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 #include "opt_inet.h"
37
38 #define IN_HISTORICAL_NETS /* include class masks */
39
40 #include <sys/param.h>
41 #include <sys/eventhandler.h>
42 #include <sys/systm.h>
43 #include <sys/sockio.h>
44 #include <sys/malloc.h>
45 #include <sys/priv.h>
46 #include <sys/socket.h>
47 #include <sys/jail.h>
48 #include <sys/kernel.h>
49 #include <sys/lock.h>
50 #include <sys/proc.h>
51 #include <sys/sysctl.h>
52 #include <sys/syslog.h>
53 #include <sys/sx.h>
54
55 #include <net/if.h>
56 #include <net/if_var.h>
57 #include <net/if_arp.h>
58 #include <net/if_dl.h>
59 #include <net/if_llatbl.h>
60 #include <net/if_private.h>
61 #include <net/if_types.h>
62 #include <net/route.h>
63 #include <net/route/nhop.h>
64 #include <net/route/route_ctl.h>
65 #include <net/vnet.h>
66
67 #include <netinet/if_ether.h>
68 #include <netinet/in.h>
69 #include <netinet/in_fib.h>
70 #include <netinet/in_var.h>
71 #include <netinet/in_pcb.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/ip_carp.h>
74 #include <netinet/igmp_var.h>
75 #include <netinet/udp.h>
76 #include <netinet/udp_var.h>
77
78 #ifdef MAC
79 #include <security/mac/mac_framework.h>
80 #endif
81
82 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
83 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
84 static int in_gifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
85
86 static void in_socktrim(struct sockaddr_in *);
87 static void in_purgemaddrs(struct ifnet *);
88
89 static bool ia_need_loopback_route(const struct in_ifaddr *);
90
91 VNET_DEFINE_STATIC(int, nosameprefix);
92 #define V_nosameprefix VNET(nosameprefix)
93 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW,
94 &VNET_NAME(nosameprefix), 0,
95 "Refuse to create same prefixes on different interfaces");
96
97 VNET_DEFINE_STATIC(bool, broadcast_lowest);
98 #define V_broadcast_lowest VNET(broadcast_lowest)
99 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, broadcast_lowest, CTLFLAG_VNET | CTLFLAG_RW,
100 &VNET_NAME(broadcast_lowest), 0,
101 "Treat lowest address on a subnet (host 0) as broadcast");
102
103 VNET_DEFINE(bool, ip_allow_net240) = false;
104 #define V_ip_allow_net240 VNET(ip_allow_net240)
105 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net240,
106 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net240), 0,
107 "Allow use of Experimental addresses, aka Class E (240/4)");
108 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-240 */
109
110 VNET_DEFINE(bool, ip_allow_net0) = false;
111 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net0,
112 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net0), 0,
113 "Allow use of addresses in network 0/8");
114 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-0 */
115
116 VNET_DEFINE(uint32_t, in_loopback_mask) = IN_LOOPBACK_MASK_DFLT;
117 #define V_in_loopback_mask VNET(in_loopback_mask)
118 static int sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS);
119 SYSCTL_PROC(_net_inet_ip, OID_AUTO, loopback_prefixlen,
120 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
121 NULL, 0, sysctl_loopback_prefixlen, "I",
122 "Prefix length of address space reserved for loopback");
123 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-127 */
124
125 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
126 #define V_ripcbinfo VNET(ripcbinfo)
127
128 static struct sx in_control_sx;
129 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control");
130
131 /*
132 * Return true if an internet address is for a ``local'' host
133 * (one to which we have a connection).
134 */
135 bool
in_localaddr(struct in_addr in)136 in_localaddr(struct in_addr in)
137 {
138 u_long i = ntohl(in.s_addr);
139 struct in_ifaddr *ia;
140
141 NET_EPOCH_ASSERT();
142
143 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
144 if ((i & ia->ia_subnetmask) == ia->ia_subnet)
145 return (true);
146 }
147
148 return (false);
149 }
150
151 /*
152 * Return true if an internet address is for the local host and configured
153 * on one of its interfaces.
154 */
155 bool
in_localip(struct in_addr in)156 in_localip(struct in_addr in)
157 {
158 struct in_ifaddr *ia;
159
160 NET_EPOCH_ASSERT();
161
162 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
163 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
164 return (true);
165
166 return (false);
167 }
168
169 /*
170 * Like in_localip(), but FIB-aware and carp(4)-aware.
171 */
172 bool
in_localip_fib(struct in_addr in,uint16_t fib)173 in_localip_fib(struct in_addr in, uint16_t fib)
174 {
175 struct in_ifaddr *ia;
176
177 NET_EPOCH_ASSERT();
178
179 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
180 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr &&
181 (ia->ia_ifa.ifa_carp == NULL ||
182 carp_master_p(&ia->ia_ifa)) &&
183 ia->ia_ifa.ifa_ifp->if_fib == fib)
184 return (true);
185
186 return (false);
187 }
188
189 /*
190 * Return true if an internet address is configured on an interface.
191 */
192 bool
in_ifhasaddr(struct ifnet * ifp,struct in_addr in)193 in_ifhasaddr(struct ifnet *ifp, struct in_addr in)
194 {
195 struct ifaddr *ifa;
196 struct in_ifaddr *ia;
197
198 NET_EPOCH_ASSERT();
199
200 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
201 if (ifa->ifa_addr->sa_family != AF_INET)
202 continue;
203 ia = (struct in_ifaddr *)ifa;
204 if (ia->ia_addr.sin_addr.s_addr == in.s_addr)
205 return (true);
206 }
207
208 return (false);
209 }
210
211 /*
212 * Return a reference to the interface address which is different to
213 * the supplied one but with same IP address value.
214 */
215 static struct in_ifaddr *
in_localip_more(struct in_ifaddr * original_ia)216 in_localip_more(struct in_ifaddr *original_ia)
217 {
218 struct epoch_tracker et;
219 in_addr_t original_addr = IA_SIN(original_ia)->sin_addr.s_addr;
220 uint32_t original_fib = original_ia->ia_ifa.ifa_ifp->if_fib;
221 struct in_ifaddr *ia;
222
223 NET_EPOCH_ENTER(et);
224 CK_LIST_FOREACH(ia, INADDR_HASH(original_addr), ia_hash) {
225 in_addr_t addr = IA_SIN(ia)->sin_addr.s_addr;
226 uint32_t fib = ia->ia_ifa.ifa_ifp->if_fib;
227 if (!V_rt_add_addr_allfibs && (original_fib != fib))
228 continue;
229 if ((original_ia != ia) && (original_addr == addr)) {
230 ifa_ref(&ia->ia_ifa);
231 NET_EPOCH_EXIT(et);
232 return (ia);
233 }
234 }
235 NET_EPOCH_EXIT(et);
236
237 return (NULL);
238 }
239
240 /*
241 * Tries to find first IPv4 address in the provided fib.
242 * Prefers non-loopback addresses and return loopback IFF
243 * @loopback_ok is set.
244 *
245 * Returns ifa or NULL.
246 */
247 struct in_ifaddr *
in_findlocal(uint32_t fibnum,bool loopback_ok)248 in_findlocal(uint32_t fibnum, bool loopback_ok)
249 {
250 struct in_ifaddr *ia = NULL, *ia_lo = NULL;
251
252 NET_EPOCH_ASSERT();
253
254 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
255 uint32_t ia_fib = ia->ia_ifa.ifa_ifp->if_fib;
256 if (!V_rt_add_addr_allfibs && (fibnum != ia_fib))
257 continue;
258
259 if (!IN_LOOPBACK(ntohl(IA_SIN(ia)->sin_addr.s_addr)))
260 break;
261 if (loopback_ok)
262 ia_lo = ia;
263 }
264
265 if (ia == NULL)
266 ia = ia_lo;
267
268 return (ia);
269 }
270
271 /*
272 * Determine whether an IP address is in a reserved set of addresses
273 * that may not be forwarded, or whether datagrams to that destination
274 * may be forwarded.
275 */
276 bool
in_canforward(struct in_addr in)277 in_canforward(struct in_addr in)
278 {
279 u_long i = ntohl(in.s_addr);
280
281 if (IN_MULTICAST(i) || IN_LINKLOCAL(i) || IN_LOOPBACK(i) ||
282 in_nullhost(in))
283 return (false);
284 if (IN_EXPERIMENTAL(i) && !V_ip_allow_net240)
285 return (false);
286 if (IN_ZERONET(i) && !V_ip_allow_net0)
287 return (false);
288 return (true);
289 }
290
291 /*
292 * Sysctl to manage prefix of reserved loopback network; translate
293 * to/from mask. The mask is always contiguous high-order 1 bits
294 * followed by all 0 bits.
295 */
296 static int
sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS)297 sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS)
298 {
299 int error, preflen;
300
301 /* ffs is 1-based; compensate. */
302 preflen = 33 - ffs(V_in_loopback_mask);
303 error = sysctl_handle_int(oidp, &preflen, 0, req);
304 if (error || !req->newptr)
305 return (error);
306 if (preflen < 8 || preflen > 31)
307 return (EINVAL);
308 V_in_loopback_mask = 0xffffffff << (32 - preflen);
309 return (0);
310 }
311
312 /*
313 * Trim a mask in a sockaddr
314 */
315 static void
in_socktrim(struct sockaddr_in * ap)316 in_socktrim(struct sockaddr_in *ap)
317 {
318 char *cplim = (char *) &ap->sin_addr;
319 char *cp = (char *) (&ap->sin_addr + 1);
320
321 ap->sin_len = 0;
322 while (--cp >= cplim)
323 if (*cp) {
324 (ap)->sin_len = cp - (char *) (ap) + 1;
325 break;
326 }
327 }
328
329 /*
330 * Generic internet control operations (ioctl's).
331 */
332 int
in_control_ioctl(u_long cmd,void * data,struct ifnet * ifp,struct ucred * cred)333 in_control_ioctl(u_long cmd, void *data, struct ifnet *ifp,
334 struct ucred *cred)
335 {
336 struct ifreq *ifr = (struct ifreq *)data;
337 struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
338 struct epoch_tracker et;
339 struct ifaddr *ifa;
340 struct in_ifaddr *ia;
341 int error;
342
343 if (ifp == NULL)
344 return (EADDRNOTAVAIL);
345
346 /*
347 * Filter out 4 ioctls we implement directly. Forward the rest
348 * to specific functions and ifp->if_ioctl().
349 */
350 switch (cmd) {
351 case SIOCGIFADDR:
352 case SIOCGIFBRDADDR:
353 case SIOCGIFDSTADDR:
354 case SIOCGIFNETMASK:
355 break;
356 case SIOCGIFALIAS:
357 sx_xlock(&in_control_sx);
358 error = in_gifaddr_ioctl(cmd, data, ifp, cred);
359 sx_xunlock(&in_control_sx);
360 return (error);
361 case SIOCDIFADDR:
362 sx_xlock(&in_control_sx);
363 error = in_difaddr_ioctl(cmd, data, ifp, cred);
364 sx_xunlock(&in_control_sx);
365 return (error);
366 case OSIOCAIFADDR: /* 9.x compat */
367 case SIOCAIFADDR:
368 sx_xlock(&in_control_sx);
369 error = in_aifaddr_ioctl(cmd, data, ifp, cred);
370 sx_xunlock(&in_control_sx);
371 return (error);
372 case SIOCSIFADDR:
373 case SIOCSIFBRDADDR:
374 case SIOCSIFDSTADDR:
375 case SIOCSIFNETMASK:
376 /* We no longer support that old commands. */
377 return (EINVAL);
378 default:
379 if (ifp->if_ioctl == NULL)
380 return (EOPNOTSUPP);
381 return ((*ifp->if_ioctl)(ifp, cmd, data));
382 }
383
384 if (addr->sin_addr.s_addr != INADDR_ANY &&
385 prison_check_ip4(cred, &addr->sin_addr) != 0)
386 return (EADDRNOTAVAIL);
387
388 /*
389 * Find address for this interface, if it exists. If an
390 * address was specified, find that one instead of the
391 * first one on the interface, if possible.
392 */
393 NET_EPOCH_ENTER(et);
394 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
395 if (ifa->ifa_addr->sa_family != AF_INET)
396 continue;
397 ia = (struct in_ifaddr *)ifa;
398 if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
399 break;
400 }
401 if (ifa == NULL)
402 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
403 if (ifa->ifa_addr->sa_family == AF_INET) {
404 ia = (struct in_ifaddr *)ifa;
405 if (prison_check_ip4(cred,
406 &ia->ia_addr.sin_addr) == 0)
407 break;
408 }
409
410 if (ifa == NULL) {
411 NET_EPOCH_EXIT(et);
412 return (EADDRNOTAVAIL);
413 }
414
415 error = 0;
416 switch (cmd) {
417 case SIOCGIFADDR:
418 *addr = ia->ia_addr;
419 break;
420
421 case SIOCGIFBRDADDR:
422 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
423 error = EINVAL;
424 break;
425 }
426 *addr = ia->ia_broadaddr;
427 break;
428
429 case SIOCGIFDSTADDR:
430 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
431 error = EINVAL;
432 break;
433 }
434 *addr = ia->ia_dstaddr;
435 break;
436
437 case SIOCGIFNETMASK:
438 *addr = ia->ia_sockmask;
439 break;
440 }
441
442 NET_EPOCH_EXIT(et);
443
444 return (error);
445 }
446
447 int
in_control(struct socket * so,u_long cmd,void * data,struct ifnet * ifp,struct thread * td)448 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
449 struct thread *td)
450 {
451 return (in_control_ioctl(cmd, data, ifp, td ? td->td_ucred : NULL));
452 }
453
454 static int
in_aifaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)455 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
456 {
457 const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
458 const struct sockaddr_in *addr = &ifra->ifra_addr;
459 const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
460 const struct sockaddr_in *mask = &ifra->ifra_mask;
461 const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
462 const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
463 struct epoch_tracker et;
464 struct ifaddr *ifa;
465 struct in_ifaddr *ia;
466 bool iaIsFirst;
467 int error = 0;
468
469 error = priv_check_cred(cred, PRIV_NET_ADDIFADDR);
470 if (error)
471 return (error);
472
473 /*
474 * ifra_addr must be present and be of INET family.
475 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
476 */
477 if (addr->sin_len != sizeof(struct sockaddr_in) ||
478 addr->sin_family != AF_INET)
479 return (EINVAL);
480 if (broadaddr->sin_len != 0 &&
481 (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
482 broadaddr->sin_family != AF_INET))
483 return (EINVAL);
484 if (mask->sin_len != 0 &&
485 (mask->sin_len != sizeof(struct sockaddr_in) ||
486 mask->sin_family != AF_INET))
487 return (EINVAL);
488 if ((ifp->if_flags & IFF_POINTOPOINT) &&
489 (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
490 dstaddr->sin_addr.s_addr == INADDR_ANY))
491 return (EDESTADDRREQ);
492 if (vhid != 0 && carp_attach_p == NULL)
493 return (EPROTONOSUPPORT);
494
495 #ifdef MAC
496 /* Check if a MAC policy disallows setting the IPv4 address. */
497 error = mac_inet_check_add_addr(cred, &addr->sin_addr, ifp);
498 if (error != 0)
499 return (error);
500 #endif
501
502 /*
503 * See whether address already exist.
504 */
505 iaIsFirst = true;
506 ia = NULL;
507 NET_EPOCH_ENTER(et);
508 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
509 struct in_ifaddr *it;
510
511 if (ifa->ifa_addr->sa_family != AF_INET)
512 continue;
513
514 it = (struct in_ifaddr *)ifa;
515 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
516 prison_check_ip4(cred, &addr->sin_addr) == 0)
517 ia = it;
518 else
519 iaIsFirst = false;
520 }
521 NET_EPOCH_EXIT(et);
522
523 if (ia != NULL)
524 (void )in_difaddr_ioctl(cmd, data, ifp, cred);
525
526 ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
527 ia = (struct in_ifaddr *)ifa;
528 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
529 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
530 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
531 callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
532 CALLOUT_RETURNUNLOCKED);
533
534 ia->ia_ifp = ifp;
535 ia->ia_addr = *addr;
536 if (mask->sin_len != 0) {
537 ia->ia_sockmask = *mask;
538 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
539 } else {
540 in_addr_t i = ntohl(addr->sin_addr.s_addr);
541
542 /*
543 * If netmask isn't supplied, use historical default.
544 * This is deprecated for interfaces other than loopback
545 * or point-to-point; warn in other cases. In the future
546 * we should return an error rather than warning.
547 */
548 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0)
549 printf("%s: set address: WARNING: network mask "
550 "should be specified; using historical default\n",
551 ifp->if_xname);
552 if (IN_CLASSA(i))
553 ia->ia_subnetmask = IN_CLASSA_NET;
554 else if (IN_CLASSB(i))
555 ia->ia_subnetmask = IN_CLASSB_NET;
556 else
557 ia->ia_subnetmask = IN_CLASSC_NET;
558 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
559 }
560 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
561 in_socktrim(&ia->ia_sockmask);
562
563 if (ifp->if_flags & IFF_BROADCAST) {
564 if (broadaddr->sin_len != 0) {
565 ia->ia_broadaddr = *broadaddr;
566 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
567 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
568 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
569 ia->ia_broadaddr.sin_family = AF_INET;
570 } else {
571 ia->ia_broadaddr.sin_addr.s_addr =
572 htonl(ia->ia_subnet | ~ia->ia_subnetmask);
573 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
574 ia->ia_broadaddr.sin_family = AF_INET;
575 }
576 }
577
578 if (ifp->if_flags & IFF_POINTOPOINT)
579 ia->ia_dstaddr = *dstaddr;
580
581 if (vhid != 0) {
582 error = (*carp_attach_p)(&ia->ia_ifa, vhid);
583 if (error)
584 return (error);
585 }
586
587 /* if_addrhead is already referenced by ifa_alloc() */
588 IF_ADDR_WLOCK(ifp);
589 CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
590 IF_ADDR_WUNLOCK(ifp);
591
592 ifa_ref(ifa); /* in_ifaddrhead */
593 sx_assert(&in_control_sx, SA_XLOCKED);
594 CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
595 CK_LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia,
596 ia_hash);
597
598 /*
599 * Give the interface a chance to initialize
600 * if this is its first address,
601 * and to validate the address if necessary.
602 */
603 if (ifp->if_ioctl != NULL) {
604 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
605 if (error)
606 goto fail1;
607 }
608
609 /*
610 * Add route for the network.
611 */
612 if (vhid == 0) {
613 error = in_addprefix(ia);
614 if (error)
615 goto fail1;
616 }
617
618 /*
619 * Add a loopback route to self.
620 */
621 if (vhid == 0 && ia_need_loopback_route(ia)) {
622 struct in_ifaddr *eia;
623
624 eia = in_localip_more(ia);
625
626 if (eia == NULL) {
627 error = ifa_add_loopback_route((struct ifaddr *)ia,
628 (struct sockaddr *)&ia->ia_addr);
629 if (error)
630 goto fail2;
631 } else
632 ifa_free(&eia->ia_ifa);
633 }
634
635 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
636 struct in_addr allhosts_addr;
637 struct in_ifinfo *ii;
638
639 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
640 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
641
642 error = in_joingroup(ifp, &allhosts_addr, NULL,
643 &ii->ii_allhosts);
644 }
645
646 /*
647 * Note: we don't need extra reference for ifa, since we called
648 * with sx lock held, and ifaddr can not be deleted in concurrent
649 * thread.
650 */
651 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
652
653 return (error);
654
655 fail2:
656 if (vhid == 0)
657 (void )in_scrubprefix(ia, LLE_STATIC);
658
659 fail1:
660 if (ia->ia_ifa.ifa_carp)
661 (*carp_detach_p)(&ia->ia_ifa, false);
662
663 IF_ADDR_WLOCK(ifp);
664 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
665 IF_ADDR_WUNLOCK(ifp);
666 ifa_free(&ia->ia_ifa); /* if_addrhead */
667
668 sx_assert(&in_control_sx, SA_XLOCKED);
669 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
670 CK_LIST_REMOVE(ia, ia_hash);
671 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */
672
673 return (error);
674 }
675
676 static int
in_difaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)677 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
678 {
679 const struct ifreq *ifr = (struct ifreq *)data;
680 const struct sockaddr_in *addr = (const struct sockaddr_in *)
681 &ifr->ifr_addr;
682 struct ifaddr *ifa;
683 struct in_ifaddr *ia;
684 bool deleteAny, iaIsLast;
685 int error;
686
687 if (cred != NULL) {
688 error = priv_check_cred(cred, PRIV_NET_DELIFADDR);
689 if (error)
690 return (error);
691 }
692
693 if (addr->sin_len != sizeof(struct sockaddr_in) ||
694 addr->sin_family != AF_INET)
695 deleteAny = true;
696 else
697 deleteAny = false;
698
699 iaIsLast = true;
700 ia = NULL;
701 IF_ADDR_WLOCK(ifp);
702 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
703 struct in_ifaddr *it;
704
705 if (ifa->ifa_addr->sa_family != AF_INET)
706 continue;
707
708 it = (struct in_ifaddr *)ifa;
709 if (deleteAny && ia == NULL && (cred == NULL ||
710 prison_check_ip4(cred, &it->ia_addr.sin_addr) == 0))
711 ia = it;
712
713 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
714 (cred == NULL || prison_check_ip4(cred,
715 &addr->sin_addr) == 0))
716 ia = it;
717
718 if (it != ia)
719 iaIsLast = false;
720 }
721
722 if (ia == NULL) {
723 IF_ADDR_WUNLOCK(ifp);
724 return (EADDRNOTAVAIL);
725 }
726
727 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
728 IF_ADDR_WUNLOCK(ifp);
729 ifa_free(&ia->ia_ifa); /* if_addrhead */
730
731 sx_assert(&in_control_sx, SA_XLOCKED);
732 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
733 CK_LIST_REMOVE(ia, ia_hash);
734
735 /*
736 * in_scrubprefix() kills the interface route.
737 */
738 in_scrubprefix(ia, LLE_STATIC);
739
740 /*
741 * in_ifadown gets rid of all the rest of
742 * the routes. This is not quite the right
743 * thing to do, but at least if we are running
744 * a routing process they will come back.
745 */
746 in_ifadown(&ia->ia_ifa, 1);
747
748 if (ia->ia_ifa.ifa_carp)
749 (*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
750
751 /*
752 * If this is the last IPv4 address configured on this
753 * interface, leave the all-hosts group.
754 * No state-change report need be transmitted.
755 */
756 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
757 struct in_ifinfo *ii;
758
759 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
760 if (ii->ii_allhosts) {
761 (void)in_leavegroup(ii->ii_allhosts, NULL);
762 ii->ii_allhosts = NULL;
763 }
764 }
765
766 IF_ADDR_WLOCK(ifp);
767 if (callout_stop(&ia->ia_garp_timer) == 1) {
768 ifa_free(&ia->ia_ifa);
769 }
770 IF_ADDR_WUNLOCK(ifp);
771
772 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
773 IFADDR_EVENT_DEL);
774 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */
775
776 return (0);
777 }
778
779 static int
in_gifaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)780 in_gifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
781 {
782 struct in_aliasreq *ifra = (struct in_aliasreq *)data;
783 const struct sockaddr_in *addr = &ifra->ifra_addr;
784 struct epoch_tracker et;
785 struct ifaddr *ifa;
786 struct in_ifaddr *ia;
787
788 /*
789 * ifra_addr must be present and be of INET family.
790 */
791 if (addr->sin_len != sizeof(struct sockaddr_in) ||
792 addr->sin_family != AF_INET)
793 return (EINVAL);
794
795 /*
796 * See whether address exist.
797 */
798 ia = NULL;
799 NET_EPOCH_ENTER(et);
800 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
801 struct in_ifaddr *it;
802
803 if (ifa->ifa_addr->sa_family != AF_INET)
804 continue;
805
806 it = (struct in_ifaddr *)ifa;
807 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
808 prison_check_ip4(cred, &addr->sin_addr) == 0) {
809 ia = it;
810 break;
811 }
812 }
813 if (ia == NULL) {
814 NET_EPOCH_EXIT(et);
815 return (EADDRNOTAVAIL);
816 }
817
818 ifra->ifra_mask = ia->ia_sockmask;
819 if ((ifp->if_flags & IFF_POINTOPOINT) &&
820 ia->ia_dstaddr.sin_family == AF_INET)
821 ifra->ifra_dstaddr = ia->ia_dstaddr;
822 else if ((ifp->if_flags & IFF_BROADCAST) &&
823 ia->ia_broadaddr.sin_family == AF_INET)
824 ifra->ifra_broadaddr = ia->ia_broadaddr;
825 else
826 memset(&ifra->ifra_broadaddr, 0,
827 sizeof(ifra->ifra_broadaddr));
828
829 NET_EPOCH_EXIT(et);
830 return (0);
831 }
832
833 static int
in_match_ifaddr(const struct rtentry * rt,const struct nhop_object * nh,void * arg)834 in_match_ifaddr(const struct rtentry *rt, const struct nhop_object *nh, void *arg)
835 {
836
837 if (nh->nh_ifa == (struct ifaddr *)arg)
838 return (1);
839
840 return (0);
841 }
842
843 static int
in_handle_prefix_route(uint32_t fibnum,int cmd,struct sockaddr_in * dst,struct sockaddr_in * netmask,struct ifaddr * ifa,struct ifnet * ifp)844 in_handle_prefix_route(uint32_t fibnum, int cmd,
845 struct sockaddr_in *dst, struct sockaddr_in *netmask, struct ifaddr *ifa,
846 struct ifnet *ifp)
847 {
848
849 NET_EPOCH_ASSERT();
850
851 /* Prepare gateway */
852 struct sockaddr_dl_short sdl = {
853 .sdl_family = AF_LINK,
854 .sdl_len = sizeof(struct sockaddr_dl_short),
855 .sdl_type = ifa->ifa_ifp->if_type,
856 .sdl_index = ifa->ifa_ifp->if_index,
857 };
858
859 struct rt_addrinfo info = {
860 .rti_ifa = ifa,
861 .rti_ifp = ifp,
862 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
863 .rti_info = {
864 [RTAX_DST] = (struct sockaddr *)dst,
865 [RTAX_NETMASK] = (struct sockaddr *)netmask,
866 [RTAX_GATEWAY] = (struct sockaddr *)&sdl,
867 },
868 /* Ensure we delete the prefix IFF prefix ifa matches */
869 .rti_filter = in_match_ifaddr,
870 .rti_filterdata = ifa,
871 };
872
873 return (rib_handle_ifaddr_info(fibnum, cmd, &info));
874 }
875
876 /*
877 * Routing table interaction with interface addresses.
878 *
879 * In general, two types of routes needs to be installed:
880 * a) "interface" or "prefix" route, telling user that the addresses
881 * behind the ifa prefix are reached directly.
882 * b) "loopback" route installed for the ifa address, telling user that
883 * the address belongs to local system.
884 *
885 * Handling for (a) and (b) differs in multi-fib aspects, hence they
886 * are implemented in different functions below.
887 *
888 * The cases above may intersect - /32 interface aliases results in
889 * the same prefix produced by (a) and (b). This blurs the definition
890 * of the "loopback" route and complicate interactions. The interaction
891 * table is defined below. The case numbers are used in the multiple
892 * functions below to refer to the particular test case.
893 *
894 * There can be multiple options:
895 * 1) Adding address with prefix on non-p2p/non-loopback interface.
896 * Example: 192.0.2.1/24. Action:
897 * * add "prefix" route towards 192.0.2.0/24 via @ia interface,
898 * using @ia as an address source.
899 * * add "loopback" route towards 192.0.2.1 via V_loif, saving
900 * @ia ifp in the gateway and using @ia as an address source.
901 *
902 * 2) Adding address with /32 mask to non-p2p/non-loopback interface.
903 * Example: 192.0.2.2/32. Action:
904 * * add "prefix" host route via V_loif, using @ia as an address source.
905 *
906 * 3) Adding address with or without prefix to p2p interface.
907 * Example: 10.0.0.1/24->10.0.0.2. Action:
908 * * add "prefix" host route towards 10.0.0.2 via this interface, using @ia
909 * as an address source. Note: no sense in installing full /24 as the interface
910 * is point-to-point.
911 * * add "loopback" route towards 10.0.9.1 via V_loif, saving
912 * @ia ifp in the gateway and using @ia as an address source.
913 *
914 * 4) Adding address with or without prefix to loopback interface.
915 * Example: 192.0.2.1/24. Action:
916 * * add "prefix" host route via @ia interface, using @ia as an address source.
917 * Note: Skip installing /24 prefix as it would introduce TTL loop
918 * for the traffic destined to these addresses.
919 */
920
921 /*
922 * Checks if @ia needs to install loopback route to @ia address via
923 * ifa_maintain_loopback_route().
924 *
925 * Return true on success.
926 */
927 static bool
ia_need_loopback_route(const struct in_ifaddr * ia)928 ia_need_loopback_route(const struct in_ifaddr *ia)
929 {
930 struct ifnet *ifp = ia->ia_ifp;
931
932 /* Case 4: Skip loopback interfaces */
933 if ((ifp->if_flags & IFF_LOOPBACK) ||
934 (ia->ia_addr.sin_addr.s_addr == INADDR_ANY))
935 return (false);
936
937 /* Clash avoidance: Skip p2p interfaces with both addresses are equal */
938 if ((ifp->if_flags & IFF_POINTOPOINT) &&
939 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
940 return (false);
941
942 /* Case 2: skip /32 prefixes */
943 if (!(ifp->if_flags & IFF_POINTOPOINT) &&
944 (ia->ia_sockmask.sin_addr.s_addr == INADDR_BROADCAST))
945 return (false);
946
947 return (true);
948 }
949
950 /*
951 * Calculate "prefix" route corresponding to @ia.
952 */
953 static void
ia_getrtprefix(const struct in_ifaddr * ia,struct in_addr * prefix,struct in_addr * mask)954 ia_getrtprefix(const struct in_ifaddr *ia, struct in_addr *prefix, struct in_addr *mask)
955 {
956
957 if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) {
958 /* Case 3: return host route for dstaddr */
959 *prefix = ia->ia_dstaddr.sin_addr;
960 mask->s_addr = INADDR_BROADCAST;
961 } else if (ia->ia_ifp->if_flags & IFF_LOOPBACK) {
962 /* Case 4: return host route for ifaddr */
963 *prefix = ia->ia_addr.sin_addr;
964 mask->s_addr = INADDR_BROADCAST;
965 } else {
966 /* Cases 1,2: return actual ia prefix */
967 *prefix = ia->ia_addr.sin_addr;
968 *mask = ia->ia_sockmask.sin_addr;
969 prefix->s_addr &= mask->s_addr;
970 }
971 }
972
973 /*
974 * Adds or delete interface "prefix" route corresponding to @ifa.
975 * Returns 0 on success or errno.
976 */
977 static int
in_handle_ifaddr_route(int cmd,struct in_ifaddr * ia)978 in_handle_ifaddr_route(int cmd, struct in_ifaddr *ia)
979 {
980 struct ifaddr *ifa = &ia->ia_ifa;
981 struct in_addr daddr, maddr;
982 struct sockaddr_in *pmask;
983 struct epoch_tracker et;
984 int error;
985
986 ia_getrtprefix(ia, &daddr, &maddr);
987
988 struct sockaddr_in mask = {
989 .sin_family = AF_INET,
990 .sin_len = sizeof(struct sockaddr_in),
991 .sin_addr = maddr,
992 };
993
994 pmask = (maddr.s_addr != INADDR_BROADCAST) ? &mask : NULL;
995
996 struct sockaddr_in dst = {
997 .sin_family = AF_INET,
998 .sin_len = sizeof(struct sockaddr_in),
999 .sin_addr.s_addr = daddr.s_addr & maddr.s_addr,
1000 };
1001
1002 struct ifnet *ifp = ia->ia_ifp;
1003
1004 if ((maddr.s_addr == INADDR_BROADCAST) &&
1005 (!(ia->ia_ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)))) {
1006 /* Case 2: host route on broadcast interface */
1007 ifp = V_loif;
1008 }
1009
1010 uint32_t fibnum = ifa->ifa_ifp->if_fib;
1011 NET_EPOCH_ENTER(et);
1012 error = in_handle_prefix_route(fibnum, cmd, &dst, pmask, ifa, ifp);
1013 NET_EPOCH_EXIT(et);
1014
1015 return (error);
1016 }
1017
1018 /*
1019 * Check if we have a route for the given prefix already.
1020 */
1021 static bool
in_hasrtprefix(struct in_ifaddr * target)1022 in_hasrtprefix(struct in_ifaddr *target)
1023 {
1024 struct epoch_tracker et;
1025 struct in_ifaddr *ia;
1026 struct in_addr prefix, mask, p, m;
1027 bool result = false;
1028
1029 ia_getrtprefix(target, &prefix, &mask);
1030
1031 /* Look for an existing address with the same prefix, mask, and fib */
1032 NET_EPOCH_ENTER(et);
1033 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1034 ia_getrtprefix(ia, &p, &m);
1035
1036 if (prefix.s_addr != p.s_addr ||
1037 mask.s_addr != m.s_addr)
1038 continue;
1039
1040 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
1041 continue;
1042
1043 /*
1044 * If we got a matching prefix route inserted by other
1045 * interface address, we are done here.
1046 */
1047 if (ia->ia_flags & IFA_ROUTE) {
1048 result = true;
1049 break;
1050 }
1051 }
1052 NET_EPOCH_EXIT(et);
1053
1054 return (result);
1055 }
1056
1057 int
in_addprefix(struct in_ifaddr * target)1058 in_addprefix(struct in_ifaddr *target)
1059 {
1060 int error;
1061
1062 if (in_hasrtprefix(target)) {
1063 if (V_nosameprefix)
1064 return (EEXIST);
1065 else {
1066 rt_addrmsg(RTM_ADD, &target->ia_ifa,
1067 target->ia_ifp->if_fib);
1068 return (0);
1069 }
1070 }
1071
1072 /*
1073 * No-one seem to have this prefix route, so we try to insert it.
1074 */
1075 rt_addrmsg(RTM_ADD, &target->ia_ifa, target->ia_ifp->if_fib);
1076 error = in_handle_ifaddr_route(RTM_ADD, target);
1077 if (!error)
1078 target->ia_flags |= IFA_ROUTE;
1079 return (error);
1080 }
1081
1082 /*
1083 * Removes either all lle entries for given @ia, or lle
1084 * corresponding to @ia address.
1085 */
1086 static void
in_scrubprefixlle(struct in_ifaddr * ia,int all,u_int flags)1087 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
1088 {
1089 struct sockaddr_in addr, mask;
1090 struct sockaddr *saddr, *smask;
1091 struct ifnet *ifp;
1092
1093 saddr = (struct sockaddr *)&addr;
1094 bzero(&addr, sizeof(addr));
1095 addr.sin_len = sizeof(addr);
1096 addr.sin_family = AF_INET;
1097 smask = (struct sockaddr *)&mask;
1098 bzero(&mask, sizeof(mask));
1099 mask.sin_len = sizeof(mask);
1100 mask.sin_family = AF_INET;
1101 mask.sin_addr.s_addr = ia->ia_subnetmask;
1102 ifp = ia->ia_ifp;
1103
1104 if (all) {
1105 /*
1106 * Remove all L2 entries matching given prefix.
1107 * Convert address to host representation to avoid
1108 * doing this on every callback. ia_subnetmask is already
1109 * stored in host representation.
1110 */
1111 addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
1112 lltable_prefix_free(AF_INET, saddr, smask, flags);
1113 } else {
1114 /* Remove interface address only */
1115 addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
1116 lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
1117 }
1118 }
1119
1120 /*
1121 * If there is no other address in the system that can serve a route to the
1122 * same prefix, remove the route. Hand over the route to the new address
1123 * otherwise.
1124 */
1125 int
in_scrubprefix(struct in_ifaddr * target,u_int flags)1126 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1127 {
1128 struct epoch_tracker et;
1129 struct in_ifaddr *ia;
1130 struct in_addr prefix, mask, p, m;
1131 int error = 0;
1132
1133 /*
1134 * Remove the loopback route to the interface address.
1135 */
1136 if (ia_need_loopback_route(target) && (flags & LLE_STATIC)) {
1137 struct in_ifaddr *eia;
1138
1139 eia = in_localip_more(target);
1140
1141 if (eia != NULL) {
1142 error = ifa_switch_loopback_route((struct ifaddr *)eia,
1143 (struct sockaddr *)&target->ia_addr);
1144 ifa_free(&eia->ia_ifa);
1145 } else {
1146 error = ifa_del_loopback_route((struct ifaddr *)target,
1147 (struct sockaddr *)&target->ia_addr);
1148 }
1149 }
1150
1151 ia_getrtprefix(target, &prefix, &mask);
1152
1153 if ((target->ia_flags & IFA_ROUTE) == 0) {
1154 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1155
1156 /*
1157 * Removing address from !IFF_UP interface or
1158 * prefix which exists on other interface (along with route).
1159 * No entries should exist here except target addr.
1160 * Given that, delete this entry only.
1161 */
1162 in_scrubprefixlle(target, 0, flags);
1163 return (0);
1164 }
1165
1166 NET_EPOCH_ENTER(et);
1167 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1168 ia_getrtprefix(ia, &p, &m);
1169
1170 if (prefix.s_addr != p.s_addr ||
1171 mask.s_addr != m.s_addr)
1172 continue;
1173
1174 if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1175 continue;
1176
1177 /*
1178 * If we got a matching prefix address, move IFA_ROUTE and
1179 * the route itself to it. Make sure that routing daemons
1180 * get a heads-up.
1181 */
1182 if ((ia->ia_flags & IFA_ROUTE) == 0) {
1183 ifa_ref(&ia->ia_ifa);
1184 NET_EPOCH_EXIT(et);
1185 error = in_handle_ifaddr_route(RTM_DELETE, target);
1186 if (error == 0)
1187 target->ia_flags &= ~IFA_ROUTE;
1188 else
1189 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1190 error);
1191 /* Scrub all entries IFF interface is different */
1192 in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
1193 flags);
1194 error = in_handle_ifaddr_route(RTM_ADD, ia);
1195 if (error == 0)
1196 ia->ia_flags |= IFA_ROUTE;
1197 else
1198 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1199 error);
1200 ifa_free(&ia->ia_ifa);
1201 return (error);
1202 }
1203 }
1204 NET_EPOCH_EXIT(et);
1205
1206 /*
1207 * remove all L2 entries on the given prefix
1208 */
1209 in_scrubprefixlle(target, 1, flags);
1210
1211 /*
1212 * As no-one seem to have this prefix, we can remove the route.
1213 */
1214 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1215 error = in_handle_ifaddr_route(RTM_DELETE, target);
1216 if (error == 0)
1217 target->ia_flags &= ~IFA_ROUTE;
1218 else
1219 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1220 return (error);
1221 }
1222
1223 void
in_ifscrub_all(void)1224 in_ifscrub_all(void)
1225 {
1226 struct ifnet *ifp;
1227 struct ifaddr *ifa, *nifa;
1228 struct ifaliasreq ifr;
1229
1230 IFNET_RLOCK();
1231 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1232 /* Cannot lock here - lock recursion. */
1233 /* NET_EPOCH_ENTER(et); */
1234 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1235 if (ifa->ifa_addr->sa_family != AF_INET)
1236 continue;
1237
1238 /*
1239 * This is ugly but the only way for legacy IP to
1240 * cleanly remove addresses and everything attached.
1241 */
1242 bzero(&ifr, sizeof(ifr));
1243 ifr.ifra_addr = *ifa->ifa_addr;
1244 if (ifa->ifa_dstaddr)
1245 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
1246 (void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
1247 ifp, NULL);
1248 }
1249 /* NET_EPOCH_EXIT(et); */
1250 in_purgemaddrs(ifp);
1251 igmp_domifdetach(ifp);
1252 }
1253 IFNET_RUNLOCK();
1254 }
1255
1256 int
in_ifaddr_broadcast(struct in_addr in,struct in_ifaddr * ia)1257 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
1258 {
1259
1260 return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1261 /*
1262 * Optionally check for old-style (host 0) broadcast, but
1263 * taking into account that RFC 3021 obsoletes it.
1264 */
1265 (V_broadcast_lowest && ia->ia_subnetmask != IN_RFC3021_MASK &&
1266 ntohl(in.s_addr) == ia->ia_subnet)) &&
1267 /*
1268 * Check for an all one subnetmask. These
1269 * only exist when an interface gets a secondary
1270 * address.
1271 */
1272 ia->ia_subnetmask != (u_long)0xffffffff);
1273 }
1274
1275 /*
1276 * Return 1 if the address might be a local broadcast address.
1277 */
1278 int
in_broadcast(struct in_addr in,struct ifnet * ifp)1279 in_broadcast(struct in_addr in, struct ifnet *ifp)
1280 {
1281 struct ifaddr *ifa;
1282 int found;
1283
1284 NET_EPOCH_ASSERT();
1285
1286 if (in.s_addr == INADDR_BROADCAST ||
1287 in.s_addr == INADDR_ANY)
1288 return (1);
1289 if ((ifp->if_flags & IFF_BROADCAST) == 0)
1290 return (0);
1291 found = 0;
1292 /*
1293 * Look through the list of addresses for a match
1294 * with a broadcast address.
1295 */
1296 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1297 if (ifa->ifa_addr->sa_family == AF_INET &&
1298 in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) {
1299 found = 1;
1300 break;
1301 }
1302 return (found);
1303 }
1304
1305 /*
1306 * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1307 */
1308 void
in_ifdetach(struct ifnet * ifp)1309 in_ifdetach(struct ifnet *ifp)
1310 {
1311 IN_MULTI_LOCK();
1312 in_pcbpurgeif0(&V_ripcbinfo, ifp);
1313 in_pcbpurgeif0(&V_udbinfo, ifp);
1314 in_pcbpurgeif0(&V_ulitecbinfo, ifp);
1315 in_purgemaddrs(ifp);
1316 IN_MULTI_UNLOCK();
1317
1318 /*
1319 * Make sure all multicast deletions invoking if_ioctl() are
1320 * completed before returning. Else we risk accessing a freed
1321 * ifnet structure pointer.
1322 */
1323 inm_release_wait(NULL);
1324 }
1325
1326 static void
in_ifnet_event(void * arg __unused,struct ifnet * ifp,int event)1327 in_ifnet_event(void *arg __unused, struct ifnet *ifp, int event)
1328 {
1329 struct epoch_tracker et;
1330 struct ifaddr *ifa;
1331 struct in_ifaddr *ia;
1332 int error;
1333
1334 NET_EPOCH_ENTER(et);
1335 switch (event) {
1336 case IFNET_EVENT_DOWN:
1337 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1338 if (ifa->ifa_addr->sa_family != AF_INET)
1339 continue;
1340 ia = (struct in_ifaddr *)ifa;
1341 if ((ia->ia_flags & IFA_ROUTE) == 0)
1342 continue;
1343 ifa_ref(ifa);
1344 /*
1345 * in_scrubprefix() kills the interface route.
1346 */
1347 in_scrubprefix(ia, 0);
1348 /*
1349 * in_ifadown gets rid of all the rest of the
1350 * routes. This is not quite the right thing
1351 * to do, but at least if we are running a
1352 * routing process they will come back.
1353 */
1354 in_ifadown(ifa, 0);
1355 ifa_free(ifa);
1356 }
1357 break;
1358
1359 case IFNET_EVENT_UP:
1360 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1361 if (ifa->ifa_addr->sa_family != AF_INET)
1362 continue;
1363 ia = (struct in_ifaddr *)ifa;
1364 if (ia->ia_flags & IFA_ROUTE)
1365 continue;
1366 ifa_ref(ifa);
1367 error = ifa_del_loopback_route(ifa, ifa->ifa_addr);
1368 rt_addrmsg(RTM_ADD, ifa, ifa->ifa_ifp->if_fib);
1369 error = in_handle_ifaddr_route(RTM_ADD, ia);
1370 if (error == 0)
1371 ia->ia_flags |= IFA_ROUTE;
1372 error = ifa_add_loopback_route(ifa, ifa->ifa_addr);
1373 ifa_free(ifa);
1374 }
1375 break;
1376 }
1377 NET_EPOCH_EXIT(et);
1378 }
1379 EVENTHANDLER_DEFINE(ifnet_event, in_ifnet_event, NULL, EVENTHANDLER_PRI_ANY);
1380
1381 /*
1382 * Delete all IPv4 multicast address records, and associated link-layer
1383 * multicast address records, associated with ifp.
1384 * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1385 * XXX This should not race with ifma_protospec being set during
1386 * a new allocation, if it does, we have bigger problems.
1387 */
1388 static void
in_purgemaddrs(struct ifnet * ifp)1389 in_purgemaddrs(struct ifnet *ifp)
1390 {
1391 struct epoch_tracker et;
1392 struct in_multi_head purgeinms;
1393 struct in_multi *inm;
1394 struct ifmultiaddr *ifma;
1395
1396 SLIST_INIT(&purgeinms);
1397 IN_MULTI_LIST_LOCK();
1398
1399 /*
1400 * Extract list of in_multi associated with the detaching ifp
1401 * which the PF_INET layer is about to release.
1402 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1403 * by code further down.
1404 */
1405 IF_ADDR_WLOCK(ifp);
1406 NET_EPOCH_ENTER(et);
1407 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1408 inm = inm_ifmultiaddr_get_inm(ifma);
1409 if (inm == NULL)
1410 continue;
1411 inm_rele_locked(&purgeinms, inm);
1412 }
1413 NET_EPOCH_EXIT(et);
1414 IF_ADDR_WUNLOCK(ifp);
1415
1416 inm_release_list_deferred(&purgeinms);
1417 igmp_ifdetach(ifp);
1418 IN_MULTI_LIST_UNLOCK();
1419 }
1420
1421 struct in_llentry {
1422 struct llentry base;
1423 };
1424
1425 #define IN_LLTBL_DEFAULT_HSIZE 32
1426 #define IN_LLTBL_HASH(k, h) \
1427 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1428
1429 /*
1430 * Do actual deallocation of @lle.
1431 */
1432 static void
in_lltable_destroy_lle_unlocked(epoch_context_t ctx)1433 in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1434 {
1435 struct llentry *lle;
1436
1437 lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1438 LLE_LOCK_DESTROY(lle);
1439 LLE_REQ_DESTROY(lle);
1440 free(lle, M_LLTABLE);
1441 }
1442
1443 /*
1444 * Called by LLE_FREE_LOCKED when number of references
1445 * drops to zero.
1446 */
1447 static void
in_lltable_destroy_lle(struct llentry * lle)1448 in_lltable_destroy_lle(struct llentry *lle)
1449 {
1450
1451 LLE_WUNLOCK(lle);
1452 NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1453 }
1454
1455 static struct llentry *
in_lltable_new(struct in_addr addr4,u_int flags)1456 in_lltable_new(struct in_addr addr4, u_int flags)
1457 {
1458 struct in_llentry *lle;
1459
1460 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1461 if (lle == NULL) /* NB: caller generates msg */
1462 return NULL;
1463
1464 /*
1465 * For IPv4 this will trigger "arpresolve" to generate
1466 * an ARP request.
1467 */
1468 lle->base.la_expire = time_uptime; /* mark expired */
1469 lle->base.r_l3addr.addr4 = addr4;
1470 lle->base.lle_refcnt = 1;
1471 lle->base.lle_free = in_lltable_destroy_lle;
1472 LLE_LOCK_INIT(&lle->base);
1473 LLE_REQ_INIT(&lle->base);
1474 callout_init(&lle->base.lle_timer, 1);
1475
1476 return (&lle->base);
1477 }
1478
1479 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \
1480 ((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 )
1481
1482 static int
in_lltable_match_prefix(const struct sockaddr * saddr,const struct sockaddr * smask,u_int flags,struct llentry * lle)1483 in_lltable_match_prefix(const struct sockaddr *saddr,
1484 const struct sockaddr *smask, u_int flags, struct llentry *lle)
1485 {
1486 struct in_addr addr, mask, lle_addr;
1487
1488 addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1489 mask = ((const struct sockaddr_in *)smask)->sin_addr;
1490 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1491
1492 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1493 return (0);
1494
1495 if (lle->la_flags & LLE_IFADDR) {
1496 /*
1497 * Delete LLE_IFADDR records IFF address & flag matches.
1498 * Note that addr is the interface address within prefix
1499 * being matched.
1500 * Note also we should handle 'ifdown' cases without removing
1501 * ifaddr macs.
1502 */
1503 if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1504 return (1);
1505 return (0);
1506 }
1507
1508 /* flags & LLE_STATIC means deleting both dynamic and static entries */
1509 if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1510 return (1);
1511
1512 return (0);
1513 }
1514
1515 static void
in_lltable_free_entry(struct lltable * llt,struct llentry * lle)1516 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1517 {
1518 size_t pkts_dropped;
1519
1520 LLE_WLOCK_ASSERT(lle);
1521 KASSERT(llt != NULL, ("lltable is NULL"));
1522
1523 /* Unlink entry from table if not already */
1524 if ((lle->la_flags & LLE_LINKED) != 0) {
1525 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1526 lltable_unlink_entry(llt, lle);
1527 }
1528
1529 /* Drop hold queue */
1530 pkts_dropped = llentry_free(lle);
1531 ARPSTAT_ADD(dropped, pkts_dropped);
1532 }
1533
1534 static int
in_lltable_rtcheck(struct ifnet * ifp,u_int flags,const struct sockaddr * l3addr)1535 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1536 {
1537 struct nhop_object *nh;
1538 struct in_addr addr;
1539
1540 KASSERT(l3addr->sa_family == AF_INET,
1541 ("sin_family %d", l3addr->sa_family));
1542
1543 addr = ((const struct sockaddr_in *)l3addr)->sin_addr;
1544
1545 nh = fib4_lookup(ifp->if_fib, addr, 0, NHR_NONE, 0);
1546 if (nh == NULL)
1547 return (EINVAL);
1548
1549 /*
1550 * If the gateway for an existing host route matches the target L3
1551 * address, which is a special route inserted by some implementation
1552 * such as MANET, and the interface is of the correct type, then
1553 * allow for ARP to proceed.
1554 */
1555 if (nh->nh_flags & NHF_GATEWAY) {
1556 if (!(nh->nh_flags & NHF_HOST) || nh->nh_ifp->if_type != IFT_ETHER ||
1557 (nh->nh_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1558 memcmp(nh->gw_sa.sa_data, l3addr->sa_data,
1559 sizeof(in_addr_t)) != 0) {
1560 return (EINVAL);
1561 }
1562 }
1563
1564 /*
1565 * Make sure that at least the destination address is covered
1566 * by the route. This is for handling the case where 2 or more
1567 * interfaces have the same prefix. An incoming packet arrives
1568 * on one interface and the corresponding outgoing packet leaves
1569 * another interface.
1570 */
1571 if ((nh->nh_ifp != ifp) && (nh->nh_flags & NHF_HOST) == 0) {
1572 struct in_ifaddr *ia = (struct in_ifaddr *)ifaof_ifpforaddr(l3addr, ifp);
1573 struct in_addr dst_addr, mask_addr;
1574
1575 if (ia == NULL)
1576 return (EINVAL);
1577
1578 /*
1579 * ifaof_ifpforaddr() returns _best matching_ IFA.
1580 * It is possible that ifa prefix does not cover our address.
1581 * Explicitly verify and fail if that's the case.
1582 */
1583 dst_addr = IA_SIN(ia)->sin_addr;
1584 mask_addr.s_addr = htonl(ia->ia_subnetmask);
1585
1586 if (!IN_ARE_MASKED_ADDR_EQUAL(dst_addr, addr, mask_addr))
1587 return (EINVAL);
1588 }
1589
1590 return (0);
1591 }
1592
1593 static inline uint32_t
in_lltable_hash_dst(const struct in_addr dst,uint32_t hsize)1594 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1595 {
1596
1597 return (IN_LLTBL_HASH(dst.s_addr, hsize));
1598 }
1599
1600 static uint32_t
in_lltable_hash(const struct llentry * lle,uint32_t hsize)1601 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1602 {
1603
1604 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1605 }
1606
1607 static void
in_lltable_fill_sa_entry(const struct llentry * lle,struct sockaddr * sa)1608 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1609 {
1610 struct sockaddr_in *sin;
1611
1612 sin = (struct sockaddr_in *)sa;
1613 bzero(sin, sizeof(*sin));
1614 sin->sin_family = AF_INET;
1615 sin->sin_len = sizeof(*sin);
1616 sin->sin_addr = lle->r_l3addr.addr4;
1617 }
1618
1619 static inline struct llentry *
in_lltable_find_dst(struct lltable * llt,struct in_addr dst)1620 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1621 {
1622 struct llentry *lle;
1623 struct llentries *lleh;
1624 u_int hashidx;
1625
1626 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1627 lleh = &llt->lle_head[hashidx];
1628 CK_LIST_FOREACH(lle, lleh, lle_next) {
1629 if (lle->la_flags & LLE_DELETED)
1630 continue;
1631 if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1632 break;
1633 }
1634
1635 return (lle);
1636 }
1637
1638 static void
in_lltable_delete_entry(struct lltable * llt,struct llentry * lle)1639 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1640 {
1641
1642 lle->la_flags |= LLE_DELETED;
1643 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1644 #ifdef DIAGNOSTIC
1645 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1646 #endif
1647 llentry_free(lle);
1648 }
1649
1650 static struct llentry *
in_lltable_alloc(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)1651 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1652 {
1653 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1654 struct ifnet *ifp = llt->llt_ifp;
1655 struct llentry *lle;
1656 char linkhdr[LLE_MAX_LINKHDR];
1657 size_t linkhdrsize;
1658 int lladdr_off;
1659
1660 KASSERT(l3addr->sa_family == AF_INET,
1661 ("sin_family %d", l3addr->sa_family));
1662
1663 /*
1664 * A route that covers the given address must have
1665 * been installed 1st because we are doing a resolution,
1666 * verify this.
1667 */
1668 if (!(flags & LLE_IFADDR) &&
1669 in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1670 return (NULL);
1671
1672 lle = in_lltable_new(sin->sin_addr, flags);
1673 if (lle == NULL) {
1674 log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1675 return (NULL);
1676 }
1677 lle->la_flags = flags;
1678 if (flags & LLE_STATIC)
1679 lle->r_flags |= RLLE_VALID;
1680 if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1681 linkhdrsize = LLE_MAX_LINKHDR;
1682 if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1683 linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1684 in_lltable_free_entry(llt, lle);
1685 return (NULL);
1686 }
1687 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1688 lladdr_off);
1689 lle->la_flags |= LLE_STATIC;
1690 lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1691 lle->la_expire = 0;
1692 }
1693
1694 return (lle);
1695 }
1696
1697 /*
1698 * Return NULL if not found or marked for deletion.
1699 * If found return lle read locked.
1700 */
1701 static struct llentry *
in_lltable_lookup(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)1702 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1703 {
1704 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1705 struct llentry *lle;
1706
1707 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1708 KASSERT(l3addr->sa_family == AF_INET,
1709 ("sin_family %d", l3addr->sa_family));
1710 KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1711 (LLE_UNLOCKED | LLE_EXCLUSIVE),
1712 ("wrong lle request flags: %#x", flags));
1713
1714 lle = in_lltable_find_dst(llt, sin->sin_addr);
1715 if (lle == NULL)
1716 return (NULL);
1717 if (flags & LLE_UNLOCKED)
1718 return (lle);
1719
1720 if (flags & LLE_EXCLUSIVE)
1721 LLE_WLOCK(lle);
1722 else
1723 LLE_RLOCK(lle);
1724
1725 /*
1726 * If the afdata lock is not held, the LLE may have been unlinked while
1727 * we were blocked on the LLE lock. Check for this case.
1728 */
1729 if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1730 if (flags & LLE_EXCLUSIVE)
1731 LLE_WUNLOCK(lle);
1732 else
1733 LLE_RUNLOCK(lle);
1734 return (NULL);
1735 }
1736 return (lle);
1737 }
1738
1739 static int
in_lltable_dump_entry(struct lltable * llt,struct llentry * lle,struct sysctl_req * wr)1740 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1741 struct sysctl_req *wr)
1742 {
1743 struct ifnet *ifp = llt->llt_ifp;
1744 /* XXX stack use */
1745 struct {
1746 struct rt_msghdr rtm;
1747 struct sockaddr_in sin;
1748 struct sockaddr_dl sdl;
1749 } arpc;
1750 struct sockaddr_dl *sdl;
1751 int error;
1752
1753 bzero(&arpc, sizeof(arpc));
1754 /* skip deleted entries */
1755 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1756 return (0);
1757 /* Skip if jailed and not a valid IP of the prison. */
1758 lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1759 if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1760 return (0);
1761 /*
1762 * produce a msg made of:
1763 * struct rt_msghdr;
1764 * struct sockaddr_in; (IPv4)
1765 * struct sockaddr_dl;
1766 */
1767 arpc.rtm.rtm_msglen = sizeof(arpc);
1768 arpc.rtm.rtm_version = RTM_VERSION;
1769 arpc.rtm.rtm_type = RTM_GET;
1770 arpc.rtm.rtm_flags = RTF_UP;
1771 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1772
1773 /* publish */
1774 if (lle->la_flags & LLE_PUB)
1775 arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1776
1777 sdl = &arpc.sdl;
1778 sdl->sdl_family = AF_LINK;
1779 sdl->sdl_len = sizeof(*sdl);
1780 sdl->sdl_index = ifp->if_index;
1781 sdl->sdl_type = ifp->if_type;
1782 if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1783 sdl->sdl_alen = ifp->if_addrlen;
1784 bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1785 } else {
1786 sdl->sdl_alen = 0;
1787 bzero(LLADDR(sdl), ifp->if_addrlen);
1788 }
1789
1790 arpc.rtm.rtm_rmx.rmx_expire =
1791 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1792 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1793 if (lle->la_flags & LLE_STATIC)
1794 arpc.rtm.rtm_flags |= RTF_STATIC;
1795 if (lle->la_flags & LLE_IFADDR)
1796 arpc.rtm.rtm_flags |= RTF_PINNED;
1797 arpc.rtm.rtm_index = ifp->if_index;
1798 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1799
1800 return (error);
1801 }
1802
1803 static void
in_lltable_post_resolved(struct lltable * llt,struct llentry * lle)1804 in_lltable_post_resolved(struct lltable *llt, struct llentry *lle)
1805 {
1806 struct ifnet *ifp = llt->llt_ifp;
1807
1808 /* gratuitous ARP */
1809 if ((lle->la_flags & LLE_PUB) != 0)
1810 arprequest(ifp, &lle->r_l3addr.addr4, &lle->r_l3addr.addr4,
1811 lle->ll_addr);
1812 }
1813
1814 static struct lltable *
in_lltattach(struct ifnet * ifp)1815 in_lltattach(struct ifnet *ifp)
1816 {
1817 struct lltable *llt;
1818
1819 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1820 llt->llt_af = AF_INET;
1821 llt->llt_ifp = ifp;
1822
1823 llt->llt_lookup = in_lltable_lookup;
1824 llt->llt_alloc_entry = in_lltable_alloc;
1825 llt->llt_delete_entry = in_lltable_delete_entry;
1826 llt->llt_dump_entry = in_lltable_dump_entry;
1827 llt->llt_hash = in_lltable_hash;
1828 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1829 llt->llt_free_entry = in_lltable_free_entry;
1830 llt->llt_match_prefix = in_lltable_match_prefix;
1831 llt->llt_mark_used = llentry_mark_used;
1832 llt->llt_post_resolved = in_lltable_post_resolved;
1833 lltable_link(llt);
1834
1835 return (llt);
1836 }
1837
1838 struct lltable *
in_lltable_get(struct ifnet * ifp)1839 in_lltable_get(struct ifnet *ifp)
1840 {
1841 struct lltable *llt = NULL;
1842
1843 void *afdata_ptr = ifp->if_afdata[AF_INET];
1844 if (afdata_ptr != NULL)
1845 llt = ((struct in_ifinfo *)afdata_ptr)->ii_llt;
1846 return (llt);
1847 }
1848
1849 void *
in_domifattach(struct ifnet * ifp)1850 in_domifattach(struct ifnet *ifp)
1851 {
1852 struct in_ifinfo *ii;
1853
1854 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1855
1856 ii->ii_llt = in_lltattach(ifp);
1857 ii->ii_igmp = igmp_domifattach(ifp);
1858
1859 return (ii);
1860 }
1861
1862 void
in_domifdetach(struct ifnet * ifp,void * aux)1863 in_domifdetach(struct ifnet *ifp, void *aux)
1864 {
1865 struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1866
1867 igmp_domifdetach(ifp);
1868 lltable_free(ii->ii_llt);
1869 free(ii, M_IFADDR);
1870 }
1871