include <sys/lock.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/priv.h>
45 #include <sys/proc.h>
46 #include <sys/protosw.h>
47 #include <sys/rmlock.h>
48 #include <sys/rwlock.h>
49 #include <sys/signalvar.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/systm.h>
54
55 #include <net/if.h>
56 #include <net/if_var.h>
57 #include <net/if_private.h>
58 #include <net/if_dl.h>
59 #include <net/if_llatbl.h>
60 #include <net/if_types.h>
61 #include <net/netisr.h>
62 #include <net/route.h>
63 #include <net/route/route_ctl.h>
64 #include <net/route/route_var.h>
65 #include <net/vnet.h>
66
67 #include <netinet/in.h>
68 #include <netinet/if_ether.h>
69 #include <netinet/ip_carp.h>
70 #ifdef INET6
71 #include <netinet6/in6_var.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/scope6_var.h>
74 #endif
75 #include <net/route/nhop.h>
76
77 #define DEBUG_MOD_NAME rtsock
78 #define DEBUG_MAX_LEVEL LOG_DEBUG
79 #include <net/route/route_debug.h>
80 _DECLARE_DEBUG(LOG_INFO);
81
82 #ifdef COMPAT_FREEBSD32
83 #include <sys/mount.h>
84 #include <compat/freebsd32/freebsd32.h>
85
86 struct if_msghdr32 {
87 uint16_t ifm_msglen;
88 uint8_t ifm_version;
89 uint8_t ifm_type;
90 int32_t ifm_addrs;
91 int32_t ifm_flags;
92 uint16_t ifm_index;
93 uint16_t _ifm_spare1;
94 struct if_data ifm_data;
95 };
96
97 struct if_msghdrl32 {
98 uint16_t ifm_msglen;
99 uint8_t ifm_version;
100 uint8_t ifm_type;
101 int32_t ifm_addrs;
102 int32_t ifm_flags;
103 uint16_t ifm_index;
104 uint16_t _ifm_spare1;
105 uint16_t ifm_len;
106 uint16_t ifm_data_off;
107 uint32_t _ifm_spare2;
108 struct if_data ifm_data;
109 };
110
111 struct ifa_msghdrl32 {
112 uint16_t ifam_msglen;
113 uint8_t ifam_version;
114 uint8_t ifam_type;
115 int32_t ifam_addrs;
116 int32_t ifam_flags;
117 uint16_t ifam_index;
118 uint16_t _ifam_spare1;
119 uint16_t ifam_len;
120 uint16_t ifam_data_off;
121 int32_t ifam_metric;
122 struct if_data ifam_data;
123 };
124
125 #define SA_SIZE32(sa) \
126 ( (((struct sockaddr *)(sa))->sa_len == 0) ? \
127 sizeof(int) : \
128 1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
129
130 #endif /* COMPAT_FREEBSD32 */
131
132 struct linear_buffer {
133 char *base; /* Base allocated memory pointer */
134 uint32_t offset; /* Currently used offset */
135 uint32_t size; /* Total buffer size */
136 };
137 #define SCRATCH_BUFFER_SIZE 1024
138
139 #define RTS_PID_LOG(_l, _fmt, ...) \
140 RT_LOG_##_l(_l, "PID %d: " _fmt, curproc ? curproc->p_pid : 0, \
141 ## __VA_ARGS__)
142
143 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
144
145 /* NB: these are not modified */
146 static struct sockaddr route_src = { 2, PF_ROUTE, };
147 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, };
148
149 /* These are external hooks for CARP. */
150 int (*carp_get_vhid_p)(struct ifaddr *);
151
152 /*
153 * Used by rtsock callback code to decide whether to filter the update
154 * notification to a socket bound to a particular FIB.
155 */
156 #define RTS_FILTER_FIB M_PROTO8
157 /*
158 * Used to store address family of the notification.
159 */
160 #define m_rtsock_family m_pkthdr.PH_loc.eight[0]
161
162 struct rcb {
163 LIST_ENTRY(rcb) list;
164 struct socket *rcb_socket;
165 sa_family_t rcb_family;
166 };
167
168 typedef struct {
169 LIST_HEAD(, rcb) cblist;
170 int ip_count; /* attached w/ AF_INET */
171 int ip6_count; /* attached w/ AF_INET6 */
172 int any_count; /* total attached */
173 } route_cb_t;
174 VNET_DEFINE_STATIC(route_cb_t, route_cb);
175 #define V_route_cb VNET(route_cb)
176
177 struct mtx rtsock_mtx;
178 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
179
180 #define RTSOCK_LOCK() mtx_lock(&rtsock_mtx)
181 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
182 #define RTSOCK_LOCK_ASSERT() mtx_assert(&rtsock_mtx, MA_OWNED)
183
184 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
185
186 struct walkarg {
187 int family;
188 int w_tmemsize;
189 int w_op, w_arg;
190 caddr_t w_tmem;
191 struct sysctl_req *w_req;
192 struct sockaddr *dst;
193 struct sockaddr *mask;
194 };
195
196 static void rts_input(struct mbuf *m);
197 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
198 static int rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
199 struct walkarg *w, int *plen);
200 static int rt_xaddrs(caddr_t cp, caddr_t cplim,
201 struct rt_addrinfo *rtinfo);
202 static int cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb);
203 static int sysctl_dumpentry(struct rtentry *rt, void *vw);
204 static int sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
205 uint32_t weight, struct walkarg *w);
206 static int sysctl_iflist(int af, struct walkarg *w);
207 static int sysctl_ifmalist(int af, struct walkarg *w);
208 static void rt_getmetrics(const struct rtentry *rt,
209 const struct nhop_object *nh, struct rt_metrics *out);
210 static void rt_dispatch(struct mbuf *, sa_family_t);
211 static void rt_ifannouncemsg(struct ifnet *ifp, int what);
212 static int handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
213 struct rt_msghdr *rtm, struct rib_cmd_info *rc);
214 static int update_rtm_from_rc(struct rt_addrinfo *info,
215 struct rt_msghdr **prtm, int alloc_len,
216 struct rib_cmd_info *rc, struct nhop_object *nh);
217 static void send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
218 struct mbuf *m, sa_family_t saf, u_int fibnum,
219 int rtm_errno);
220 static void rtsock_notify_event(uint32_t fibnum, const struct rib_cmd_info *rc);
221 static void rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask);
222
223 static struct netisr_handler rtsock_nh = {
224 .nh_name = "rtsock",
225 .nh_handler = rts_input,
226 .nh_proto = NETISR_ROUTE,
227 .nh_policy = NETISR_POLICY_SOURCE,
228 };
229
230 static int
sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)231 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
232 {
233 int error, qlimit;
234
235 netisr_getqlimit(&rtsock_nh, &qlimit);
236 error = sysctl_handle_int(oidp, &qlimit, 0, req);
237 if (error || !req->newptr)
238 return (error);
239 if (qlimit < 1)
240 return (EINVAL);
241 return (netisr_setqlimit(&rtsock_nh, qlimit));
242 }
243 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
244 CTLTYPE_INT | CTLFLAG_RWTUN | CTLFLAG_NOFETCH | CTLFLAG_MPSAFE,
245 0, 0, sysctl_route_netisr_maxqlen, "I",
246 "maximum routing socket dispatch queue length");
247
248 static void
vnet_rts_init(void)249 vnet_rts_init(void)
250 {
251 int tmp;
252
253 if (IS_DEFAULT_VNET(curvnet)) {
254 if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
255 rtsock_nh.nh_qlimit = tmp;
256 netisr_register(&rtsock_nh);
257 }
258 #ifdef VIMAGE
259 else
260 netisr_register_vnet(&rtsock_nh);
261 #endif
262 }
263 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
264 vnet_rts_init, 0);
265
266 #ifdef VIMAGE
267 static void
vnet_rts_uninit(void)268 vnet_rts_uninit(void)
269 {
270
271 netisr_unregister_vnet(&rtsock_nh);
272 }
273 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
274 vnet_rts_uninit, 0);
275 #endif
276
277 static void
report_route_event(const struct rib_cmd_info * rc,void * _cbdata)278 report_route_event(const struct rib_cmd_info *rc, void *_cbdata)
279 {
280 uint32_t fibnum = (uint32_t)(uintptr_t)_cbdata;
281 struct nhop_object *nh;
282
283 nh = rc->rc_cmd == RTM_DELETE ? rc->rc_nh_old : rc->rc_nh_new;
284 rt_routemsg(rc->rc_cmd, rc->rc_rt, nh, fibnum);
285 }
286
287 static void
rts_handle_route_event(uint32_t fibnum,const struct rib_cmd_info * rc)288 rts_handle_route_event(uint32_t fibnum, const struct rib_cmd_info *rc)
289 {
290 #ifdef ROUTE_MPATH
291 if ((rc->rc_nh_new && NH_IS_NHGRP(rc->rc_nh_new)) ||
292 (rc->rc_nh_old && NH_IS_NHGRP(rc->rc_nh_old))) {
293 rib_decompose_notification(rc, report_route_event,
294 (void *)(uintptr_t)fibnum);
295 } else
296 #endif
297 report_route_event(rc, (void *)(uintptr_t)fibnum);
298 }
299 static struct rtbridge rtsbridge = {
300 .route_f = rts_handle_route_event,
301 .ifmsg_f = rtsock_ifmsg,
302 };
303 static struct rtbridge *rtsbridge_orig_p;
304
305 static void
rtsock_notify_event(uint32_t fibnum,const struct rib_cmd_info * rc)306 rtsock_notify_event(uint32_t fibnum, const struct rib_cmd_info *rc)
307 {
308 netlink_callback_p->route_f(fibnum, rc);
309 }
310
311 static void
rtsock_init(void)312 rtsock_init(void)
313 {
314 rtsbridge_orig_p = rtsock_callback_p;
315 rtsock_callback_p = &rtsbridge;
316 }
317 SYSINIT(rtsock_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rtsock_init, NULL);
318
319 static void
rts_handle_ifnet_arrival(void * arg __unused,struct ifnet * ifp)320 rts_handle_ifnet_arrival(void *arg __unused, struct ifnet *ifp)
321 {
322 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
323 }
324 EVENTHANDLER_DEFINE(ifnet_arrival_event, rts_handle_ifnet_arrival, NULL, 0);
325
326 static void
rts_handle_ifnet_departure(void * arg __unused,struct ifnet * ifp)327 rts_handle_ifnet_departure(void *arg __unused, struct ifnet *ifp)
328 {
329 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
330 }
331 EVENTHANDLER_DEFINE(ifnet_departure_event, rts_handle_ifnet_departure, NULL, 0);
332
333 static void
rts_append_data(struct socket * so,struct mbuf * m)334 rts_append_data(struct socket *so, struct mbuf *m)
335 {
336
337 if (sbappendaddr(&so->so_rcv, &route_src, m, NULL) == 0) {
338 soroverflow(so);
339 m_freem(m);
340 } else
341 sorwakeup(so);
342 }
343
344 static void
rts_input(struct mbuf * m)345 rts_input(struct mbuf *m)
346 {
347 struct rcb *rcb;
348 struct socket *last;
349
350 last = NULL;
351 RTSOCK_LOCK();
352 LIST_FOREACH(rcb, &V_route_cb.cblist, list) {
353 if (rcb->rcb_family != AF_UNSPEC &&
354 rcb->rcb_family != m->m_rtsock_family)
355 continue;
356 if ((m->m_flags & RTS_FILTER_FIB) &&
357 M_GETFIB(m) != rcb->rcb_socket->so_fibnum)
358 continue;
359 if (last != NULL) {
360 struct mbuf *n;
361
362 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
363 if (n != NULL)
364 rts_append_data(last, n);
365 }
366 last = rcb->rcb_socket;
367 }
368 if (last != NULL)
369 rts_append_data(last, m);
370 else
371 m_freem(m);
372 RTSOCK_UNLOCK();
373 }
374
375 static void
rts_close(struct socket * so)376 rts_close(struct socket *so)
377 {
378
379 soisdisconnected(so);
380 }
381
382 static SYSCTL_NODE(_net, OID_AUTO, rtsock, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
383 "Routing socket infrastructure");
384 static u_long rts_sendspace = 8192;
385 SYSCTL_ULONG(_net_rtsock, OID_AUTO, sendspace, CTLFLAG_RW, &rts_sendspace, 0,
386 "Default routing socket send space");
387 static u_long rts_recvspace = 8192;
388 SYSCTL_ULONG(_net_rtsock, OID_AUTO, recvspace, CTLFLAG_RW, &rts_recvspace, 0,
389 "Default routing socket receive space");
390
391 static int
rts_attach(struct socket * so,int proto,struct thread * td)392 rts_attach(struct socket *so, int proto, struct thread *td)
393 {
394 struct rcb *rcb;
395 int error;
396
397 error = soreserve(so, rts_sendspace, rts_recvspace);
398 if (error)
399 return (error);
400
401 rcb = malloc(sizeof(*rcb), M_PCB, M_WAITOK);
402 rcb->rcb_socket = so;
403 rcb->rcb_family = proto;
404
405 so->so_pcb = rcb;
406 so->so_fibnum = td->td_proc->p_fibnum;
407 so->so_options |= SO_USELOOPBACK;
408
409 RTSOCK_LOCK();
410 LIST_INSERT_HEAD(&V_route_cb.cblist, rcb, list);
411 switch (proto) {
412 case AF_INET:
413 V_route_cb.ip_count++;
414 break;
415 case AF_INET6:
416 V_route_cb.ip6_count++;
417 break;
418 }
419 V_route_cb.any_count++;
420 RTSOCK_UNLOCK();
421 soisconnected(so);
422
423 return (0);
424 }
425
426 static void
rts_detach(struct socket * so)427 rts_detach(struct socket *so)
428 {
429 struct rcb *rcb = so->so_pcb;
430
431 RTSOCK_LOCK();
432 LIST_REMOVE(rcb, list);
433 switch(rcb->rcb_family) {
434 case AF_INET:
435 V_route_cb.ip_count--;
436 break;
437 case AF_INET6:
438 V_route_cb.ip6_count--;
439 break;
440 }
441 V_route_cb.any_count--;
442 RTSOCK_UNLOCK();
443 free(rcb, M_PCB);
444 so->so_pcb = NULL;
445 }
446
447 static int
rts_disconnect(struct socket * so)448 rts_disconnect(struct socket *so)
449 {
450
451 return (ENOTCONN);
452 }
453
454 static int
rts_shutdown(struct socket * so,enum shutdown_how how)455 rts_shutdown(struct socket *so, enum shutdown_how how)
456 {
457 /*
458 * Note: route socket marks itself as connected through its lifetime.
459 */
460 switch (how) {
461 case SHUT_RD:
462 sorflush(so);
463 break;
464 case SHUT_RDWR:
465 sorflush(so);
466 /* FALLTHROUGH */
467 case SHUT_WR:
468 socantsendmore(so);
469 }
470
471 return (0);
472 }
473
474 #ifndef _SOCKADDR_UNION_DEFINED
475 #define _SOCKADDR_UNION_DEFINED
476 /*
477 * The union of all possible address formats we handle.
478 */
479 union sockaddr_union {
480 struct sockaddr sa;
481 struct sockaddr_in sin;
482 struct sockaddr_in6 sin6;
483 };
484 #endif /* _SOCKADDR_UNION_DEFINED */
485
486 static int
rtm_get_jailed(struct rt_addrinfo * info,struct ifnet * ifp,struct nhop_object * nh,union sockaddr_union * saun,struct ucred * cred)487 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
488 struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
489 {
490 #if defined(INET) || defined(INET6)
491 struct epoch_tracker et;
492 #endif
493
494 /* First, see if the returned address is part of the jail. */
495 if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
496 info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
497 return (0);
498 }
499
500 switch (info->rti_info[RTAX_DST]->sa_family) {
501 #ifdef INET
502 case AF_INET:
503 {
504 struct in_addr ia;
505 struct ifaddr *ifa;
506 int found;
507
508 found = 0;
509 /*
510 * Try to find an address on the given outgoing interface
511 * that belongs to the jail.
512 */
513 NET_EPOCH_ENTER(et);
514 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
515 struct sockaddr *sa;
516 sa = ifa->ifa_addr;
517 if (sa->sa_family != AF_INET)
518 continue;
519 ia = ((struct sockaddr_in *)sa)->sin_addr;
520 if (prison_check_ip4(cred, &ia) == 0) {
521 found = 1;
522 break;
523 }
524 }
525 NET_EPOCH_EXIT(et);
526 if (!found) {
527 /*
528 * As a last resort return the 'default' jail address.
529 */
530 ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
531 sin_addr;
532 if (prison_get_ip4(cred, &ia) != 0)
533 return (ESRCH);
534 }
535 bzero(&saun->sin, sizeof(struct sockaddr_in));
536 saun->sin.sin_len = sizeof(struct sockaddr_in);
537 saun->sin.sin_family = AF_INET;
538 saun->sin.sin_addr.s_addr = ia.s_addr;
539 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
540 break;
541 }
542 #endif
543 #ifdef INET6
544 case AF_INET6:
545 {
546 struct in6_addr ia6;
547 struct ifaddr *ifa;
548 int found;
549
550 found = 0;
551 /*
552 * Try to find an address on the given outgoing interface
553 * that belongs to the jail.
554 */
555 NET_EPOCH_ENTER(et);
556 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
557 struct sockaddr *sa;
558 sa = ifa->ifa_addr;
559 if (sa->sa_family != AF_INET6)
560 continue;
561 bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
562 &ia6, sizeof(struct in6_addr));
563 if (prison_check_ip6(cred, &ia6) == 0) {
564 found = 1;
565 break;
566 }
567 }
568 NET_EPOCH_EXIT(et);
569 if (!found) {
570 /*
571 * As a last resort return the 'default' jail address.
572 */
573 ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
574 sin6_addr;
575 if (prison_get_ip6(cred, &ia6) != 0)
576 return (ESRCH);
577 }
578 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
579 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
580 saun->sin6.sin6_family = AF_INET6;
581 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
582 if (sa6_recoverscope(&saun->sin6) != 0)
583 return (ESRCH);
584 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
585 break;
586 }
587 #endif
588 default:
589 return (ESRCH);
590 }
591 return (0);
592 }
593
594 static int
fill_blackholeinfo(struct rt_addrinfo * info,union sockaddr_union * saun)595 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
596 {
597 struct ifaddr *ifa;
598 sa_family_t saf;
599
600 if (V_loif == NULL) {
601 RTS_PID_LOG(LOG_INFO, "Unable to add blackhole/reject nhop without loopback");
602 return (ENOTSUP);
603 }
604 info->rti_ifp = V_loif;
605
606 saf = info->rti_info[RTAX_DST]->sa_family;
607
608 CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
609 if (ifa->ifa_addr->sa_family == saf) {
610 info->rti_ifa = ifa;
611 break;
612 }
613 }
614 if (info->rti_ifa == NULL) {
615 RTS_PID_LOG(LOG_INFO, "Unable to find ifa for blackhole/reject nhop");
616 return (ENOTSUP);
617 }
618
619 bzero(saun, sizeof(union sockaddr_union));
620 switch (saf) {
621 #ifdef INET
622 case AF_INET:
623 saun->sin.sin_family = AF_INET;
624 saun->sin.sin_len = sizeof(struct sockaddr_in);
625 saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
626 break;
627 #endif
628 #ifdef INET6
629 case AF_INET6:
630 saun->sin6.sin6_family = AF_INET6;
631 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
632 saun->sin6.sin6_addr = in6addr_loopback;
633 break;
634 #endif
635 default:
636 RTS_PID_LOG(LOG_INFO, "unsupported family: %d", saf);
637 return (ENOTSUP);
638 }
639 info->rti_info[RTAX_GATEWAY] = &saun->sa;
640 info->rti_flags |= RTF_GATEWAY;
641
642 return (0);
643 }
644
645 /*
646 * Fills in @info based on userland-provided @rtm message.
647 *
648 * Returns 0 on success.
649 */
650 static int
fill_addrinfo(struct rt_msghdr * rtm,int len,struct linear_buffer * lb,u_int fibnum,struct rt_addrinfo * info)651 fill_addrinfo(struct rt_msghdr *rtm, int len, struct linear_buffer *lb, u_int fibnum,
652 struct rt_addrinfo *info)
653 {
654 int error;
655
656 rtm->rtm_pid = curproc->p_pid;
657 info->rti_addrs = rtm->rtm_addrs;
658
659 info->rti_mflags = rtm->rtm_inits;
660 info->rti_rmx = &rtm->rtm_rmx;
661
662 /*
663 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
664 * link-local address because rtrequest requires addresses with
665 * embedded scope id.
666 */
667 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
668 return (EINVAL);
669
670 info->rti_flags = rtm->rtm_flags;
671 error = cleanup_xaddrs(info, lb);
672 if (error != 0)
673 return (error);
674 /*
675 * Verify that the caller has the appropriate privilege; RTM_GET
676 * is the only operation the non-superuser is allowed.
677 */
678 if (rtm->rtm_type != RTM_GET) {
679 error = priv_check(curthread, PRIV_NET_ROUTE);
680 if (error != 0)
681 return (error);
682 }
683
684 /*
685 * The given gateway address may be an interface address.
686 * For example, issuing a "route change" command on a route
687 * entry that was created from a tunnel, and the gateway
688 * address given is the local end point. In this case the
689 * RTF_GATEWAY flag must be cleared or the destination will
690 * not be reachable even though there is no error message.
691 */
692 if (info->rti_info[RTAX_GATEWAY] != NULL &&
693 info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
694 struct nhop_object *nh;
695
696 /*
697 * A host route through the loopback interface is
698 * installed for each interface address. In pre 8.0
699 * releases the interface address of a PPP link type
700 * is not reachable locally. This behavior is fixed as
701 * part of the new L2/L3 redesign and rewrite work. The
702 * signature of this interface address route is the
703 * AF_LINK sa_family type of the gateway, and the
704 * rt_ifp has the IFF_LOOPBACK flag set.
705 */
706 nh = rib_lookup(fibnum, info->rti_info[RTAX_GATEWAY], NHR_NONE, 0);
707 if (nh != NULL && nh->gw_sa.sa_family == AF_LINK &&
708 nh->nh_ifp->if_flags & IFF_LOOPBACK) {
709 info->rti_flags &= ~RTF_GATEWAY;
710 info->rti_flags |= RTF_GWFLAG_COMPAT;
711 }
712 }
713
714 return (0);
715 }
716
717 static struct nhop_object *
select_nhop(struct nhop_object * nh,const struct sockaddr * gw)718 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
719 {
720 if (!NH_IS_NHGRP(nh))
721 return (nh);
722 #ifdef ROUTE_MPATH
723 const struct weightened_nhop *wn;
724 uint32_t num_nhops;
725 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
726 if (gw == NULL)
727 return (wn[0].nh);
728 for (int i = 0; i < num_nhops; i++) {
729 if (match_nhop_gw(wn[i].nh, gw))
730 return (wn[i].nh);
731 }
732 #endif
733 return (NULL);
734 }
735
736 /*
737 * Handles RTM_GET message from routing socket, returning matching rt.
738 *
739 * Returns:
740 * 0 on success, with locked and referenced matching rt in @rt_nrt
741 * errno of failure
742 */
743 static int
handle_rtm_get(struct rt_addrinfo * info,u_int fibnum,struct rt_msghdr * rtm,struct rib_cmd_info * rc)744 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
745 struct rt_msghdr *rtm, struct rib_cmd_info *rc)
746 {
747 RIB_RLOCK_TRACKER;
748 struct rib_head *rnh;
749 struct nhop_object *nh;
750 sa_family_t saf;
751
752 saf = info->rti_info[RTAX_DST]->sa_family;
753
754 rnh = rt_tables_get_rnh(fibnum, saf);
755 if (rnh == NULL)
756 return (EAFNOSUPPORT);
757
758 RIB_RLOCK(rnh);
759
760 /*
761 * By (implicit) convention host route (one without netmask)
762 * means longest-prefix-match request and the route with netmask
763 * means exact-match lookup.
764 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128
765 * prefixes, use original data to check for the netmask presence.
766 */
767 if ((rtm->rtm_addrs & RTA_NETMASK) == 0) {
768 /*
769 * Provide longest prefix match for
770 * address lookup (no mask).
771 * 'route -n get addr'
772 */
773 rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
774 info->rti_info[RTAX_DST], &rnh->head);
775 } else
776 rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
777 info->rti_info[RTAX_DST],
778 info->rti_info[RTAX_NETMASK], &rnh->head);
779
780 if (rc->rc_rt == NULL) {
781 RIB_RUNLOCK(rnh);
782 return (ESRCH);
783 }
784
785 nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
786 if (nh == NULL) {
787 RIB_RUNLOCK(rnh);
788 return (ESRCH);
789 }
790 /*
791 * If performing proxied L2 entry insertion, and
792 * the actual PPP host entry is found, perform
793 * another search to retrieve the prefix route of
794 * the local end point of the PPP link.
795 * TODO: move this logic to userland.
796 */
797 if (rtm->rtm_flags & RTF_ANNOUNCE) {
798 struct sockaddr_storage laddr;
799
800 if (nh->nh_ifp != NULL &&
801 nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
802 struct ifaddr *ifa;
803
804 ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
805 RT_ALL_FIBS);
806 if (ifa != NULL)
807 rt_maskedcopy(ifa->ifa_addr,
808 (struct sockaddr *)&laddr,
809 ifa->ifa_netmask);
810 } else
811 rt_maskedcopy(nh->nh_ifa->ifa_addr,
812 (struct sockaddr *)&laddr,
813 nh->nh_ifa->ifa_netmask);
814 /*
815 * refactor rt and no lock operation necessary
816 */
817 rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(
818 (struct sockaddr *)&laddr, &rnh->head);
819 if (rc->rc_rt == NULL) {
820 RIB_RUNLOCK(rnh);
821 return (ESRCH);
822 }
823 nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
824 if (nh == NULL) {
825 RIB_RUNLOCK(rnh);
826 return (ESRCH);
827 }
828 }
829 rc->rc_nh_new = nh;
830 rc->rc_nh_weight = rc->rc_rt->rt_weight;
831 RIB_RUNLOCK(rnh);
832
833 return (0);
834 }
835
836 static void
init_sockaddrs_family(int family,struct sockaddr * dst,struct sockaddr * mask)837 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
838 {
839 #ifdef INET
840 if (family == AF_INET) {
841 struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
842 struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
843
844 bzero(dst4, sizeof(struct sockaddr_in));
845 bzero(mask4, sizeof(struct sockaddr_in));
846
847 dst4->sin_family = AF_INET;
848 dst4->sin_len = sizeof(struct sockaddr_in);
849 mask4->sin_family = AF_INET;
850 mask4->sin_len = sizeof(struct sockaddr_in);
851 }
852 #endif
853 #ifdef INET6
854 if (family == AF_INET6) {
855 struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
856 struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
857
858 bzero(dst6, sizeof(struct sockaddr_in6));
859 bzero(mask6, sizeof(struct sockaddr_in6));
860
861 dst6->sin6_family = AF_INET6;
862 dst6->sin6_len = sizeof(struct sockaddr_in6);
863 mask6->sin6_family = AF_INET6;
864 mask6->sin6_len = sizeof(struct sockaddr_in6);
865 }
866 #endif
867 }
868
869 static void
export_rtaddrs(const struct rtentry * rt,struct sockaddr * dst,struct sockaddr * mask)870 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
871 struct sockaddr *mask)
872 {
873 #ifdef INET
874 if (dst->sa_family == AF_INET) {
875 struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
876 struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
877 uint32_t scopeid = 0;
878 rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
879 &scopeid);
880 return;
881 }
882 #endif
883 #ifdef INET6
884 if (dst->sa_family == AF_INET6) {
885 struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
886 struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
887 uint32_t scopeid = 0;
888 rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
889 &mask6->sin6_addr, &scopeid);
890 dst6->sin6_scope_id = scopeid;
891 return;
892 }
893 #endif
894 }
895
896 static int
update_rtm_from_info(struct rt_addrinfo * info,struct rt_msghdr ** prtm,int alloc_len)897 update_rtm_from_info(struct rt_addrinfo *info, struct rt_msghdr **prtm,
898 int alloc_len)
899 {
900 struct rt_msghdr *rtm, *orig_rtm = NULL;
901 struct walkarg w;
902 int len;
903
904 rtm = *prtm;
905 /* Check if we need to realloc storage */
906 rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
907 if (len > alloc_len) {
908 struct rt_msghdr *tmp_rtm;
909
910 tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
911 if (tmp_rtm == NULL)
912 return (ENOBUFS);
913 bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
914 orig_rtm = rtm;
915 rtm = tmp_rtm;
916 alloc_len = len;
917
918 /*
919 * Delay freeing original rtm as info contains
920 * data referencing it.
921 */
922 }
923
924 w.w_tmem = (caddr_t)rtm;
925 w.w_tmemsize = alloc_len;
926 rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
927 rtm->rtm_addrs = info->rti_addrs;
928
929 if (orig_rtm != NULL)
930 free(orig_rtm, M_TEMP);
931 *prtm = rtm;
932 return (0);
933 }
934
935
936 /*
937 * Update sockaddrs, flags, etc in @prtm based on @rc data.
938 * rtm can be reallocated.
939 *
940 * Returns 0 on success, along with pointer to (potentially reallocated)
941 * rtm.
942 *
943 */
944 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)945 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
946 int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
947 {
948 union sockaddr_union saun;
949 struct rt_msghdr *rtm;
950 struct ifnet *ifp;
951 int error;
952
953 rtm = *prtm;
954 union sockaddr_union sa_dst, sa_mask;
955 int family = info->rti_info[RTAX_DST]->sa_family;
956 init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
957 export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
958
959 info->rti_info[RTAX_DST] = &sa_dst.sa;
960 info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
961 info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
962 info->rti_info[RTAX_GENMASK] = 0;
963 ifp = nh->nh_ifp;
964 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
965 if (ifp) {
966 info->rti_info[RTAX_IFP] =
967 ifp->if_addr->ifa_addr;
968 error = rtm_get_jailed(info, ifp, nh,
969 &saun, curthread->td_ucred);
970 if (error != 0)
971 return (error);
972 if (ifp->if_flags & IFF_POINTOPOINT)
973 info->rti_info[RTAX_BRD] =
974 nh->nh_ifa->ifa_dstaddr;
975 rtm->rtm_index = ifp->if_index;
976 } else {
977 info->rti_info[RTAX_IFP] = NULL;
978 info->rti_info[RTAX_IFA] = NULL;
979 }
980 } else if (ifp != NULL)
981 rtm->rtm_index = ifp->if_index;
982
983 if ((error = update_rtm_from_info(info, prtm, alloc_len)) != 0)
984 return (error);
985
986 rtm = *prtm;
987 rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
988 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
989 rtm->rtm_flags = RTF_GATEWAY |
990 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
991 rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
992 rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
993
994 return (0);
995 }
996
997 #ifdef ROUTE_MPATH
998 static void
save_del_notification(const struct rib_cmd_info * rc,void * _cbdata)999 save_del_notification(const struct rib_cmd_info *rc, void *_cbdata)
1000 {
1001 struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1002
1003 if (rc->rc_cmd == RTM_DELETE)
1004 *rc_new = *rc;
1005 }
1006
1007 static void
save_add_notification(const struct rib_cmd_info * rc,void * _cbdata)1008 save_add_notification(const struct rib_cmd_info *rc, void *_cbdata)
1009 {
1010 struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
1011
1012 if (rc->rc_cmd == RTM_ADD)
1013 *rc_new = *rc;
1014 }
1015 #endif
1016
1017 #if defined(INET6) || defined(INET)
1018 static struct sockaddr *
alloc_sockaddr_aligned(struct linear_buffer * lb,int len)1019 alloc_sockaddr_aligned(struct linear_buffer *lb, int len)
1020 {
1021 len = roundup2(len, sizeof(uint64_t));
1022 if (lb->offset + len > lb->size)
1023 return (NULL);
1024 struct sockaddr *sa = (struct sockaddr *)(lb->base + lb->offset);
1025 lb->offset += len;
1026 return (sa);
1027 }
1028 #endif
1029
1030 static int
rts_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)1031 rts_send(struct socket *so, int flags, struct mbuf *m,
1032 struct sockaddr *nam, struct mbuf *control, struct thread *td)
1033 {
1034 struct rt_msghdr *rtm = NULL;
1035 struct rt_addrinfo info;
1036 struct epoch_tracker et;
1037 #ifdef INET6
1038 struct sockaddr_storage ss;
1039 struct sockaddr_in6 *sin6;
1040 int i, rti_need_deembed = 0;
1041 #endif
1042 int alloc_len = 0, len, error = 0, fibnum;
1043 sa_family_t saf = AF_UNSPEC;
1044 struct rib_cmd_info rc;
1045 struct nhop_object *nh;
1046
1047 if ((flags & PRUS_OOB) || control != NULL) {
1048 m_freem(m);
1049 if (control != NULL)
1050 m_freem(control);
1051 return (EOPNOTSUPP);
1052 }
1053
1054 fibnum = so->so_fibnum;
1055 #define senderr(e) { error = e; goto flush;}
1056 if (m == NULL || ((m->m_len < sizeof(long)) &&
1057 (m = m_pullup(m, sizeof(long))) == NULL))
1058 return (ENOBUFS);
1059 if ((m->m_flags & M_PKTHDR) == 0)
1060 panic("route_output");
1061 NET_EPOCH_ENTER(et);
1062 len = m->m_pkthdr.len;
1063 if (len < sizeof(*rtm) ||
1064 len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1065 senderr(EINVAL);
1066
1067 /*
1068 * Most of current messages are in range 200-240 bytes,
1069 * minimize possible re-allocation on reply using larger size
1070 * buffer aligned on 1k boundaty.
1071 */
1072 alloc_len = roundup2(len, 1024);
1073 int total_len = alloc_len + SCRATCH_BUFFER_SIZE;
1074 if ((rtm = malloc(total_len, M_TEMP, M_NOWAIT)) == NULL)
1075 senderr(ENOBUFS);
1076
1077 m_copydata(m, 0, len, (caddr_t)rtm);
1078 bzero(&info, sizeof(info));
1079 nh = NULL;
1080 struct linear_buffer lb = {
1081 .base = (char *)rtm + alloc_len,
1082 .size = SCRATCH_BUFFER_SIZE,
1083 };
1084
1085 if (rtm->rtm_version != RTM_VERSION) {
1086 /* Do not touch message since format is unknown */
1087 free(rtm, M_TEMP);
1088 rtm = NULL;
1089 senderr(EPROTONOSUPPORT);
1090 }
1091
1092 /*
1093 * Starting from here, it is possible
1094 * to alter original message and insert
1095 * caller PID and error value.
1096 */
1097
1098 if ((error = fill_addrinfo(rtm, len, &lb, fibnum, &info)) != 0) {
1099 senderr(error);
1100 }
1101 /* fill_addringo() embeds scope into IPv6 addresses */
1102 #ifdef INET6
1103 rti_need_deembed = 1;
1104 #endif
1105
1106 saf = info.rti_info[RTAX_DST]->sa_family;
1107
1108 /* support for new ARP code */
1109 if (rtm->rtm_flags & RTF_LLDATA) {
1110 error = lla_rt_output(rtm, &info);
1111 goto flush;
1112 }
1113
1114 union sockaddr_union gw_saun;
1115 int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1116 if (blackhole_flags != 0) {
1117 if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1118 error = fill_blackholeinfo(&info, &gw_saun);
1119 else {
1120 RTS_PID_LOG(LOG_DEBUG, "both BLACKHOLE and REJECT flags specifiied");
1121 error = EINVAL;
1122 }
1123 if (error != 0)
1124 senderr(error);
1125 }
1126
1127 switch (rtm->rtm_type) {
1128 case RTM_ADD:
1129 case RTM_CHANGE:
1130 if (rtm->rtm_type == RTM_ADD) {
1131 if (info.rti_info[RTAX_GATEWAY] == NULL) {
1132 RTS_PID_LOG(LOG_DEBUG, "RTM_ADD w/o gateway");
1133 senderr(EINVAL);
1134 }
1135 }
1136 error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1137 if (error == 0) {
1138 rtsock_notify_event(fibnum, &rc);
1139 #ifdef ROUTE_MPATH
1140 if (NH_IS_NHGRP(rc.rc_nh_new) ||
1141 (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1142 struct rib_cmd_info rc_simple = {};
1143 rib_decompose_notification(&rc,
1144 save_add_notification, (void *)&rc_simple);
1145 rc = rc_simple;
1146 }
1147 #endif
1148 /* nh MAY be empty if RTM_CHANGE request is no-op */
1149 nh = rc.rc_nh_new;
1150 if (nh != NULL) {
1151 rtm->rtm_index = nh->nh_ifp->if_index;
1152 rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1153 }
1154 }
1155 break;
1156
1157 case RTM_DELETE:
1158 error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1159 if (error == 0) {
1160 rtsock_notify_event(fibnum, &rc);
1161 #ifdef ROUTE_MPATH
1162 if (NH_IS_NHGRP(rc.rc_nh_old) ||
1163 (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1164 struct rib_cmd_info rc_simple = {};
1165 rib_decompose_notification(&rc,
1166 save_del_notification, (void *)&rc_simple);
1167 rc = rc_simple;
1168 }
1169 #endif
1170 nh = rc.rc_nh_old;
1171 }
1172 break;
1173
1174 case RTM_GET:
1175 error = handle_rtm_get(&info, fibnum, rtm, &rc);
1176 if (error != 0)
1177 senderr(error);
1178 nh = rc.rc_nh_new;
1179
1180 if (!rt_is_exportable(rc.rc_rt, curthread->td_ucred))
1181 senderr(ESRCH);
1182 break;
1183
1184 default:
1185 senderr(EOPNOTSUPP);
1186 }
1187
1188 if (error == 0 && nh != NULL) {
1189 error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1190 /*
1191 * Note that some sockaddr pointers may have changed to
1192 * point to memory outsize @rtm. Some may be pointing
1193 * to the on-stack variables.
1194 * Given that, any pointer in @info CANNOT BE USED.
1195 */
1196
1197 /*
1198 * scopeid deembedding has been performed while
1199 * writing updated rtm in rtsock_msg_buffer().
1200 * With that in mind, skip deembedding procedure below.
1201 */
1202 #ifdef INET6
1203 rti_need_deembed = 0;
1204 #endif
1205 }
1206
1207 flush:
1208 NET_EPOCH_EXIT(et);
1209
1210 #ifdef INET6
1211 if (rtm != NULL) {
1212 if (rti_need_deembed) {
1213 /* sin6_scope_id is recovered before sending rtm. */
1214 sin6 = (struct sockaddr_in6 *)&ss;
1215 for (i = 0; i < RTAX_MAX; i++) {
1216 if (info.rti_info[i] == NULL)
1217 continue;
1218 if (info.rti_info[i]->sa_family != AF_INET6)
1219 continue;
1220 bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1221 if (sa6_recoverscope(sin6) == 0)
1222 bcopy(sin6, info.rti_info[i],
1223 sizeof(*sin6));
1224 }
1225 if (update_rtm_from_info(&info, &rtm, alloc_len) != 0) {
1226 if (error != 0)
1227 error = ENOBUFS;
1228 }
1229 }
1230 }
1231 #endif
1232 send_rtm_reply(so, rtm, m, saf, fibnum, error);
1233
1234 return (error);
1235 }
1236
1237 /*
1238 * Sends the prepared reply message in @rtm to all rtsock clients.
1239 * Frees @m and @rtm.
1240 *
1241 */
1242 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)1243 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1244 sa_family_t saf, u_int fibnum, int rtm_errno)
1245 {
1246 struct rcb *rcb = NULL;
1247
1248 /*
1249 * Check to see if we don't want our own messages.
1250 */
1251 if ((so->so_options & SO_USELOOPBACK) == 0) {
1252 if (V_route_cb.any_count <= 1) {
1253 if (rtm != NULL)
1254 free(rtm, M_TEMP);
1255 m_freem(m);
1256 return;
1257 }
1258 /* There is another listener, so construct message */
1259 rcb = so->so_pcb;
1260 }
1261
1262 if (rtm != NULL) {
1263 if (rtm_errno!= 0)
1264 rtm->rtm_errno = rtm_errno;
1265 else
1266 rtm->rtm_flags |= RTF_DONE;
1267
1268 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1269 if (m->m_pkthdr.len < rtm->rtm_msglen) {
1270 m_freem(m);
1271 m = NULL;
1272 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
1273 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1274
1275 free(rtm, M_TEMP);
1276 }
1277 if (m != NULL) {
1278 M_SETFIB(m, fibnum);
1279 m->m_flags |= RTS_FILTER_FIB;
1280 if (rcb) {
1281 /*
1282 * XXX insure we don't get a copy by
1283 * invalidating our protocol
1284 */
1285 sa_family_t family = rcb->rcb_family;
1286 rcb->rcb_family = AF_UNSPEC;
1287 rt_dispatch(m, saf);
1288 rcb->rcb_family = family;
1289 } else
1290 rt_dispatch(m, saf);
1291 }
1292 }
1293
1294 static void
rt_getmetrics(const struct rtentry * rt,const struct nhop_object * nh,struct rt_metrics * out)1295 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1296 struct rt_metrics *out)
1297 {
1298
1299 bzero(out, sizeof(*out));
1300 out->rmx_mtu = nh->nh_mtu;
1301 out->rmx_weight = rt->rt_weight;
1302 out->rmx_nhidx = nhop_get_idx(nh);
1303 /* Kernel -> userland timebase conversion. */
1304 out->rmx_expire = nhop_get_expire(nh) ?
1305 nhop_get_expire(nh) - time_uptime + time_second : 0;
1306 }
1307
1308 /*
1309 * Extract the addresses of the passed sockaddrs.
1310 * Do a little sanity checking so as to avoid bad memory references.
1311 * This data is derived straight from userland.
1312 */
1313 static int
rt_xaddrs(caddr_t cp,caddr_t cplim,struct rt_addrinfo * rtinfo)1314 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1315 {
1316 struct sockaddr *sa;
1317 int i;
1318
1319 for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1320 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1321 continue;
1322 sa = (struct sockaddr *)cp;
1323 /*
1324 * It won't fit.
1325 */
1326 if (cp + sa->sa_len > cplim) {
1327 RTS_PID_LOG(LOG_DEBUG, "sa_len too big for sa type %d", i);
1328 return (EINVAL);
1329 }
1330 /*
1331 * there are no more.. quit now
1332 * If there are more bits, they are in error.
1333 * I've seen this. route(1) can evidently generate these.
1334 * This causes kernel to core dump.
1335 * for compatibility, If we see this, point to a safe address.
1336 */
1337 if (sa->sa_len == 0) {
1338 rtinfo->rti_info[i] = &sa_zero;
1339 return (0); /* should be EINVAL but for compat */
1340 }
1341 /* accept it */
1342 #ifdef INET6
1343 if (sa->sa_family == AF_INET6)
1344 sa6_embedscope((struct sockaddr_in6 *)sa,
1345 V_ip6_use_defzone);
1346 #endif
1347 rtinfo->rti_info[i] = sa;
1348 cp += SA_SIZE(sa);
1349 }
1350 return (0);
1351 }
1352
1353 #ifdef INET
1354 static inline void
fill_sockaddr_inet(struct sockaddr_in * sin,struct in_addr addr)1355 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1356 {
1357
1358 const struct sockaddr_in nsin = {
1359 .sin_family = AF_INET,
1360 .sin_len = sizeof(struct sockaddr_in),
1361 .sin_addr = addr,
1362 };
1363 *sin = nsin;
1364 }
1365 #endif
1366
1367 #ifdef INET6
1368 static inline void
fill_sockaddr_inet6(struct sockaddr_in6 * sin6,const struct in6_addr * addr6,uint32_t scopeid)1369 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1370 uint32_t scopeid)
1371 {
1372
1373 const struct sockaddr_in6 nsin6 = {
1374 .sin6_family = AF_INET6,
1375 .sin6_len = sizeof(struct sockaddr_in6),
1376 .sin6_addr = *addr6,
1377 .sin6_scope_id = scopeid,
1378 };
1379 *sin6 = nsin6;
1380 }
1381 #endif
1382
1383 #if defined(INET6) || defined(INET)
1384 /*
1385 * Checks if gateway is suitable for lltable operations.
1386 * Lltable code requires AF_LINK gateway with ifindex
1387 * and mac address specified.
1388 * Returns 0 on success.
1389 */
1390 static int
cleanup_xaddrs_lladdr(struct rt_addrinfo * info)1391 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1392 {
1393 struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1394
1395 if (sdl->sdl_family != AF_LINK)
1396 return (EINVAL);
1397
1398 if (sdl->sdl_index == 0) {
1399 RTS_PID_LOG(LOG_DEBUG, "AF_LINK gateway w/o ifindex");
1400 return (EINVAL);
1401 }
1402
1403 if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len) {
1404 RTS_PID_LOG(LOG_DEBUG, "AF_LINK gw: sdl_nlen/sdl_alen too large");
1405 return (EINVAL);
1406 }
1407
1408 return (0);
1409 }
1410
1411 static int
cleanup_xaddrs_gateway(struct rt_addrinfo * info,struct linear_buffer * lb)1412 cleanup_xaddrs_gateway(struct rt_addrinfo *info, struct linear_buffer *lb)
1413 {
1414 struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1415 struct sockaddr *sa;
1416
1417 if (info->rti_flags & RTF_LLDATA)
1418 return (cleanup_xaddrs_lladdr(info));
1419
1420 switch (gw->sa_family) {
1421 #ifdef INET
1422 case AF_INET:
1423 {
1424 struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw;
1425
1426 /* Ensure reads do not go beyoud SA boundary */
1427 if (SA_SIZE(gw) < offsetof(struct sockaddr_in, sin_zero)) {
1428 RTS_PID_LOG(LOG_DEBUG, "gateway sin_len too small: %d",
1429 gw->sa_len);
1430 return (EINVAL);
1431 }
1432 sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_in));
1433 if (sa == NULL)
1434 return (ENOBUFS);
1435 fill_sockaddr_inet((struct sockaddr_in *)sa, gw_sin->sin_addr);
1436 info->rti_info[RTAX_GATEWAY] = sa;
1437 }
1438 break;
1439 #endif
1440 #ifdef INET6
1441 case AF_INET6:
1442 {
1443 struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1444 if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1445 RTS_PID_LOG(LOG_DEBUG, "gateway sin6_len too small: %d",
1446 gw->sa_len);
1447 return (EINVAL);
1448 }
1449 fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1450 break;
1451 }
1452 #endif
1453 case AF_LINK:
1454 {
1455 struct sockaddr_dl *gw_sdl;
1456
1457 size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data);
1458 gw_sdl = (struct sockaddr_dl *)gw;
1459 if (gw_sdl->sdl_len < sdl_min_len) {
1460 RTS_PID_LOG(LOG_DEBUG, "gateway sdl_len too small: %d",
1461 gw_sdl->sdl_len);
1462 return (EINVAL);
1463 }
1464 sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_dl_short));
1465 if (sa == NULL)
1466 return (ENOBUFS);
1467
1468 const struct sockaddr_dl_short sdl = {
1469 .sdl_family = AF_LINK,
1470 .sdl_len = sizeof(struct sockaddr_dl_short),
1471 .sdl_index = gw_sdl->sdl_index,
1472 };
1473 *((struct sockaddr_dl_short *)sa) = sdl;
1474 info->rti_info[RTAX_GATEWAY] = sa;
1475 break;
1476 }
1477 }
1478
1479 return (0);
1480 }
1481 #endif
1482
1483 static void
remove_netmask(struct rt_addrinfo * info)1484 remove_netmask(struct rt_addrinfo *info)
1485 {
1486 info->rti_info[RTAX_NETMASK] = NULL;
1487 info->rti_flags |= RTF_HOST;
1488 info->rti_addrs &= ~RTA_NETMASK;
1489 }
1490
1491 #ifdef INET
1492 static int
cleanup_xaddrs_inet(struct rt_addrinfo * info,struct linear_buffer * lb)1493 cleanup_xaddrs_inet(struct rt_addrinfo *info, struct linear_buffer *lb)
1494 {
1495 struct sockaddr_in *dst_sa, *mask_sa;
1496 const int sa_len = sizeof(struct sockaddr_in);
1497 struct in_addr dst, mask;
1498
1499 /* Check & fixup dst/netmask combination first */
1500 dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1501 mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1502
1503 /* Ensure reads do not go beyound the buffer size */
1504 if (SA_SIZE(dst_sa) < offsetof(struct sockaddr_in, sin_zero)) {
1505 RTS_PID_LOG(LOG_DEBUG, "prefix dst sin_len too small: %d",
1506 dst_sa->sin_len);
1507 return (EINVAL);
1508 }
1509
1510 if ((mask_sa != NULL) && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1511 /*
1512 * Some older routing software encode mask length into the
1513 * sin_len, thus resulting in "truncated" sockaddr.
1514 */
1515 int len = mask_sa->sin_len - offsetof(struct sockaddr_in, sin_addr);
1516 if (len >= 0) {
1517 mask.s_addr = 0;
1518 if (len > sizeof(struct in_addr))
1519 len = sizeof(struct in_addr);
1520 memcpy(&mask, &mask_sa->sin_addr, len);
1521 } else {
1522 RTS_PID_LOG(LOG_DEBUG, "prefix mask sin_len too small: %d",
1523 mask_sa->sin_len);
1524 return (EINVAL);
1525 }
1526 } else
1527 mask.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST;
1528
1529 dst.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr));
1530
1531 /* Construct new "clean" dst/mask sockaddresses */
1532 if ((dst_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1533 return (ENOBUFS);
1534 fill_sockaddr_inet(dst_sa, dst);
1535 info->rti_info[RTAX_DST] = (struct sockaddr *)dst_sa;
1536
1537 if (mask.s_addr != INADDR_BROADCAST) {
1538 if ((mask_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1539 return (ENOBUFS);
1540 fill_sockaddr_inet(mask_sa, mask);
1541 info->rti_info[RTAX_NETMASK] = (struct sockaddr *)mask_sa;
1542 info->rti_flags &= ~RTF_HOST;
1543 } else
1544 remove_netmask(info);
1545
1546 /* Check gateway */
1547 if (info->rti_info[RTAX_GATEWAY] != NULL)
1548 return (cleanup_xaddrs_gateway(info, lb));
1549
1550 return (0);
1551 }
1552 #endif
1553
1554 #ifdef INET6
1555 static int
cleanup_xaddrs_inet6(struct rt_addrinfo * info,struct linear_buffer * lb)1556 cleanup_xaddrs_inet6(struct rt_addrinfo *info, struct linear_buffer *lb)
1557 {
1558 struct sockaddr *sa;
1559 struct sockaddr_in6 *dst_sa, *mask_sa;
1560 struct in6_addr mask, *dst;
1561 const int sa_len = sizeof(struct sockaddr_in6);
1562
1563 /* Check & fixup dst/netmask combination first */
1564 dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1565 mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1566
1567 if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1568 RTS_PID_LOG(LOG_DEBUG, "prefix dst sin6_len too small: %d",
1569 dst_sa->sin6_len);
1570 return (EINVAL);
1571 }
1572
1573 if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1574 /*
1575 * Some older routing software encode mask length into the
1576 * sin6_len, thus resulting in "truncated" sockaddr.
1577 */
1578 int len = mask_sa->sin6_len - offsetof(struct sockaddr_in6, sin6_addr);
1579 if (len >= 0) {
1580 bzero(&mask, sizeof(mask));
1581 if (len > sizeof(struct in6_addr))
1582 len = sizeof(struct in6_addr);
1583 memcpy(&mask, &mask_sa->sin6_addr, len);
1584 } else {
1585 RTS_PID_LOG(LOG_DEBUG, "rtsock: prefix mask sin6_len too small: %d",
1586 mask_sa->sin6_len);
1587 return (EINVAL);
1588 }
1589 } else
1590 mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1591
1592 dst = &dst_sa->sin6_addr;
1593 IN6_MASK_ADDR(dst, &mask);
1594
1595 if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1596 return (ENOBUFS);
1597 fill_sockaddr_inet6((struct sockaddr_in6 *)sa, dst, 0);
1598 info->rti_info[RTAX_DST] = sa;
1599
1600 if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128)) {
1601 if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1602 return (ENOBUFS);
1603 fill_sockaddr_inet6((struct sockaddr_in6 *)sa, &mask, 0);
1604 info->rti_info[RTAX_NETMASK] = sa;
1605 info->rti_flags &= ~RTF_HOST;
1606 } else
1607 remove_netmask(info);
1608
1609 /* Check gateway */
1610 if (info->rti_info[RTAX_GATEWAY] != NULL)
1611 return (cleanup_xaddrs_gateway(info, lb));
1612
1613 return (0);
1614 }
1615 #endif
1616
1617 static int
cleanup_xaddrs(struct rt_addrinfo * info,struct linear_buffer * lb)1618 cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb)
1619 {
1620 int error = EAFNOSUPPORT;
1621
1622 if (info->rti_info[RTAX_DST] == NULL) {
1623 RTS_PID_LOG(LOG_DEBUG, "prefix dst is not set");
1624 return (EINVAL);
1625 }
1626
1627 if (info->rti_flags & RTF_LLDATA) {
1628 /*
1629 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1630 * prefix along with the actual address in RTA_DST.
1631 * Remove netmask to avoid unnecessary address masking.
1632 */
1633 remove_netmask(info);
1634 }
1635
1636 switch (info->rti_info[RTAX_DST]->sa_family) {
1637 #ifdef INET
1638 case AF_INET:
1639 error = cleanup_xaddrs_inet(info, lb);
1640 break;
1641 #endif
1642 #ifdef INET6
1643 case AF_INET6:
1644 error = cleanup_xaddrs_inet6(info, lb);
1645 break;
1646 #endif
1647 }
1648
1649 return (error);
1650 }
1651
1652 /*
1653 * Fill in @dmask with valid netmask leaving original @smask
1654 * intact. Mostly used with radix netmasks.
1655 */
1656 struct sockaddr *
rtsock_fix_netmask(const struct sockaddr * dst,const struct sockaddr * smask,struct sockaddr_storage * dmask)1657 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1658 struct sockaddr_storage *dmask)
1659 {
1660 if (dst == NULL || smask == NULL)
1661 return (NULL);
1662
1663 memset(dmask, 0, dst->sa_len);
1664 memcpy(dmask, smask, smask->sa_len);
1665 dmask->ss_len = dst->sa_len;
1666 dmask->ss_family = dst->sa_family;
1667
1668 return ((struct sockaddr *)dmask);
1669 }
1670
1671 /*
1672 * Writes information related to @rtinfo object to newly-allocated mbuf.
1673 * Assumes MCLBYTES is enough to construct any message.
1674 * Used for OS notifications of vaious events (if/ifa announces,etc)
1675 *
1676 * Returns allocated mbuf or NULL on failure.
1677 */
1678 static struct mbuf *
rtsock_msg_mbuf(int type,struct rt_addrinfo * rtinfo)1679 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1680 {
1681 struct sockaddr_storage ss;
1682 struct rt_msghdr *rtm;
1683 struct mbuf *m;
1684 int i;
1685 struct sockaddr *sa;
1686 #ifdef INET6
1687 struct sockaddr_in6 *sin6;
1688 #endif
1689 int len, dlen;
1690
1691 switch (type) {
1692 case RTM_DELADDR:
1693 case RTM_NEWADDR:
1694 len = sizeof(struct ifa_msghdr);
1695 break;
1696
1697 case RTM_DELMADDR:
1698 case RTM_NEWMADDR:
1699 len = sizeof(struct ifma_msghdr);
1700 break;
1701
1702 case RTM_IFINFO:
1703 len = sizeof(struct if_msghdr);
1704 break;
1705
1706 case RTM_IFANNOUNCE:
1707 case RTM_IEEE80211:
1708 len = sizeof(struct if_announcemsghdr);
1709 break;
1710
1711 default:
1712 len = sizeof(struct rt_msghdr);
1713 }
1714
1715 /* XXXGL: can we use MJUMPAGESIZE cluster here? */
1716 KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1717 if (len > MHLEN)
1718 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1719 else
1720 m = m_gethdr(M_NOWAIT, MT_DATA);
1721 if (m == NULL)
1722 return (m);
1723
1724 m->m_pkthdr.len = m->m_len = len;
1725 rtm = mtod(m, struct rt_msghdr *);
1726 bzero((caddr_t)rtm, len);
1727 for (i = 0; i < RTAX_MAX; i++) {
1728 if ((sa = rtinfo->rti_info[i]) == NULL)
1729 continue;
1730 rtinfo->rti_addrs |= (1 << i);
1731
1732 dlen = SA_SIZE(sa);
1733 KASSERT(dlen <= sizeof(ss),
1734 ("%s: sockaddr size overflow", __func__));
1735 bzero(&ss, sizeof(ss));
1736 bcopy(sa, &ss, sa->sa_len);
1737 sa = (struct sockaddr *)&ss;
1738 #ifdef INET6
1739 if (sa->sa_family == AF_INET6) {
1740 sin6 = (struct sockaddr_in6 *)sa;
1741 (void)sa6_recoverscope(sin6);
1742 }
1743 #endif
1744 m_copyback(m, len, dlen, (caddr_t)sa);
1745 len += dlen;
1746 }
1747 if (m->m_pkthdr.len != len) {
1748 m_freem(m);
1749 return (NULL);
1750 }
1751 rtm->rtm_msglen = len;
1752 rtm->rtm_version = RTM_VERSION;
1753 rtm->rtm_type = type;
1754 return (m);
1755 }
1756
1757 /*
1758 * Writes information related to @rtinfo object to preallocated buffer.
1759 * Stores needed size in @plen. If @w is NULL, calculates size without
1760 * writing.
1761 * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1762 *
1763 * Returns 0 on success.
1764 *
1765 */
1766 static int
rtsock_msg_buffer(int type,struct rt_addrinfo * rtinfo,struct walkarg * w,int * plen)1767 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1768 {
1769 struct sockaddr_storage ss;
1770 int len, buflen = 0, dlen, i;
1771 caddr_t cp = NULL;
1772 struct rt_msghdr *rtm = NULL;
1773 #ifdef INET6
1774 struct sockaddr_in6 *sin6;
1775 #endif
1776 #ifdef COMPAT_FREEBSD32
1777 bool compat32 = false;
1778 #endif
1779
1780 switch (type) {
1781 case RTM_DELADDR:
1782 case RTM_NEWADDR:
1783 if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1784 #ifdef COMPAT_FREEBSD32
1785 if (w->w_req->flags & SCTL_MASK32) {
1786 len = sizeof(struct ifa_msghdrl32);
1787 compat32 = true;
1788 } else
1789 #endif
1790 len = sizeof(struct ifa_msghdrl);
1791 } else
1792 len = sizeof(struct ifa_msghdr);
1793 break;
1794
1795 case RTM_IFINFO:
1796 #ifdef COMPAT_FREEBSD32
1797 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1798 if (w->w_op == NET_RT_IFLISTL)
1799 len = sizeof(struct if_msghdrl32);
1800 else
1801 len = sizeof(struct if_msghdr32);
1802 compat32 = true;
1803 break;
1804 }
1805 #endif
1806 if (w != NULL && w->w_op == NET_RT_IFLISTL)
1807 len = sizeof(struct if_msghdrl);
1808 else
1809 len = sizeof(struct if_msghdr);
1810 break;
1811
1812 case RTM_NEWMADDR:
1813 len = sizeof(struct ifma_msghdr);
1814 break;
1815
1816 default:
1817 len = sizeof(struct rt_msghdr);
1818 }
1819
1820 if (w != NULL) {
1821 rtm = (struct rt_msghdr *)w->w_tmem;
1822 buflen = w->w_tmemsize - len;
1823 cp = (caddr_t)w->w_tmem + len;
1824 }
1825
1826 rtinfo->rti_addrs = 0;
1827 for (i = 0; i < RTAX_MAX; i++) {
1828 struct sockaddr *sa;
1829
1830 if ((sa = rtinfo->rti_info[i]) == NULL)
1831 continue;
1832 rtinfo->rti_addrs |= (1 << i);
1833 #ifdef COMPAT_FREEBSD32
1834 if (compat32)
1835 dlen = SA_SIZE32(sa);
1836 else
1837 #endif
1838 dlen = SA_SIZE(sa);
1839 if (cp != NULL && buflen >= dlen) {
1840 KASSERT(dlen <= sizeof(ss),
1841 ("%s: sockaddr size overflow", __func__));
1842 bzero(&ss, sizeof(ss));
1843 bcopy(sa, &ss, sa->sa_len);
1844 sa = (struct sockaddr *)&ss;
1845 #ifdef INET6
1846 if (sa->sa_family == AF_INET6) {
1847 sin6 = (struct sockaddr_in6 *)sa;
1848 (void)sa6_recoverscope(sin6);
1849 }
1850 #endif
1851 bcopy((caddr_t)sa, cp, (unsigned)dlen);
1852 cp += dlen;
1853 buflen -= dlen;
1854 } else if (cp != NULL) {
1855 /*
1856 * Buffer too small. Count needed size
1857 * and return with error.
1858 */
1859 cp = NULL;
1860 }
1861
1862 len += dlen;
1863 }
1864
1865 if (cp != NULL) {
1866 dlen = ALIGN(len) - len;
1867 if (buflen < dlen)
1868 cp = NULL;
1869 else {
1870 bzero(cp, dlen);
1871 cp += dlen;
1872 buflen -= dlen;
1873 }
1874 }
1875 len = ALIGN(len);
1876
1877 if (cp != NULL) {
1878 /* fill header iff buffer is large enough */
1879 rtm->rtm_version = RTM_VERSION;
1880 rtm->rtm_type = type;
1881 rtm->rtm_msglen = len;
1882 }
1883
1884 *plen = len;
1885
1886 if (w != NULL && cp == NULL)
1887 return (ENOBUFS);
1888
1889 return (0);
1890 }
1891
1892 /*
1893 * This routine is called to generate a message from the routing
1894 * socket indicating that a redirect has occurred, a routing lookup
1895 * has failed, or that a protocol has detected timeouts to a particular
1896 * destination.
1897 */
1898 void
rt_missmsg_fib(int type,struct rt_addrinfo * rtinfo,int flags,int error,int fibnum)1899 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1900 int fibnum)
1901 {
1902 struct rt_msghdr *rtm;
1903 struct mbuf *m;
1904 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1905
1906 if (V_route_cb.any_count == 0)
1907 return;
1908 m = rtsock_msg_mbuf(type, rtinfo);
1909 if (m == NULL)
1910 return;
1911
1912 if (fibnum != RT_ALL_FIBS) {
1913 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1914 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1915 M_SETFIB(m, fibnum);
1916 m->m_flags |= RTS_FILTER_FIB;
1917 }
1918
1919 rtm = mtod(m, struct rt_msghdr *);
1920 rtm->rtm_flags = RTF_DONE | flags;
1921 rtm->rtm_errno = error;
1922 rtm->rtm_addrs = rtinfo->rti_addrs;
1923 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1924 }
1925
1926 void
rt_missmsg(int type,struct rt_addrinfo * rtinfo,int flags,int error)1927 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1928 {
1929
1930 rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1931 }
1932
1933 /*
1934 * This routine is called to generate a message from the routing
1935 * socket indicating that the status of a network interface has changed.
1936 */
1937 static void
rtsock_ifmsg(struct ifnet * ifp,int if_flags_mask __unused)1938 rtsock_ifmsg(struct ifnet *ifp, int if_flags_mask __unused)
1939 {
1940 struct if_msghdr *ifm;
1941 struct mbuf *m;
1942 struct rt_addrinfo info;
1943
1944 if (V_route_cb.any_count == 0)
1945 return;
1946 bzero((caddr_t)&info, sizeof(info));
1947 m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1948 if (m == NULL)
1949 return;
1950 ifm = mtod(m, struct if_msghdr *);
1951 ifm->ifm_index = ifp->if_index;
1952 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1953 if_data_copy(ifp, &ifm->ifm_data);
1954 ifm->ifm_addrs = 0;
1955 rt_dispatch(m, AF_UNSPEC);
1956 }
1957
1958 /*
1959 * Announce interface address arrival/withdraw.
1960 * Please do not call directly, use rt_addrmsg().
1961 * Assume input data to be valid.
1962 * Returns 0 on success.
1963 */
1964 int
rtsock_addrmsg(int cmd,struct ifaddr * ifa,int fibnum)1965 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1966 {
1967 struct rt_addrinfo info;
1968 struct sockaddr *sa;
1969 int ncmd;
1970 struct mbuf *m;
1971 struct ifa_msghdr *ifam;
1972 struct ifnet *ifp = ifa->ifa_ifp;
1973 struct sockaddr_storage ss;
1974
1975 if (V_route_cb.any_count == 0)
1976 return (0);
1977
1978 ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1979
1980 bzero((caddr_t)&info, sizeof(info));
1981 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1982 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1983 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1984 info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1985 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1986 if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1987 return (ENOBUFS);
1988 ifam = mtod(m, struct ifa_msghdr *);
1989 ifam->ifam_index = ifp->if_index;
1990 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1991 ifam->ifam_flags = ifa->ifa_flags;
1992 ifam->ifam_addrs = info.rti_addrs;
1993
1994 if (fibnum != RT_ALL_FIBS) {
1995 M_SETFIB(m, fibnum);
1996 m->m_flags |= RTS_FILTER_FIB;
1997 }
1998
1999 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2000
2001 return (0);
2002 }
2003
2004 /*
2005 * Announce route addition/removal to rtsock based on @rt data.
2006 * Callers are advives to use rt_routemsg() instead of using this
2007 * function directly.
2008 * Assume @rt data is consistent.
2009 *
2010 * Returns 0 on success.
2011 */
2012 int
rtsock_routemsg(int cmd,struct rtentry * rt,struct nhop_object * nh,int fibnum)2013 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
2014 int fibnum)
2015 {
2016 union sockaddr_union dst, mask;
2017 struct rt_addrinfo info;
2018
2019 if (V_route_cb.any_count == 0)
2020 return (0);
2021
2022 int family = rt_get_family(rt);
2023 init_sockaddrs_family(family, &dst.sa, &mask.sa);
2024 export_rtaddrs(rt, &dst.sa, &mask.sa);
2025
2026 bzero((caddr_t)&info, sizeof(info));
2027 info.rti_info[RTAX_DST] = &dst.sa;
2028 info.rti_info[RTAX_NETMASK] = &mask.sa;
2029 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2030 info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
2031 info.rti_ifp = nh->nh_ifp;
2032
2033 return (rtsock_routemsg_info(cmd, &info, fibnum));
2034 }
2035
2036 int
rtsock_routemsg_info(int cmd,struct rt_addrinfo * info,int fibnum)2037 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
2038 {
2039 struct rt_msghdr *rtm;
2040 struct sockaddr *sa;
2041 struct mbuf *m;
2042
2043 if (V_route_cb.any_count == 0)
2044 return (0);
2045
2046 if (info->rti_flags & RTF_HOST)
2047 info->rti_info[RTAX_NETMASK] = NULL;
2048
2049 m = rtsock_msg_mbuf(cmd, info);
2050 if (m == NULL)
2051 return (ENOBUFS);
2052
2053 if (fibnum != RT_ALL_FIBS) {
2054 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
2055 "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
2056 M_SETFIB(m, fibnum);
2057 m->m_flags |= RTS_FILTER_FIB;
2058 }
2059
2060 rtm = mtod(m, struct rt_msghdr *);
2061 rtm->rtm_addrs = info->rti_addrs;
2062 if (info->rti_ifp != NULL)
2063 rtm->rtm_index = info->rti_ifp->if_index;
2064 /* Add RTF_DONE to indicate command 'completion' required by API */
2065 info->rti_flags |= RTF_DONE;
2066 /* Reported routes has to be up */
2067 if (cmd == RTM_ADD || cmd == RTM_CHANGE)
2068 info->rti_flags |= RTF_UP;
2069 rtm->rtm_flags = info->rti_flags;
2070
2071 sa = info->rti_info[RTAX_DST];
2072 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2073
2074 return (0);
2075 }
2076
2077 /*
2078 * This is the analogue to the rt_newaddrmsg which performs the same
2079 * function but for multicast group memberhips. This is easier since
2080 * there is no route state to worry about.
2081 */
2082 void
rt_newmaddrmsg(int cmd,struct ifmultiaddr * ifma)2083 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
2084 {
2085 struct rt_addrinfo info;
2086 struct mbuf *m = NULL;
2087 struct ifnet *ifp = ifma->ifma_ifp;
2088 struct ifma_msghdr *ifmam;
2089
2090 if (V_route_cb.any_count == 0)
2091 return;
2092
2093 bzero((caddr_t)&info, sizeof(info));
2094 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2095 if (ifp && ifp->if_addr)
2096 info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2097 else
2098 info.rti_info[RTAX_IFP] = NULL;
2099 /*
2100 * If a link-layer address is present, present it as a ``gateway''
2101 * (similarly to how ARP entries, e.g., are presented).
2102 */
2103 info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
2104 m = rtsock_msg_mbuf(cmd, &info);
2105 if (m == NULL)
2106 return;
2107 ifmam = mtod(m, struct ifma_msghdr *);
2108 KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
2109 __func__));
2110 ifmam->ifmam_index = ifp->if_index;
2111 ifmam->ifmam_addrs = info.rti_addrs;
2112 rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
2113 }
2114
2115 static struct mbuf *
rt_makeifannouncemsg(struct ifnet * ifp,int type,int what,struct rt_addrinfo * info)2116 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
2117 struct rt_addrinfo *info)
2118 {
2119 struct if_announcemsghdr *ifan;
2120 struct mbuf *m;
2121
2122 if (V_route_cb.any_count == 0)
2123 return NULL;
2124 bzero((caddr_t)info, sizeof(*info));
2125 m = rtsock_msg_mbuf(type, info);
2126 if (m != NULL) {
2127 ifan = mtod(m, struct if_announcemsghdr *);
2128 ifan->ifan_index = ifp->if_index;
2129 strlcpy(ifan->ifan_name, ifp->if_xname,
2130 sizeof(ifan->ifan_name));
2131 ifan->ifan_what = what;
2132 }
2133 return m;
2134 }
2135
2136 /*
2137 * This is called to generate routing socket messages indicating
2138 * IEEE80211 wireless events.
2139 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2140 */
2141 void
rt_ieee80211msg(struct ifnet * ifp,int what,void * data,size_t data_len)2142 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2143 {
2144 struct mbuf *m;
2145 struct rt_addrinfo info;
2146
2147 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2148 if (m != NULL) {
2149 /*
2150 * Append the ieee80211 data. Try to stick it in the
2151 * mbuf containing the ifannounce msg; otherwise allocate
2152 * a new mbuf and append.
2153 *
2154 * NB: we assume m is a single mbuf.
2155 */
2156 if (data_len > M_TRAILINGSPACE(m)) {
2157 struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2158 if (n == NULL) {
2159 m_freem(m);
2160 return;
2161 }
2162 bcopy(data, mtod(n, void *), data_len);
2163 n->m_len = data_len;
2164 m->m_next = n;
2165 } else if (data_len > 0) {
2166 bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2167 m->m_len += data_len;
2168 }
2169 if (m->m_flags & M_PKTHDR)
2170 m->m_pkthdr.len += data_len;
2171 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2172 rt_dispatch(m, AF_UNSPEC);
2173 }
2174 }
2175
2176 /*
2177 * This is called to generate routing socket messages indicating
2178 * network interface arrival and departure.
2179 */
2180 static void
rt_ifannouncemsg(struct ifnet * ifp,int what)2181 rt_ifannouncemsg(struct ifnet *ifp, int what)
2182 {
2183 struct mbuf *m;
2184 struct rt_addrinfo info;
2185
2186 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2187 if (m != NULL)
2188 rt_dispatch(m, AF_UNSPEC);
2189 }
2190
2191 static void
rt_dispatch(struct mbuf * m,sa_family_t saf)2192 rt_dispatch(struct mbuf *m, sa_family_t saf)
2193 {
2194
2195 M_ASSERTPKTHDR(m);
2196
2197 m->m_rtsock_family = saf;
2198 if (V_loif)
2199 m->m_pkthdr.rcvif = V_loif;
2200 else {
2201 m_freem(m);
2202 return;
2203 }
2204 netisr_queue(NETISR_ROUTE, m); /* mbuf is free'd on failure. */
2205 }
2206
2207 /*
2208 * This is used in dumping the kernel table via sysctl().
2209 */
2210 static int
sysctl_dumpentry(struct rtentry * rt,void * vw)2211 sysctl_dumpentry(struct rtentry *rt, void *vw)
2212 {
2213 struct walkarg *w = vw;
2214 struct nhop_object *nh;
2215
2216 NET_EPOCH_ASSERT();
2217
2218 if (!rt_is_exportable(rt, w->w_req->td->td_ucred))
2219 return (0);
2220
2221 export_rtaddrs(rt, w->dst, w->mask);
2222 nh = rt_get_raw_nhop(rt);
2223 #ifdef ROUTE_MPATH
2224 if (NH_IS_NHGRP(nh)) {
2225 const struct weightened_nhop *wn;
2226 uint32_t num_nhops;
2227 int error;
2228 wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2229 for (int i = 0; i < num_nhops; i++) {
2230 error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2231 if (error != 0)
2232 return (error);
2233 }
2234 } else
2235 #endif
2236 sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2237
2238 return (0);
2239 }
2240
2241
2242 static int
sysctl_dumpnhop(struct rtentry * rt,struct nhop_object * nh,uint32_t weight,struct walkarg * w)2243 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2244 struct walkarg *w)
2245 {
2246 struct rt_addrinfo info;
2247 int error = 0, size;
2248 uint32_t rtflags;
2249
2250 rtflags = nhop_get_rtflags(nh);
2251
2252 if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2253 return (0);
2254
2255 bzero((caddr_t)&info, sizeof(info));
2256 info.rti_info[RTAX_DST] = w->dst;
2257 info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2258 info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2259 info.rti_info[RTAX_GENMASK] = 0;
2260 if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2261 info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2262 info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2263 if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2264 info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2265 }
2266 if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2267 return (error);
2268 if (w->w_req && w->w_tmem) {
2269 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2270
2271 bzero(&rtm->rtm_index,
2272 sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2273
2274 /*
2275 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2276 * and RTF_UP (if entry is linked, which is always true here).
2277 * Given that, use nhop rtflags & add RTF_UP.
2278 */
2279 rtm->rtm_flags = rtflags | RTF_UP;
2280 if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2281 rtm->rtm_flags = RTF_GATEWAY |
2282 (rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2283 rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2284 rtm->rtm_rmx.rmx_weight = weight;
2285 rtm->rtm_index = nh->nh_ifp->if_index;
2286 rtm->rtm_addrs = info.rti_addrs;
2287 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2288 return (error);
2289 }
2290 return (error);
2291 }
2292
2293 static int
sysctl_iflist_ifml(struct ifnet * ifp,const struct if_data * src_ifd,struct rt_addrinfo * info,struct walkarg * w,int len)2294 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2295 struct rt_addrinfo *info, struct walkarg *w, int len)
2296 {
2297 struct if_msghdrl *ifm;
2298 struct if_data *ifd;
2299
2300 ifm = (struct if_msghdrl *)w->w_tmem;
2301
2302 #ifdef COMPAT_FREEBSD32
2303 if (w->w_req->flags & SCTL_MASK32) {
2304 struct if_msghdrl32 *ifm32;
2305
2306 ifm32 = (struct if_msghdrl32 *)ifm;
2307 ifm32->ifm_addrs = info->rti_addrs;
2308 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2309 ifm32->ifm_index = ifp->if_index;
2310 ifm32->_ifm_spare1 = 0;
2311 ifm32->ifm_len = sizeof(*ifm32);
2312 ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2313 ifm32->_ifm_spare2 = 0;
2314 ifd = &ifm32->ifm_data;
2315 } else
2316 #endif
2317 {
2318 ifm->ifm_addrs = info->rti_addrs;
2319 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2320 ifm->ifm_index = ifp->if_index;
2321 ifm->_ifm_spare1 = 0;
2322 ifm->ifm_len = sizeof(*ifm);
2323 ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2324 ifm->_ifm_spare2 = 0;
2325 ifd = &ifm->ifm_data;
2326 }
2327
2328 memcpy(ifd, src_ifd, sizeof(*ifd));
2329
2330 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2331 }
2332
2333 static int
sysctl_iflist_ifm(struct ifnet * ifp,const struct if_data * src_ifd,struct rt_addrinfo * info,struct walkarg * w,int len)2334 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2335 struct rt_addrinfo *info, struct walkarg *w, int len)
2336 {
2337 struct if_msghdr *ifm;
2338 struct if_data *ifd;
2339
2340 ifm = (struct if_msghdr *)w->w_tmem;
2341
2342 #ifdef COMPAT_FREEBSD32
2343 if (w->w_req->flags & SCTL_MASK32) {
2344 struct if_msghdr32 *ifm32;
2345
2346 ifm32 = (struct if_msghdr32 *)ifm;
2347 ifm32->ifm_addrs = info->rti_addrs;
2348 ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2349 ifm32->ifm_index = ifp->if_index;
2350 ifm32->_ifm_spare1 = 0;
2351 ifd = &ifm32->ifm_data;
2352 } else
2353 #endif
2354 {
2355 ifm->ifm_addrs = info->rti_addrs;
2356 ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2357 ifm->ifm_index = ifp->if_index;
2358 ifm->_ifm_spare1 = 0;
2359 ifd = &ifm->ifm_data;
2360 }
2361
2362 memcpy(ifd, src_ifd, sizeof(*ifd));
2363
2364 return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2365 }
2366
2367 static int
sysctl_iflist_ifaml(struct ifaddr * ifa,struct rt_addrinfo * info,struct walkarg * w,int len)2368 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2369 struct walkarg *w, int len)
2370 {
2371 struct ifa_msghdrl *ifam;
2372 struct if_data *ifd;
2373
2374 ifam = (struct ifa_msghdrl *)w->w_tmem;
2375
2376 #ifdef COMPAT_FREEBSD32
2377 if (w->w_req->flags & SCTL_MASK32) {
2378 struct ifa_msghdrl32 *ifam32;
2379
2380 ifam32 = (struct ifa_msghdrl32 *)ifam;
2381 ifam32->ifam_addrs = info->rti_addrs;
2382 ifam32->ifam_flags = ifa->ifa_flags;
2383 ifam32->ifam_index = ifa->ifa_ifp->if_index;
2384 ifam32->_ifam_spare1 = 0;
2385 ifam32->ifam_len = sizeof(*ifam32);
2386 ifam32->ifam_data_off =
2387 offsetof(struct ifa_msghdrl32, ifam_data);
2388 ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2389 ifd = &ifam32->ifam_data;
2390 } else
2391 #endif
2392 {
2393 ifam->ifam_addrs = info->rti_addrs;
2394 ifam->ifam_flags = ifa->ifa_flags;
2395 ifam->ifam_index = ifa->ifa_ifp->if_index;
2396 ifam->_ifam_spare1 = 0;
2397 ifam->ifam_len = sizeof(*ifam);
2398 ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2399 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2400 ifd = &ifam->ifam_data;
2401 }
2402
2403 bzero(ifd, sizeof(*ifd));
2404 ifd->ifi_datalen = sizeof(struct if_data);
2405 ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2406 ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2407 ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2408 ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2409
2410 /* Fixup if_data carp(4) vhid. */
2411 if (carp_get_vhid_p != NULL)
2412 ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2413
2414 return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2415 }
2416
2417 static int
sysctl_iflist_ifam(struct ifaddr * ifa,struct rt_addrinfo * info,struct walkarg * w,int len)2418 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2419 struct walkarg *w, int len)
2420 {
2421 struct ifa_msghdr *ifam;
2422
2423 ifam = (struct ifa_msghdr *)w->w_tmem;
2424 ifam->ifam_addrs = info->rti_addrs;
2425 ifam->ifam_flags = ifa->ifa_flags;
2426 ifam->ifam_index = ifa->ifa_ifp->if_index;
2427 ifam->_ifam_spare1 = 0;
2428 ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2429
2430 return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2431 }
2432
2433 static int
sysctl_iflist(int af,struct walkarg * w)2434 sysctl_iflist(int af, struct walkarg *w)
2435 {
2436 struct ifnet *ifp;
2437 struct ifaddr *ifa;
2438 struct if_data ifd;
2439 struct rt_addrinfo info;
2440 int len, error = 0;
2441 struct sockaddr_storage ss;
2442
2443 bzero((caddr_t)&info, sizeof(info));
2444 bzero(&ifd, sizeof(ifd));
2445 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2446 if (w->w_arg && w->w_arg != ifp->if_index)
2447 continue;
2448 if_data_copy(ifp, &ifd);
2449 ifa = ifp->if_addr;
2450 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2451 error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2452 if (error != 0)
2453 goto done;
2454 info.rti_info[RTAX_IFP] = NULL;
2455 if (w->w_req && w->w_tmem) {
2456 if (w->w_op == NET_RT_IFLISTL)
2457 error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2458 len);
2459 else
2460 error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2461 len);
2462 if (error)
2463 goto done;
2464 }
2465 while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2466 if (af && af != ifa->ifa_addr->sa_family)
2467 continue;
2468 if (prison_if(w->w_req->td->td_ucred,
2469 ifa->ifa_addr) != 0)
2470 continue;
2471 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2472 info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2473 ifa->ifa_addr, ifa->ifa_netmask, &ss);
2474 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2475 error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2476 if (error != 0)
2477 goto done;
2478 if (w->w_req && w->w_tmem) {
2479 if (w->w_op == NET_RT_IFLISTL)
2480 error = sysctl_iflist_ifaml(ifa, &info,
2481 w, len);
2482 else
2483 error = sysctl_iflist_ifam(ifa, &info,
2484 w, len);
2485 if (error)
2486 goto done;
2487 }
2488 }
2489 info.rti_info[RTAX_IFA] = NULL;
2490 info.rti_info[RTAX_NETMASK] = NULL;
2491 info.rti_info[RTAX_BRD] = NULL;
2492 }
2493 done:
2494 return (error);
2495 }
2496
2497 static int
sysctl_ifmalist(int af,struct walkarg * w)2498 sysctl_ifmalist(int af, struct walkarg *w)
2499 {
2500 struct rt_addrinfo info;
2501 struct ifaddr *ifa;
2502 struct ifmultiaddr *ifma;
2503 struct ifnet *ifp;
2504 int error, len;
2505
2506 NET_EPOCH_ASSERT();
2507
2508 error = 0;
2509 bzero((caddr_t)&info, sizeof(info));
2510
2511 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2512 if (w->w_arg && w->w_arg != ifp->if_index)
2513 continue;
2514 ifa = ifp->if_addr;
2515 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2516 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2517 if (af && af != ifma->ifma_addr->sa_family)
2518 continue;
2519 if (prison_if(w->w_req->td->td_ucred,
2520 ifma->ifma_addr) != 0)
2521 continue;
2522 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2523 info.rti_info[RTAX_GATEWAY] =
2524 (ifma->ifma_addr->sa_family != AF_LINK) ?
2525 ifma->ifma_lladdr : NULL;
2526 error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2527 if (error != 0)
2528 break;
2529 if (w->w_req && w->w_tmem) {
2530 struct ifma_msghdr *ifmam;
2531
2532 ifmam = (struct ifma_msghdr *)w->w_tmem;
2533 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2534 ifmam->ifmam_flags = 0;
2535 ifmam->ifmam_addrs = info.rti_addrs;
2536 ifmam->_ifmam_spare1 = 0;
2537 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2538 if (error != 0)
2539 break;
2540 }
2541 }
2542 if (error != 0)
2543 break;
2544 }
2545 return (error);
2546 }
2547
2548 static void
rtable_sysctl_dump(uint32_t fibnum,int family,struct walkarg * w)2549 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2550 {
2551 union sockaddr_union sa_dst, sa_mask;
2552
2553 w->family = family;
2554 w->dst = (struct sockaddr *)&sa_dst;
2555 w->mask = (struct sockaddr *)&sa_mask;
2556
2557 init_sockaddrs_family(family, w->dst, w->mask);
2558
2559 rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2560 }
2561
2562 static int
sysctl_rtsock(SYSCTL_HANDLER_ARGS)2563 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2564 {
2565 struct epoch_tracker et;
2566 int *name = (int *)arg1;
2567 u_int namelen = arg2;
2568 struct rib_head *rnh = NULL; /* silence compiler. */
2569 int i, lim, error = EINVAL;
2570 int fib = 0;
2571 u_char af;
2572 struct walkarg w;
2573
2574 if (namelen < 3)
2575 return (EINVAL);
2576
2577 name++;
2578 namelen--;
2579 if (req->newptr)
2580 return (EPERM);
2581 if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2582 if (namelen == 3)
2583 fib = req->td->td_proc->p_fibnum;
2584 else if (namelen == 4)
2585 fib = (name[3] == RT_ALL_FIBS) ?
2586 req->td->td_proc->p_fibnum : name[3];
2587 else
2588 return ((namelen < 3) ? EISDIR : ENOTDIR);
2589 if (fib < 0 || fib >= rt_numfibs)
2590 return (EINVAL);
2591 } else if (namelen != 3)
2592 return ((namelen < 3) ? EISDIR : ENOTDIR);
2593 af = name[0];
2594 if (af > AF_MAX)
2595 return (EINVAL);
2596 bzero(&w, sizeof(w));
2597 w.w_op = name[1];
2598 w.w_arg = name[2];
2599 w.w_req = req;
2600
2601 error = sysctl_wire_old_buffer(req, 0);
2602 if (error)
2603 return (error);
2604
2605 /*
2606 * Allocate reply buffer in advance.
2607 * All rtsock messages has maximum length of u_short.
2608 */
2609 w.w_tmemsize = 65536;
2610 w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2611
2612 NET_EPOCH_ENTER(et);
2613 switch (w.w_op) {
2614 case NET_RT_DUMP:
2615 case NET_RT_FLAGS:
2616 if (af == 0) { /* dump all tables */
2617 i = 1;
2618 lim = AF_MAX;
2619 } else /* dump only one table */
2620 i = lim = af;
2621
2622 /*
2623 * take care of llinfo entries, the caller must
2624 * specify an AF
2625 */
2626 if (w.w_op == NET_RT_FLAGS &&
2627 (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2628 if (af != 0)
2629 error = lltable_sysctl_dumparp(af, w.w_req);
2630 else
2631 error = EINVAL;
2632 break;
2633 }
2634 /*
2635 * take care of routing entries
2636 */
2637 for (error = 0; error == 0 && i <= lim; i++) {
2638 rnh = rt_tables_get_rnh(fib, i);
2639 if (rnh != NULL) {
2640 rtable_sysctl_dump(fib, i, &w);
2641 } else if (af != 0)
2642 error = EAFNOSUPPORT;
2643 }
2644 break;
2645 case NET_RT_NHOP:
2646 case NET_RT_NHGRP:
2647 /* Allow dumping one specific af/fib at a time */
2648 if (namelen < 4) {
2649 error = EINVAL;
2650 break;
2651 }
2652 fib = name[3];
2653 if (fib < 0 || fib > rt_numfibs) {
2654 error = EINVAL;
2655 break;
2656 }
2657 rnh = rt_tables_get_rnh(fib, af);
2658 if (rnh == NULL) {
2659 error = EAFNOSUPPORT;
2660 break;
2661 }
2662 if (w.w_op == NET_RT_NHOP)
2663 error = nhops_dump_sysctl(rnh, w.w_req);
2664 else
2665 #ifdef ROUTE_MPATH
2666 error = nhgrp_dump_sysctl(rnh, w.w_req);
2667 #else
2668 error = ENOTSUP;
2669 #endif
2670 break;
2671 case NET_RT_IFLIST:
2672 case NET_RT_IFLISTL:
2673 error = sysctl_iflist(af, &w);
2674 break;
2675
2676 case NET_RT_IFMALIST:
2677 error = sysctl_ifmalist(af, &w);
2678 break;
2679 }
2680 NET_EPOCH_EXIT(et);
2681
2682 free(w.w_tmem, M_TEMP);
2683 return (error);
2684 }
2685
2686 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2687 sysctl_rtsock, "Return route tables and interface/address lists");
2688
2689 /*
2690 * Definitions of protocols supported in the ROUTE domain.
2691 */
2692
2693 static struct domain routedomain; /* or at least forward */
2694
2695 static struct protosw routesw = {
2696 .pr_type = SOCK_RAW,
2697 .pr_flags = PR_ATOMIC|PR_ADDR,
2698 .pr_abort = rts_close,
2699 .pr_attach = rts_attach,
2700 .pr_detach = rts_detach,
2701 .pr_send = rts_send,
2702 .pr_shutdown = rts_shutdown,
2703 .pr_disconnect = rts_disconnect,
2704 .pr_close = rts_close,
2705 };
2706
2707 static struct domain routedomain = {
2708 .dom_family = PF_ROUTE,
2709 .dom_name = "route",
2710 .dom_nprotosw = 1,
2711 .dom_protosw = { &routesw },
2712 };
2713
2714 DOMAIN_SET(route);
2715