rtsock.c revision 1.217 1 /* $NetBSD: rtsock.c,v 1.217 2017/06/22 09:58:04 ozaki-r Exp $ */
2
3 /*
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * 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 project 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 PROJECT 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 PROJECT 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
32 /*
33 * Copyright (c) 1988, 1991, 1993
34 * The Regents of the University of California. All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 3. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)rtsock.c 8.7 (Berkeley) 10/12/95
61 */
62
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: rtsock.c,v 1.217 2017/06/22 09:58:04 ozaki-r Exp $");
65
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_mpls.h"
69 #include "opt_compat_netbsd.h"
70 #include "opt_sctp.h"
71 #include "opt_net_mpsafe.h"
72 #endif
73
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/proc.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/domain.h>
80 #include <sys/protosw.h>
81 #include <sys/sysctl.h>
82 #include <sys/kauth.h>
83 #include <sys/kmem.h>
84 #include <sys/intr.h>
85
86 #include <net/if.h>
87 #include <net/if_llatbl.h>
88 #include <net/if_types.h>
89 #include <net/route.h>
90 #include <net/raw_cb.h>
91
92 #include <netinet/in_var.h>
93 #include <netinet/if_inarp.h>
94
95 #include <netmpls/mpls.h>
96
97 #ifdef SCTP
98 extern void sctp_add_ip_address(struct ifaddr *);
99 extern void sctp_delete_ip_address(struct ifaddr *);
100 #endif
101
102 #if defined(COMPAT_14) || defined(COMPAT_50) || defined(COMPAT_70)
103 #include <compat/net/if.h>
104 #include <compat/net/route.h>
105 #endif
106 #ifdef COMPAT_RTSOCK
107 #define RTM_XVERSION RTM_OVERSION
108 #define RTM_XNEWADDR RTM_ONEWADDR
109 #define RTM_XDELADDR RTM_ODELADDR
110 #define RTM_XCHGADDR RTM_OCHGADDR
111 #define RT_XADVANCE(a,b) RT_OADVANCE(a,b)
112 #define RT_XROUNDUP(n) RT_OROUNDUP(n)
113 #define PF_XROUTE PF_OROUTE
114 #define rt_xmsghdr rt_msghdr50
115 #define if_xmsghdr if_msghdr /* if_msghdr50 is for RTM_OIFINFO */
116 #define ifa_xmsghdr ifa_msghdr50
117 #define if_xannouncemsghdr if_announcemsghdr50
118 #define COMPATNAME(x) compat_50_ ## x
119 #define DOMAINNAME "oroute"
120 CTASSERT(sizeof(struct ifa_xmsghdr) == 20);
121 DOMAIN_DEFINE(compat_50_routedomain); /* forward declare and add to link set */
122 #undef COMPAT_70
123 #else /* COMPAT_RTSOCK */
124 #define RTM_XVERSION RTM_VERSION
125 #define RTM_XNEWADDR RTM_NEWADDR
126 #define RTM_XDELADDR RTM_DELADDR
127 #define RTM_XCHGADDR RTM_CHGADDR
128 #define RT_XADVANCE(a,b) RT_ADVANCE(a,b)
129 #define RT_XROUNDUP(n) RT_ROUNDUP(n)
130 #define PF_XROUTE PF_ROUTE
131 #define rt_xmsghdr rt_msghdr
132 #define if_xmsghdr if_msghdr
133 #define ifa_xmsghdr ifa_msghdr
134 #define if_xannouncemsghdr if_announcemsghdr
135 #define COMPATNAME(x) x
136 #define DOMAINNAME "route"
137 CTASSERT(sizeof(struct ifa_xmsghdr) == 32);
138 #ifdef COMPAT_50
139 #define COMPATCALL(name, args) compat_50_ ## name args
140 #endif
141 DOMAIN_DEFINE(routedomain); /* forward declare and add to link set */
142 #undef COMPAT_50
143 #undef COMPAT_14
144 #endif /* COMPAT_RTSOCK */
145
146 #ifndef COMPATCALL
147 #define COMPATCALL(name, args) do { } while (/*CONSTCOND*/ 0)
148 #endif
149
150 #ifdef RTSOCK_DEBUG
151 #define RT_IN_PRINT(info, b, a) (in_print((b), sizeof(b), \
152 &((const struct sockaddr_in *)(info)->rti_info[(a)])->sin_addr), (b))
153 #endif /* RTSOCK_DEBUG */
154
155 struct route_info COMPATNAME(route_info) = {
156 .ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, },
157 .ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, },
158 .ri_maxqlen = IFQ_MAXLEN,
159 };
160
161 #define PRESERVED_RTF (RTF_UP | RTF_GATEWAY | RTF_HOST | RTF_DONE | RTF_MASK)
162
163 static void COMPATNAME(route_init)(void);
164 static int COMPATNAME(route_output)(struct mbuf *, struct socket *);
165
166 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
167 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
168 struct rt_addrinfo *);
169 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
170 static void rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *);
171 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *);
172 static void sysctl_net_route_setup(struct sysctllog **);
173 static int sysctl_dumpentry(struct rtentry *, void *);
174 static int sysctl_iflist(int, struct rt_walkarg *, int);
175 static int sysctl_rtable(SYSCTLFN_PROTO);
176 static void rt_adjustcount(int, int);
177
178 static const struct protosw COMPATNAME(route_protosw)[];
179
180 struct routecb {
181 struct rawcb rocb_rcb;
182 unsigned int rocb_msgfilter;
183 #define RTMSGFILTER(m) (1U << (m))
184 };
185 #define sotoroutecb(so) ((struct routecb *)(so)->so_pcb)
186
187 static void
188 rt_adjustcount(int af, int cnt)
189 {
190 struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
191
192 cb->any_count += cnt;
193
194 switch (af) {
195 case AF_INET:
196 cb->ip_count += cnt;
197 return;
198 #ifdef INET6
199 case AF_INET6:
200 cb->ip6_count += cnt;
201 return;
202 #endif
203 case AF_MPLS:
204 cb->mpls_count += cnt;
205 return;
206 }
207 }
208
209 static int
210 COMPATNAME(route_filter)(struct mbuf *m, struct sockproto *proto,
211 struct rawcb *rp)
212 {
213 struct routecb *rop = (struct routecb *)rp;
214 struct rt_xmsghdr *rtm;
215
216 KASSERT(m != NULL);
217 KASSERT(proto != NULL);
218 KASSERT(rp != NULL);
219
220 /* Wrong family for this socket. */
221 if (proto->sp_family != PF_ROUTE)
222 return ENOPROTOOPT;
223
224 /* If no filter set, just return. */
225 if (rop->rocb_msgfilter == 0)
226 return 0;
227
228 /* Ensure we can access rtm_type */
229 if (m->m_len <
230 offsetof(struct rt_xmsghdr, rtm_type) + sizeof(rtm->rtm_type))
231 return EINVAL;
232
233 rtm = mtod(m, struct rt_xmsghdr *);
234 /* If the rtm type is filtered out, return a positive. */
235 if (!(rop->rocb_msgfilter & RTMSGFILTER(rtm->rtm_type)))
236 return EEXIST;
237
238 /* Passed the filter. */
239 return 0;
240 }
241
242 static int
243 COMPATNAME(route_attach)(struct socket *so, int proto)
244 {
245 struct rawcb *rp;
246 struct routecb *rop;
247 int s, error;
248
249 KASSERT(sotorawcb(so) == NULL);
250 rop = kmem_zalloc(sizeof(*rop), KM_SLEEP);
251 rp = &rop->rocb_rcb;
252 rp->rcb_len = sizeof(*rop);
253 so->so_pcb = rp;
254
255 s = splsoftnet();
256 if ((error = raw_attach(so, proto)) == 0) {
257 rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
258 rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
259 rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
260 rp->rcb_filter = COMPATNAME(route_filter);
261 }
262 splx(s);
263
264 if (error) {
265 kmem_free(rop, sizeof(*rop));
266 so->so_pcb = NULL;
267 return error;
268 }
269
270 soisconnected(so);
271 so->so_options |= SO_USELOOPBACK;
272 KASSERT(solocked(so));
273
274 return error;
275 }
276
277 static void
278 COMPATNAME(route_detach)(struct socket *so)
279 {
280 struct rawcb *rp = sotorawcb(so);
281 int s;
282
283 KASSERT(rp != NULL);
284 KASSERT(solocked(so));
285
286 s = splsoftnet();
287 rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
288 raw_detach(so);
289 splx(s);
290 }
291
292 static int
293 COMPATNAME(route_accept)(struct socket *so, struct sockaddr *nam)
294 {
295 KASSERT(solocked(so));
296
297 panic("route_accept");
298
299 return EOPNOTSUPP;
300 }
301
302 static int
303 COMPATNAME(route_bind)(struct socket *so, struct sockaddr *nam, struct lwp *l)
304 {
305 KASSERT(solocked(so));
306
307 return EOPNOTSUPP;
308 }
309
310 static int
311 COMPATNAME(route_listen)(struct socket *so, struct lwp *l)
312 {
313 KASSERT(solocked(so));
314
315 return EOPNOTSUPP;
316 }
317
318 static int
319 COMPATNAME(route_connect)(struct socket *so, struct sockaddr *nam, struct lwp *l)
320 {
321 KASSERT(solocked(so));
322
323 return EOPNOTSUPP;
324 }
325
326 static int
327 COMPATNAME(route_connect2)(struct socket *so, struct socket *so2)
328 {
329 KASSERT(solocked(so));
330
331 return EOPNOTSUPP;
332 }
333
334 static int
335 COMPATNAME(route_disconnect)(struct socket *so)
336 {
337 struct rawcb *rp = sotorawcb(so);
338 int s;
339
340 KASSERT(solocked(so));
341 KASSERT(rp != NULL);
342
343 s = splsoftnet();
344 soisdisconnected(so);
345 raw_disconnect(rp);
346 splx(s);
347
348 return 0;
349 }
350
351 static int
352 COMPATNAME(route_shutdown)(struct socket *so)
353 {
354 int s;
355
356 KASSERT(solocked(so));
357
358 /*
359 * Mark the connection as being incapable of further input.
360 */
361 s = splsoftnet();
362 socantsendmore(so);
363 splx(s);
364 return 0;
365 }
366
367 static int
368 COMPATNAME(route_abort)(struct socket *so)
369 {
370 KASSERT(solocked(so));
371
372 panic("route_abort");
373
374 return EOPNOTSUPP;
375 }
376
377 static int
378 COMPATNAME(route_ioctl)(struct socket *so, u_long cmd, void *nam,
379 struct ifnet * ifp)
380 {
381 return EOPNOTSUPP;
382 }
383
384 static int
385 COMPATNAME(route_stat)(struct socket *so, struct stat *ub)
386 {
387 KASSERT(solocked(so));
388
389 return 0;
390 }
391
392 static int
393 COMPATNAME(route_peeraddr)(struct socket *so, struct sockaddr *nam)
394 {
395 struct rawcb *rp = sotorawcb(so);
396
397 KASSERT(solocked(so));
398 KASSERT(rp != NULL);
399 KASSERT(nam != NULL);
400
401 if (rp->rcb_faddr == NULL)
402 return ENOTCONN;
403
404 raw_setpeeraddr(rp, nam);
405 return 0;
406 }
407
408 static int
409 COMPATNAME(route_sockaddr)(struct socket *so, struct sockaddr *nam)
410 {
411 struct rawcb *rp = sotorawcb(so);
412
413 KASSERT(solocked(so));
414 KASSERT(rp != NULL);
415 KASSERT(nam != NULL);
416
417 if (rp->rcb_faddr == NULL)
418 return ENOTCONN;
419
420 raw_setsockaddr(rp, nam);
421 return 0;
422 }
423
424 static int
425 COMPATNAME(route_rcvd)(struct socket *so, int flags, struct lwp *l)
426 {
427 KASSERT(solocked(so));
428
429 return EOPNOTSUPP;
430 }
431
432 static int
433 COMPATNAME(route_recvoob)(struct socket *so, struct mbuf *m, int flags)
434 {
435 KASSERT(solocked(so));
436
437 return EOPNOTSUPP;
438 }
439
440 static int
441 COMPATNAME(route_send)(struct socket *so, struct mbuf *m,
442 struct sockaddr *nam, struct mbuf *control, struct lwp *l)
443 {
444 int error = 0;
445 int s;
446
447 KASSERT(solocked(so));
448 KASSERT(so->so_proto == &COMPATNAME(route_protosw)[0]);
449
450 s = splsoftnet();
451 error = raw_send(so, m, nam, control, l, &COMPATNAME(route_output));
452 splx(s);
453
454 return error;
455 }
456
457 static int
458 COMPATNAME(route_sendoob)(struct socket *so, struct mbuf *m,
459 struct mbuf *control)
460 {
461 KASSERT(solocked(so));
462
463 m_freem(m);
464 m_freem(control);
465
466 return EOPNOTSUPP;
467 }
468 static int
469 COMPATNAME(route_purgeif)(struct socket *so, struct ifnet *ifp)
470 {
471
472 panic("route_purgeif");
473
474 return EOPNOTSUPP;
475 }
476
477 #ifdef INET
478 static int
479 route_get_sdl_index(struct rt_addrinfo *info, int *sdl_index)
480 {
481 struct rtentry *nrt;
482 int error;
483
484 error = rtrequest1(RTM_GET, info, &nrt);
485 if (error != 0)
486 return error;
487 /*
488 * nrt->rt_ifp->if_index may not be correct
489 * due to changing to ifplo0.
490 */
491 *sdl_index = satosdl(nrt->rt_gateway)->sdl_index;
492 rt_unref(nrt);
493
494 return 0;
495 }
496 #endif /* INET */
497
498 static void
499 route_get_sdl(const struct ifnet *ifp, const struct sockaddr *dst,
500 struct sockaddr_dl *sdl, int *flags)
501 {
502 struct llentry *la;
503
504 KASSERT(ifp != NULL);
505
506 IF_AFDATA_RLOCK(ifp);
507 switch (dst->sa_family) {
508 case AF_INET:
509 la = lla_lookup(LLTABLE(ifp), 0, dst);
510 break;
511 case AF_INET6:
512 la = lla_lookup(LLTABLE6(ifp), 0, dst);
513 break;
514 default:
515 la = NULL;
516 KASSERTMSG(0, "Invalid AF=%d\n", dst->sa_family);
517 break;
518 }
519 IF_AFDATA_RUNLOCK(ifp);
520
521 void *a = (LLE_IS_VALID(la) && (la->la_flags & LLE_VALID) == LLE_VALID)
522 ? &la->ll_addr : NULL;
523
524 a = sockaddr_dl_init(sdl, sizeof(*sdl), ifp->if_index, ifp->if_type,
525 NULL, 0, a, ifp->if_addrlen);
526 KASSERT(a != NULL);
527
528 if (la != NULL) {
529 *flags = la->la_flags;
530 LLE_RUNLOCK(la);
531 }
532 }
533
534 static int
535 route_output_report(struct rtentry *rt, struct rt_addrinfo *info,
536 struct rt_xmsghdr *rtm, struct rt_xmsghdr **new_rtm)
537 {
538 int len;
539
540 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
541 const struct ifaddr *rtifa;
542 const struct ifnet *ifp = rt->rt_ifp;
543
544 info->rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
545 /* rtifa used to be simply rt->rt_ifa.
546 * If rt->rt_ifa != NULL, then
547 * rt_get_ifa() != NULL. So this
548 * ought to still be safe. --dyoung
549 */
550 rtifa = rt_get_ifa(rt);
551 info->rti_info[RTAX_IFA] = rtifa->ifa_addr;
552 #ifdef RTSOCK_DEBUG
553 if (info->rti_info[RTAX_IFA]->sa_family == AF_INET) {
554 char ibuf[INET_ADDRSTRLEN];
555 char abuf[INET_ADDRSTRLEN];
556 printf("%s: copying out RTAX_IFA %s "
557 "for info->rti_info[RTAX_DST] %s "
558 "ifa_getifa %p ifa_seqno %p\n",
559 __func__,
560 RT_IN_PRINT(info, ibuf, RTAX_IFA),
561 RT_IN_PRINT(info, abuf, RTAX_DST),
562 (void *)rtifa->ifa_getifa,
563 rtifa->ifa_seqno);
564 }
565 #endif /* RTSOCK_DEBUG */
566 if (ifp->if_flags & IFF_POINTOPOINT)
567 info->rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
568 else
569 info->rti_info[RTAX_BRD] = NULL;
570 rtm->rtm_index = ifp->if_index;
571 }
572 (void)rt_msg2(rtm->rtm_type, info, NULL, NULL, &len);
573 if (len > rtm->rtm_msglen) {
574 struct rt_xmsghdr *old_rtm = rtm;
575 R_Malloc(*new_rtm, struct rt_xmsghdr *, len);
576 if (*new_rtm == NULL)
577 return ENOBUFS;
578 (void)memcpy(*new_rtm, old_rtm, old_rtm->rtm_msglen);
579 rtm = *new_rtm;
580 }
581 (void)rt_msg2(rtm->rtm_type, info, rtm, NULL, 0);
582 rtm->rtm_flags = rt->rt_flags;
583 rtm_setmetrics(rt, rtm);
584 rtm->rtm_addrs = info->rti_addrs;
585
586 return 0;
587 }
588
589 static struct ifaddr *
590 route_output_get_ifa(const struct rt_addrinfo info, const struct rtentry *rt,
591 struct ifnet **ifp, struct psref *psref)
592 {
593 struct ifaddr *ifa = NULL;
594
595 *ifp = NULL;
596 if (info.rti_info[RTAX_IFP] != NULL) {
597 ifa = ifa_ifwithnet_psref(info.rti_info[RTAX_IFP], psref);
598 if (ifa == NULL)
599 goto next;
600 *ifp = ifa->ifa_ifp;
601 if (info.rti_info[RTAX_IFA] == NULL &&
602 info.rti_info[RTAX_GATEWAY] == NULL)
603 goto next;
604 if (info.rti_info[RTAX_IFA] == NULL) {
605 /* route change <dst> <gw> -ifp <if> */
606 ifa = ifaof_ifpforaddr_psref(info.rti_info[RTAX_GATEWAY],
607 *ifp, psref);
608 } else {
609 /* route change <dst> -ifp <if> -ifa <addr> */
610 ifa = ifa_ifwithaddr_psref(info.rti_info[RTAX_IFA], psref);
611 if (ifa != NULL)
612 goto out;
613 ifa = ifaof_ifpforaddr_psref(info.rti_info[RTAX_IFA],
614 *ifp, psref);
615 }
616 goto out;
617 }
618 next:
619 if (info.rti_info[RTAX_IFA] != NULL) {
620 /* route change <dst> <gw> -ifa <addr> */
621 ifa = ifa_ifwithaddr_psref(info.rti_info[RTAX_IFA], psref);
622 if (ifa != NULL)
623 goto out;
624 }
625 if (info.rti_info[RTAX_GATEWAY] != NULL) {
626 /* route change <dst> <gw> */
627 ifa = ifa_ifwithroute_psref(rt->rt_flags, rt_getkey(rt),
628 info.rti_info[RTAX_GATEWAY], psref);
629 }
630 out:
631 if (ifa != NULL && *ifp == NULL)
632 *ifp = ifa->ifa_ifp;
633 return ifa;
634 }
635
636 static int
637 route_output_change(struct rtentry *rt, struct rt_addrinfo *info,
638 struct rt_xmsghdr *rtm)
639 {
640 int error = 0;
641 struct ifnet *ifp = NULL, *new_ifp;
642 struct ifaddr *ifa = NULL, *new_ifa;
643 struct psref psref_ifa, psref_new_ifa, psref_ifp;
644 bool newgw, ifp_changed = false;
645
646 /*
647 * New gateway could require new ifaddr, ifp;
648 * flags may also be different; ifp may be specified
649 * by ll sockaddr when protocol address is ambiguous
650 */
651 newgw = info->rti_info[RTAX_GATEWAY] != NULL &&
652 sockaddr_cmp(info->rti_info[RTAX_GATEWAY], rt->rt_gateway) != 0;
653
654 if (newgw || info->rti_info[RTAX_IFP] != NULL ||
655 info->rti_info[RTAX_IFA] != NULL) {
656 ifp = rt_getifp(info, &psref_ifp);
657 ifa = rt_getifa(info, &psref_ifa);
658 if (ifa == NULL) {
659 error = ENETUNREACH;
660 goto out;
661 }
662 }
663 if (newgw) {
664 error = rt_setgate(rt, info->rti_info[RTAX_GATEWAY]);
665 if (error != 0)
666 goto out;
667 }
668 if (info->rti_info[RTAX_TAG]) {
669 const struct sockaddr *tag;
670 tag = rt_settag(rt, info->rti_info[RTAX_TAG]);
671 if (tag == NULL) {
672 error = ENOBUFS;
673 goto out;
674 }
675 }
676 /*
677 * New gateway could require new ifaddr, ifp;
678 * flags may also be different; ifp may be specified
679 * by ll sockaddr when protocol address is ambiguous
680 */
681 new_ifa = route_output_get_ifa(*info, rt, &new_ifp, &psref_new_ifa);
682 if (new_ifa != NULL) {
683 ifa_release(ifa, &psref_ifa);
684 ifa = new_ifa;
685 }
686 if (ifa) {
687 struct ifaddr *oifa = rt->rt_ifa;
688 if (oifa != ifa && !ifa_is_destroying(ifa) &&
689 new_ifp != NULL && !if_is_deactivated(new_ifp)) {
690 if (oifa && oifa->ifa_rtrequest)
691 oifa->ifa_rtrequest(RTM_DELETE, rt, info);
692 rt_replace_ifa(rt, ifa);
693 rt->rt_ifp = new_ifp;
694 ifp_changed = true;
695 }
696 if (new_ifa == NULL)
697 ifa_release(ifa, &psref_ifa);
698 }
699 ifa_release(new_ifa, &psref_new_ifa);
700 if (new_ifp && rt->rt_ifp != new_ifp && !if_is_deactivated(new_ifp)) {
701 rt->rt_ifp = new_ifp;
702 ifp_changed = true;
703 }
704 rt_setmetrics(rtm->rtm_inits, rtm, rt);
705 if (rt->rt_flags != info->rti_flags) {
706 rt->rt_flags = (info->rti_flags & ~PRESERVED_RTF) |
707 (rt->rt_flags & PRESERVED_RTF);
708 }
709 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
710 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
711 #if defined(INET) || defined(INET6)
712 if (ifp_changed && rt_mask(rt) != NULL)
713 lltable_prefix_free(rt_getkey(rt)->sa_family, rt_getkey(rt),
714 rt_mask(rt), 0);
715 #endif
716 out:
717 if_put(ifp, &psref_ifp);
718
719 return error;
720 }
721
722 /*ARGSUSED*/
723 int
724 COMPATNAME(route_output)(struct mbuf *m, struct socket *so)
725 {
726 struct sockproto proto = { .sp_family = PF_XROUTE, };
727 struct rt_xmsghdr *rtm = NULL;
728 struct rt_xmsghdr *old_rtm = NULL, *new_rtm = NULL;
729 struct rtentry *rt = NULL;
730 struct rtentry *saved_nrt = NULL;
731 struct rt_addrinfo info;
732 int len, error = 0;
733 sa_family_t family;
734 struct sockaddr_dl sdl;
735 int bound = curlwp_bind();
736 bool do_rt_free = false;
737
738 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
739 if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
740 (m = m_pullup(m, sizeof(int32_t))) == NULL)) {
741 error = ENOBUFS;
742 goto out;
743 }
744 if ((m->m_flags & M_PKTHDR) == 0)
745 panic("%s", __func__);
746 len = m->m_pkthdr.len;
747 if (len < sizeof(*rtm) ||
748 len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
749 info.rti_info[RTAX_DST] = NULL;
750 senderr(EINVAL);
751 }
752 R_Malloc(rtm, struct rt_xmsghdr *, len);
753 if (rtm == NULL) {
754 info.rti_info[RTAX_DST] = NULL;
755 senderr(ENOBUFS);
756 }
757 m_copydata(m, 0, len, rtm);
758 if (rtm->rtm_version != RTM_XVERSION) {
759 info.rti_info[RTAX_DST] = NULL;
760 senderr(EPROTONOSUPPORT);
761 }
762 rtm->rtm_pid = curproc->p_pid;
763 memset(&info, 0, sizeof(info));
764 info.rti_addrs = rtm->rtm_addrs;
765 if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
766 &info)) {
767 senderr(EINVAL);
768 }
769 info.rti_flags = rtm->rtm_flags;
770 #ifdef RTSOCK_DEBUG
771 if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
772 char abuf[INET_ADDRSTRLEN];
773 printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
774 RT_IN_PRINT(&info, abuf, RTAX_DST));
775 }
776 #endif /* RTSOCK_DEBUG */
777 if (info.rti_info[RTAX_DST] == NULL ||
778 (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
779 senderr(EINVAL);
780 }
781 if (info.rti_info[RTAX_GATEWAY] != NULL &&
782 (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
783 senderr(EINVAL);
784 }
785
786 /*
787 * Verify that the caller has the appropriate privilege; RTM_GET
788 * is the only operation the non-superuser is allowed.
789 */
790 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
791 0, rtm, NULL, NULL) != 0)
792 senderr(EACCES);
793
794 switch (rtm->rtm_type) {
795
796 case RTM_ADD:
797 if (info.rti_info[RTAX_GATEWAY] == NULL) {
798 senderr(EINVAL);
799 }
800 #ifdef INET
801 /* support for new ARP code with keeping backcompat */
802 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) {
803 const struct sockaddr_dl *sdlp =
804 satocsdl(info.rti_info[RTAX_GATEWAY]);
805
806 /* Allow routing requests by interface index */
807 if (sdlp->sdl_nlen == 0 && sdlp->sdl_alen == 0
808 && sdlp->sdl_slen == 0)
809 goto fallback;
810 /*
811 * Old arp binaries don't set the sdl_index
812 * so we have to complement it.
813 */
814 int sdl_index = sdlp->sdl_index;
815 if (sdl_index == 0) {
816 error = route_get_sdl_index(&info, &sdl_index);
817 if (error != 0)
818 goto fallback;
819 } else if (
820 info.rti_info[RTAX_DST]->sa_family == AF_INET) {
821 /*
822 * XXX workaround for SIN_PROXY case; proxy arp
823 * entry should be in an interface that has
824 * a network route including the destination,
825 * not a local (link) route that may not be a
826 * desired place, for example a tap.
827 */
828 const struct sockaddr_inarp *sina =
829 (const struct sockaddr_inarp *)
830 info.rti_info[RTAX_DST];
831 if (sina->sin_other & SIN_PROXY) {
832 error = route_get_sdl_index(&info,
833 &sdl_index);
834 if (error != 0)
835 goto fallback;
836 }
837 }
838 error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
839 rtm->rtm_rmx.rmx_expire, &info, sdl_index);
840 break;
841 }
842 fallback:
843 #endif /* INET */
844 error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
845 if (error == 0) {
846 rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
847 rt_unref(saved_nrt);
848 }
849 break;
850
851 case RTM_DELETE:
852 #ifdef INET
853 /* support for new ARP code */
854 if (info.rti_info[RTAX_GATEWAY] &&
855 (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
856 (rtm->rtm_flags & RTF_LLDATA) != 0) {
857 error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
858 rtm->rtm_rmx.rmx_expire, &info, 0);
859 break;
860 }
861 #endif /* INET */
862 error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
863 if (error != 0)
864 break;
865
866 rt = saved_nrt;
867 do_rt_free = true;
868 info.rti_info[RTAX_DST] = rt_getkey(rt);
869 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
870 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
871 info.rti_info[RTAX_TAG] = rt_gettag(rt);
872 error = route_output_report(rt, &info, rtm, &new_rtm);
873 if (error)
874 senderr(error);
875 if (new_rtm != NULL) {
876 old_rtm = rtm;
877 rtm = new_rtm;
878 }
879 break;
880
881 case RTM_GET:
882 case RTM_CHANGE:
883 case RTM_LOCK:
884 /* XXX This will mask info.rti_info[RTAX_DST] with
885 * info.rti_info[RTAX_NETMASK] before
886 * searching. It did not used to do that. --dyoung
887 */
888 rt = NULL;
889 error = rtrequest1(RTM_GET, &info, &rt);
890 if (error != 0)
891 senderr(error);
892 if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
893 if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
894 info.rti_info[RTAX_DST]->sa_len) != 0)
895 senderr(ESRCH);
896 if (info.rti_info[RTAX_NETMASK] == NULL &&
897 rt_mask(rt) != NULL)
898 senderr(ETOOMANYREFS);
899 }
900
901 /*
902 * XXX if arp/ndp requests an L2 entry, we have to obtain
903 * it from lltable while for the route command we have to
904 * return a route as it is. How to distinguish them?
905 * For newer arp/ndp, RTF_LLDATA flag set by arp/ndp
906 * indicates an L2 entry is requested. For old arp/ndp
907 * binaries, we check RTF_UP flag is NOT set; it works
908 * by the fact that arp/ndp don't set it while the route
909 * command sets it.
910 */
911 if (((rtm->rtm_flags & RTF_LLDATA) != 0 ||
912 (rtm->rtm_flags & RTF_UP) == 0) &&
913 rtm->rtm_type == RTM_GET &&
914 sockaddr_cmp(rt_getkey(rt), info.rti_info[RTAX_DST]) != 0) {
915 int ll_flags = 0;
916 route_get_sdl(rt->rt_ifp, info.rti_info[RTAX_DST], &sdl,
917 &ll_flags);
918 info.rti_info[RTAX_GATEWAY] = sstocsa(&sdl);
919 error = route_output_report(rt, &info, rtm, &new_rtm);
920 if (error)
921 senderr(error);
922 if (new_rtm != NULL) {
923 old_rtm = rtm;
924 rtm = new_rtm;
925 }
926 rtm->rtm_flags |= RTF_LLDATA;
927 rtm->rtm_flags &= ~RTF_CONNECTED;
928 rtm->rtm_flags |= (ll_flags & LLE_STATIC) ? RTF_STATIC : 0;
929 break;
930 }
931
932 switch (rtm->rtm_type) {
933 case RTM_GET:
934 info.rti_info[RTAX_DST] = rt_getkey(rt);
935 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
936 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
937 info.rti_info[RTAX_TAG] = rt_gettag(rt);
938 error = route_output_report(rt, &info, rtm, &new_rtm);
939 if (error)
940 senderr(error);
941 if (new_rtm != NULL) {
942 old_rtm = rtm;
943 rtm = new_rtm;
944 }
945 break;
946
947 case RTM_CHANGE:
948 #ifdef NET_MPSAFE
949 error = rt_update_prepare(rt);
950 if (error == 0) {
951 error = route_output_change(rt, &info, rtm);
952 rt_update_finish(rt);
953 }
954 #else
955 error = route_output_change(rt, &info, rtm);
956 #endif
957 if (error != 0)
958 goto flush;
959 /*FALLTHROUGH*/
960 case RTM_LOCK:
961 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
962 rt->rt_rmx.rmx_locks |=
963 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
964 break;
965 }
966 break;
967
968 default:
969 senderr(EOPNOTSUPP);
970 }
971
972 flush:
973 if (rtm) {
974 if (error)
975 rtm->rtm_errno = error;
976 else
977 rtm->rtm_flags |= RTF_DONE;
978 }
979 family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
980 0;
981 /* We cannot free old_rtm until we have stopped using the
982 * pointers in info, some of which may point to sockaddrs
983 * in old_rtm.
984 */
985 if (old_rtm != NULL)
986 Free(old_rtm);
987 if (rt) {
988 if (do_rt_free)
989 rt_free(rt);
990 else
991 rt_unref(rt);
992 }
993 {
994 struct rawcb *rp = NULL;
995 /*
996 * Check to see if we don't want our own messages.
997 */
998 if ((so->so_options & SO_USELOOPBACK) == 0) {
999 if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
1000 if (rtm)
1001 Free(rtm);
1002 m_freem(m);
1003 goto out;
1004 }
1005 /* There is another listener, so construct message */
1006 rp = sotorawcb(so);
1007 }
1008 if (rtm) {
1009 m_copyback(m, 0, rtm->rtm_msglen, rtm);
1010 if (m->m_pkthdr.len < rtm->rtm_msglen) {
1011 m_freem(m);
1012 m = NULL;
1013 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
1014 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1015 Free(rtm);
1016 }
1017 if (rp)
1018 rp->rcb_proto.sp_family = 0; /* Avoid us */
1019 if (family)
1020 proto.sp_protocol = family;
1021 if (m)
1022 raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
1023 &COMPATNAME(route_info).ri_dst);
1024 if (rp)
1025 rp->rcb_proto.sp_family = PF_XROUTE;
1026 }
1027 out:
1028 curlwp_bindx(bound);
1029 return error;
1030 }
1031
1032 static int
1033 route_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1034 {
1035 struct routecb *rop = sotoroutecb(so);
1036 int error = 0;
1037 unsigned char *rtm_type;
1038 size_t len;
1039 unsigned int msgfilter;
1040
1041 KASSERT(solocked(so));
1042
1043 if (sopt->sopt_level != AF_ROUTE) {
1044 error = ENOPROTOOPT;
1045 } else switch (op) {
1046 case PRCO_SETOPT:
1047 switch (sopt->sopt_name) {
1048 case RO_MSGFILTER:
1049 msgfilter = 0;
1050 for (rtm_type = sopt->sopt_data, len = sopt->sopt_size;
1051 len != 0;
1052 rtm_type++, len -= sizeof(*rtm_type))
1053 {
1054 /* Guard against overflowing our storage. */
1055 if (*rtm_type >= sizeof(msgfilter) * CHAR_BIT) {
1056 error = EOVERFLOW;
1057 break;
1058 }
1059 msgfilter |= RTMSGFILTER(*rtm_type);
1060 }
1061 if (error == 0)
1062 rop->rocb_msgfilter = msgfilter;
1063 break;
1064 default:
1065 error = ENOPROTOOPT;
1066 break;
1067 }
1068 break;
1069 case PRCO_GETOPT:
1070 switch (sopt->sopt_name) {
1071 case RO_MSGFILTER:
1072 error = ENOTSUP;
1073 break;
1074 default:
1075 error = ENOPROTOOPT;
1076 break;
1077 }
1078 }
1079 return error;
1080 }
1081
1082 static void
1083 rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
1084 {
1085 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
1086 metric(RTV_RPIPE, rmx_recvpipe);
1087 metric(RTV_SPIPE, rmx_sendpipe);
1088 metric(RTV_SSTHRESH, rmx_ssthresh);
1089 metric(RTV_RTT, rmx_rtt);
1090 metric(RTV_RTTVAR, rmx_rttvar);
1091 metric(RTV_HOPCOUNT, rmx_hopcount);
1092 metric(RTV_MTU, rmx_mtu);
1093 #undef metric
1094 if (which & RTV_EXPIRE) {
1095 out->rt_rmx.rmx_expire = in->rtm_rmx.rmx_expire ?
1096 time_wall_to_mono(in->rtm_rmx.rmx_expire) : 0;
1097 }
1098 }
1099
1100 static void
1101 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
1102 {
1103 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
1104 metric(rmx_recvpipe);
1105 metric(rmx_sendpipe);
1106 metric(rmx_ssthresh);
1107 metric(rmx_rtt);
1108 metric(rmx_rttvar);
1109 metric(rmx_hopcount);
1110 metric(rmx_mtu);
1111 metric(rmx_locks);
1112 #undef metric
1113 out->rtm_rmx.rmx_expire = in->rt_rmx.rmx_expire ?
1114 time_mono_to_wall(in->rt_rmx.rmx_expire) : 0;
1115 }
1116
1117 static int
1118 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
1119 struct rt_addrinfo *rtinfo)
1120 {
1121 const struct sockaddr *sa = NULL; /* Quell compiler warning */
1122 int i;
1123
1124 for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1125 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1126 continue;
1127 rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
1128 RT_XADVANCE(cp, sa);
1129 }
1130
1131 /*
1132 * Check for extra addresses specified, except RTM_GET asking
1133 * for interface info.
1134 */
1135 if (rtmtype == RTM_GET) {
1136 if (((rtinfo->rti_addrs &
1137 (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0U << i)) != 0)
1138 return 1;
1139 } else if ((rtinfo->rti_addrs & (~0U << i)) != 0)
1140 return 1;
1141 /* Check for bad data length. */
1142 if (cp != cplim) {
1143 if (i == RTAX_NETMASK + 1 && sa != NULL &&
1144 cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
1145 /*
1146 * The last sockaddr was info.rti_info[RTAX_NETMASK].
1147 * We accept this for now for the sake of old
1148 * binaries or third party softwares.
1149 */
1150 ;
1151 else
1152 return 1;
1153 }
1154 return 0;
1155 }
1156
1157 static int
1158 rt_getlen(int type)
1159 {
1160 #ifndef COMPAT_RTSOCK
1161 CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
1162 CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
1163 CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
1164 CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
1165 #endif
1166
1167 switch (type) {
1168 case RTM_ODELADDR:
1169 case RTM_ONEWADDR:
1170 case RTM_OCHGADDR:
1171 #ifdef COMPAT_70
1172 return sizeof(struct ifa_msghdr70);
1173 #else
1174 #ifdef RTSOCK_DEBUG
1175 printf("%s: unsupported RTM type %d\n", __func__, type);
1176 #endif
1177 return -1;
1178 #endif
1179 case RTM_DELADDR:
1180 case RTM_NEWADDR:
1181 case RTM_CHGADDR:
1182 return sizeof(struct ifa_xmsghdr);
1183
1184 case RTM_OOIFINFO:
1185 #ifdef COMPAT_14
1186 return sizeof(struct if_msghdr14);
1187 #else
1188 #ifdef RTSOCK_DEBUG
1189 printf("%s: unsupported RTM type RTM_OOIFINFO\n", __func__);
1190 #endif
1191 return -1;
1192 #endif
1193 case RTM_OIFINFO:
1194 #ifdef COMPAT_50
1195 return sizeof(struct if_msghdr50);
1196 #else
1197 #ifdef RTSOCK_DEBUG
1198 printf("%s: unsupported RTM type RTM_OIFINFO\n", __func__);
1199 #endif
1200 return -1;
1201 #endif
1202
1203 case RTM_IFINFO:
1204 return sizeof(struct if_xmsghdr);
1205
1206 case RTM_IFANNOUNCE:
1207 case RTM_IEEE80211:
1208 return sizeof(struct if_xannouncemsghdr);
1209
1210 default:
1211 return sizeof(struct rt_xmsghdr);
1212 }
1213 }
1214
1215
1216 struct mbuf *
1217 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
1218 {
1219 struct rt_xmsghdr *rtm;
1220 struct mbuf *m;
1221 int i;
1222 const struct sockaddr *sa;
1223 int len, dlen;
1224
1225 m = m_gethdr(M_DONTWAIT, MT_DATA);
1226 if (m == NULL)
1227 return m;
1228 MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
1229
1230 if ((len = rt_getlen(type)) == -1)
1231 goto out;
1232 if (len > MHLEN + MLEN)
1233 panic("%s: message too long", __func__);
1234 else if (len > MHLEN) {
1235 m->m_next = m_get(M_DONTWAIT, MT_DATA);
1236 if (m->m_next == NULL)
1237 goto out;
1238 MCLAIM(m->m_next, m->m_owner);
1239 m->m_pkthdr.len = len;
1240 m->m_len = MHLEN;
1241 m->m_next->m_len = len - MHLEN;
1242 } else {
1243 m->m_pkthdr.len = m->m_len = len;
1244 }
1245 m_reset_rcvif(m);
1246 m_copyback(m, 0, datalen, data);
1247 if (len > datalen)
1248 (void)memset(mtod(m, char *) + datalen, 0, len - datalen);
1249 rtm = mtod(m, struct rt_xmsghdr *);
1250 for (i = 0; i < RTAX_MAX; i++) {
1251 if ((sa = rtinfo->rti_info[i]) == NULL)
1252 continue;
1253 rtinfo->rti_addrs |= (1 << i);
1254 dlen = RT_XROUNDUP(sa->sa_len);
1255 m_copyback(m, len, sa->sa_len, sa);
1256 if (dlen != sa->sa_len) {
1257 /*
1258 * Up to 6 + 1 nul's since roundup is to
1259 * sizeof(uint64_t) (8 bytes)
1260 */
1261 m_copyback(m, len + sa->sa_len,
1262 dlen - sa->sa_len, "\0\0\0\0\0\0");
1263 }
1264 len += dlen;
1265 }
1266 if (m->m_pkthdr.len != len)
1267 goto out;
1268 rtm->rtm_msglen = len;
1269 rtm->rtm_version = RTM_XVERSION;
1270 rtm->rtm_type = type;
1271 return m;
1272 out:
1273 m_freem(m);
1274 return NULL;
1275 }
1276
1277 /*
1278 * rt_msg2
1279 *
1280 * fills 'cp' or 'w'.w_tmem with the routing socket message and
1281 * returns the length of the message in 'lenp'.
1282 *
1283 * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
1284 * the message
1285 * otherwise walkarg's w_needed is updated and if the user buffer is
1286 * specified and w_needed indicates space exists the information is copied
1287 * into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
1288 * if the allocation fails ENOBUFS is returned.
1289 */
1290 static int
1291 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
1292 int *lenp)
1293 {
1294 int i;
1295 int len, dlen, second_time = 0;
1296 char *cp0, *cp = cpv;
1297
1298 rtinfo->rti_addrs = 0;
1299 again:
1300 if ((len = rt_getlen(type)) == -1)
1301 return EINVAL;
1302
1303 if ((cp0 = cp) != NULL)
1304 cp += len;
1305 for (i = 0; i < RTAX_MAX; i++) {
1306 const struct sockaddr *sa;
1307
1308 if ((sa = rtinfo->rti_info[i]) == NULL)
1309 continue;
1310 rtinfo->rti_addrs |= (1 << i);
1311 dlen = RT_XROUNDUP(sa->sa_len);
1312 if (cp) {
1313 int diff = dlen - sa->sa_len;
1314 (void)memcpy(cp, sa, (size_t)sa->sa_len);
1315 cp += sa->sa_len;
1316 if (diff > 0) {
1317 (void)memset(cp, 0, (size_t)diff);
1318 cp += diff;
1319 }
1320 }
1321 len += dlen;
1322 }
1323 if (cp == NULL && w != NULL && !second_time) {
1324 struct rt_walkarg *rw = w;
1325
1326 rw->w_needed += len;
1327 if (rw->w_needed <= 0 && rw->w_where) {
1328 if (rw->w_tmemsize < len) {
1329 if (rw->w_tmem)
1330 kmem_free(rw->w_tmem, rw->w_tmemsize);
1331 rw->w_tmem = kmem_alloc(len, KM_SLEEP);
1332 rw->w_tmemsize = len;
1333 }
1334 if (rw->w_tmem) {
1335 cp = rw->w_tmem;
1336 second_time = 1;
1337 goto again;
1338 } else {
1339 rw->w_tmemneeded = len;
1340 return ENOBUFS;
1341 }
1342 }
1343 }
1344 if (cp) {
1345 struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
1346
1347 rtm->rtm_version = RTM_XVERSION;
1348 rtm->rtm_type = type;
1349 rtm->rtm_msglen = len;
1350 }
1351 if (lenp)
1352 *lenp = len;
1353 return 0;
1354 }
1355
1356 #ifndef COMPAT_RTSOCK
1357 int
1358 rt_msg3(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
1359 int *lenp)
1360 {
1361 return rt_msg2(type, rtinfo, cpv, w, lenp);
1362 }
1363 #endif
1364
1365 /*
1366 * This routine is called to generate a message from the routing
1367 * socket indicating that a redirect has occurred, a routing lookup
1368 * has failed, or that a protocol has detected timeouts to a particular
1369 * destination.
1370 */
1371 void
1372 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
1373 int error)
1374 {
1375 struct rt_xmsghdr rtm;
1376 struct mbuf *m;
1377 const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1378 struct rt_addrinfo info = *rtinfo;
1379
1380 COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
1381 if (COMPATNAME(route_info).ri_cb.any_count == 0)
1382 return;
1383 memset(&rtm, 0, sizeof(rtm));
1384 rtm.rtm_pid = curproc->p_pid;
1385 rtm.rtm_flags = RTF_DONE | flags;
1386 rtm.rtm_errno = error;
1387 m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
1388 if (m == NULL)
1389 return;
1390 mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1391 COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1392 }
1393
1394 /*
1395 * This routine is called to generate a message from the routing
1396 * socket indicating that the status of a network interface has changed.
1397 */
1398 void
1399 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
1400 {
1401 struct if_xmsghdr ifm;
1402 struct mbuf *m;
1403 struct rt_addrinfo info;
1404
1405 COMPATCALL(rt_ifmsg, (ifp));
1406 if (COMPATNAME(route_info).ri_cb.any_count == 0)
1407 return;
1408 (void)memset(&info, 0, sizeof(info));
1409 (void)memset(&ifm, 0, sizeof(ifm));
1410 ifm.ifm_index = ifp->if_index;
1411 ifm.ifm_flags = ifp->if_flags;
1412 ifm.ifm_data = ifp->if_data;
1413 ifm.ifm_addrs = 0;
1414 m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
1415 if (m == NULL)
1416 return;
1417 COMPATNAME(route_enqueue)(m, 0);
1418 #ifdef COMPAT_14
1419 compat_14_rt_oifmsg(ifp);
1420 #endif
1421 #ifdef COMPAT_50
1422 compat_50_rt_oifmsg(ifp);
1423 #endif
1424 }
1425
1426 #ifndef COMPAT_RTSOCK
1427 static int
1428 if_addrflags(struct ifaddr *ifa)
1429 {
1430
1431 switch (ifa->ifa_addr->sa_family) {
1432 #ifdef INET
1433 case AF_INET:
1434 return ((struct in_ifaddr *)ifa)->ia4_flags;
1435 #endif
1436 #ifdef INET6
1437 case AF_INET6:
1438 return ((struct in6_ifaddr *)ifa)->ia6_flags;
1439 #endif
1440 default:
1441 return 0;
1442 }
1443 }
1444 #endif
1445
1446 /*
1447 * This is called to generate messages from the routing socket
1448 * indicating a network interface has had addresses associated with it.
1449 * if we ever reverse the logic and replace messages TO the routing
1450 * socket indicate a request to configure interfaces, then it will
1451 * be unnecessary as the routing socket will automatically generate
1452 * copies of it.
1453 */
1454 void
1455 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
1456 struct rtentry *rt)
1457 {
1458 #define cmdpass(__cmd, __pass) (((__cmd) << 2) | (__pass))
1459 struct rt_addrinfo info;
1460 const struct sockaddr *sa;
1461 int pass;
1462 struct mbuf *m;
1463 struct ifnet *ifp;
1464 struct rt_xmsghdr rtm;
1465 struct ifa_xmsghdr ifam;
1466 int ncmd;
1467
1468 KASSERT(ifa != NULL);
1469 KASSERT(ifa->ifa_addr != NULL);
1470 ifp = ifa->ifa_ifp;
1471 #ifdef SCTP
1472 if (cmd == RTM_ADD) {
1473 sctp_add_ip_address(ifa);
1474 } else if (cmd == RTM_DELETE) {
1475 sctp_delete_ip_address(ifa);
1476 }
1477 #endif
1478
1479 COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
1480 if (COMPATNAME(route_info).ri_cb.any_count == 0)
1481 return;
1482 for (pass = 1; pass < 3; pass++) {
1483 memset(&info, 0, sizeof(info));
1484 switch (cmdpass(cmd, pass)) {
1485 case cmdpass(RTM_ADD, 1):
1486 case cmdpass(RTM_CHANGE, 1):
1487 case cmdpass(RTM_DELETE, 2):
1488 case cmdpass(RTM_NEWADDR, 1):
1489 case cmdpass(RTM_DELADDR, 1):
1490 case cmdpass(RTM_CHGADDR, 1):
1491 switch (cmd) {
1492 case RTM_ADD:
1493 ncmd = RTM_XNEWADDR;
1494 break;
1495 case RTM_DELETE:
1496 ncmd = RTM_XDELADDR;
1497 break;
1498 case RTM_CHANGE:
1499 ncmd = RTM_XCHGADDR;
1500 break;
1501 case RTM_NEWADDR:
1502 ncmd = RTM_XNEWADDR;
1503 break;
1504 case RTM_DELADDR:
1505 ncmd = RTM_XDELADDR;
1506 break;
1507 case RTM_CHGADDR:
1508 ncmd = RTM_XCHGADDR;
1509 break;
1510 default:
1511 panic("%s: unknown command %d", __func__, cmd);
1512 }
1513 #ifdef COMPAT_70
1514 compat_70_rt_newaddrmsg1(ncmd, ifa);
1515 #endif
1516 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1517 KASSERT(ifp->if_dl != NULL);
1518 info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1519 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1520 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1521 memset(&ifam, 0, sizeof(ifam));
1522 ifam.ifam_index = ifp->if_index;
1523 ifam.ifam_metric = ifa->ifa_metric;
1524 ifam.ifam_flags = ifa->ifa_flags;
1525 #ifndef COMPAT_RTSOCK
1526 ifam.ifam_pid = curproc->p_pid;
1527 ifam.ifam_addrflags = if_addrflags(ifa);
1528 #endif
1529 m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
1530 if (m == NULL)
1531 continue;
1532 mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
1533 info.rti_addrs;
1534 break;
1535 case cmdpass(RTM_ADD, 2):
1536 case cmdpass(RTM_CHANGE, 2):
1537 case cmdpass(RTM_DELETE, 1):
1538 if (rt == NULL)
1539 continue;
1540 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1541 info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
1542 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1543 memset(&rtm, 0, sizeof(rtm));
1544 rtm.rtm_pid = curproc->p_pid;
1545 rtm.rtm_index = ifp->if_index;
1546 rtm.rtm_flags |= rt->rt_flags;
1547 rtm.rtm_errno = error;
1548 m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
1549 if (m == NULL)
1550 continue;
1551 mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1552 break;
1553 default:
1554 continue;
1555 }
1556 KASSERTMSG(m != NULL, "called with wrong command");
1557 COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1558 }
1559 #undef cmdpass
1560
1561 }
1562
1563 static struct mbuf *
1564 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1565 struct rt_addrinfo *info)
1566 {
1567 struct if_xannouncemsghdr ifan;
1568
1569 memset(info, 0, sizeof(*info));
1570 memset(&ifan, 0, sizeof(ifan));
1571 ifan.ifan_index = ifp->if_index;
1572 strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
1573 ifan.ifan_what = what;
1574 return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
1575 }
1576
1577 /*
1578 * This is called to generate routing socket messages indicating
1579 * network interface arrival and departure.
1580 */
1581 void
1582 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
1583 {
1584 struct mbuf *m;
1585 struct rt_addrinfo info;
1586
1587 COMPATCALL(rt_ifannouncemsg, (ifp, what));
1588 if (COMPATNAME(route_info).ri_cb.any_count == 0)
1589 return;
1590 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1591 if (m == NULL)
1592 return;
1593 COMPATNAME(route_enqueue)(m, 0);
1594 }
1595
1596 /*
1597 * This is called to generate routing socket messages indicating
1598 * IEEE80211 wireless events.
1599 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1600 */
1601 void
1602 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
1603 size_t data_len)
1604 {
1605 struct mbuf *m;
1606 struct rt_addrinfo info;
1607
1608 COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
1609 if (COMPATNAME(route_info).ri_cb.any_count == 0)
1610 return;
1611 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1612 if (m == NULL)
1613 return;
1614 /*
1615 * Append the ieee80211 data. Try to stick it in the
1616 * mbuf containing the ifannounce msg; otherwise allocate
1617 * a new mbuf and append.
1618 *
1619 * NB: we assume m is a single mbuf.
1620 */
1621 if (data_len > M_TRAILINGSPACE(m)) {
1622 struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1623 if (n == NULL) {
1624 m_freem(m);
1625 return;
1626 }
1627 (void)memcpy(mtod(n, void *), data, data_len);
1628 n->m_len = data_len;
1629 m->m_next = n;
1630 } else if (data_len > 0) {
1631 (void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
1632 m->m_len += data_len;
1633 }
1634 if (m->m_flags & M_PKTHDR)
1635 m->m_pkthdr.len += data_len;
1636 mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
1637 COMPATNAME(route_enqueue)(m, 0);
1638 }
1639
1640 #ifndef COMPAT_RTSOCK
1641 /*
1642 * Send a routing message as mimicing that a cloned route is added.
1643 */
1644 void
1645 rt_clonedmsg(const struct sockaddr *dst, const struct ifnet *ifp,
1646 const struct rtentry *rt)
1647 {
1648 struct rt_addrinfo info;
1649 /* Mimic flags exactly */
1650 #define RTF_LLINFO 0x400
1651 #define RTF_CLONED 0x2000
1652 int flags = RTF_UP | RTF_HOST | RTF_DONE | RTF_LLINFO | RTF_CLONED;
1653 union {
1654 struct sockaddr sa;
1655 struct sockaddr_storage ss;
1656 struct sockaddr_dl sdl;
1657 } u;
1658 uint8_t namelen = strlen(ifp->if_xname);
1659 uint8_t addrlen = ifp->if_addrlen;
1660
1661 if (rt == NULL)
1662 return; /* XXX */
1663
1664 memset(&info, 0, sizeof(info));
1665 info.rti_info[RTAX_DST] = dst;
1666 sockaddr_dl_init(&u.sdl, sizeof(u.ss), ifp->if_index, ifp->if_type,
1667 NULL, namelen, NULL, addrlen);
1668 info.rti_info[RTAX_GATEWAY] = &u.sa;
1669 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
1670 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1671
1672 rt_missmsg(RTM_ADD, &info, flags, 0);
1673 #undef RTF_LLINFO
1674 #undef RTF_CLONED
1675 }
1676 #endif /* COMPAT_RTSOCK */
1677
1678 /*
1679 * This is used in dumping the kernel table via sysctl().
1680 */
1681 static int
1682 sysctl_dumpentry(struct rtentry *rt, void *v)
1683 {
1684 struct rt_walkarg *w = v;
1685 int error = 0, size;
1686 struct rt_addrinfo info;
1687
1688 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1689 return 0;
1690 memset(&info, 0, sizeof(info));
1691 info.rti_info[RTAX_DST] = rt_getkey(rt);
1692 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1693 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1694 info.rti_info[RTAX_TAG] = rt_gettag(rt);
1695 if (rt->rt_ifp) {
1696 const struct ifaddr *rtifa;
1697 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
1698 /* rtifa used to be simply rt->rt_ifa. If rt->rt_ifa != NULL,
1699 * then rt_get_ifa() != NULL. So this ought to still be safe.
1700 * --dyoung
1701 */
1702 rtifa = rt_get_ifa(rt);
1703 info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
1704 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1705 info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
1706 }
1707 if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
1708 return error;
1709 if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1710 struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
1711
1712 rtm->rtm_flags = rt->rt_flags;
1713 rtm->rtm_use = rt->rt_use;
1714 rtm_setmetrics(rt, rtm);
1715 KASSERT(rt->rt_ifp != NULL);
1716 rtm->rtm_index = rt->rt_ifp->if_index;
1717 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1718 rtm->rtm_addrs = info.rti_addrs;
1719 if ((error = copyout(rtm, w->w_where, size)) != 0)
1720 w->w_where = NULL;
1721 else
1722 w->w_where = (char *)w->w_where + size;
1723 }
1724 return error;
1725 }
1726
1727 static int
1728 sysctl_iflist_if(struct ifnet *ifp, struct rt_walkarg *w,
1729 struct rt_addrinfo *info, size_t len)
1730 {
1731 struct if_xmsghdr *ifm;
1732 int error;
1733
1734 ifm = (struct if_xmsghdr *)w->w_tmem;
1735 ifm->ifm_index = ifp->if_index;
1736 ifm->ifm_flags = ifp->if_flags;
1737 ifm->ifm_data = ifp->if_data;
1738 ifm->ifm_addrs = info->rti_addrs;
1739 if ((error = copyout(ifm, w->w_where, len)) == 0)
1740 w->w_where = (char *)w->w_where + len;
1741 return error;
1742 }
1743
1744 static int
1745 sysctl_iflist_addr(struct rt_walkarg *w, struct ifaddr *ifa,
1746 struct rt_addrinfo *info)
1747 {
1748 int len, error;
1749
1750 if ((error = rt_msg2(RTM_XNEWADDR, info, 0, w, &len)))
1751 return error;
1752 if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1753 struct ifa_xmsghdr *ifam;
1754
1755 ifam = (struct ifa_xmsghdr *)w->w_tmem;
1756 ifam->ifam_index = ifa->ifa_ifp->if_index;
1757 ifam->ifam_flags = ifa->ifa_flags;
1758 ifam->ifam_metric = ifa->ifa_metric;
1759 ifam->ifam_addrs = info->rti_addrs;
1760 #ifndef COMPAT_RTSOCK
1761 ifam->ifam_pid = 0;
1762 ifam->ifam_addrflags = if_addrflags(ifa);
1763 #endif
1764 if ((error = copyout(w->w_tmem, w->w_where, len)) == 0)
1765 w->w_where = (char *)w->w_where + len;
1766 }
1767 return error;
1768 }
1769
1770 static int
1771 sysctl_iflist(int af, struct rt_walkarg *w, int type)
1772 {
1773 struct ifnet *ifp;
1774 struct ifaddr *ifa;
1775 struct rt_addrinfo info;
1776 int cmd, len, error = 0;
1777 int (*iflist_if)(struct ifnet *, struct rt_walkarg *,
1778 struct rt_addrinfo *, size_t);
1779 int (*iflist_addr)(struct rt_walkarg *, struct ifaddr *,
1780 struct rt_addrinfo *);
1781 int s;
1782 struct psref psref;
1783 int bound;
1784
1785 switch (type) {
1786 case NET_RT_IFLIST:
1787 cmd = RTM_IFINFO;
1788 iflist_if = sysctl_iflist_if;
1789 iflist_addr = sysctl_iflist_addr;
1790 break;
1791 #ifdef COMPAT_14
1792 case NET_RT_OOOIFLIST:
1793 cmd = RTM_OOIFINFO;
1794 iflist_if = compat_14_iflist;
1795 iflist_addr = compat_70_iflist_addr;
1796 break;
1797 #endif
1798 #ifdef COMPAT_50
1799 case NET_RT_OOIFLIST:
1800 cmd = RTM_OIFINFO;
1801 iflist_if = compat_50_iflist;
1802 iflist_addr = compat_70_iflist_addr;
1803 break;
1804 #endif
1805 #ifdef COMPAT_70
1806 case NET_RT_OIFLIST:
1807 cmd = RTM_IFINFO;
1808 iflist_if = sysctl_iflist_if;
1809 iflist_addr = compat_70_iflist_addr;
1810 break;
1811 #endif
1812 default:
1813 #ifdef RTSOCK_DEBUG
1814 printf("%s: unsupported IFLIST type %d\n", __func__, type);
1815 #endif
1816 return EINVAL;
1817 }
1818
1819 memset(&info, 0, sizeof(info));
1820
1821 bound = curlwp_bind();
1822 s = pserialize_read_enter();
1823 IFNET_READER_FOREACH(ifp) {
1824 int _s;
1825 if (w->w_arg && w->w_arg != ifp->if_index)
1826 continue;
1827 if (IFADDR_READER_EMPTY(ifp))
1828 continue;
1829
1830 if_acquire(ifp, &psref);
1831 pserialize_read_exit(s);
1832
1833 info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1834 if ((error = rt_msg2(cmd, &info, NULL, w, &len)) != 0)
1835 goto release_exit;
1836 info.rti_info[RTAX_IFP] = NULL;
1837 if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1838 if ((error = iflist_if(ifp, w, &info, len)) != 0)
1839 goto release_exit;
1840 }
1841 _s = pserialize_read_enter();
1842 IFADDR_READER_FOREACH(ifa, ifp) {
1843 struct psref _psref;
1844 if (af && af != ifa->ifa_addr->sa_family)
1845 continue;
1846 ifa_acquire(ifa, &_psref);
1847 pserialize_read_exit(_s);
1848
1849 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1850 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1851 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1852 error = iflist_addr(w, ifa, &info);
1853
1854 _s = pserialize_read_enter();
1855 ifa_release(ifa, &_psref);
1856 if (error != 0) {
1857 pserialize_read_exit(_s);
1858 goto release_exit;
1859 }
1860 }
1861 pserialize_read_exit(_s);
1862 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1863 info.rti_info[RTAX_BRD] = NULL;
1864
1865 s = pserialize_read_enter();
1866 if_release(ifp, &psref);
1867 }
1868 pserialize_read_exit(s);
1869 curlwp_bindx(bound);
1870
1871 return 0;
1872
1873 release_exit:
1874 if_release(ifp, &psref);
1875 curlwp_bindx(bound);
1876 return error;
1877 }
1878
1879 static int
1880 sysctl_rtable(SYSCTLFN_ARGS)
1881 {
1882 void *where = oldp;
1883 size_t *given = oldlenp;
1884 int i, s, error = EINVAL;
1885 u_char af;
1886 struct rt_walkarg w;
1887
1888 if (namelen == 1 && name[0] == CTL_QUERY)
1889 return sysctl_query(SYSCTLFN_CALL(rnode));
1890
1891 if (newp)
1892 return EPERM;
1893 if (namelen != 3)
1894 return EINVAL;
1895 af = name[0];
1896 w.w_tmemneeded = 0;
1897 w.w_tmemsize = 0;
1898 w.w_tmem = NULL;
1899 again:
1900 /* we may return here if a later [re]alloc of the t_mem buffer fails */
1901 if (w.w_tmemneeded) {
1902 w.w_tmem = kmem_alloc(w.w_tmemneeded, KM_SLEEP);
1903 w.w_tmemsize = w.w_tmemneeded;
1904 w.w_tmemneeded = 0;
1905 }
1906 w.w_op = name[1];
1907 w.w_arg = name[2];
1908 w.w_given = *given;
1909 w.w_needed = 0 - w.w_given;
1910 w.w_where = where;
1911
1912 s = splsoftnet();
1913 switch (w.w_op) {
1914
1915 case NET_RT_DUMP:
1916 case NET_RT_FLAGS:
1917 #ifdef INET
1918 /*
1919 * take care of llinfo entries, the caller must
1920 * specify an AF
1921 */
1922 if (w.w_op == NET_RT_FLAGS &&
1923 (w.w_arg == 0 || w.w_arg & RTF_LLDATA)) {
1924 if (af != 0)
1925 error = lltable_sysctl_dumparp(af, &w);
1926 else
1927 error = EINVAL;
1928 break;
1929 }
1930 #endif /* INET */
1931
1932 for (i = 1; i <= AF_MAX; i++)
1933 if ((af == 0 || af == i) &&
1934 (error = rt_walktree(i, sysctl_dumpentry, &w)))
1935 break;
1936 break;
1937
1938 #ifdef COMPAT_14
1939 case NET_RT_OOOIFLIST:
1940 error = sysctl_iflist(af, &w, w.w_op);
1941 break;
1942 #endif
1943 #ifdef COMPAT_50
1944 case NET_RT_OOIFLIST:
1945 error = sysctl_iflist(af, &w, w.w_op);
1946 break;
1947 #endif
1948 #ifdef COMPAT_70
1949 case NET_RT_OIFLIST:
1950 error = sysctl_iflist(af, &w, w.w_op);
1951 break;
1952 #endif
1953 case NET_RT_IFLIST:
1954 error = sysctl_iflist(af, &w, w.w_op);
1955 break;
1956 }
1957 splx(s);
1958
1959 /* check to see if we couldn't allocate memory with NOWAIT */
1960 if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
1961 goto again;
1962
1963 if (w.w_tmem)
1964 kmem_free(w.w_tmem, w.w_tmemsize);
1965 w.w_needed += w.w_given;
1966 if (where) {
1967 *given = (char *)w.w_where - (char *)where;
1968 if (*given < w.w_needed)
1969 return ENOMEM;
1970 } else {
1971 *given = (11 * w.w_needed) / 10;
1972 }
1973 return error;
1974 }
1975
1976 /*
1977 * Routing message software interrupt routine
1978 */
1979 static void
1980 COMPATNAME(route_intr)(void *cookie)
1981 {
1982 struct sockproto proto = { .sp_family = PF_XROUTE, };
1983 struct route_info * const ri = &COMPATNAME(route_info);
1984 struct mbuf *m;
1985
1986 mutex_enter(softnet_lock);
1987 KERNEL_LOCK(1, NULL);
1988 for (;;) {
1989 IFQ_LOCK(&ri->ri_intrq);
1990 IF_DEQUEUE(&ri->ri_intrq, m);
1991 IFQ_UNLOCK(&ri->ri_intrq);
1992 if (m == NULL)
1993 break;
1994 proto.sp_protocol = M_GETCTX(m, uintptr_t);
1995 raw_input(m, &proto, &ri->ri_src, &ri->ri_dst);
1996 }
1997 KERNEL_UNLOCK_ONE(NULL);
1998 mutex_exit(softnet_lock);
1999 }
2000
2001 /*
2002 * Enqueue a message to the software interrupt routine.
2003 */
2004 void
2005 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
2006 {
2007 struct route_info * const ri = &COMPATNAME(route_info);
2008 int wasempty;
2009
2010 IFQ_LOCK(&ri->ri_intrq);
2011 if (IF_QFULL(&ri->ri_intrq)) {
2012 IF_DROP(&ri->ri_intrq);
2013 IFQ_UNLOCK(&ri->ri_intrq);
2014 m_freem(m);
2015 } else {
2016 wasempty = IF_IS_EMPTY(&ri->ri_intrq);
2017 M_SETCTX(m, (uintptr_t)family);
2018 IF_ENQUEUE(&ri->ri_intrq, m);
2019 IFQ_UNLOCK(&ri->ri_intrq);
2020 if (wasempty) {
2021 kpreempt_disable();
2022 softint_schedule(ri->ri_sih);
2023 kpreempt_enable();
2024 }
2025 }
2026 }
2027
2028 static void
2029 COMPATNAME(route_init)(void)
2030 {
2031 struct route_info * const ri = &COMPATNAME(route_info);
2032
2033 #ifndef COMPAT_RTSOCK
2034 rt_init();
2035 #endif
2036
2037 sysctl_net_route_setup(NULL);
2038 ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
2039 ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
2040 COMPATNAME(route_intr), NULL);
2041 IFQ_LOCK_INIT(&ri->ri_intrq);
2042 }
2043
2044 /*
2045 * Definitions of protocols supported in the ROUTE domain.
2046 */
2047 #ifndef COMPAT_RTSOCK
2048 PR_WRAP_USRREQS(route);
2049 #else
2050 PR_WRAP_USRREQS(compat_50_route);
2051 #endif
2052
2053 static const struct pr_usrreqs route_usrreqs = {
2054 .pr_attach = COMPATNAME(route_attach_wrapper),
2055 .pr_detach = COMPATNAME(route_detach_wrapper),
2056 .pr_accept = COMPATNAME(route_accept_wrapper),
2057 .pr_bind = COMPATNAME(route_bind_wrapper),
2058 .pr_listen = COMPATNAME(route_listen_wrapper),
2059 .pr_connect = COMPATNAME(route_connect_wrapper),
2060 .pr_connect2 = COMPATNAME(route_connect2_wrapper),
2061 .pr_disconnect = COMPATNAME(route_disconnect_wrapper),
2062 .pr_shutdown = COMPATNAME(route_shutdown_wrapper),
2063 .pr_abort = COMPATNAME(route_abort_wrapper),
2064 .pr_ioctl = COMPATNAME(route_ioctl_wrapper),
2065 .pr_stat = COMPATNAME(route_stat_wrapper),
2066 .pr_peeraddr = COMPATNAME(route_peeraddr_wrapper),
2067 .pr_sockaddr = COMPATNAME(route_sockaddr_wrapper),
2068 .pr_rcvd = COMPATNAME(route_rcvd_wrapper),
2069 .pr_recvoob = COMPATNAME(route_recvoob_wrapper),
2070 .pr_send = COMPATNAME(route_send_wrapper),
2071 .pr_sendoob = COMPATNAME(route_sendoob_wrapper),
2072 .pr_purgeif = COMPATNAME(route_purgeif_wrapper),
2073 };
2074
2075 static const struct protosw COMPATNAME(route_protosw)[] = {
2076 {
2077 .pr_type = SOCK_RAW,
2078 .pr_domain = &COMPATNAME(routedomain),
2079 .pr_flags = PR_ATOMIC|PR_ADDR,
2080 .pr_input = raw_input,
2081 .pr_ctlinput = raw_ctlinput,
2082 .pr_ctloutput = route_ctloutput,
2083 .pr_usrreqs = &route_usrreqs,
2084 .pr_init = raw_init,
2085 },
2086 };
2087
2088 struct domain COMPATNAME(routedomain) = {
2089 .dom_family = PF_XROUTE,
2090 .dom_name = DOMAINNAME,
2091 .dom_init = COMPATNAME(route_init),
2092 .dom_protosw = COMPATNAME(route_protosw),
2093 .dom_protoswNPROTOSW =
2094 &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
2095 };
2096
2097 static void
2098 sysctl_net_route_setup(struct sysctllog **clog)
2099 {
2100 const struct sysctlnode *rnode = NULL;
2101
2102 sysctl_createv(clog, 0, NULL, &rnode,
2103 CTLFLAG_PERMANENT,
2104 CTLTYPE_NODE, DOMAINNAME,
2105 SYSCTL_DESCR("PF_ROUTE information"),
2106 NULL, 0, NULL, 0,
2107 CTL_NET, PF_XROUTE, CTL_EOL);
2108
2109 sysctl_createv(clog, 0, NULL, NULL,
2110 CTLFLAG_PERMANENT,
2111 CTLTYPE_NODE, "rtable",
2112 SYSCTL_DESCR("Routing table information"),
2113 sysctl_rtable, 0, NULL, 0,
2114 CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
2115
2116 sysctl_createv(clog, 0, &rnode, NULL,
2117 CTLFLAG_PERMANENT,
2118 CTLTYPE_STRUCT, "stats",
2119 SYSCTL_DESCR("Routing statistics"),
2120 NULL, 0, &rtstat, sizeof(rtstat),
2121 CTL_CREATE, CTL_EOL);
2122 }
2123