FreeBSD/Linux Kernel Cross Reference
sys/netinet/in.c
1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1982, 1986, 1991, 1993
5 * The Regents of the University of California. All rights reserved.
6 * Copyright (C) 2001 WIDE Project. All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 * @(#)in.c 8.4 (Berkeley) 1/9/95
33 */
34
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37
38 #include "opt_inet.h"
39
40 #define IN_HISTORICAL_NETS /* include class masks */
41
42 #include <sys/param.h>
43 #include <sys/eventhandler.h>
44 #include <sys/systm.h>
45 #include <sys/sockio.h>
46 #include <sys/malloc.h>
47 #include <sys/priv.h>
48 #include <sys/socket.h>
49 #include <sys/jail.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/proc.h>
53 #include <sys/sysctl.h>
54 #include <sys/syslog.h>
55 #include <sys/sx.h>
56
57 #include <net/if.h>
58 #include <net/if_var.h>
59 #include <net/if_arp.h>
60 #include <net/if_dl.h>
61 #include <net/if_llatbl.h>
62 #include <net/if_types.h>
63 #include <net/route.h>
64 #include <net/route/nhop.h>
65 #include <net/route/route_ctl.h>
66 #include <net/vnet.h>
67
68 #include <netinet/if_ether.h>
69 #include <netinet/in.h>
70 #include <netinet/in_fib.h>
71 #include <netinet/in_var.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/ip_carp.h>
75 #include <netinet/igmp_var.h>
76 #include <netinet/udp.h>
77 #include <netinet/udp_var.h>
78
79 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
80 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
81 static int in_gifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
82
83 static void in_socktrim(struct sockaddr_in *);
84 static void in_purgemaddrs(struct ifnet *);
85
86 static bool ia_need_loopback_route(const struct in_ifaddr *);
87
88 VNET_DEFINE_STATIC(int, nosameprefix);
89 #define V_nosameprefix VNET(nosameprefix)
90 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW,
91 &VNET_NAME(nosameprefix), 0,
92 "Refuse to create same prefixes on different interfaces");
93
94 VNET_DEFINE_STATIC(bool, broadcast_lowest);
95 #define V_broadcast_lowest VNET(broadcast_lowest)
96 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, broadcast_lowest, CTLFLAG_VNET | CTLFLAG_RW,
97 &VNET_NAME(broadcast_lowest), 0,
98 "Treat lowest address on a subnet (host 0) as broadcast");
99
100 VNET_DEFINE(bool, ip_allow_net240) = false;
101 #define V_ip_allow_net240 VNET(ip_allow_net240)
102 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net240,
103 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net240), 0,
104 "Allow use of Experimental addresses, aka Class E (240/4)");
105 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-240 */
106
107 VNET_DEFINE(bool, ip_allow_net0) = false;
108 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net0,
109 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net0), 0,
110 "Allow use of addresses in network 0/8");
111 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-0 */
112
113 VNET_DEFINE(uint32_t, in_loopback_mask) = IN_LOOPBACK_MASK_DFLT;
114 #define V_in_loopback_mask VNET(in_loopback_mask)
115 static int sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS);
116 SYSCTL_PROC(_net_inet_ip, OID_AUTO, loopback_prefixlen,
117 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
118 NULL, 0, sysctl_loopback_prefixlen, "I",
119 "Prefix length of address space reserved for loopback");
120 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-127 */
121
122 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
123 #define V_ripcbinfo VNET(ripcbinfo)
124
125 static struct sx in_control_sx;
126 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control");
127
128 /*
129 * Return 1 if an internet address is for a ``local'' host
130 * (one to which we have a connection).
131 */
132 int
133 in_localaddr(struct in_addr in)
134 {
135 u_long i = ntohl(in.s_addr);
136 struct in_ifaddr *ia;
137
138 NET_EPOCH_ASSERT();
139
140 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
141 if ((i & ia->ia_subnetmask) == ia->ia_subnet)
142 return (1);
143 }
144
145 return (0);
146 }
147
148 /*
149 * Return 1 if an internet address is for the local host and configured
150 * on one of its interfaces.
151 */
152 bool
153 in_localip(struct in_addr in)
154 {
155 struct in_ifaddr *ia;
156
157 NET_EPOCH_ASSERT();
158
159 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
160 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
161 return (true);
162
163 return (false);
164 }
165
166 /*
167 * Like in_localip(), but FIB-aware.
168 */
169 bool
170 in_localip_fib(struct in_addr in, uint16_t fib)
171 {
172 struct in_ifaddr *ia;
173
174 NET_EPOCH_ASSERT();
175
176 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
177 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr &&
178 ia->ia_ifa.ifa_ifp->if_fib == fib)
179 return (true);
180
181 return (false);
182 }
183
184 /*
185 * Return 1 if an internet address is configured on an interface.
186 */
187 int
188 in_ifhasaddr(struct ifnet *ifp, struct in_addr in)
189 {
190 struct ifaddr *ifa;
191 struct in_ifaddr *ia;
192
193 NET_EPOCH_ASSERT();
194
195 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
196 if (ifa->ifa_addr->sa_family != AF_INET)
197 continue;
198 ia = (struct in_ifaddr *)ifa;
199 if (ia->ia_addr.sin_addr.s_addr == in.s_addr)
200 return (1);
201 }
202
203 return (0);
204 }
205
206 /*
207 * Return a reference to the interface address which is different to
208 * the supplied one but with same IP address value.
209 */
210 static struct in_ifaddr *
211 in_localip_more(struct in_ifaddr *original_ia)
212 {
213 struct epoch_tracker et;
214 in_addr_t original_addr = IA_SIN(original_ia)->sin_addr.s_addr;
215 uint32_t original_fib = original_ia->ia_ifa.ifa_ifp->if_fib;
216 struct in_ifaddr *ia;
217
218 NET_EPOCH_ENTER(et);
219 CK_LIST_FOREACH(ia, INADDR_HASH(original_addr), ia_hash) {
220 in_addr_t addr = IA_SIN(ia)->sin_addr.s_addr;
221 uint32_t fib = ia->ia_ifa.ifa_ifp->if_fib;
222 if (!V_rt_add_addr_allfibs && (original_fib != fib))
223 continue;
224 if ((original_ia != ia) && (original_addr == addr)) {
225 ifa_ref(&ia->ia_ifa);
226 NET_EPOCH_EXIT(et);
227 return (ia);
228 }
229 }
230 NET_EPOCH_EXIT(et);
231
232 return (NULL);
233 }
234
235 /*
236 * Tries to find first IPv4 address in the provided fib.
237 * Prefers non-loopback addresses and return loopback IFF
238 * @loopback_ok is set.
239 *
240 * Returns ifa or NULL.
241 */
242 struct in_ifaddr *
243 in_findlocal(uint32_t fibnum, bool loopback_ok)
244 {
245 struct in_ifaddr *ia = NULL, *ia_lo = NULL;
246
247 NET_EPOCH_ASSERT();
248
249 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
250 uint32_t ia_fib = ia->ia_ifa.ifa_ifp->if_fib;
251 if (!V_rt_add_addr_allfibs && (fibnum != ia_fib))
252 continue;
253
254 if (!IN_LOOPBACK(ntohl(IA_SIN(ia)->sin_addr.s_addr)))
255 break;
256 if (loopback_ok)
257 ia_lo = ia;
258 }
259
260 if (ia == NULL)
261 ia = ia_lo;
262
263 return (ia);
264 }
265
266 /*
267 * Determine whether an IP address is in a reserved set of addresses
268 * that may not be forwarded, or whether datagrams to that destination
269 * may be forwarded.
270 */
271 int
272 in_canforward(struct in_addr in)
273 {
274 u_long i = ntohl(in.s_addr);
275
276 if (IN_MULTICAST(i) || IN_LINKLOCAL(i) || IN_LOOPBACK(i))
277 return (0);
278 if (IN_EXPERIMENTAL(i) && !V_ip_allow_net240)
279 return (0);
280 if (IN_ZERONET(i) && !V_ip_allow_net0)
281 return (0);
282 return (1);
283 }
284
285 /*
286 * Sysctl to manage prefix of reserved loopback network; translate
287 * to/from mask. The mask is always contiguous high-order 1 bits
288 * followed by all 0 bits.
289 */
290 static int
291 sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS)
292 {
293 int error, preflen;
294
295 /* ffs is 1-based; compensate. */
296 preflen = 33 - ffs(V_in_loopback_mask);
297 error = sysctl_handle_int(oidp, &preflen, 0, req);
298 if (error || !req->newptr)
299 return (error);
300 if (preflen < 8 || preflen > 31)
301 return (EINVAL);
302 V_in_loopback_mask = 0xffffffff << (32 - preflen);
303 return (0);
304 }
305
306 /*
307 * Trim a mask in a sockaddr
308 */
309 static void
310 in_socktrim(struct sockaddr_in *ap)
311 {
312 char *cplim = (char *) &ap->sin_addr;
313 char *cp = (char *) (&ap->sin_addr + 1);
314
315 ap->sin_len = 0;
316 while (--cp >= cplim)
317 if (*cp) {
318 (ap)->sin_len = cp - (char *) (ap) + 1;
319 break;
320 }
321 }
322
323 /*
324 * Generic internet control operations (ioctl's).
325 */
326 int
327 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
328 struct thread *td)
329 {
330 struct ifreq *ifr = (struct ifreq *)data;
331 struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
332 struct epoch_tracker et;
333 struct ifaddr *ifa;
334 struct in_ifaddr *ia;
335 int error;
336
337 if (ifp == NULL)
338 return (EADDRNOTAVAIL);
339
340 struct ucred *cred = (td != NULL) ? td->td_ucred : NULL;
341
342 /*
343 * Filter out 4 ioctls we implement directly. Forward the rest
344 * to specific functions and ifp->if_ioctl().
345 */
346 switch (cmd) {
347 case SIOCGIFADDR:
348 case SIOCGIFBRDADDR:
349 case SIOCGIFDSTADDR:
350 case SIOCGIFNETMASK:
351 break;
352 case SIOCGIFALIAS:
353 sx_xlock(&in_control_sx);
354 error = in_gifaddr_ioctl(cmd, data, ifp, cred);
355 sx_xunlock(&in_control_sx);
356 return (error);
357 case SIOCDIFADDR:
358 sx_xlock(&in_control_sx);
359 error = in_difaddr_ioctl(cmd, data, ifp, cred);
360 sx_xunlock(&in_control_sx);
361 return (error);
362 case OSIOCAIFADDR: /* 9.x compat */
363 case SIOCAIFADDR:
364 sx_xlock(&in_control_sx);
365 error = in_aifaddr_ioctl(cmd, data, ifp, cred);
366 sx_xunlock(&in_control_sx);
367 return (error);
368 case SIOCSIFADDR:
369 case SIOCSIFBRDADDR:
370 case SIOCSIFDSTADDR:
371 case SIOCSIFNETMASK:
372 /* We no longer support that old commands. */
373 return (EINVAL);
374 default:
375 if (ifp->if_ioctl == NULL)
376 return (EOPNOTSUPP);
377 return ((*ifp->if_ioctl)(ifp, cmd, data));
378 }
379
380 if (addr->sin_addr.s_addr != INADDR_ANY &&
381 prison_check_ip4(cred, &addr->sin_addr) != 0)
382 return (EADDRNOTAVAIL);
383
384 /*
385 * Find address for this interface, if it exists. If an
386 * address was specified, find that one instead of the
387 * first one on the interface, if possible.
388 */
389 NET_EPOCH_ENTER(et);
390 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
391 if (ifa->ifa_addr->sa_family != AF_INET)
392 continue;
393 ia = (struct in_ifaddr *)ifa;
394 if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
395 break;
396 }
397 if (ifa == NULL)
398 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
399 if (ifa->ifa_addr->sa_family == AF_INET) {
400 ia = (struct in_ifaddr *)ifa;
401 if (prison_check_ip4(cred,
402 &ia->ia_addr.sin_addr) == 0)
403 break;
404 }
405
406 if (ifa == NULL) {
407 NET_EPOCH_EXIT(et);
408 return (EADDRNOTAVAIL);
409 }
410
411 error = 0;
412 switch (cmd) {
413 case SIOCGIFADDR:
414 *addr = ia->ia_addr;
415 break;
416
417 case SIOCGIFBRDADDR:
418 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
419 error = EINVAL;
420 break;
421 }
422 *addr = ia->ia_broadaddr;
423 break;
424
425 case SIOCGIFDSTADDR:
426 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
427 error = EINVAL;
428 break;
429 }
430 *addr = ia->ia_dstaddr;
431 break;
432
433 case SIOCGIFNETMASK:
434 *addr = ia->ia_sockmask;
435 break;
436 }
437
438 NET_EPOCH_EXIT(et);
439
440 return (error);
441 }
442
443 static int
444 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
445 {
446 const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
447 const struct sockaddr_in *addr = &ifra->ifra_addr;
448 const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
449 const struct sockaddr_in *mask = &ifra->ifra_mask;
450 const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
451 const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
452 struct epoch_tracker et;
453 struct ifaddr *ifa;
454 struct in_ifaddr *ia;
455 bool iaIsFirst;
456 int error = 0;
457
458 error = priv_check_cred(cred, PRIV_NET_ADDIFADDR);
459 if (error)
460 return (error);
461
462 /*
463 * ifra_addr must be present and be of INET family.
464 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
465 */
466 if (addr->sin_len != sizeof(struct sockaddr_in) ||
467 addr->sin_family != AF_INET)
468 return (EINVAL);
469 if (broadaddr->sin_len != 0 &&
470 (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
471 broadaddr->sin_family != AF_INET))
472 return (EINVAL);
473 if (mask->sin_len != 0 &&
474 (mask->sin_len != sizeof(struct sockaddr_in) ||
475 mask->sin_family != AF_INET))
476 return (EINVAL);
477 if ((ifp->if_flags & IFF_POINTOPOINT) &&
478 (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
479 dstaddr->sin_addr.s_addr == INADDR_ANY))
480 return (EDESTADDRREQ);
481 if (vhid != 0 && carp_attach_p == NULL)
482 return (EPROTONOSUPPORT);
483
484 /*
485 * See whether address already exist.
486 */
487 iaIsFirst = true;
488 ia = NULL;
489 NET_EPOCH_ENTER(et);
490 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
491 struct in_ifaddr *it;
492
493 if (ifa->ifa_addr->sa_family != AF_INET)
494 continue;
495
496 it = (struct in_ifaddr *)ifa;
497 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
498 prison_check_ip4(cred, &addr->sin_addr) == 0)
499 ia = it;
500 else
501 iaIsFirst = false;
502 }
503 NET_EPOCH_EXIT(et);
504
505 if (ia != NULL)
506 (void )in_difaddr_ioctl(cmd, data, ifp, cred);
507
508 ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
509 ia = (struct in_ifaddr *)ifa;
510 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
511 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
512 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
513 callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
514 CALLOUT_RETURNUNLOCKED);
515
516 ia->ia_ifp = ifp;
517 ia->ia_addr = *addr;
518 if (mask->sin_len != 0) {
519 ia->ia_sockmask = *mask;
520 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
521 } else {
522 in_addr_t i = ntohl(addr->sin_addr.s_addr);
523
524 /*
525 * If netmask isn't supplied, use historical default.
526 * This is deprecated for interfaces other than loopback
527 * or point-to-point; warn in other cases. In the future
528 * we should return an error rather than warning.
529 */
530 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0)
531 printf("%s: set address: WARNING: network mask "
532 "should be specified; using historical default\n",
533 ifp->if_xname);
534 if (IN_CLASSA(i))
535 ia->ia_subnetmask = IN_CLASSA_NET;
536 else if (IN_CLASSB(i))
537 ia->ia_subnetmask = IN_CLASSB_NET;
538 else
539 ia->ia_subnetmask = IN_CLASSC_NET;
540 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
541 }
542 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
543 in_socktrim(&ia->ia_sockmask);
544
545 if (ifp->if_flags & IFF_BROADCAST) {
546 if (broadaddr->sin_len != 0) {
547 ia->ia_broadaddr = *broadaddr;
548 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
549 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
550 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
551 ia->ia_broadaddr.sin_family = AF_INET;
552 } else {
553 ia->ia_broadaddr.sin_addr.s_addr =
554 htonl(ia->ia_subnet | ~ia->ia_subnetmask);
555 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
556 ia->ia_broadaddr.sin_family = AF_INET;
557 }
558 }
559
560 if (ifp->if_flags & IFF_POINTOPOINT)
561 ia->ia_dstaddr = *dstaddr;
562
563 if (vhid != 0) {
564 error = (*carp_attach_p)(&ia->ia_ifa, vhid);
565 if (error)
566 return (error);
567 }
568
569 /* if_addrhead is already referenced by ifa_alloc() */
570 IF_ADDR_WLOCK(ifp);
571 CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
572 IF_ADDR_WUNLOCK(ifp);
573
574 ifa_ref(ifa); /* in_ifaddrhead */
575 sx_assert(&in_control_sx, SA_XLOCKED);
576 CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
577 CK_LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia,
578 ia_hash);
579
580 /*
581 * Give the interface a chance to initialize
582 * if this is its first address,
583 * and to validate the address if necessary.
584 */
585 if (ifp->if_ioctl != NULL) {
586 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
587 if (error)
588 goto fail1;
589 }
590
591 /*
592 * Add route for the network.
593 */
594 if (vhid == 0) {
595 error = in_addprefix(ia);
596 if (error)
597 goto fail1;
598 }
599
600 /*
601 * Add a loopback route to self.
602 */
603 if (vhid == 0 && ia_need_loopback_route(ia)) {
604 struct in_ifaddr *eia;
605
606 eia = in_localip_more(ia);
607
608 if (eia == NULL) {
609 error = ifa_add_loopback_route((struct ifaddr *)ia,
610 (struct sockaddr *)&ia->ia_addr);
611 if (error)
612 goto fail2;
613 } else
614 ifa_free(&eia->ia_ifa);
615 }
616
617 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
618 struct in_addr allhosts_addr;
619 struct in_ifinfo *ii;
620
621 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
622 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
623
624 error = in_joingroup(ifp, &allhosts_addr, NULL,
625 &ii->ii_allhosts);
626 }
627
628 /*
629 * Note: we don't need extra reference for ifa, since we called
630 * with sx lock held, and ifaddr can not be deleted in concurrent
631 * thread.
632 */
633 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
634
635 return (error);
636
637 fail2:
638 if (vhid == 0)
639 (void )in_scrubprefix(ia, LLE_STATIC);
640
641 fail1:
642 if (ia->ia_ifa.ifa_carp)
643 (*carp_detach_p)(&ia->ia_ifa, false);
644
645 IF_ADDR_WLOCK(ifp);
646 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
647 IF_ADDR_WUNLOCK(ifp);
648 ifa_free(&ia->ia_ifa); /* if_addrhead */
649
650 sx_assert(&in_control_sx, SA_XLOCKED);
651 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
652 CK_LIST_REMOVE(ia, ia_hash);
653 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */
654
655 return (error);
656 }
657
658 static int
659 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
660 {
661 const struct ifreq *ifr = (struct ifreq *)data;
662 const struct sockaddr_in *addr = (const struct sockaddr_in *)
663 &ifr->ifr_addr;
664 struct ifaddr *ifa;
665 struct in_ifaddr *ia;
666 bool deleteAny, iaIsLast;
667 int error;
668
669 if (cred != NULL) {
670 error = priv_check_cred(cred, PRIV_NET_DELIFADDR);
671 if (error)
672 return (error);
673 }
674
675 if (addr->sin_len != sizeof(struct sockaddr_in) ||
676 addr->sin_family != AF_INET)
677 deleteAny = true;
678 else
679 deleteAny = false;
680
681 iaIsLast = true;
682 ia = NULL;
683 IF_ADDR_WLOCK(ifp);
684 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
685 struct in_ifaddr *it;
686
687 if (ifa->ifa_addr->sa_family != AF_INET)
688 continue;
689
690 it = (struct in_ifaddr *)ifa;
691 if (deleteAny && ia == NULL && (cred == NULL ||
692 prison_check_ip4(cred, &it->ia_addr.sin_addr) == 0))
693 ia = it;
694
695 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
696 (cred == NULL || prison_check_ip4(cred,
697 &addr->sin_addr) == 0))
698 ia = it;
699
700 if (it != ia)
701 iaIsLast = false;
702 }
703
704 if (ia == NULL) {
705 IF_ADDR_WUNLOCK(ifp);
706 return (EADDRNOTAVAIL);
707 }
708
709 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
710 IF_ADDR_WUNLOCK(ifp);
711 ifa_free(&ia->ia_ifa); /* if_addrhead */
712
713 sx_assert(&in_control_sx, SA_XLOCKED);
714 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
715 CK_LIST_REMOVE(ia, ia_hash);
716
717 /*
718 * in_scrubprefix() kills the interface route.
719 */
720 in_scrubprefix(ia, LLE_STATIC);
721
722 /*
723 * in_ifadown gets rid of all the rest of
724 * the routes. This is not quite the right
725 * thing to do, but at least if we are running
726 * a routing process they will come back.
727 */
728 in_ifadown(&ia->ia_ifa, 1);
729
730 if (ia->ia_ifa.ifa_carp)
731 (*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
732
733 /*
734 * If this is the last IPv4 address configured on this
735 * interface, leave the all-hosts group.
736 * No state-change report need be transmitted.
737 */
738 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
739 struct in_ifinfo *ii;
740
741 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
742 if (ii->ii_allhosts) {
743 (void)in_leavegroup(ii->ii_allhosts, NULL);
744 ii->ii_allhosts = NULL;
745 }
746 }
747
748 IF_ADDR_WLOCK(ifp);
749 if (callout_stop(&ia->ia_garp_timer) == 1) {
750 ifa_free(&ia->ia_ifa);
751 }
752 IF_ADDR_WUNLOCK(ifp);
753
754 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
755 IFADDR_EVENT_DEL);
756 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */
757
758 return (0);
759 }
760
761 static int
762 in_gifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
763 {
764 struct in_aliasreq *ifra = (struct in_aliasreq *)data;
765 const struct sockaddr_in *addr = &ifra->ifra_addr;
766 struct epoch_tracker et;
767 struct ifaddr *ifa;
768 struct in_ifaddr *ia;
769
770 /*
771 * ifra_addr must be present and be of INET family.
772 */
773 if (addr->sin_len != sizeof(struct sockaddr_in) ||
774 addr->sin_family != AF_INET)
775 return (EINVAL);
776
777 /*
778 * See whether address exist.
779 */
780 ia = NULL;
781 NET_EPOCH_ENTER(et);
782 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
783 struct in_ifaddr *it;
784
785 if (ifa->ifa_addr->sa_family != AF_INET)
786 continue;
787
788 it = (struct in_ifaddr *)ifa;
789 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
790 prison_check_ip4(cred, &addr->sin_addr) == 0) {
791 ia = it;
792 break;
793 }
794 }
795 if (ia == NULL) {
796 NET_EPOCH_EXIT(et);
797 return (EADDRNOTAVAIL);
798 }
799
800 ifra->ifra_mask = ia->ia_sockmask;
801 if ((ifp->if_flags & IFF_POINTOPOINT) &&
802 ia->ia_dstaddr.sin_family == AF_INET)
803 ifra->ifra_dstaddr = ia->ia_dstaddr;
804 else if ((ifp->if_flags & IFF_BROADCAST) &&
805 ia->ia_broadaddr.sin_family == AF_INET)
806 ifra->ifra_broadaddr = ia->ia_broadaddr;
807 else
808 memset(&ifra->ifra_broadaddr, 0,
809 sizeof(ifra->ifra_broadaddr));
810
811 NET_EPOCH_EXIT(et);
812 return (0);
813 }
814
815 static int
816 in_match_ifaddr(const struct rtentry *rt, const struct nhop_object *nh, void *arg)
817 {
818
819 if (nh->nh_ifa == (struct ifaddr *)arg)
820 return (1);
821
822 return (0);
823 }
824
825 static int
826 in_handle_prefix_route(uint32_t fibnum, int cmd,
827 struct sockaddr_in *dst, struct sockaddr_in *netmask, struct ifaddr *ifa,
828 struct ifnet *ifp)
829 {
830
831 NET_EPOCH_ASSERT();
832
833 /* Prepare gateway */
834 struct sockaddr_dl_short sdl = {
835 .sdl_family = AF_LINK,
836 .sdl_len = sizeof(struct sockaddr_dl_short),
837 .sdl_type = ifa->ifa_ifp->if_type,
838 .sdl_index = ifa->ifa_ifp->if_index,
839 };
840
841 struct rt_addrinfo info = {
842 .rti_ifa = ifa,
843 .rti_ifp = ifp,
844 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
845 .rti_info = {
846 [RTAX_DST] = (struct sockaddr *)dst,
847 [RTAX_NETMASK] = (struct sockaddr *)netmask,
848 [RTAX_GATEWAY] = (struct sockaddr *)&sdl,
849 },
850 /* Ensure we delete the prefix IFF prefix ifa matches */
851 .rti_filter = in_match_ifaddr,
852 .rti_filterdata = ifa,
853 };
854
855 return (rib_handle_ifaddr_info(fibnum, cmd, &info));
856 }
857
858 /*
859 * Routing table interaction with interface addresses.
860 *
861 * In general, two types of routes needs to be installed:
862 * a) "interface" or "prefix" route, telling user that the addresses
863 * behind the ifa prefix are reached directly.
864 * b) "loopback" route installed for the ifa address, telling user that
865 * the address belongs to local system.
866 *
867 * Handling for (a) and (b) differs in multi-fib aspects, hence they
868 * are implemented in different functions below.
869 *
870 * The cases above may intersect - /32 interface aliases results in
871 * the same prefix produced by (a) and (b). This blurs the definition
872 * of the "loopback" route and complicate interactions. The interaction
873 * table is defined below. The case numbers are used in the multiple
874 * functions below to refer to the particular test case.
875 *
876 * There can be multiple options:
877 * 1) Adding address with prefix on non-p2p/non-loopback interface.
878 * Example: 192.0.2.1/24. Action:
879 * * add "prefix" route towards 192.0.2.0/24 via @ia interface,
880 * using @ia as an address source.
881 * * add "loopback" route towards 192.0.2.1 via V_loif, saving
882 * @ia ifp in the gateway and using @ia as an address source.
883 *
884 * 2) Adding address with /32 mask to non-p2p/non-loopback interface.
885 * Example: 192.0.2.2/32. Action:
886 * * add "prefix" host route via V_loif, using @ia as an address source.
887 *
888 * 3) Adding address with or without prefix to p2p interface.
889 * Example: 10.0.0.1/24->10.0.0.2. Action:
890 * * add "prefix" host route towards 10.0.0.2 via this interface, using @ia
891 * as an address source. Note: no sense in installing full /24 as the interface
892 * is point-to-point.
893 * * add "loopback" route towards 10.0.9.1 via V_loif, saving
894 * @ia ifp in the gateway and using @ia as an address source.
895 *
896 * 4) Adding address with or without prefix to loopback interface.
897 * Example: 192.0.2.1/24. Action:
898 * * add "prefix" host route via @ia interface, using @ia as an address source.
899 * Note: Skip installing /24 prefix as it would introduce TTL loop
900 * for the traffic destined to these addresses.
901 */
902
903 /*
904 * Checks if @ia needs to install loopback route to @ia address via
905 * ifa_maintain_loopback_route().
906 *
907 * Return true on success.
908 */
909 static bool
910 ia_need_loopback_route(const struct in_ifaddr *ia)
911 {
912 struct ifnet *ifp = ia->ia_ifp;
913
914 /* Case 4: Skip loopback interfaces */
915 if ((ifp->if_flags & IFF_LOOPBACK) ||
916 (ia->ia_addr.sin_addr.s_addr == INADDR_ANY))
917 return (false);
918
919 /* Clash avoidance: Skip p2p interfaces with both addresses are equal */
920 if ((ifp->if_flags & IFF_POINTOPOINT) &&
921 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
922 return (false);
923
924 /* Case 2: skip /32 prefixes */
925 if (!(ifp->if_flags & IFF_POINTOPOINT) &&
926 (ia->ia_sockmask.sin_addr.s_addr == INADDR_BROADCAST))
927 return (false);
928
929 return (true);
930 }
931
932 /*
933 * Calculate "prefix" route corresponding to @ia.
934 */
935 static void
936 ia_getrtprefix(const struct in_ifaddr *ia, struct in_addr *prefix, struct in_addr *mask)
937 {
938
939 if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) {
940 /* Case 3: return host route for dstaddr */
941 *prefix = ia->ia_dstaddr.sin_addr;
942 mask->s_addr = INADDR_BROADCAST;
943 } else if (ia->ia_ifp->if_flags & IFF_LOOPBACK) {
944 /* Case 4: return host route for ifaddr */
945 *prefix = ia->ia_addr.sin_addr;
946 mask->s_addr = INADDR_BROADCAST;
947 } else {
948 /* Cases 1,2: return actual ia prefix */
949 *prefix = ia->ia_addr.sin_addr;
950 *mask = ia->ia_sockmask.sin_addr;
951 prefix->s_addr &= mask->s_addr;
952 }
953 }
954
955 /*
956 * Adds or delete interface "prefix" route corresponding to @ifa.
957 * Returns 0 on success or errno.
958 */
959 static int
960 in_handle_ifaddr_route(int cmd, struct in_ifaddr *ia)
961 {
962 struct ifaddr *ifa = &ia->ia_ifa;
963 struct in_addr daddr, maddr;
964 struct sockaddr_in *pmask;
965 struct epoch_tracker et;
966 int error;
967
968 ia_getrtprefix(ia, &daddr, &maddr);
969
970 struct sockaddr_in mask = {
971 .sin_family = AF_INET,
972 .sin_len = sizeof(struct sockaddr_in),
973 .sin_addr = maddr,
974 };
975
976 pmask = (maddr.s_addr != INADDR_BROADCAST) ? &mask : NULL;
977
978 struct sockaddr_in dst = {
979 .sin_family = AF_INET,
980 .sin_len = sizeof(struct sockaddr_in),
981 .sin_addr.s_addr = daddr.s_addr & maddr.s_addr,
982 };
983
984 struct ifnet *ifp = ia->ia_ifp;
985
986 if ((maddr.s_addr == INADDR_BROADCAST) &&
987 (!(ia->ia_ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)))) {
988 /* Case 2: host route on broadcast interface */
989 ifp = V_loif;
990 }
991
992 uint32_t fibnum = ifa->ifa_ifp->if_fib;
993 NET_EPOCH_ENTER(et);
994 error = in_handle_prefix_route(fibnum, cmd, &dst, pmask, ifa, ifp);
995 NET_EPOCH_EXIT(et);
996
997 return (error);
998 }
999
1000 /*
1001 * Check if we have a route for the given prefix already.
1002 */
1003 static bool
1004 in_hasrtprefix(struct in_ifaddr *target)
1005 {
1006 struct epoch_tracker et;
1007 struct in_ifaddr *ia;
1008 struct in_addr prefix, mask, p, m;
1009 bool result = false;
1010
1011 ia_getrtprefix(target, &prefix, &mask);
1012
1013 /* Look for an existing address with the same prefix, mask, and fib */
1014 NET_EPOCH_ENTER(et);
1015 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1016 ia_getrtprefix(ia, &p, &m);
1017
1018 if (prefix.s_addr != p.s_addr ||
1019 mask.s_addr != m.s_addr)
1020 continue;
1021
1022 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
1023 continue;
1024
1025 /*
1026 * If we got a matching prefix route inserted by other
1027 * interface address, we are done here.
1028 */
1029 if (ia->ia_flags & IFA_ROUTE) {
1030 result = true;
1031 break;
1032 }
1033 }
1034 NET_EPOCH_EXIT(et);
1035
1036 return (result);
1037 }
1038
1039 int
1040 in_addprefix(struct in_ifaddr *target)
1041 {
1042 int error;
1043
1044 if (in_hasrtprefix(target)) {
1045 if (V_nosameprefix)
1046 return (EEXIST);
1047 else {
1048 rt_addrmsg(RTM_ADD, &target->ia_ifa,
1049 target->ia_ifp->if_fib);
1050 return (0);
1051 }
1052 }
1053
1054 /*
1055 * No-one seem to have this prefix route, so we try to insert it.
1056 */
1057 rt_addrmsg(RTM_ADD, &target->ia_ifa, target->ia_ifp->if_fib);
1058 error = in_handle_ifaddr_route(RTM_ADD, target);
1059 if (!error)
1060 target->ia_flags |= IFA_ROUTE;
1061 return (error);
1062 }
1063
1064 /*
1065 * Removes either all lle entries for given @ia, or lle
1066 * corresponding to @ia address.
1067 */
1068 static void
1069 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
1070 {
1071 struct sockaddr_in addr, mask;
1072 struct sockaddr *saddr, *smask;
1073 struct ifnet *ifp;
1074
1075 saddr = (struct sockaddr *)&addr;
1076 bzero(&addr, sizeof(addr));
1077 addr.sin_len = sizeof(addr);
1078 addr.sin_family = AF_INET;
1079 smask = (struct sockaddr *)&mask;
1080 bzero(&mask, sizeof(mask));
1081 mask.sin_len = sizeof(mask);
1082 mask.sin_family = AF_INET;
1083 mask.sin_addr.s_addr = ia->ia_subnetmask;
1084 ifp = ia->ia_ifp;
1085
1086 if (all) {
1087 /*
1088 * Remove all L2 entries matching given prefix.
1089 * Convert address to host representation to avoid
1090 * doing this on every callback. ia_subnetmask is already
1091 * stored in host representation.
1092 */
1093 addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
1094 lltable_prefix_free(AF_INET, saddr, smask, flags);
1095 } else {
1096 /* Remove interface address only */
1097 addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
1098 lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
1099 }
1100 }
1101
1102 /*
1103 * If there is no other address in the system that can serve a route to the
1104 * same prefix, remove the route. Hand over the route to the new address
1105 * otherwise.
1106 */
1107 int
1108 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1109 {
1110 struct epoch_tracker et;
1111 struct in_ifaddr *ia;
1112 struct in_addr prefix, mask, p, m;
1113 int error = 0;
1114
1115 /*
1116 * Remove the loopback route to the interface address.
1117 */
1118 if (ia_need_loopback_route(target) && (flags & LLE_STATIC)) {
1119 struct in_ifaddr *eia;
1120
1121 eia = in_localip_more(target);
1122
1123 if (eia != NULL) {
1124 error = ifa_switch_loopback_route((struct ifaddr *)eia,
1125 (struct sockaddr *)&target->ia_addr);
1126 ifa_free(&eia->ia_ifa);
1127 } else {
1128 error = ifa_del_loopback_route((struct ifaddr *)target,
1129 (struct sockaddr *)&target->ia_addr);
1130 }
1131 }
1132
1133 ia_getrtprefix(target, &prefix, &mask);
1134
1135 if ((target->ia_flags & IFA_ROUTE) == 0) {
1136 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1137
1138 /*
1139 * Removing address from !IFF_UP interface or
1140 * prefix which exists on other interface (along with route).
1141 * No entries should exist here except target addr.
1142 * Given that, delete this entry only.
1143 */
1144 in_scrubprefixlle(target, 0, flags);
1145 return (0);
1146 }
1147
1148 NET_EPOCH_ENTER(et);
1149 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1150 ia_getrtprefix(ia, &p, &m);
1151
1152 if (prefix.s_addr != p.s_addr ||
1153 mask.s_addr != m.s_addr)
1154 continue;
1155
1156 if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1157 continue;
1158
1159 /*
1160 * If we got a matching prefix address, move IFA_ROUTE and
1161 * the route itself to it. Make sure that routing daemons
1162 * get a heads-up.
1163 */
1164 if ((ia->ia_flags & IFA_ROUTE) == 0) {
1165 ifa_ref(&ia->ia_ifa);
1166 NET_EPOCH_EXIT(et);
1167 error = in_handle_ifaddr_route(RTM_DELETE, target);
1168 if (error == 0)
1169 target->ia_flags &= ~IFA_ROUTE;
1170 else
1171 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1172 error);
1173 /* Scrub all entries IFF interface is different */
1174 in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
1175 flags);
1176 error = in_handle_ifaddr_route(RTM_ADD, ia);
1177 if (error == 0)
1178 ia->ia_flags |= IFA_ROUTE;
1179 else
1180 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1181 error);
1182 ifa_free(&ia->ia_ifa);
1183 return (error);
1184 }
1185 }
1186 NET_EPOCH_EXIT(et);
1187
1188 /*
1189 * remove all L2 entries on the given prefix
1190 */
1191 in_scrubprefixlle(target, 1, flags);
1192
1193 /*
1194 * As no-one seem to have this prefix, we can remove the route.
1195 */
1196 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1197 error = in_handle_ifaddr_route(RTM_DELETE, target);
1198 if (error == 0)
1199 target->ia_flags &= ~IFA_ROUTE;
1200 else
1201 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1202 return (error);
1203 }
1204
1205 void
1206 in_ifscrub_all(void)
1207 {
1208 struct ifnet *ifp;
1209 struct ifaddr *ifa, *nifa;
1210 struct ifaliasreq ifr;
1211
1212 IFNET_RLOCK();
1213 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1214 /* Cannot lock here - lock recursion. */
1215 /* NET_EPOCH_ENTER(et); */
1216 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1217 if (ifa->ifa_addr->sa_family != AF_INET)
1218 continue;
1219
1220 /*
1221 * This is ugly but the only way for legacy IP to
1222 * cleanly remove addresses and everything attached.
1223 */
1224 bzero(&ifr, sizeof(ifr));
1225 ifr.ifra_addr = *ifa->ifa_addr;
1226 if (ifa->ifa_dstaddr)
1227 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1228 (void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
1229 ifp, NULL);
1230 }
1231 /* NET_EPOCH_EXIT(et); */
1232 in_purgemaddrs(ifp);
1233 igmp_domifdetach(ifp);
1234 }
1235 IFNET_RUNLOCK();
1236 }
1237
1238 int
1239 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
1240 {
1241
1242 return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1243 /*
1244 * Optionally check for old-style (host 0) broadcast, but
1245 * taking into account that RFC 3021 obsoletes it.
1246 */
1247 (V_broadcast_lowest && ia->ia_subnetmask != IN_RFC3021_MASK &&
1248 ntohl(in.s_addr) == ia->ia_subnet)) &&
1249 /*
1250 * Check for an all one subnetmask. These
1251 * only exist when an interface gets a secondary
1252 * address.
1253 */
1254 ia->ia_subnetmask != (u_long)0xffffffff);
1255 }
1256
1257 /*
1258 * Return 1 if the address might be a local broadcast address.
1259 */
1260 int
1261 in_broadcast(struct in_addr in, struct ifnet *ifp)
1262 {
1263 struct ifaddr *ifa;
1264 int found;
1265
1266 NET_EPOCH_ASSERT();
1267
1268 if (in.s_addr == INADDR_BROADCAST ||
1269 in.s_addr == INADDR_ANY)
1270 return (1);
1271 if ((ifp->if_flags & IFF_BROADCAST) == 0)
1272 return (0);
1273 found = 0;
1274 /*
1275 * Look through the list of addresses for a match
1276 * with a broadcast address.
1277 */
1278 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1279 if (ifa->ifa_addr->sa_family == AF_INET &&
1280 in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) {
1281 found = 1;
1282 break;
1283 }
1284 return (found);
1285 }
1286
1287 /*
1288 * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1289 */
1290 void
1291 in_ifdetach(struct ifnet *ifp)
1292 {
1293 IN_MULTI_LOCK();
1294 in_pcbpurgeif0(&V_ripcbinfo, ifp);
1295 in_pcbpurgeif0(&V_udbinfo, ifp);
1296 in_pcbpurgeif0(&V_ulitecbinfo, ifp);
1297 in_purgemaddrs(ifp);
1298 IN_MULTI_UNLOCK();
1299
1300 /*
1301 * Make sure all multicast deletions invoking if_ioctl() are
1302 * completed before returning. Else we risk accessing a freed
1303 * ifnet structure pointer.
1304 */
1305 inm_release_wait(NULL);
1306 }
1307
1308 static void
1309 in_ifnet_event(void *arg __unused, struct ifnet *ifp, int event)
1310 {
1311 struct epoch_tracker et;
1312 struct ifaddr *ifa;
1313 struct in_ifaddr *ia;
1314 int error;
1315
1316 NET_EPOCH_ENTER(et);
1317 switch (event) {
1318 case IFNET_EVENT_DOWN:
1319 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1320 if (ifa->ifa_addr->sa_family != AF_INET)
1321 continue;
1322 ia = (struct in_ifaddr *)ifa;
1323 if ((ia->ia_flags & IFA_ROUTE) == 0)
1324 continue;
1325 ifa_ref(ifa);
1326 /*
1327 * in_scrubprefix() kills the interface route.
1328 */
1329 in_scrubprefix(ia, 0);
1330 /*
1331 * in_ifadown gets rid of all the rest of the
1332 * routes. This is not quite the right thing
1333 * to do, but at least if we are running a
1334 * routing process they will come back.
1335 */
1336 in_ifadown(ifa, 0);
1337 ifa_free(ifa);
1338 }
1339 break;
1340
1341 case IFNET_EVENT_UP:
1342 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1343 if (ifa->ifa_addr->sa_family != AF_INET)
1344 continue;
1345 ia = (struct in_ifaddr *)ifa;
1346 if (ia->ia_flags & IFA_ROUTE)
1347 continue;
1348 ifa_ref(ifa);
1349 error = ifa_del_loopback_route(ifa, ifa->ifa_addr);
1350 rt_addrmsg(RTM_ADD, ifa, ifa->ifa_ifp->if_fib);
1351 error = in_handle_ifaddr_route(RTM_ADD, ia);
1352 if (error == 0)
1353 ia->ia_flags |= IFA_ROUTE;
1354 error = ifa_add_loopback_route(ifa, ifa->ifa_addr);
1355 ifa_free(ifa);
1356 }
1357 break;
1358 }
1359 NET_EPOCH_EXIT(et);
1360 }
1361 EVENTHANDLER_DEFINE(ifnet_event, in_ifnet_event, NULL, EVENTHANDLER_PRI_ANY);
1362
1363 /*
1364 * Delete all IPv4 multicast address records, and associated link-layer
1365 * multicast address records, associated with ifp.
1366 * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1367 * XXX This should not race with ifma_protospec being set during
1368 * a new allocation, if it does, we have bigger problems.
1369 */
1370 static void
1371 in_purgemaddrs(struct ifnet *ifp)
1372 {
1373 struct epoch_tracker et;
1374 struct in_multi_head purgeinms;
1375 struct in_multi *inm;
1376 struct ifmultiaddr *ifma;
1377
1378 SLIST_INIT(&purgeinms);
1379 IN_MULTI_LIST_LOCK();
1380
1381 /*
1382 * Extract list of in_multi associated with the detaching ifp
1383 * which the PF_INET layer is about to release.
1384 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1385 * by code further down.
1386 */
1387 IF_ADDR_WLOCK(ifp);
1388 NET_EPOCH_ENTER(et);
1389 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1390 inm = inm_ifmultiaddr_get_inm(ifma);
1391 if (inm == NULL)
1392 continue;
1393 inm_rele_locked(&purgeinms, inm);
1394 }
1395 NET_EPOCH_EXIT(et);
1396 IF_ADDR_WUNLOCK(ifp);
1397
1398 inm_release_list_deferred(&purgeinms);
1399 igmp_ifdetach(ifp);
1400 IN_MULTI_LIST_UNLOCK();
1401 }
1402
1403 struct in_llentry {
1404 struct llentry base;
1405 };
1406
1407 #define IN_LLTBL_DEFAULT_HSIZE 32
1408 #define IN_LLTBL_HASH(k, h) \
1409 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1410
1411 /*
1412 * Do actual deallocation of @lle.
1413 */
1414 static void
1415 in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1416 {
1417 struct llentry *lle;
1418
1419 lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1420 LLE_LOCK_DESTROY(lle);
1421 LLE_REQ_DESTROY(lle);
1422 free(lle, M_LLTABLE);
1423 }
1424
1425 /*
1426 * Called by LLE_FREE_LOCKED when number of references
1427 * drops to zero.
1428 */
1429 static void
1430 in_lltable_destroy_lle(struct llentry *lle)
1431 {
1432
1433 LLE_WUNLOCK(lle);
1434 NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1435 }
1436
1437 static struct llentry *
1438 in_lltable_new(struct in_addr addr4, u_int flags)
1439 {
1440 struct in_llentry *lle;
1441
1442 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1443 if (lle == NULL) /* NB: caller generates msg */
1444 return NULL;
1445
1446 /*
1447 * For IPv4 this will trigger "arpresolve" to generate
1448 * an ARP request.
1449 */
1450 lle->base.la_expire = time_uptime; /* mark expired */
1451 lle->base.r_l3addr.addr4 = addr4;
1452 lle->base.lle_refcnt = 1;
1453 lle->base.lle_free = in_lltable_destroy_lle;
1454 LLE_LOCK_INIT(&lle->base);
1455 LLE_REQ_INIT(&lle->base);
1456 callout_init(&lle->base.lle_timer, 1);
1457
1458 return (&lle->base);
1459 }
1460
1461 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \
1462 ((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 )
1463
1464 static int
1465 in_lltable_match_prefix(const struct sockaddr *saddr,
1466 const struct sockaddr *smask, u_int flags, struct llentry *lle)
1467 {
1468 struct in_addr addr, mask, lle_addr;
1469
1470 addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1471 mask = ((const struct sockaddr_in *)smask)->sin_addr;
1472 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1473
1474 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1475 return (0);
1476
1477 if (lle->la_flags & LLE_IFADDR) {
1478 /*
1479 * Delete LLE_IFADDR records IFF address & flag matches.
1480 * Note that addr is the interface address within prefix
1481 * being matched.
1482 * Note also we should handle 'ifdown' cases without removing
1483 * ifaddr macs.
1484 */
1485 if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1486 return (1);
1487 return (0);
1488 }
1489
1490 /* flags & LLE_STATIC means deleting both dynamic and static entries */
1491 if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1492 return (1);
1493
1494 return (0);
1495 }
1496
1497 static void
1498 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1499 {
1500 size_t pkts_dropped;
1501
1502 LLE_WLOCK_ASSERT(lle);
1503 KASSERT(llt != NULL, ("lltable is NULL"));
1504
1505 /* Unlink entry from table if not already */
1506 if ((lle->la_flags & LLE_LINKED) != 0) {
1507 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1508 lltable_unlink_entry(llt, lle);
1509 }
1510
1511 /* Drop hold queue */
1512 pkts_dropped = llentry_free(lle);
1513 ARPSTAT_ADD(dropped, pkts_dropped);
1514 }
1515
1516 static int
1517 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1518 {
1519 struct nhop_object *nh;
1520 struct in_addr addr;
1521
1522 KASSERT(l3addr->sa_family == AF_INET,
1523 ("sin_family %d", l3addr->sa_family));
1524
1525 addr = ((const struct sockaddr_in *)l3addr)->sin_addr;
1526
1527 nh = fib4_lookup(ifp->if_fib, addr, 0, NHR_NONE, 0);
1528 if (nh == NULL)
1529 return (EINVAL);
1530
1531 /*
1532 * If the gateway for an existing host route matches the target L3
1533 * address, which is a special route inserted by some implementation
1534 * such as MANET, and the interface is of the correct type, then
1535 * allow for ARP to proceed.
1536 */
1537 if (nh->nh_flags & NHF_GATEWAY) {
1538 if (!(nh->nh_flags & NHF_HOST) || nh->nh_ifp->if_type != IFT_ETHER ||
1539 (nh->nh_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1540 memcmp(nh->gw_sa.sa_data, l3addr->sa_data,
1541 sizeof(in_addr_t)) != 0) {
1542 return (EINVAL);
1543 }
1544 }
1545
1546 /*
1547 * Make sure that at least the destination address is covered
1548 * by the route. This is for handling the case where 2 or more
1549 * interfaces have the same prefix. An incoming packet arrives
1550 * on one interface and the corresponding outgoing packet leaves
1551 * another interface.
1552 */
1553 if ((nh->nh_ifp != ifp) && (nh->nh_flags & NHF_HOST) == 0) {
1554 struct in_ifaddr *ia = (struct in_ifaddr *)ifaof_ifpforaddr(l3addr, ifp);
1555 struct in_addr dst_addr, mask_addr;
1556
1557 if (ia == NULL)
1558 return (EINVAL);
1559
1560 /*
1561 * ifaof_ifpforaddr() returns _best matching_ IFA.
1562 * It is possible that ifa prefix does not cover our address.
1563 * Explicitly verify and fail if that's the case.
1564 */
1565 dst_addr = IA_SIN(ia)->sin_addr;
1566 mask_addr.s_addr = htonl(ia->ia_subnetmask);
1567
1568 if (!IN_ARE_MASKED_ADDR_EQUAL(dst_addr, addr, mask_addr))
1569 return (EINVAL);
1570 }
1571
1572 return (0);
1573 }
1574
1575 static inline uint32_t
1576 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1577 {
1578
1579 return (IN_LLTBL_HASH(dst.s_addr, hsize));
1580 }
1581
1582 static uint32_t
1583 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1584 {
1585
1586 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1587 }
1588
1589 static void
1590 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1591 {
1592 struct sockaddr_in *sin;
1593
1594 sin = (struct sockaddr_in *)sa;
1595 bzero(sin, sizeof(*sin));
1596 sin->sin_family = AF_INET;
1597 sin->sin_len = sizeof(*sin);
1598 sin->sin_addr = lle->r_l3addr.addr4;
1599 }
1600
1601 static inline struct llentry *
1602 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1603 {
1604 struct llentry *lle;
1605 struct llentries *lleh;
1606 u_int hashidx;
1607
1608 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1609 lleh = &llt->lle_head[hashidx];
1610 CK_LIST_FOREACH(lle, lleh, lle_next) {
1611 if (lle->la_flags & LLE_DELETED)
1612 continue;
1613 if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1614 break;
1615 }
1616
1617 return (lle);
1618 }
1619
1620 static void
1621 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1622 {
1623
1624 lle->la_flags |= LLE_DELETED;
1625 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1626 #ifdef DIAGNOSTIC
1627 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1628 #endif
1629 llentry_free(lle);
1630 }
1631
1632 static struct llentry *
1633 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1634 {
1635 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1636 struct ifnet *ifp = llt->llt_ifp;
1637 struct llentry *lle;
1638 char linkhdr[LLE_MAX_LINKHDR];
1639 size_t linkhdrsize;
1640 int lladdr_off;
1641
1642 KASSERT(l3addr->sa_family == AF_INET,
1643 ("sin_family %d", l3addr->sa_family));
1644
1645 /*
1646 * A route that covers the given address must have
1647 * been installed 1st because we are doing a resolution,
1648 * verify this.
1649 */
1650 if (!(flags & LLE_IFADDR) &&
1651 in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1652 return (NULL);
1653
1654 lle = in_lltable_new(sin->sin_addr, flags);
1655 if (lle == NULL) {
1656 log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1657 return (NULL);
1658 }
1659 lle->la_flags = flags;
1660 if (flags & LLE_STATIC)
1661 lle->r_flags |= RLLE_VALID;
1662 if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1663 linkhdrsize = LLE_MAX_LINKHDR;
1664 if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1665 linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1666 in_lltable_free_entry(llt, lle);
1667 return (NULL);
1668 }
1669 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1670 lladdr_off);
1671 lle->la_flags |= LLE_STATIC;
1672 lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1673 }
1674
1675 return (lle);
1676 }
1677
1678 /*
1679 * Return NULL if not found or marked for deletion.
1680 * If found return lle read locked.
1681 */
1682 static struct llentry *
1683 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1684 {
1685 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1686 struct llentry *lle;
1687
1688 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1689 KASSERT(l3addr->sa_family == AF_INET,
1690 ("sin_family %d", l3addr->sa_family));
1691 KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1692 (LLE_UNLOCKED | LLE_EXCLUSIVE),
1693 ("wrong lle request flags: %#x", flags));
1694
1695 lle = in_lltable_find_dst(llt, sin->sin_addr);
1696 if (lle == NULL)
1697 return (NULL);
1698 if (flags & LLE_UNLOCKED)
1699 return (lle);
1700
1701 if (flags & LLE_EXCLUSIVE)
1702 LLE_WLOCK(lle);
1703 else
1704 LLE_RLOCK(lle);
1705
1706 /*
1707 * If the afdata lock is not held, the LLE may have been unlinked while
1708 * we were blocked on the LLE lock. Check for this case.
1709 */
1710 if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1711 if (flags & LLE_EXCLUSIVE)
1712 LLE_WUNLOCK(lle);
1713 else
1714 LLE_RUNLOCK(lle);
1715 return (NULL);
1716 }
1717 return (lle);
1718 }
1719
1720 static int
1721 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1722 struct sysctl_req *wr)
1723 {
1724 struct ifnet *ifp = llt->llt_ifp;
1725 /* XXX stack use */
1726 struct {
1727 struct rt_msghdr rtm;
1728 struct sockaddr_in sin;
1729 struct sockaddr_dl sdl;
1730 } arpc;
1731 struct sockaddr_dl *sdl;
1732 int error;
1733
1734 bzero(&arpc, sizeof(arpc));
1735 /* skip deleted entries */
1736 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1737 return (0);
1738 /* Skip if jailed and not a valid IP of the prison. */
1739 lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1740 if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1741 return (0);
1742 /*
1743 * produce a msg made of:
1744 * struct rt_msghdr;
1745 * struct sockaddr_in; (IPv4)
1746 * struct sockaddr_dl;
1747 */
1748 arpc.rtm.rtm_msglen = sizeof(arpc);
1749 arpc.rtm.rtm_version = RTM_VERSION;
1750 arpc.rtm.rtm_type = RTM_GET;
1751 arpc.rtm.rtm_flags = RTF_UP;
1752 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1753
1754 /* publish */
1755 if (lle->la_flags & LLE_PUB)
1756 arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1757
1758 sdl = &arpc.sdl;
1759 sdl->sdl_family = AF_LINK;
1760 sdl->sdl_len = sizeof(*sdl);
1761 sdl->sdl_index = ifp->if_index;
1762 sdl->sdl_type = ifp->if_type;
1763 if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1764 sdl->sdl_alen = ifp->if_addrlen;
1765 bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1766 } else {
1767 sdl->sdl_alen = 0;
1768 bzero(LLADDR(sdl), ifp->if_addrlen);
1769 }
1770
1771 arpc.rtm.rtm_rmx.rmx_expire =
1772 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1773 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1774 if (lle->la_flags & LLE_STATIC)
1775 arpc.rtm.rtm_flags |= RTF_STATIC;
1776 if (lle->la_flags & LLE_IFADDR)
1777 arpc.rtm.rtm_flags |= RTF_PINNED;
1778 arpc.rtm.rtm_index = ifp->if_index;
1779 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1780
1781 return (error);
1782 }
1783
1784 static void
1785 in_lltable_post_resolved(struct lltable *llt, struct llentry *lle)
1786 {
1787 struct ifnet *ifp = llt->llt_ifp;
1788
1789 /* gratuitous ARP */
1790 if ((lle->la_flags & LLE_PUB) != 0)
1791 arprequest(ifp, &lle->r_l3addr.addr4, &lle->r_l3addr.addr4,
1792 lle->ll_addr);
1793 }
1794
1795 static struct lltable *
1796 in_lltattach(struct ifnet *ifp)
1797 {
1798 struct lltable *llt;
1799
1800 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1801 llt->llt_af = AF_INET;
1802 llt->llt_ifp = ifp;
1803
1804 llt->llt_lookup = in_lltable_lookup;
1805 llt->llt_alloc_entry = in_lltable_alloc;
1806 llt->llt_delete_entry = in_lltable_delete_entry;
1807 llt->llt_dump_entry = in_lltable_dump_entry;
1808 llt->llt_hash = in_lltable_hash;
1809 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1810 llt->llt_free_entry = in_lltable_free_entry;
1811 llt->llt_match_prefix = in_lltable_match_prefix;
1812 llt->llt_mark_used = llentry_mark_used;
1813 llt->llt_post_resolved = in_lltable_post_resolved;
1814 lltable_link(llt);
1815
1816 return (llt);
1817 }
1818
1819 struct lltable *
1820 in_lltable_get(struct ifnet *ifp)
1821 {
1822 struct lltable *llt = NULL;
1823
1824 void *afdata_ptr = ifp->if_afdata[AF_INET];
1825 if (afdata_ptr != NULL)
1826 llt = ((struct in_ifinfo *)afdata_ptr)->ii_llt;
1827 return (llt);
1828 }
1829
1830 void *
1831 in_domifattach(struct ifnet *ifp)
1832 {
1833 struct in_ifinfo *ii;
1834
1835 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1836
1837 ii->ii_llt = in_lltattach(ifp);
1838 ii->ii_igmp = igmp_domifattach(ifp);
1839
1840 return (ii);
1841 }
1842
1843 void
1844 in_domifdetach(struct ifnet *ifp, void *aux)
1845 {
1846 struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1847
1848 igmp_domifdetach(ifp);
1849 lltable_free(ii->ii_llt);
1850 free(ii, M_IFADDR);
1851 }
Cache object: b5f9265e232b857b8a15de7c0a916fe5
|