]> git.saurik.com Git - apple/xnu.git/blob - bsd/net/rtsock.c
projects / apple / xnu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
xnu-1699.22.73.tar.gz
[apple/xnu.git] / bsd / net / rtsock.c
1 /*
2 * Copyright (c) 2000-2011 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28 /*
29 * Copyright (c) 1988, 1991, 1993
30 * The Regents of the University of California. All rights reserved.
31 *
32 * Redistribution and use in source and binary forms, with or without
33 * modification, are permitted provided that the following conditions
34 * are met:
35 * 1. Redistributions of source code must retain the above copyright
36 * notice, this list of conditions and the following disclaimer.
37 * 2. Redistributions in binary form must reproduce the above copyright
38 * notice, this list of conditions and the following disclaimer in the
39 * documentation and/or other materials provided with the distribution.
40 * 3. All advertising materials mentioning features or use of this software
41 * must display the following acknowledgement:
42 * This product includes software developed by the University of
43 * California, Berkeley and its contributors.
44 * 4. 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 8.5 (Berkeley) 11/2/94
61 */
62
63
64 #include <sys/param.h>
65 #include <sys/systm.h>
66 #include <sys/kernel.h>
67 #include <sys/sysctl.h>
68 #include <sys/proc.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/socket.h>
72 #include <sys/socketvar.h>
73 #include <sys/domain.h>
74 #include <sys/protosw.h>
75 #include <sys/syslog.h>
76 #include <sys/mcache.h>
77 #include <kern/lock.h>
78
79 #include <net/if.h>
80 #include <net/route.h>
81 #include <net/dlil.h>
82 #include <net/raw_cb.h>
83 #include <netinet/in.h>
84 #include <netinet/in_var.h>
85 #include <netinet/in_arp.h>
86 #include <netinet6/nd6.h>
87
88 #include <machine/spl.h>
89
90 extern struct rtstat rtstat;
91 extern int check_routeselfref;
92 extern struct domain routedomain;
93
94 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
95
96 static struct sockaddr route_dst = { 2, PF_ROUTE, { 0, } };
97 static struct sockaddr route_src = { 2, PF_ROUTE, { 0, } };
98 static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, { 0, } };
99
100 struct walkarg {
101 int w_tmemsize;
102 int w_op, w_arg;
103 caddr_t w_tmem;
104 struct sysctl_req *w_req;
105 };
106
107 static struct mbuf *rt_msg1(int, struct rt_addrinfo *);
108 static int rt_msg2(int, struct rt_addrinfo *, caddr_t, struct walkarg *);
109 static int rt_xaddrs(caddr_t, caddr_t, struct rt_addrinfo *);
110 static int sysctl_dumpentry(struct radix_node *rn, void *vw);
111 static int sysctl_dumpentry_ext(struct radix_node *rn, void *vw);
112 static int sysctl_iflist(int af, struct walkarg *w);
113 static int sysctl_iflist2(int af, struct walkarg *w);
114 static int route_output(struct mbuf *, struct socket *);
115 static void rt_setmetrics(u_int32_t, struct rt_metrics *, struct rtentry *);
116 static void rt_getmetrics(struct rtentry *, struct rt_metrics *);
117 static void rt_setif(struct rtentry *, struct sockaddr *, struct sockaddr *,
118 struct sockaddr *, unsigned int);
119 static void rt_drainall(void);
120
121 #define SIN(sa) ((struct sockaddr_in *)(size_t)(sa))
122
123
124 SYSCTL_NODE(_net, OID_AUTO, idle, CTLFLAG_RW, 0, "idle network monitoring");
125
126 static struct timeval last_ts;
127
128 SYSCTL_NODE(_net_idle, OID_AUTO, route, CTLFLAG_RW, 0, "idle route monitoring");
129
130 static int rt_if_idle_drain_interval = RT_IF_IDLE_DRAIN_INTERVAL;
131 SYSCTL_INT(_net_idle_route, OID_AUTO, drain_interval, CTLFLAG_RW,
132 &rt_if_idle_drain_interval, 0, "Default interval for draining "
133 "routes when doing interface idle reference counting.");
134
135 /*
136 * This macro calculates skew in wall clock, just in case the user changes the
137 * system time. This skew adjustment is required because we now keep the route
138 * expiration times in uptime terms in the kernel, but the userland still
139 * expects expiration times in terms of calendar times.
140 */
141 #define CALCULATE_CLOCKSKEW(cc, ic, cu, iu)\
142 ((cc.tv_sec - ic) - (cu - iu))
143
144 /*
145 * It really doesn't make any sense at all for this code to share much
146 * with raw_usrreq.c, since its functionality is so restricted. XXX
147 */
148 static int
149 rts_abort(struct socket *so)
150 {
151 int error;
152
153 error = raw_usrreqs.pru_abort(so);
154 return error;
155 }
156
157 /* pru_accept is EOPNOTSUPP */
158
159 static int
160 rts_attach(struct socket *so, int proto, __unused struct proc *p)
161 {
162 struct rawcb *rp;
163 int error;
164
165 if (sotorawcb(so) != 0)
166 return EISCONN; /* XXX panic? */
167 MALLOC(rp, struct rawcb *, sizeof *rp, M_PCB, M_WAITOK); /* XXX */
168 if (rp == 0)
169 return ENOBUFS;
170 bzero(rp, sizeof *rp);
171
172 /*
173 * The splnet() is necessary to block protocols from sending
174 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
175 * this PCB is extant but incompletely initialized.
176 * Probably we should try to do more of this work beforehand and
177 * eliminate the spl.
178 */
179 so->so_pcb = (caddr_t)rp;
180 error = raw_attach(so, proto); /* don't use raw_usrreqs.pru_attach, it checks for SS_PRIV */
181 rp = sotorawcb(so);
182 if (error) {
183 FREE(rp, M_PCB);
184 so->so_pcb = NULL;
185 so->so_flags |= SOF_PCBCLEARING;
186 return error;
187 }
188
189 switch(rp->rcb_proto.sp_protocol) {
190 //####LD route_cb needs looking
191 case AF_INET:
192 route_cb.ip_count++;
193 break;
194 case AF_INET6:
195 route_cb.ip6_count++;
196 break;
197 case AF_IPX:
198 route_cb.ipx_count++;
199 break;
200 case AF_NS:
201 route_cb.ns_count++;
202 break;
203 }
204 rp->rcb_faddr = &route_src;
205 route_cb.any_count++;
206 /* the socket is already locked when we enter rts_attach */
207 soisconnected(so);
208 so->so_options |= SO_USELOOPBACK;
209 return 0;
210 }
211
212 static int
213 rts_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
214 {
215 int error;
216 error = raw_usrreqs.pru_bind(so, nam, p); /* xxx just EINVAL */
217 return error;
218 }
219
220 static int
221 rts_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
222 {
223 int error;
224 error = raw_usrreqs.pru_connect(so, nam, p); /* XXX just EINVAL */
225 return error;
226 }
227
228 /* pru_connect2 is EOPNOTSUPP */
229 /* pru_control is EOPNOTSUPP */
230
231 static int
232 rts_detach(struct socket *so)
233 {
234 struct rawcb *rp = sotorawcb(so);
235 int error;
236
237 if (rp != 0) {
238 switch(rp->rcb_proto.sp_protocol) {
239 case AF_INET:
240 route_cb.ip_count--;
241 break;
242 case AF_INET6:
243 route_cb.ip6_count--;
244 break;
245 case AF_IPX:
246 route_cb.ipx_count--;
247 break;
248 case AF_NS:
249 route_cb.ns_count--;
250 break;
251 }
252 route_cb.any_count--;
253 }
254 error = raw_usrreqs.pru_detach(so);
255 return error;
256 }
257
258 static int
259 rts_disconnect(struct socket *so)
260 {
261 int error;
262 error = raw_usrreqs.pru_disconnect(so);
263 return error;
264 }
265
266 /* pru_listen is EOPNOTSUPP */
267
268 static int
269 rts_peeraddr(struct socket *so, struct sockaddr **nam)
270 {
271 int error;
272 error = raw_usrreqs.pru_peeraddr(so, nam);
273 return error;
274 }
275
276 /* pru_rcvd is EOPNOTSUPP */
277 /* pru_rcvoob is EOPNOTSUPP */
278
279 static int
280 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
281 struct mbuf *control, struct proc *p)
282 {
283 int error;
284 error = raw_usrreqs.pru_send(so, flags, m, nam, control, p);
285 return error;
286 }
287
288 /* pru_sense is null */
289
290 static int
291 rts_shutdown(struct socket *so)
292 {
293 int error;
294 error = raw_usrreqs.pru_shutdown(so);
295 return error;
296 }
297
298 static int
299 rts_sockaddr(struct socket *so, struct sockaddr **nam)
300 {
301 int error;
302 error = raw_usrreqs.pru_sockaddr(so, nam);
303 return error;
304 }
305
306 static struct pr_usrreqs route_usrreqs = {
307 rts_abort, pru_accept_notsupp, rts_attach, rts_bind,
308 rts_connect, pru_connect2_notsupp, pru_control_notsupp,
309 rts_detach, rts_disconnect, pru_listen_notsupp, rts_peeraddr,
310 pru_rcvd_notsupp, pru_rcvoob_notsupp, rts_send, pru_sense_null,
311 rts_shutdown, rts_sockaddr, sosend, soreceive, pru_sopoll_notsupp
312 };
313
314 /*ARGSUSED*/
315 static int
316 route_output(struct mbuf *m, struct socket *so)
317 {
318 struct rt_msghdr *rtm = NULL;
319 struct rtentry *rt = NULL;
320 struct rtentry *saved_nrt = NULL;
321 struct radix_node_head *rnh;
322 struct rt_addrinfo info;
323 int len, error = 0;
324 sa_family_t dst_sa_family = 0;
325 struct ifnet *ifp = NULL;
326 #ifndef __APPLE__
327 struct proc *curproc = current_proc();
328 #endif
329 struct sockaddr_in dst_in, gate_in;
330 int sendonlytoself = 0;
331 unsigned int ifscope = IFSCOPE_NONE;
332
333 #define senderr(e) { error = e; goto flush;}
334 if (m == NULL ||
335 ((m->m_len < sizeof(intptr_t)) && (m = m_pullup(m, sizeof(intptr_t))) == 0))
336 return (ENOBUFS);
337 if ((m->m_flags & M_PKTHDR) == 0)
338 panic("route_output");
339
340 /* unlock the socket (but keep a reference) it won't be accessed until raw_input appends to it. */
341 socket_unlock(so, 0);
342 lck_mtx_lock(rnh_lock);
343
344 len = m->m_pkthdr.len;
345 if (len < sizeof(*rtm) ||
346 len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
347 info.rti_info[RTAX_DST] = NULL;
348 senderr(EINVAL);
349 }
350 R_Malloc(rtm, struct rt_msghdr *, len);
351 if (rtm == NULL) {
352 info.rti_info[RTAX_DST] = NULL;
353 senderr(ENOBUFS);
354 }
355 m_copydata(m, 0, len, (caddr_t)rtm);
356 if (rtm->rtm_version != RTM_VERSION) {
357 info.rti_info[RTAX_DST] = NULL;
358 senderr(EPROTONOSUPPORT);
359 }
360
361 /*
362 * Silent version of RTM_GET for Reachabiltiy APIs. We may change
363 * all RTM_GETs to be silent in the future, so this is private for now.
364 */
365 if (rtm->rtm_type == RTM_GET_SILENT) {
366 if ((so->so_options & SO_USELOOPBACK) == 0)
367 senderr(EINVAL);
368 sendonlytoself = 1;
369 rtm->rtm_type = RTM_GET;
370 }
371
372 /*
373 * Perform permission checking, only privileged sockets
374 * may perform operations other than RTM_GET
375 */
376 if (rtm->rtm_type != RTM_GET && (so->so_state & SS_PRIV) == 0) {
377 info.rti_info[RTAX_DST] = NULL;
378 senderr(EPERM);
379 }
380
381 rtm->rtm_pid = proc_selfpid();
382 info.rti_addrs = rtm->rtm_addrs;
383 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
384 info.rti_info[RTAX_DST] = NULL;
385 senderr(EINVAL);
386 }
387 if (info.rti_info[RTAX_DST] == NULL || (info.rti_info[RTAX_DST]->sa_family >= AF_MAX) ||
388 (info.rti_info[RTAX_GATEWAY] != NULL && (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))) {
389 senderr(EINVAL);
390 }
391
392 if (info.rti_info[RTAX_DST]->sa_family == AF_INET && info.rti_info[RTAX_DST]->sa_len != sizeof (dst_in)) {
393 /* At minimum, we need up to sin_addr */
394 if (info.rti_info[RTAX_DST]->sa_len < offsetof(struct sockaddr_in, sin_zero))
395 senderr(EINVAL);
396 bzero(&dst_in, sizeof (dst_in));
397 dst_in.sin_len = sizeof (dst_in);
398 dst_in.sin_family = AF_INET;
399 dst_in.sin_port = SIN(info.rti_info[RTAX_DST])->sin_port;
400 dst_in.sin_addr = SIN(info.rti_info[RTAX_DST])->sin_addr;
401 info.rti_info[RTAX_DST] = (struct sockaddr *)&dst_in;
402 dst_sa_family = info.rti_info[RTAX_DST]->sa_family;
403 }
404
405 if (info.rti_info[RTAX_GATEWAY] != NULL &&
406 info.rti_info[RTAX_GATEWAY]->sa_family == AF_INET && info.rti_info[RTAX_GATEWAY]->sa_len != sizeof (gate_in)) {
407 /* At minimum, we need up to sin_addr */
408 if (info.rti_info[RTAX_GATEWAY]->sa_len < offsetof(struct sockaddr_in, sin_zero))
409 senderr(EINVAL);
410 bzero(&gate_in, sizeof (gate_in));
411 gate_in.sin_len = sizeof (gate_in);
412 gate_in.sin_family = AF_INET;
413 gate_in.sin_port = SIN(info.rti_info[RTAX_GATEWAY])->sin_port;
414 gate_in.sin_addr = SIN(info.rti_info[RTAX_GATEWAY])->sin_addr;
415 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gate_in;
416 }
417
418 if (info.rti_info[RTAX_GENMASK]) {
419 struct radix_node *t;
420 t = rn_addmask((caddr_t)info.rti_info[RTAX_GENMASK], 0, 1);
421 if (t && Bcmp(info.rti_info[RTAX_GENMASK], t->rn_key, *(u_char *)info.rti_info[RTAX_GENMASK]) == 0)
422 info.rti_info[RTAX_GENMASK] = (struct sockaddr *)(t->rn_key);
423 else
424 senderr(ENOBUFS);
425 }
426
427 /*
428 * If RTF_IFSCOPE flag is set, then rtm_index specifies the scope.
429 */
430 if (rtm->rtm_flags & RTF_IFSCOPE) {
431 if (info.rti_info[RTAX_DST]->sa_family != AF_INET && info.rti_info[RTAX_DST]->sa_family != AF_INET6)
432 senderr(EINVAL);
433 ifscope = rtm->rtm_index;
434 }
435
436 /*
437 * For AF_INET, always zero out the embedded scope ID. If this is
438 * a scoped request, it must be done explicitly by setting RTF_IFSCOPE
439 * flag and the corresponding rtm_index value. This is to prevent
440 * false interpretation of the scope ID because it's using the sin_zero
441 * field, which might not be properly cleared by the requestor.
442 */
443 if (info.rti_info[RTAX_DST]->sa_family == AF_INET)
444 sin_set_ifscope(info.rti_info[RTAX_DST], IFSCOPE_NONE);
445 if (info.rti_info[RTAX_GATEWAY] != NULL && info.rti_info[RTAX_GATEWAY]->sa_family == AF_INET)
446 sin_set_ifscope(info.rti_info[RTAX_GATEWAY], IFSCOPE_NONE);
447
448 switch (rtm->rtm_type) {
449
450 case RTM_ADD:
451 if (info.rti_info[RTAX_GATEWAY] == NULL)
452 senderr(EINVAL);
453
454 #ifdef __APPLE__
455 /* XXX LD11JUL02 Special case for AOL 5.1.2 connectivity issue to AirPort BS (Radar 2969954)
456 * AOL is adding a circular route ("10.0.1.1/32 10.0.1.1") when establishing its ppp tunnel
457 * to the AP BaseStation by removing the default gateway and replacing it with their tunnel entry point.
458 * There is no apparent reason to add this route as there is a valid 10.0.1.1/24 route to the BS.
459 * That circular route was ignored on previous version of MacOS X because of a routing bug
460 * corrected with the merge to FreeBSD4.4 (a route generated from an RTF_CLONING route had the RTF_WASCLONED
461 * flag set but did not have a reference to the parent route) and that entry was left in the RT. This workaround is
462 * made in order to provide binary compatibility with AOL.
463 * If we catch a process adding a circular route with a /32 from the routing socket, we error it out instead of
464 * confusing the routing table with a wrong route to the previous default gateway
465 */
466 {
467 #define satosinaddr(sa) (((struct sockaddr_in *)sa)->sin_addr.s_addr)
468
469 if (check_routeselfref && (info.rti_info[RTAX_DST] && info.rti_info[RTAX_DST]->sa_family == AF_INET) &&
470 (info.rti_info[RTAX_NETMASK] && satosinaddr(info.rti_info[RTAX_NETMASK]) == INADDR_BROADCAST) &&
471 (info.rti_info[RTAX_GATEWAY] && satosinaddr(info.rti_info[RTAX_DST]) == satosinaddr(info.rti_info[RTAX_GATEWAY]))) {
472 log(LOG_WARNING, "route_output: circular route %ld.%ld.%ld.%ld/32 ignored\n",
473 (ntohl(satosinaddr(info.rti_info[RTAX_GATEWAY])>>24))&0xff,
474 (ntohl(satosinaddr(info.rti_info[RTAX_GATEWAY])>>16))&0xff,
475 (ntohl(satosinaddr(info.rti_info[RTAX_GATEWAY])>>8))&0xff,
476 (ntohl(satosinaddr(info.rti_info[RTAX_GATEWAY])))&0xff);
477
478 senderr(EINVAL);
479 }
480 }
481 #endif
482 error = rtrequest_scoped_locked(RTM_ADD, info.rti_info[RTAX_DST], info.rti_info[RTAX_GATEWAY],
483 info.rti_info[RTAX_NETMASK], rtm->rtm_flags, &saved_nrt, ifscope);
484 if (error == 0 && saved_nrt) {
485 RT_LOCK(saved_nrt);
486 #ifdef __APPLE__
487 /*
488 * If the route request specified an interface with
489 * IFA and/or IFP, we set the requested interface on
490 * the route with rt_setif. It would be much better
491 * to do this inside rtrequest, but that would
492 * require passing the desired interface, in some
493 * form, to rtrequest. Since rtrequest is called in
494 * so many places (roughly 40 in our source), adding
495 * a parameter is to much for us to swallow; this is
496 * something for the FreeBSD developers to tackle.
497 * Instead, we let rtrequest compute whatever
498 * interface it wants, then come in behind it and
499 * stick in the interface that we really want. This
500 * works reasonably well except when rtrequest can't
501 * figure out what interface to use (with
502 * ifa_withroute) and returns ENETUNREACH. Ideally
503 * it shouldn't matter if rtrequest can't figure out
504 * the interface if we're going to explicitly set it
505 * ourselves anyway. But practically we can't
506 * recover here because rtrequest will not do any of
507 * the work necessary to add the route if it can't
508 * find an interface. As long as there is a default
509 * route that leads to some interface, rtrequest will
510 * find an interface, so this problem should be
511 * rarely encountered.
512 * dwiggins@bbn.com
513 */
514
515 rt_setif(saved_nrt, info.rti_info[RTAX_IFP], info.rti_info[RTAX_IFA], info.rti_info[RTAX_GATEWAY],
516 ifscope);
517 #endif
518 rt_setmetrics(rtm->rtm_inits,
519 &rtm->rtm_rmx, saved_nrt);
520 saved_nrt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
521 saved_nrt->rt_rmx.rmx_locks |=
522 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
523 saved_nrt->rt_genmask = info.rti_info[RTAX_GENMASK];
524 RT_REMREF_LOCKED(saved_nrt);
525 RT_UNLOCK(saved_nrt);
526 }
527 break;
528
529 case RTM_DELETE:
530 error = rtrequest_scoped_locked(RTM_DELETE, info.rti_info[RTAX_DST],
531 info.rti_info[RTAX_GATEWAY], info.rti_info[RTAX_NETMASK], rtm->rtm_flags, &saved_nrt, ifscope);
532 if (error == 0) {
533 rt = saved_nrt;
534 RT_LOCK(rt);
535 goto report;
536 }
537 break;
538
539 case RTM_GET:
540 case RTM_CHANGE:
541 case RTM_LOCK:
542 if ((rnh = rt_tables[info.rti_info[RTAX_DST]->sa_family]) == NULL)
543 senderr(EAFNOSUPPORT);
544
545 /*
546 * Lookup the best match based on the key-mask pair;
547 * callee adds a reference and checks for root node.
548 */
549 rt = rt_lookup(TRUE, info.rti_info[RTAX_DST], info.rti_info[RTAX_NETMASK], rnh, ifscope);
550 if (rt == NULL)
551 senderr(ESRCH);
552 RT_LOCK(rt);
553
554 /*
555 * Holding rnh_lock here prevents the possibility of
556 * ifa from changing (e.g. in_ifinit), so it is safe
557 * to access its ifa_addr (down below) without locking.
558 */
559 switch(rtm->rtm_type) {
560
561 case RTM_GET: {
562 struct ifaddr *ifa2;
563 report:
564 ifa2 = NULL;
565 RT_LOCK_ASSERT_HELD(rt);
566 info.rti_info[RTAX_DST] = rt_key(rt);
567 dst_sa_family = info.rti_info[RTAX_DST]->sa_family;
568 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
569 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
570 info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
571 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
572 ifp = rt->rt_ifp;
573 if (ifp) {
574 ifnet_lock_shared(ifp);
575 ifa2 = ifp->if_lladdr;
576 info.rti_info[RTAX_IFP] = ifa2->ifa_addr;
577 IFA_ADDREF(ifa2);
578 ifnet_lock_done(ifp);
579 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
580 rtm->rtm_index = ifp->if_index;
581 } else {
582 info.rti_info[RTAX_IFP] = NULL;
583 info.rti_info[RTAX_IFA] = NULL;
584 }
585 } else if ((ifp = rt->rt_ifp) != NULL) {
586 rtm->rtm_index = ifp->if_index;
587 }
588 if (ifa2 != NULL)
589 IFA_LOCK(ifa2);
590 len = rt_msg2(rtm->rtm_type, &info, (caddr_t)0,
591 (struct walkarg *)0);
592 if (ifa2 != NULL)
593 IFA_UNLOCK(ifa2);
594 if (len > rtm->rtm_msglen) {
595 struct rt_msghdr *new_rtm;
596 R_Malloc(new_rtm, struct rt_msghdr *, len);
597 if (new_rtm == 0) {
598 RT_UNLOCK(rt);
599 if (ifa2 != NULL)
600 IFA_REMREF(ifa2);
601 senderr(ENOBUFS);
602 }
603 Bcopy(rtm, new_rtm, rtm->rtm_msglen);
604 R_Free(rtm); rtm = new_rtm;
605 }
606 if (ifa2 != NULL)
607 IFA_LOCK(ifa2);
608 (void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm,
609 (struct walkarg *)0);
610 if (ifa2 != NULL)
611 IFA_UNLOCK(ifa2);
612 rtm->rtm_flags = rt->rt_flags;
613 rt_getmetrics(rt, &rtm->rtm_rmx);
614 rtm->rtm_addrs = info.rti_addrs;
615 if (ifa2 != NULL)
616 IFA_REMREF(ifa2);
617 }
618 break;
619
620 case RTM_CHANGE:
621 if (info.rti_info[RTAX_GATEWAY] && (error = rt_setgate(rt,
622 rt_key(rt), info.rti_info[RTAX_GATEWAY]))) {
623 RT_UNLOCK(rt);
624 senderr(error);
625 }
626 /*
627 * If they tried to change things but didn't specify
628 * the required gateway, then just use the old one.
629 * This can happen if the user tries to change the
630 * flags on the default route without changing the
631 * default gateway. Changing flags still doesn't work.
632 */
633 if ((rt->rt_flags & RTF_GATEWAY) && !info.rti_info[RTAX_GATEWAY])
634 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
635
636 #ifdef __APPLE__
637 /*
638 * On Darwin, we call rt_setif which contains the
639 * equivalent to the code found at this very spot
640 * in BSD.
641 */
642 rt_setif(rt, info.rti_info[RTAX_IFP], info.rti_info[RTAX_IFA], info.rti_info[RTAX_GATEWAY],
643 ifscope);
644 #endif
645
646 rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
647 rt);
648 #ifndef __APPLE__
649 /* rt_setif, called above does this for us on darwin */
650 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
651 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info.rti_info[RTAX_GATEWAY]);
652 #endif
653 if (info.rti_info[RTAX_GENMASK])
654 rt->rt_genmask = info.rti_info[RTAX_GENMASK];
655 /*
656 * Fall into
657 */
658 case RTM_LOCK:
659 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
660 rt->rt_rmx.rmx_locks |=
661 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
662 break;
663 }
664 RT_UNLOCK(rt);
665 break;
666
667 default:
668 senderr(EOPNOTSUPP);
669 }
670 flush:
671 if (rtm) {
672 if (error)
673 rtm->rtm_errno = error;
674 else
675 rtm->rtm_flags |= RTF_DONE;
676 }
677 if (rt != NULL) {
678 RT_LOCK_ASSERT_NOTHELD(rt);
679 rtfree_locked(rt);
680 }
681 lck_mtx_unlock(rnh_lock);
682 socket_lock(so, 0); /* relock the socket now */
683 {
684 struct rawcb *rp = 0;
685 /*
686 * Check to see if we don't want our own messages.
687 */
688 if ((so->so_options & SO_USELOOPBACK) == 0) {
689 if (route_cb.any_count <= 1) {
690 if (rtm)
691 R_Free(rtm);
692 m_freem(m);
693 return (error);
694 }
695 /* There is another listener, so construct message */
696 rp = sotorawcb(so);
697 }
698 if (rtm) {
699 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
700 if (m->m_pkthdr.len < rtm->rtm_msglen) {
701 m_freem(m);
702 m = NULL;
703 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
704 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
705 R_Free(rtm);
706 }
707 if (sendonlytoself && m) {
708 error = 0;
709 if (sbappendaddr(&so->so_rcv, &route_src, m, (struct mbuf*)0, &error) != 0) {
710 sorwakeup(so);
711 }
712 if (error)
713 return error;
714 } else {
715 struct sockproto route_proto = {PF_ROUTE, 0};
716 if (rp)
717 rp->rcb_proto.sp_family = 0; /* Avoid us */
718 if (dst_sa_family != 0)
719 route_proto.sp_protocol = dst_sa_family;
720 if (m) {
721 socket_unlock(so, 0);
722 raw_input(m, &route_proto, &route_src, &route_dst);
723 socket_lock(so, 0);
724 }
725 if (rp)
726 rp->rcb_proto.sp_family = PF_ROUTE;
727 }
728 }
729 return (error);
730 }
731
732 void
733 rt_setexpire(struct rtentry *rt, uint64_t expiry)
734 {
735 /* set both rt_expire and rmx_expire */
736 rt->rt_expire = expiry;
737 if (expiry) {
738 rt->rt_rmx.rmx_expire = expiry + rt->base_calendartime -
739 rt->base_uptime;
740 } else
741 rt->rt_rmx.rmx_expire = 0;
742 }
743
744 static void
745 rt_setmetrics(u_int32_t which, struct rt_metrics *in, struct rtentry *out)
746 {
747 struct timeval curr_calendar_time;
748 uint64_t curr_uptime;
749
750 getmicrotime(&curr_calendar_time);
751 curr_uptime = net_uptime();
752
753 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->e;
754 metric(RTV_RPIPE, rmx_recvpipe);
755 metric(RTV_SPIPE, rmx_sendpipe);
756 metric(RTV_SSTHRESH, rmx_ssthresh);
757 metric(RTV_RTT, rmx_rtt);
758 metric(RTV_RTTVAR, rmx_rttvar);
759 metric(RTV_HOPCOUNT, rmx_hopcount);
760 metric(RTV_MTU, rmx_mtu);
761 metric(RTV_EXPIRE, rmx_expire);
762 #undef metric
763
764 if (out->rt_rmx.rmx_expire > 0) {
765 /* account for system time change */
766 curr_uptime = net_uptime();
767 getmicrotime(&curr_calendar_time);
768 out->base_calendartime +=
769 CALCULATE_CLOCKSKEW(curr_calendar_time,
770 out->base_calendartime,
771 curr_uptime, out->base_uptime);
772 rt_setexpire(out,
773 out->rt_rmx.rmx_expire -
774 out->base_calendartime +
775 out->base_uptime);
776 } else {
777 rt_setexpire(out, 0);
778 }
779
780 VERIFY(out->rt_expire == 0 || out->rt_rmx.rmx_expire != 0);
781 VERIFY(out->rt_expire != 0 || out->rt_rmx.rmx_expire == 0);
782 }
783
784 static void
785 rt_getmetrics(struct rtentry *in, struct rt_metrics *out)
786 {
787 struct timeval curr_calendar_time;
788 uint64_t curr_uptime;
789
790 VERIFY(in->rt_expire == 0 || in->rt_rmx.rmx_expire != 0);
791 VERIFY(in->rt_expire != 0 || in->rt_rmx.rmx_expire == 0);
792
793 *out = in->rt_rmx;
794
795 if (in->rt_expire) {
796 /* account for system time change */
797 getmicrotime(&curr_calendar_time);
798 curr_uptime = net_uptime();
799
800 in->base_calendartime +=
801 CALCULATE_CLOCKSKEW(curr_calendar_time,
802 in->base_calendartime,
803 curr_uptime, in->base_uptime);
804
805 out->rmx_expire = in->base_calendartime +
806 in->rt_expire - in->base_uptime;
807 } else
808 out->rmx_expire = 0;
809 }
810
811 /*
812 * Set route's interface given info.rti_info[RTAX_IFP], info.rti_info[RTAX_IFA], and gateway.
813 */
814 static void
815 rt_setif(struct rtentry *rt, struct sockaddr *Ifpaddr, struct sockaddr *Ifaaddr,
816 struct sockaddr *Gate, unsigned int ifscope)
817 {
818 struct ifaddr *ifa = NULL;
819 struct ifnet *ifp = NULL;
820 void (*ifa_rtrequest)
821 (int, struct rtentry *, struct sockaddr *);
822
823 lck_mtx_assert(rnh_lock, LCK_MTX_ASSERT_OWNED);
824
825 RT_LOCK_ASSERT_HELD(rt);
826
827 /* trigger route cache reevaluation */
828 if (use_routegenid)
829 routegenid_update();
830
831 /* Don't update a defunct route */
832 if (rt->rt_flags & RTF_CONDEMNED)
833 return;
834
835 /* Add an extra ref for ourselves */
836 RT_ADDREF_LOCKED(rt);
837
838 /* Become a regular mutex, just in case */
839 RT_CONVERT_LOCK(rt);
840
841 /*
842 * New gateway could require new ifaddr, ifp; flags may also
843 * be different; ifp may be specified by ll sockaddr when
844 * protocol address is ambiguous.
845 */
846 if (Ifpaddr && (ifa = ifa_ifwithnet_scoped(Ifpaddr, ifscope)) &&
847 (ifp = ifa->ifa_ifp) && (Ifaaddr || Gate)) {
848 IFA_REMREF(ifa);
849 ifa = ifaof_ifpforaddr(Ifaaddr ? Ifaaddr : Gate, ifp);
850 } else {
851 if (ifa) {
852 IFA_REMREF(ifa);
853 ifa = 0;
854 }
855 if (Ifpaddr && (ifp = if_withname(Ifpaddr)) ) {
856 if (Gate) {
857 ifa = ifaof_ifpforaddr(Gate, ifp);
858 } else {
859 ifnet_lock_shared(ifp);
860 ifa = TAILQ_FIRST(&ifp->if_addrhead);
861 if (ifa != NULL)
862 IFA_ADDREF(ifa);
863 ifnet_lock_done(ifp);
864 }
865 } else if (Ifaaddr &&
866 (ifa = ifa_ifwithaddr_scoped(Ifaaddr, ifscope))) {
867 ifp = ifa->ifa_ifp;
868 } else if (Gate != NULL) {
869 /*
870 * Safe to drop rt_lock and use rt_key, since holding
871 * rnh_lock here prevents another thread from calling
872 * rt_setgate() on this route. We cannot hold the
873 * lock across ifa_ifwithroute since the lookup done
874 * by that routine may point to the same route.
875 */
876 RT_UNLOCK(rt);
877 if ((ifa = ifa_ifwithroute_scoped_locked(rt->rt_flags,
878 rt_key(rt), Gate, ifscope)) != NULL)
879 ifp = ifa->ifa_ifp;
880 RT_LOCK(rt);
881 /* Don't update a defunct route */
882 if (rt->rt_flags & RTF_CONDEMNED) {
883 if (ifa != NULL)
884 IFA_REMREF(ifa);
885 /* Release extra ref */
886 RT_REMREF_LOCKED(rt);
887 return;
888 }
889 }
890 }
891 if (ifa) {
892 struct ifaddr *oifa = rt->rt_ifa;
893 if (oifa != ifa) {
894 if (oifa != NULL) {
895 IFA_LOCK_SPIN(oifa);
896 ifa_rtrequest = oifa->ifa_rtrequest;
897 IFA_UNLOCK(oifa);
898 if (ifa_rtrequest != NULL)
899 ifa_rtrequest(RTM_DELETE, rt, Gate);
900 }
901 rtsetifa(rt, ifa);
902
903 if (rt->rt_ifp != ifp) {
904 /*
905 * Purge any link-layer info caching.
906 */
907 if (rt->rt_llinfo_purge != NULL)
908 rt->rt_llinfo_purge(rt);
909
910 /*
911 * Adjust route ref count for the interfaces.
912 */
913 if (rt->rt_if_ref_fn != NULL) {
914 rt->rt_if_ref_fn(ifp, 1);
915 rt->rt_if_ref_fn(rt->rt_ifp, -1);
916 }
917 }
918 rt->rt_ifp = ifp;
919 /*
920 * If this is the (non-scoped) default route, record
921 * the interface index used for the primary ifscope.
922 */
923 if (rt_primary_default(rt, rt_key(rt))) {
924 set_primary_ifscope(rt_key(rt)->sa_family,
925 rt->rt_ifp->if_index);
926 }
927 rt->rt_rmx.rmx_mtu = ifp->if_mtu;
928 if (rt->rt_ifa != NULL) {
929 IFA_LOCK_SPIN(rt->rt_ifa);
930 ifa_rtrequest = rt->rt_ifa->ifa_rtrequest;
931 IFA_UNLOCK(rt->rt_ifa);
932 if (ifa_rtrequest != NULL)
933 ifa_rtrequest(RTM_ADD, rt, Gate);
934 }
935 IFA_REMREF(ifa);
936 /* Release extra ref */
937 RT_REMREF_LOCKED(rt);
938 return;
939 }
940 IFA_REMREF(ifa);
941 }
942
943 /* XXX: to reset gateway to correct value, at RTM_CHANGE */
944 if (rt->rt_ifa != NULL) {
945 IFA_LOCK_SPIN(rt->rt_ifa);
946 ifa_rtrequest = rt->rt_ifa->ifa_rtrequest;
947 IFA_UNLOCK(rt->rt_ifa);
948 if (ifa_rtrequest != NULL)
949 ifa_rtrequest(RTM_ADD, rt, Gate);
950 }
951
952 /* Release extra ref */
953 RT_REMREF_LOCKED(rt);
954 }
955
956 #define ROUNDUP32(a) \
957 ((a) > 0 ? (1 + (((a) - 1) | (sizeof(uint32_t) - 1))) : sizeof(uint32_t))
958 #define ADVANCE32(x, n) (x += ROUNDUP32((n)->sa_len))
959
960
961 /*
962 * Extract the addresses of the passed sockaddrs.
963 * Do a little sanity checking so as to avoid bad memory references.
964 * This data is derived straight from userland.
965 */
966 static int
967 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
968 {
969 struct sockaddr *sa;
970 int i;
971
972 bzero(rtinfo->rti_info, sizeof(rtinfo->rti_info));
973 for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
974 if ((rtinfo->rti_addrs & (1 << i)) == 0)
975 continue;
976 sa = (struct sockaddr *)cp;
977 /*
978 * It won't fit.
979 */
980 if ( (cp + sa->sa_len) > cplim ) {
981 return (EINVAL);
982 }
983
984 /*
985 * there are no more.. quit now
986 * If there are more bits, they are in error.
987 * I've seen this. route(1) can evidently generate these.
988 * This causes kernel to core dump.
989 * for compatibility, If we see this, point to a safe address.
990 */
991 if (sa->sa_len == 0) {
992 rtinfo->rti_info[i] = &sa_zero;
993 return (0); /* should be EINVAL but for compat */
994 }
995
996 /* accept it */
997 rtinfo->rti_info[i] = sa;
998 ADVANCE32(cp, sa);
999 }
1000 return (0);
1001 }
1002
1003 static struct mbuf *
1004 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1005 {
1006 struct rt_msghdr *rtm;
1007 struct mbuf *m;
1008 int i;
1009 int len, dlen;
1010
1011 switch (type) {
1012
1013 case RTM_DELADDR:
1014 case RTM_NEWADDR:
1015 len = sizeof(struct ifa_msghdr);
1016 break;
1017
1018 case RTM_DELMADDR:
1019 case RTM_NEWMADDR:
1020 len = sizeof(struct ifma_msghdr);
1021 break;
1022
1023 case RTM_IFINFO:
1024 len = sizeof(struct if_msghdr);
1025 break;
1026
1027 default:
1028 len = sizeof(struct rt_msghdr);
1029 }
1030 if (len > MCLBYTES)
1031 panic("rt_msg1");
1032 m = m_gethdr(M_DONTWAIT, MT_DATA);
1033 if (m && len > MHLEN) {
1034 MCLGET(m, M_DONTWAIT);
1035 if ((m->m_flags & M_EXT) == 0) {
1036 m_free(m);
1037 m = NULL;
1038 }
1039 }
1040 if (m == 0)
1041 return (m);
1042 m->m_pkthdr.len = m->m_len = len;
1043 m->m_pkthdr.rcvif = 0;
1044 rtm = mtod(m, struct rt_msghdr *);
1045 bzero((caddr_t)rtm, len);
1046 for (i = 0; i < RTAX_MAX; i++) {
1047 struct sockaddr *sa, *hint;
1048 struct sockaddr_storage ss;
1049
1050 if ((sa = rtinfo->rti_info[i]) == NULL)
1051 continue;
1052
1053 switch (i) {
1054 case RTAX_DST:
1055 case RTAX_NETMASK:
1056 if ((hint = rtinfo->rti_info[RTAX_DST]) == NULL)
1057 hint = rtinfo->rti_info[RTAX_IFA];
1058
1059 /* Scrub away any trace of embedded interface scope */
1060 sa = rtm_scrub_ifscope(type, i, hint, sa, &ss);
1061 break;
1062
1063 default:
1064 break;
1065 }
1066
1067 rtinfo->rti_addrs |= (1 << i);
1068 dlen = ROUNDUP32(sa->sa_len);
1069 m_copyback(m, len, dlen, (caddr_t)sa);
1070 len += dlen;
1071 }
1072 if (m->m_pkthdr.len != len) {
1073 m_freem(m);
1074 return (NULL);
1075 }
1076 rtm->rtm_msglen = len;
1077 rtm->rtm_version = RTM_VERSION;
1078 rtm->rtm_type = type;
1079 return (m);
1080 }
1081
1082 static int
1083 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1084 {
1085 int i;
1086 int len, dlen, second_time = 0;
1087 caddr_t cp0;
1088
1089 rtinfo->rti_addrs = 0;
1090 again:
1091 switch (type) {
1092
1093 case RTM_DELADDR:
1094 case RTM_NEWADDR:
1095 len = sizeof(struct ifa_msghdr);
1096 break;
1097
1098 case RTM_DELMADDR:
1099 case RTM_NEWMADDR:
1100 len = sizeof(struct ifma_msghdr);
1101 break;
1102
1103 case RTM_IFINFO:
1104 len = sizeof(struct if_msghdr);
1105 break;
1106
1107 case RTM_IFINFO2:
1108 len = sizeof(struct if_msghdr2);
1109 break;
1110
1111 case RTM_NEWMADDR2:
1112 len = sizeof(struct ifma_msghdr2);
1113 break;
1114
1115 case RTM_GET_EXT:
1116 len = sizeof (struct rt_msghdr_ext);
1117 break;
1118
1119 case RTM_GET2:
1120 len = sizeof(struct rt_msghdr2);
1121 break;
1122
1123 default:
1124 len = sizeof(struct rt_msghdr);
1125 }
1126 cp0 = cp;
1127 if (cp0)
1128 cp += len;
1129 for (i = 0; i < RTAX_MAX; i++) {
1130 struct sockaddr *sa, *hint;
1131 struct sockaddr_storage ss;
1132
1133 if ((sa = rtinfo->rti_info[i]) == 0)
1134 continue;
1135
1136 switch (i) {
1137 case RTAX_DST:
1138 case RTAX_NETMASK:
1139 if ((hint = rtinfo->rti_info[RTAX_DST]) == NULL)
1140 hint = rtinfo->rti_info[RTAX_IFA];
1141
1142 /* Scrub away any trace of embedded interface scope */
1143 sa = rtm_scrub_ifscope(type, i, hint, sa, &ss);
1144 break;
1145
1146 default:
1147 break;
1148 }
1149
1150 rtinfo->rti_addrs |= (1 << i);
1151 dlen = ROUNDUP32(sa->sa_len);
1152 if (cp) {
1153 bcopy((caddr_t)sa, cp, (unsigned)dlen);
1154 cp += dlen;
1155 }
1156 len += dlen;
1157 }
1158 if (cp == 0 && w != NULL && !second_time) {
1159 struct walkarg *rw = w;
1160
1161 if (rw->w_req) {
1162 if (rw->w_tmemsize < len) {
1163 if (rw->w_tmem)
1164 FREE(rw->w_tmem, M_RTABLE);
1165 rw->w_tmem = (caddr_t)
1166 _MALLOC(len, M_RTABLE, M_WAITOK); /*###LD0412 was NOWAIT */
1167 if (rw->w_tmem)
1168 rw->w_tmemsize = len;
1169 }
1170 if (rw->w_tmem) {
1171 cp = rw->w_tmem;
1172 second_time = 1;
1173 goto again;
1174 }
1175 }
1176 }
1177 if (cp) {
1178 struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1179
1180 rtm->rtm_version = RTM_VERSION;
1181 rtm->rtm_type = type;
1182 rtm->rtm_msglen = len;
1183 }
1184 return (len);
1185 }
1186
1187 /*
1188 * This routine is called to generate a message from the routing
1189 * socket indicating that a redirect has occurred, a routing lookup
1190 * has failed, or that a protocol has detected timeouts to a particular
1191 * destination.
1192 */
1193 void
1194 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1195 {
1196 struct rt_msghdr *rtm;
1197 struct mbuf *m;
1198 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1199 struct sockproto route_proto = {PF_ROUTE, 0};
1200
1201 if (route_cb.any_count == 0)
1202 return;
1203 m = rt_msg1(type, rtinfo);
1204 if (m == 0)
1205 return;
1206 rtm = mtod(m, struct rt_msghdr *);
1207 rtm->rtm_flags = RTF_DONE | flags;
1208 rtm->rtm_errno = error;
1209 rtm->rtm_addrs = rtinfo->rti_addrs;
1210 route_proto.sp_family = sa ? sa->sa_family : 0;
1211 raw_input(m, &route_proto, &route_src, &route_dst);
1212 }
1213
1214 /*
1215 * This routine is called to generate a message from the routing
1216 * socket indicating that the status of a network interface has changed.
1217 */
1218 void
1219 rt_ifmsg(
1220 struct ifnet *ifp)
1221 {
1222 struct if_msghdr *ifm;
1223 struct mbuf *m;
1224 struct rt_addrinfo info;
1225 struct sockproto route_proto = {PF_ROUTE, 0};
1226
1227 if (route_cb.any_count == 0)
1228 return;
1229 bzero((caddr_t)&info, sizeof(info));
1230 m = rt_msg1(RTM_IFINFO, &info);
1231 if (m == 0)
1232 return;
1233 ifm = mtod(m, struct if_msghdr *);
1234 ifm->ifm_index = ifp->if_index;
1235 ifm->ifm_flags = (u_short)ifp->if_flags;
1236 if_data_internal_to_if_data(ifp, &ifp->if_data, &ifm->ifm_data);
1237 ifm->ifm_addrs = 0;
1238 raw_input(m, &route_proto, &route_src, &route_dst);
1239 }
1240
1241 /*
1242 * This is called to generate messages from the routing socket
1243 * indicating a network interface has had addresses associated with it.
1244 * if we ever reverse the logic and replace messages TO the routing
1245 * socket indicate a request to configure interfaces, then it will
1246 * be unnecessary as the routing socket will automatically generate
1247 * copies of it.
1248 *
1249 * Since this is coming from the interface, it is expected that the
1250 * interface will be locked. Caller must hold rnh_lock and rt_lock.
1251 */
1252 void
1253 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1254 {
1255 struct rt_addrinfo info;
1256 struct sockaddr *sa = 0;
1257 int pass;
1258 struct mbuf *m = 0;
1259 struct ifnet *ifp = ifa->ifa_ifp;
1260 struct sockproto route_proto = {PF_ROUTE, 0};
1261
1262 lck_mtx_assert(rnh_lock, LCK_MTX_ASSERT_OWNED);
1263 RT_LOCK_ASSERT_HELD(rt);
1264
1265 if (route_cb.any_count == 0)
1266 return;
1267
1268 /* Become a regular mutex, just in case */
1269 RT_CONVERT_LOCK(rt);
1270 for (pass = 1; pass < 3; pass++) {
1271 bzero((caddr_t)&info, sizeof(info));
1272 if ((cmd == RTM_ADD && pass == 1) ||
1273 (cmd == RTM_DELETE && pass == 2)) {
1274 struct ifa_msghdr *ifam;
1275 int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1276
1277 /* Lock ifp for if_lladdr */
1278 ifnet_lock_shared(ifp);
1279 IFA_LOCK(ifa);
1280 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1281 /*
1282 * Holding ifnet lock here prevents the link address
1283 * from changing contents, so no need to hold its
1284 * lock. The link address is always present; it's
1285 * never freed.
1286 */
1287 info.rti_info[RTAX_IFP] = ifp->if_lladdr->ifa_addr;
1288 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1289 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1290 if ((m = rt_msg1(ncmd, &info)) == NULL) {
1291 IFA_UNLOCK(ifa);
1292 ifnet_lock_done(ifp);
1293 continue;
1294 }
1295 IFA_UNLOCK(ifa);
1296 ifnet_lock_done(ifp);
1297 ifam = mtod(m, struct ifa_msghdr *);
1298 ifam->ifam_index = ifp->if_index;
1299 IFA_LOCK_SPIN(ifa);
1300 ifam->ifam_metric = ifa->ifa_metric;
1301 ifam->ifam_flags = ifa->ifa_flags;
1302 IFA_UNLOCK(ifa);
1303 ifam->ifam_addrs = info.rti_addrs;
1304 }
1305 if ((cmd == RTM_ADD && pass == 2) ||
1306 (cmd == RTM_DELETE && pass == 1)) {
1307 struct rt_msghdr *rtm;
1308
1309 if (rt == 0)
1310 continue;
1311 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1312 info.rti_info[RTAX_DST] = sa = rt_key(rt);
1313 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1314 if ((m = rt_msg1(cmd, &info)) == NULL)
1315 continue;
1316 rtm = mtod(m, struct rt_msghdr *);
1317 rtm->rtm_index = ifp->if_index;
1318 rtm->rtm_flags |= rt->rt_flags;
1319 rtm->rtm_errno = error;
1320 rtm->rtm_addrs = info.rti_addrs;
1321 }
1322 route_proto.sp_protocol = sa ? sa->sa_family : 0;
1323 raw_input(m, &route_proto, &route_src, &route_dst);
1324 }
1325 }
1326
1327 /*
1328 * This is the analogue to the rt_newaddrmsg which performs the same
1329 * function but for multicast group memberhips. This is easier since
1330 * there is no route state to worry about.
1331 */
1332 void
1333 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1334 {
1335 struct rt_addrinfo info;
1336 struct mbuf *m = 0;
1337 struct ifnet *ifp = ifma->ifma_ifp;
1338 struct ifma_msghdr *ifmam;
1339 struct sockproto route_proto = {PF_ROUTE, 0};
1340
1341 if (route_cb.any_count == 0)
1342 return;
1343
1344 /* Lock ifp for if_lladdr */
1345 ifnet_lock_shared(ifp);
1346 bzero((caddr_t)&info, sizeof(info));
1347 IFMA_LOCK(ifma);
1348 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1349 info.rti_info[RTAX_IFP] = ifp->if_lladdr->ifa_addr; /* lladdr doesn't need lock */
1350
1351 /*
1352 * If a link-layer address is present, present it as a ``gateway''
1353 * (similarly to how ARP entries, e.g., are presented).
1354 */
1355 info.rti_info[RTAX_GATEWAY] = (ifma->ifma_ll != NULL) ? ifma->ifma_ll->ifma_addr : NULL;
1356 if ((m = rt_msg1(cmd, &info)) == NULL) {
1357 IFMA_UNLOCK(ifma);
1358 ifnet_lock_done(ifp);
1359 return;
1360 }
1361 ifmam = mtod(m, struct ifma_msghdr *);
1362 ifmam->ifmam_index = ifp->if_index;
1363 ifmam->ifmam_addrs = info.rti_addrs;
1364 route_proto.sp_protocol = ifma->ifma_addr->sa_family;
1365 IFMA_UNLOCK(ifma);
1366 ifnet_lock_done(ifp);
1367 raw_input(m, &route_proto, &route_src, &route_dst);
1368 }
1369
1370 /*
1371 * This is used in dumping the kernel table via sysctl().
1372 */
1373 int
1374 sysctl_dumpentry(struct radix_node *rn, void *vw)
1375 {
1376 struct walkarg *w = vw;
1377 struct rtentry *rt = (struct rtentry *)rn;
1378 int error = 0, size;
1379 struct rt_addrinfo info;
1380
1381 RT_LOCK(rt);
1382 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg)) {
1383 RT_UNLOCK(rt);
1384 return 0;
1385 }
1386 bzero((caddr_t)&info, sizeof(info));
1387 info.rti_info[RTAX_DST] = rt_key(rt);
1388 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1389 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1390 info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1391
1392 if (w->w_op != NET_RT_DUMP2) {
1393 size = rt_msg2(RTM_GET, &info, 0, w);
1394 if (w->w_req && w->w_tmem) {
1395 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1396
1397 rtm->rtm_flags = rt->rt_flags;
1398 rtm->rtm_use = rt->rt_use;
1399 rt_getmetrics(rt, &rtm->rtm_rmx);
1400 rtm->rtm_index = rt->rt_ifp->if_index;
1401 rtm->rtm_pid = 0;
1402 rtm->rtm_seq = 0;
1403 rtm->rtm_errno = 0;
1404 rtm->rtm_addrs = info.rti_addrs;
1405 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1406 RT_UNLOCK(rt);
1407 return (error);
1408 }
1409 } else {
1410 size = rt_msg2(RTM_GET2, &info, 0, w);
1411 if (w->w_req && w->w_tmem) {
1412 struct rt_msghdr2 *rtm = (struct rt_msghdr2 *)w->w_tmem;
1413
1414 rtm->rtm_flags = rt->rt_flags;
1415 rtm->rtm_use = rt->rt_use;
1416 rt_getmetrics(rt, &rtm->rtm_rmx);
1417 rtm->rtm_index = rt->rt_ifp->if_index;
1418 rtm->rtm_refcnt = rt->rt_refcnt;
1419 if (rt->rt_parent)
1420 rtm->rtm_parentflags = rt->rt_parent->rt_flags;
1421 else
1422 rtm->rtm_parentflags = 0;
1423 rtm->rtm_reserved = 0;
1424 rtm->rtm_addrs = info.rti_addrs;
1425 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1426 RT_UNLOCK(rt);
1427 return (error);
1428 }
1429 }
1430 RT_UNLOCK(rt);
1431 return (error);
1432 }
1433
1434 /*
1435 * This is used for dumping extended information from route entries.
1436 */
1437 int
1438 sysctl_dumpentry_ext(struct radix_node *rn, void *vw)
1439 {
1440 struct walkarg *w = vw;
1441 struct rtentry *rt = (struct rtentry *)rn;
1442 int error = 0, size;
1443 struct rt_addrinfo info;
1444
1445 RT_LOCK(rt);
1446 if (w->w_op == NET_RT_DUMPX_FLAGS && !(rt->rt_flags & w->w_arg)) {
1447 RT_UNLOCK(rt);
1448 return (0);
1449 }
1450 bzero(&info, sizeof (info));
1451 info.rti_info[RTAX_DST] = rt_key(rt);
1452 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1453 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1454 info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1455
1456 size = rt_msg2(RTM_GET_EXT, &info, 0, w);
1457 if (w->w_req && w->w_tmem) {
1458 struct rt_msghdr_ext *ertm = (struct rt_msghdr_ext *)w->w_tmem;
1459
1460 ertm->rtm_flags = rt->rt_flags;
1461 ertm->rtm_use = rt->rt_use;
1462 rt_getmetrics(rt, &ertm->rtm_rmx);
1463 ertm->rtm_index = rt->rt_ifp->if_index;
1464 ertm->rtm_pid = 0;
1465 ertm->rtm_seq = 0;
1466 ertm->rtm_errno = 0;
1467 ertm->rtm_addrs = info.rti_addrs;
1468 if (rt->rt_llinfo_get_ri == NULL)
1469 bzero(&ertm->rtm_ri, sizeof (ertm->rtm_ri));
1470 else
1471 rt->rt_llinfo_get_ri(rt, &ertm->rtm_ri);
1472
1473 error = SYSCTL_OUT(w->w_req, (caddr_t)ertm, size);
1474 RT_UNLOCK(rt);
1475 return (error);
1476 }
1477 RT_UNLOCK(rt);
1478 return (error);
1479 }
1480
1481 /*
1482 * rdar://9307819
1483 * To avoid to call copyout() while holding locks and to cause problems
1484 * in the paging path, sysctl_iflist() and sysctl_iflist2() contstruct
1485 * the list in two passes. In the first pass we compute the total
1486 * length of the data we are going to copyout, then we release
1487 * all locks to allocate a temporary buffer that gets filled
1488 * in the second pass.
1489 *
1490 * Note that we are verifying the assumption that _MALLOC returns a buffer
1491 * that is at least 32 bits aligned and that the messages and addresses are
1492 * 32 bits aligned.
1493 */
1494
1495 int
1496 sysctl_iflist(int af, struct walkarg *w)
1497 {
1498 struct ifnet *ifp;
1499 struct ifaddr *ifa;
1500 struct rt_addrinfo info;
1501 int len, error = 0;
1502 int pass = 0;
1503 int total_len = 0, current_len = 0;
1504 char *total_buffer = NULL, *cp = NULL;
1505
1506 bzero((caddr_t)&info, sizeof(info));
1507
1508 for (pass = 0; pass < 2; pass++) {
1509 ifnet_head_lock_shared();
1510
1511 TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
1512 if (error)
1513 break;
1514 if (w->w_arg && w->w_arg != ifp->if_index)
1515 continue;
1516 ifnet_lock_shared(ifp);
1517 /*
1518 * Holding ifnet lock here prevents the link address from
1519 * changing contents, so no need to hold the ifa lock.
1520 * The link address is always present; it's never freed.
1521 */
1522 ifa = ifp->if_lladdr;
1523 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1524 len = rt_msg2(RTM_IFINFO, &info, (caddr_t)0, NULL);
1525 if (pass == 0) {
1526 total_len += len;
1527 } else {
1528 struct if_msghdr *ifm;
1529
1530 if (current_len + len > total_len) {
1531 ifnet_lock_done(ifp);
1532 printf("sysctl_iflist: current_len (%d) + len (%d) > total_len (%d)\n",
1533 current_len, len, total_len);
1534 error = ENOBUFS;
1535 break;
1536 }
1537 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1538 len = rt_msg2(RTM_IFINFO, &info, (caddr_t)cp, NULL);
1539 info.rti_info[RTAX_IFP] = NULL;
1540
1541 ifm = (struct if_msghdr *)cp;
1542 ifm->ifm_index = ifp->if_index;
1543 ifm->ifm_flags = (u_short)ifp->if_flags;
1544 if_data_internal_to_if_data(ifp, &ifp->if_data,
1545 &ifm->ifm_data);
1546 ifm->ifm_addrs = info.rti_addrs;
1547
1548 cp += len;
1549 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1550 current_len += len;
1551 }
1552 while ((ifa = ifa->ifa_link.tqe_next) != 0) {
1553 IFA_LOCK(ifa);
1554 if (af && af != ifa->ifa_addr->sa_family) {
1555 IFA_UNLOCK(ifa);
1556 continue;
1557 }
1558 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1559 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1560 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1561 len = rt_msg2(RTM_NEWADDR, &info, 0, 0);
1562 if (pass == 0) {
1563 total_len += len;
1564 } else {
1565 struct ifa_msghdr *ifam;
1566
1567 if (current_len + len > total_len) {
1568 IFA_UNLOCK(ifa);
1569 printf("sysctl_iflist: current_len (%d) + len (%d) > total_len (%d)\n",
1570 current_len, len, total_len);
1571 error = ENOBUFS;
1572 break;
1573 }
1574 len = rt_msg2(RTM_NEWADDR, &info, (caddr_t)cp, NULL);
1575
1576 ifam = (struct ifa_msghdr *)cp;
1577 ifam->ifam_index = ifa->ifa_ifp->if_index;
1578 ifam->ifam_flags = ifa->ifa_flags;
1579 ifam->ifam_metric = ifa->ifa_metric;
1580 ifam->ifam_addrs = info.rti_addrs;
1581
1582 cp += len;
1583 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1584 current_len += len;
1585 }
1586 IFA_UNLOCK(ifa);
1587 }
1588 ifnet_lock_done(ifp);
1589 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1590 info.rti_info[RTAX_BRD] = NULL;
1591 }
1592
1593 ifnet_head_done();
1594
1595 if (error)
1596 break;
1597
1598 if (pass == 0) {
1599 /* Better to return zero length buffer than ENOBUFS */
1600 if (total_len == 0)
1601 total_len = 1;
1602 total_len += total_len >> 3;
1603 total_buffer = _MALLOC(total_len, M_RTABLE, M_ZERO | M_WAITOK);
1604 if (total_buffer == NULL) {
1605 printf("sysctl_iflist: _MALLOC(%d) failed\n", total_len);
1606 error = ENOBUFS;
1607 break;
1608 }
1609 cp = total_buffer;
1610 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1611 } else {
1612 error = SYSCTL_OUT(w->w_req, total_buffer, current_len);
1613 if (error)
1614 break;
1615 }
1616 }
1617
1618 if (total_buffer != NULL)
1619 _FREE(total_buffer, M_RTABLE);
1620
1621 return error;
1622 }
1623
1624 int
1625 sysctl_iflist2(int af, struct walkarg *w)
1626 {
1627 struct ifnet *ifp;
1628 struct ifaddr *ifa;
1629 struct rt_addrinfo info;
1630 int len, error = 0;
1631 int pass = 0;
1632 int total_len = 0, current_len = 0;
1633 char *total_buffer = NULL, *cp = NULL;
1634
1635 bzero((caddr_t)&info, sizeof(info));
1636
1637 for (pass = 0; pass < 2; pass++) {
1638 ifnet_head_lock_shared();
1639
1640 TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
1641 if (error)
1642 break;
1643 if (w->w_arg && w->w_arg != ifp->if_index)
1644 continue;
1645 ifnet_lock_shared(ifp);
1646 /*
1647 * Holding ifnet lock here prevents the link address from
1648 * changing contents, so no need to hold the ifa lock.
1649 * The link address is always present; it's never freed.
1650 */
1651 ifa = ifp->if_lladdr;
1652 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1653 len = rt_msg2(RTM_IFINFO2, &info, (caddr_t)0, NULL);
1654 if (pass == 0) {
1655 total_len += len;
1656 } else {
1657 struct if_msghdr2 *ifm;
1658
1659 if (current_len + len > total_len) {
1660 ifnet_lock_done(ifp);
1661 printf("sysctl_iflist2: current_len (%d) + len (%d) > total_len (%d)\n",
1662 current_len, len, total_len);
1663 error = ENOBUFS;
1664 break;
1665 }
1666 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1667 len = rt_msg2(RTM_IFINFO2, &info, (caddr_t)cp, NULL);
1668 info.rti_info[RTAX_IFP] = NULL;
1669
1670 ifm = (struct if_msghdr2 *)cp;
1671 ifm->ifm_addrs = info.rti_addrs;
1672 ifm->ifm_flags = (u_short)ifp->if_flags;
1673 ifm->ifm_index = ifp->if_index;
1674 ifm->ifm_snd_len = ifp->if_snd.ifq_len;
1675 ifm->ifm_snd_maxlen = ifp->if_snd.ifq_maxlen;
1676 ifm->ifm_snd_drops = ifp->if_snd.ifq_drops;
1677 ifm->ifm_timer = ifp->if_timer;
1678 if_data_internal_to_if_data64(ifp, &ifp->if_data,
1679 &ifm->ifm_data);
1680
1681 cp += len;
1682 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1683 current_len += len;
1684 }
1685 while ((ifa = ifa->ifa_link.tqe_next) != 0) {
1686 IFA_LOCK(ifa);
1687 if (af && af != ifa->ifa_addr->sa_family) {
1688 IFA_UNLOCK(ifa);
1689 continue;
1690 }
1691 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1692 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1693 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1694 len = rt_msg2(RTM_NEWADDR, &info, 0, 0);
1695 if (pass == 0) {
1696 total_len += len;
1697 } else {
1698 struct ifa_msghdr *ifam;
1699
1700 if (current_len + len > total_len) {
1701 IFA_UNLOCK(ifa);
1702 printf("sysctl_iflist2: current_len (%d) + len (%d) > total_len (%d)\n",
1703 current_len, len, total_len);
1704 error = ENOBUFS;
1705 break;
1706 }
1707 len = rt_msg2(RTM_NEWADDR, &info, (caddr_t)cp, 0);
1708
1709 ifam = (struct ifa_msghdr *)cp;
1710 ifam->ifam_index = ifa->ifa_ifp->if_index;
1711 ifam->ifam_flags = ifa->ifa_flags;
1712 ifam->ifam_metric = ifa->ifa_metric;
1713 ifam->ifam_addrs = info.rti_addrs;
1714
1715 cp += len;
1716 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1717 current_len += len;
1718 }
1719 IFA_UNLOCK(ifa);
1720 }
1721 if (error) {
1722 ifnet_lock_done(ifp);
1723 break;
1724 }
1725 {
1726 struct ifmultiaddr *ifma;
1727
1728 for (ifma = LIST_FIRST(&ifp->if_multiaddrs);
1729 ifma != NULL; ifma = LIST_NEXT(ifma, ifma_link)) {
1730 struct ifaddr *ifa0;
1731
1732 IFMA_LOCK(ifma);
1733 if (af && af != ifma->ifma_addr->sa_family) {
1734 IFMA_UNLOCK(ifma);
1735 continue;
1736 }
1737 bzero((caddr_t)&info, sizeof(info));
1738 info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1739 /*
1740 * Holding ifnet lock here prevents the link
1741 * address from changing contents, so no need
1742 * to hold the ifa0 lock. The link address is
1743 * always present; it's never freed.
1744 */
1745 ifa0 = ifp->if_lladdr;
1746 info.rti_info[RTAX_IFP] = ifa0->ifa_addr;
1747 if (ifma->ifma_ll != NULL)
1748 info.rti_info[RTAX_GATEWAY] = ifma->ifma_ll->ifma_addr;
1749 len = rt_msg2(RTM_NEWMADDR2, &info, 0, 0);
1750 if (pass == 0) {
1751 total_len += len;
1752 } else {
1753 struct ifma_msghdr2 *ifmam;
1754
1755 if (current_len + len > total_len) {
1756 IFMA_UNLOCK(ifma);
1757 printf("sysctl_iflist2: current_len (%d) + len (%d) > total_len (%d)\n",
1758 current_len, len, total_len);
1759 error = ENOBUFS;
1760 break;
1761 }
1762 len = rt_msg2(RTM_NEWMADDR2, &info, (caddr_t)cp, 0);
1763
1764 ifmam = (struct ifma_msghdr2 *)cp;
1765 ifmam->ifmam_addrs = info.rti_addrs;
1766 ifmam->ifmam_flags = 0;
1767 ifmam->ifmam_index =
1768 ifma->ifma_ifp->if_index;
1769 ifmam->ifmam_refcount =
1770 ifma->ifma_reqcnt;
1771
1772 cp += len;
1773 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1774 current_len += len;
1775 }
1776 IFMA_UNLOCK(ifma);
1777 }
1778 }
1779 ifnet_lock_done(ifp);
1780 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1781 info.rti_info[RTAX_BRD] = NULL;
1782 }
1783 ifnet_head_done();
1784
1785 if (error)
1786 break;
1787
1788 if (pass == 0) {
1789 /* Better to return zero length buffer than ENOBUFS */
1790 if (total_len == 0)
1791 total_len = 1;
1792 total_len += total_len >> 3;
1793 total_buffer = _MALLOC(total_len, M_RTABLE, M_ZERO | M_WAITOK);
1794 if (total_buffer == NULL) {
1795 printf("sysctl_iflist2: _MALLOC(%d) failed\n", total_len);
1796 error = ENOBUFS;
1797 break;
1798 }
1799 cp = total_buffer;
1800 VERIFY(IS_P2ALIGNED(cp, sizeof(u_int32_t)));
1801 } else {
1802 error = SYSCTL_OUT(w->w_req, total_buffer, current_len);
1803 if (error)
1804 break;
1805 }
1806 }
1807
1808 if (total_buffer != NULL)
1809 _FREE(total_buffer, M_RTABLE);
1810
1811 return error;
1812 }
1813
1814
1815 static int
1816 sysctl_rtstat(struct sysctl_req *req)
1817 {
1818 int error;
1819
1820 error = SYSCTL_OUT(req, &rtstat, sizeof(struct rtstat));
1821 if (error)
1822 return (error);
1823
1824 return 0;
1825 }
1826
1827 static int
1828 sysctl_rttrash(struct sysctl_req *req)
1829 {
1830 int error;
1831
1832 error = SYSCTL_OUT(req, &rttrash, sizeof(rttrash));
1833 if (error)
1834 return (error);
1835
1836 return 0;
1837 }
1838
1839 /*
1840 * Called from pfslowtimo(), protected by domain_proto_mtx
1841 */
1842 static void
1843 rt_drainall(void)
1844 {
1845 struct timeval delta_ts, current_ts;
1846
1847 /*
1848 * This test is done without holding rnh_lock; in the even that
1849 * we read stale value, it will only cause an extra (or miss)
1850 * drain and is therefore harmless.
1851 */
1852 if (ifnet_aggressive_drainers == 0) {
1853 if (timerisset(&last_ts))
1854 timerclear(&last_ts);
1855 return;
1856 }
1857
1858 microuptime(&current_ts);
1859 timersub(&current_ts, &last_ts, &delta_ts);
1860
1861 if (delta_ts.tv_sec >= rt_if_idle_drain_interval) {
1862 timerclear(&last_ts);
1863
1864 in_rtqdrain(); /* protocol cloned routes: INET */
1865 in_arpdrain(NULL); /* cloned routes: ARP */
1866 #if INET6
1867 in6_rtqdrain(); /* protocol cloned routes: INET6 */
1868 nd6_drain(NULL); /* cloned routes: ND6 */
1869 #endif /* INET6 */
1870
1871 last_ts.tv_sec = current_ts.tv_sec;
1872 last_ts.tv_usec = current_ts.tv_usec;
1873 }
1874 }
1875
1876 void
1877 rt_aggdrain(int on)
1878 {
1879 lck_mtx_assert(rnh_lock, LCK_MTX_ASSERT_OWNED);
1880
1881 if (on)
1882 routedomain.dom_protosw->pr_flags |= PR_AGGDRAIN;
1883 else
1884 routedomain.dom_protosw->pr_flags &= ~PR_AGGDRAIN;
1885 }
1886
1887 static int
1888 sysctl_rtsock SYSCTL_HANDLER_ARGS
1889 {
1890 #pragma unused(oidp)
1891 int *name = (int *)arg1;
1892 u_int namelen = arg2;
1893 struct radix_node_head *rnh;
1894 int i, error = EINVAL;
1895 u_char af;
1896 struct walkarg w;
1897
1898 name ++;
1899 namelen--;
1900 if (req->newptr)
1901 return (EPERM);
1902 if (namelen != 3)
1903 return (EINVAL);
1904 af = name[0];
1905 Bzero(&w, sizeof(w));
1906 w.w_op = name[1];
1907 w.w_arg = name[2];
1908 w.w_req = req;
1909
1910 switch (w.w_op) {
1911
1912 case NET_RT_DUMP:
1913 case NET_RT_DUMP2:
1914 case NET_RT_FLAGS:
1915 lck_mtx_lock(rnh_lock);
1916 for (i = 1; i <= AF_MAX; i++)
1917 if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
1918 (error = rnh->rnh_walktree(rnh,
1919 sysctl_dumpentry, &w)))
1920 break;
1921 lck_mtx_unlock(rnh_lock);
1922 break;
1923 case NET_RT_DUMPX:
1924 case NET_RT_DUMPX_FLAGS:
1925 lck_mtx_lock(rnh_lock);
1926 for (i = 1; i <= AF_MAX; i++)
1927 if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
1928 (error = rnh->rnh_walktree(rnh,
1929 sysctl_dumpentry_ext, &w)))
1930 break;
1931 lck_mtx_unlock(rnh_lock);
1932 break;
1933 case NET_RT_IFLIST:
1934 error = sysctl_iflist(af, &w);
1935 break;
1936 case NET_RT_IFLIST2:
1937 error = sysctl_iflist2(af, &w);
1938 break;
1939 case NET_RT_STAT:
1940 error = sysctl_rtstat(req);
1941 break;
1942 case NET_RT_TRASH:
1943 error = sysctl_rttrash(req);
1944 break;
1945 }
1946 if (w.w_tmem)
1947 FREE(w.w_tmem, M_RTABLE);
1948 return (error);
1949 }
1950
1951 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_LOCKED, sysctl_rtsock, "");
1952
1953 /*
1954 * Definitions of protocols supported in the ROUTE domain.
1955 */
1956 static struct protosw routesw[] = {
1957 { SOCK_RAW, &routedomain, 0, PR_ATOMIC|PR_ADDR,
1958 0, route_output, raw_ctlinput, 0,
1959 0,
1960 raw_init, 0, 0, rt_drainall,
1961 0,
1962 &route_usrreqs,
1963 0, 0, 0,
1964 { 0, 0 }, 0, { 0 }
1965 }
1966 };
1967
1968 struct domain routedomain =
1969 { PF_ROUTE, "route", route_init, 0, 0,
1970 routesw,
1971 NULL, NULL, 0, 0, 0, 0, NULL, 0,
1972 { 0, 0 } };
1973
1974 DOMAIN_SET(route);
1975
mirror of XNU