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
33 #include <sys/cdefs.h>
34 #include "opt_inet.h"
35
36 #define IN_HISTORICAL_NETS /* include class masks */
37
38 #include <sys/param.h>
39 #include <sys/eventhandler.h>
40 #include <sys/systm.h>
41 #include <sys/sockio.h>
42 #include <sys/malloc.h>
43 #include <sys/priv.h>
44 #include <sys/socket.h>
45 #include <sys/jail.h>
46 #include <sys/kernel.h>
47 #include <sys/lock.h>
48 #include <sys/proc.h>
49 #include <sys/sysctl.h>
50 #include <sys/syslog.h>
51 #include <sys/sx.h>
52
53 #include <net/if.h>
54 #include <net/if_var.h>
55 #include <net/if_arp.h>
56 #include <net/if_dl.h>
57 #include <net/if_llatbl.h>
58 #include <net/if_private.h>
59 #include <net/if_types.h>
60 #include <net/if_bridgevar.h>
61 #include <net/route.h>
62 #include <net/route/nhop.h>
63 #include <net/route/route_ctl.h>
64 #include <net/vnet.h>
65
66 #include <netinet/if_ether.h>
67 #include <netinet/in.h>
68 #include <netinet/in_fib.h>
69 #include <netinet/in_var.h>
70 #include <netinet/in_pcb.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/ip_carp.h>
73 #include <netinet/igmp_var.h>
74 #include <netinet/udp.h>
75 #include <netinet/udp_var.h>
76
77 #ifdef MAC
78 #include <security/mac/mac_framework.h>
79 #endif
80
81 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
82 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
83 static int in_gifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
84
85 static void in_socktrim(struct sockaddr_in *);
86 static void in_purgemaddrs(struct ifnet *);
87
88 static bool ia_need_loopback_route(const struct in_ifaddr *);
89
90 VNET_DEFINE_STATIC(int, nosameprefix);
91 #define V_nosameprefix VNET(nosameprefix)
92 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW,
93 &VNET_NAME(nosameprefix), 0,
94 "Refuse to create same prefixes on different interfaces");
95
96 VNET_DEFINE_STATIC(bool, broadcast_lowest);
97 #define V_broadcast_lowest VNET(broadcast_lowest)
98 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, broadcast_lowest, CTLFLAG_VNET | CTLFLAG_RW,
99 &VNET_NAME(broadcast_lowest), 0,
100 "Treat lowest address on a subnet (host 0) as broadcast");
101
102 VNET_DEFINE(bool, ip_allow_net240) = false;
103 #define V_ip_allow_net240 VNET(ip_allow_net240)
104 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net240,
105 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net240), 0,
106 "Allow forwarding of and ICMP response to Experimental addresses, aka Class E (240/4)");
107 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-240 */
108
109 VNET_DEFINE(bool, ip_allow_net0) = false;
110 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net0,
111 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net0), 0,
112 "Allow forwarding of and ICMP response to addresses in network 0/8");
113 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-0 */
114
115 VNET_DEFINE(uint32_t, in_loopback_mask) = IN_LOOPBACK_MASK_DFLT;
116 #define V_in_loopback_mask VNET(in_loopback_mask)
117 static int sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS);
118 SYSCTL_PROC(_net_inet_ip, OID_AUTO, loopback_prefixlen,
119 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
120 NULL, 0, sysctl_loopback_prefixlen, "I",
121 "Prefix length of address space reserved for loopback");
122 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-127 */
123
124 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
125 #define V_ripcbinfo VNET(ripcbinfo)
126
127 static struct sx in_control_sx;
128 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control");
129
130 /*
131 * Return true if an internet address is for a ``local'' host
132 * (one to which we have a connection).
133 */
134 bool
in_localaddr(struct in_addr in)135 in_localaddr(struct in_addr in)
136 {
137 u_long i = ntohl(in.s_addr);
138 struct in_ifaddr *ia;
139
140 NET_EPOCH_ASSERT();
141
142 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
143 if ((i & ia->ia_subnetmask) == ia->ia_subnet)
144 return (true);
145 }
146
147 return (false);
148 }
149
150 /*
151 * Return true if an internet address is for the local host and configured
152 * on one of its interfaces.
153 */
154 bool
in_localip(struct in_addr in)155 in_localip(struct in_addr in)
156 {
157 struct in_ifaddr *ia;
158
159 NET_EPOCH_ASSERT();
160
161 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
162 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
163 return (true);
164
165 return (false);
166 }
167
168 /*
169 * Like in_localip(), but FIB-aware and carp(4)-aware.
170 */
171 bool
in_localip_fib(struct in_addr in,uint16_t fib)172 in_localip_fib(struct in_addr in, uint16_t fib)
173 {
174 struct in_ifaddr *ia;
175
176 NET_EPOCH_ASSERT();
177
178 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
179 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr &&
180 (ia->ia_ifa.ifa_carp == NULL ||
181 carp_master_p(&ia->ia_ifa)) &&
182 ia->ia_ifa.ifa_ifp->if_fib == fib)
183 return (true);
184
185 return (false);
186 }
187
188 /*
189 * Return true if an internet address is configured on an interface.
190 */
191 bool
in_ifhasaddr(struct ifnet * ifp,struct in_addr in)192 in_ifhasaddr(struct ifnet *ifp, struct in_addr in)
193 {
194 struct ifaddr *ifa;
195 struct in_ifaddr *ia;
196
197 NET_EPOCH_ASSERT();
198
199 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
200 if (ifa->ifa_addr->sa_family != AF_INET)
201 continue;
202 ia = (struct in_ifaddr *)ifa;
203 if (ia->ia_addr.sin_addr.s_addr == in.s_addr)
204 return (true);
205 }
206
207 return (false);
208 }
209
210 /*
211 * Return a reference to the interface address which is different to
212 * the supplied one but with same IP address value.
213 */
214 static struct in_ifaddr *
in_localip_more(struct in_ifaddr * original_ia)215 in_localip_more(struct in_ifaddr *original_ia)
216 {
217 struct epoch_tracker et;
218 in_addr_t original_addr = IA_SIN(original_ia)->sin_addr.s_addr;
219 uint32_t original_fib = original_ia->ia_ifa.ifa_ifp->if_fib;
220 struct in_ifaddr *ia;
221
222 NET_EPOCH_ENTER(et);
223 CK_LIST_FOREACH(ia, INADDR_HASH(original_addr), ia_hash) {
224 in_addr_t addr = IA_SIN(ia)->sin_addr.s_addr;
225 uint32_t fib = ia->ia_ifa.ifa_ifp->if_fib;
226 if (!V_rt_add_addr_allfibs && (original_fib != fib))
227 continue;
228 if ((original_ia != ia) && (original_addr == addr)) {
229 ifa_ref(&ia->ia_ifa);
230 NET_EPOCH_EXIT(et);
231 return (ia);
232 }
233 }
234 NET_EPOCH_EXIT(et);
235
236 return (NULL);
237 }
238
239 /*
240 * Tries to find first IPv4 address in the provided fib.
241 * Prefers non-loopback addresses and return loopback IFF
242 * @loopback_ok is set.
243 *
244 * Returns ifa or NULL.
245 */
246 struct in_ifaddr *
in_findlocal(uint32_t fibnum,bool loopback_ok)247 in_findlocal(uint32_t fibnum, bool loopback_ok)
248 {
249 struct in_ifaddr *ia = NULL, *ia_lo = NULL;
250
251 NET_EPOCH_ASSERT();
252
253 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
254 uint32_t ia_fib = ia->ia_ifa.ifa_ifp->if_fib;
255 if (!V_rt_add_addr_allfibs && (fibnum != ia_fib))
256 continue;
257
258 if (!IN_LOOPBACK(ntohl(IA_SIN(ia)->sin_addr.s_addr)))
259 break;
260 if (loopback_ok)
261 ia_lo = ia;
262 }
263
264 if (ia == NULL)
265 ia = ia_lo;
266
267 return (ia);
268 }
269
270 /*
271 * Determine whether an IP address is in a reserved set of addresses
272 * that may not be forwarded, or whether datagrams to that destination
273 * may be forwarded.
274 */
275 bool
in_canforward(struct in_addr in)276 in_canforward(struct in_addr in)
277 {
278 u_long i = ntohl(in.s_addr);
279
280 if (IN_MULTICAST(i) || IN_LINKLOCAL(i) || IN_LOOPBACK(i) ||
281 in_nullhost(in))
282 return (false);
283 if (IN_EXPERIMENTAL(i) && !V_ip_allow_net240)
284 return (false);
285 if (IN_ZERONET(i) && !V_ip_allow_net0)
286 return (false);
287 return (true);
288 }
289
290 /*
291 * Sysctl to manage prefix of reserved loopback network; translate
292 * to/from mask. The mask is always contiguous high-order 1 bits
293 * followed by all 0 bits.
294 */
295 static int
sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS)296 sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS)
297 {
298 int error, preflen;
299
300 /* ffs is 1-based; compensate. */
301 preflen = 33 - ffs(V_in_loopback_mask);
302 error = sysctl_handle_int(oidp, &preflen, 0, req);
303 if (error || !req->newptr)
304 return (error);
305 if (preflen < 8 || preflen > 31)
306 return (EINVAL);
307 V_in_loopback_mask = 0xffffffff << (32 - preflen);
308 return (0);
309 }
310
311 /*
312 * Trim a mask in a sockaddr
313 */
314 static void
in_socktrim(struct sockaddr_in * ap)315 in_socktrim(struct sockaddr_in *ap)
316 {
317 char *cplim = (char *) &ap->sin_addr;
318 char *cp = (char *) (&ap->sin_addr + 1);
319
320 ap->sin_len = 0;
321 while (--cp >= cplim)
322 if (*cp) {
323 (ap)->sin_len = cp - (char *) (ap) + 1;
324 break;
325 }
326 }
327
328 /*
329 * Generic internet control operations (ioctl's).
330 */
331 int
in_control_ioctl(u_long cmd,void * data,struct ifnet * ifp,struct ucred * cred)332 in_control_ioctl(u_long cmd, void *data, struct ifnet *ifp,
333 struct ucred *cred)
334 {
335 struct ifreq *ifr = (struct ifreq *)data;
336 struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
337 struct epoch_tracker et;
338 struct ifaddr *ifa;
339 struct in_ifaddr *ia;
340 int error;
341
342 if (ifp == NULL)
343 return (EADDRNOTAVAIL);
344
345 /*
346 * Filter out 4 ioctls we implement directly. Forward the rest
347 * to specific functions and ifp->if_ioctl().
348 */
349 switch (cmd) {
350 case SIOCGIFADDR:
351 case SIOCGIFBRDADDR:
352 case SIOCGIFDSTADDR:
353 case SIOCGIFNETMASK:
354 break;
355 case SIOCGIFALIAS:
356 sx_xlock(&in_control_sx);
357 error = in_gifaddr_ioctl(cmd, data, ifp, cred);
358 sx_xunlock(&in_control_sx);
359 return (error);
360 case SIOCDIFADDR:
361 sx_xlock(&in_control_sx);
362 error = in_difaddr_ioctl(cmd, data, ifp, cred);
363 sx_xunlock(&in_control_sx);
364 return (error);
365 case OSIOCAIFADDR: /* 9.x compat */
366 case SIOCAIFADDR:
367 sx_xlock(&in_control_sx);
368 error = in_aifaddr_ioctl(cmd, data, ifp, cred);
369 sx_xunlock(&in_control_sx);
370 return (error);
371 case SIOCSIFADDR:
372 case SIOCSIFBRDADDR:
373 case SIOCSIFDSTADDR:
374 case SIOCSIFNETMASK:
375 /* We no longer support that old commands. */
376 return (EINVAL);
377 default:
378 if (ifp->if_ioctl == NULL)
379 return (EOPNOTSUPP);
380 return ((*ifp->if_ioctl)(ifp, cmd, data));
381 }
382
383 if (addr->sin_addr.s_addr != INADDR_ANY &&
384 prison_check_ip4(cred, &addr->sin_addr) != 0)
385 return (EADDRNOTAVAIL);
386
387 /*
388 * Find address for this interface, if it exists. If an
389 * address was specified, find that one instead of the
390 * first one on the interface, if possible.
391 */
392 NET_EPOCH_ENTER(et);
393 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
394 if (ifa->ifa_addr->sa_family != AF_INET)
395 continue;
396 ia = (struct in_ifaddr *)ifa;
397 if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
398 break;
399 }
400 if (ifa == NULL)
401 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
402 if (ifa->ifa_addr->sa_family == AF_INET) {
403 ia = (struct in_ifaddr *)ifa;
404 if (prison_check_ip4(cred,
405 &ia->ia_addr.sin_addr) == 0)
406 break;
407 }
408
409 if (ifa == NULL) {
410 NET_EPOCH_EXIT(et);
411 return (EADDRNOTAVAIL);
412 }
413
414 error = 0;
415 switch (cmd) {
416 case SIOCGIFADDR:
417 *addr = ia->ia_addr;
418 break;
419
420 case SIOCGIFBRDADDR:
421 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
422 error = EINVAL;
423 break;
424 }
425 *addr = ia->ia_broadaddr;
426 break;
427
428 case SIOCGIFDSTADDR:
429 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
430 error = EINVAL;
431 break;
432 }
433 *addr = ia->ia_dstaddr;
434 break;
435
436 case SIOCGIFNETMASK:
437 *addr = ia->ia_sockmask;
438 break;
439 }
440
441 NET_EPOCH_EXIT(et);
442
443 return (error);
444 }
445
446 int
in_mask2len(struct in_addr * mask)447 in_mask2len(struct in_addr *mask)
448 {
449 int x, y;
450 u_char *p;
451
452 p = (u_char *)mask;
453 for (x = 0; x < sizeof(*mask); x++) {
454 if (p[x] != 0xff)
455 break;
456 }
457 y = 0;
458 if (x < sizeof(*mask)) {
459 for (y = 0; y < 8; y++) {
460 if ((p[x] & (0x80 >> y)) == 0)
461 break;
462 }
463 }
464 return (x * 8 + y);
465 }
466
467 int
in_control(struct socket * so,u_long cmd,void * data,struct ifnet * ifp,struct thread * td)468 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
469 struct thread *td)
470 {
471 return (in_control_ioctl(cmd, data, ifp, td ? td->td_ucred : NULL));
472 }
473
474 static int
in_aifaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)475 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
476 {
477 const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
478 const struct sockaddr_in *addr = &ifra->ifra_addr;
479 const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
480 const struct sockaddr_in *mask = &ifra->ifra_mask;
481 const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
482 const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
483 struct epoch_tracker et;
484 struct ifaddr *ifa;
485 struct in_ifaddr *ia;
486 bool iaIsFirst;
487 int error = 0;
488
489 error = priv_check_cred(cred, PRIV_NET_ADDIFADDR);
490 if (error)
491 return (error);
492
493 /*
494 * ifra_addr must be present and be of INET family.
495 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
496 */
497 if (addr->sin_len != sizeof(struct sockaddr_in) ||
498 addr->sin_family != AF_INET)
499 return (EINVAL);
500 if (broadaddr->sin_len != 0 &&
501 (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
502 broadaddr->sin_family != AF_INET))
503 return (EINVAL);
504 if (mask->sin_len != 0 &&
505 (mask->sin_len != sizeof(struct sockaddr_in) ||
506 mask->sin_family != AF_INET))
507 return (EINVAL);
508 if ((ifp->if_flags & IFF_POINTOPOINT) &&
509 (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
510 dstaddr->sin_addr.s_addr == INADDR_ANY))
511 return (EDESTADDRREQ);
512 if (vhid != 0 && carp_attach_p == NULL)
513 return (EPROTONOSUPPORT);
514
515 #ifdef MAC
516 /* Check if a MAC policy disallows setting the IPv4 address. */
517 error = mac_inet_check_add_addr(cred, &addr->sin_addr, ifp);
518 if (error != 0)
519 return (error);
520 #endif
521
522 /*
523 * Check if bridge wants to allow adding addrs to member interfaces.
524 */
525 if (ifp->if_bridge != NULL && ifp->if_type != IFT_GIF &&
526 bridge_member_ifaddrs_p != NULL) {
527 if (bridge_member_ifaddrs_p())
528 if_printf(ifp, "WARNING: Assigning an IP address to "
529 "an interface which is also a bridge member is "
530 "deprecated and will be unsupported in a future "
531 "release.\n");
532 else
533 return (EINVAL);
534 }
535
536 /*
537 * See whether address already exist.
538 */
539 iaIsFirst = true;
540 ia = NULL;
541 NET_EPOCH_ENTER(et);
542 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
543 struct in_ifaddr *it;
544
545 if (ifa->ifa_addr->sa_family != AF_INET)
546 continue;
547
548 it = (struct in_ifaddr *)ifa;
549 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
550 prison_check_ip4(cred, &addr->sin_addr) == 0)
551 ia = it;
552 else
553 iaIsFirst = false;
554 }
555 NET_EPOCH_EXIT(et);
556
557 if (ia != NULL)
558 (void )in_difaddr_ioctl(cmd, data, ifp, cred);
559
560 ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
561 ia = (struct in_ifaddr *)ifa;
562 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
563 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
564 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
565 callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
566 CALLOUT_RETURNUNLOCKED);
567
568 ia->ia_ifp = ifp;
569 ia->ia_addr = *addr;
570 if (mask->sin_len != 0) {
571 ia->ia_sockmask = *mask;
572 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
573 } else {
574 in_addr_t i = ntohl(addr->sin_addr.s_addr);
575
576 /*
577 * If netmask isn't supplied, use historical default.
578 * This is deprecated for interfaces other than loopback
579 * or point-to-point; warn in other cases. In the future
580 * we should return an error rather than warning.
581 */
582 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0)
583 printf("%s: set address: WARNING: network mask "
584 "should be specified; using historical default\n",
585 ifp->if_xname);
586 if (IN_CLASSA(i))
587 ia->ia_subnetmask = IN_CLASSA_NET;
588 else if (IN_CLASSB(i))
589 ia->ia_subnetmask = IN_CLASSB_NET;
590 else
591 ia->ia_subnetmask = IN_CLASSC_NET;
592 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
593 }
594 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
595 in_socktrim(&ia->ia_sockmask);
596
597 if (ifp->if_flags & IFF_BROADCAST) {
598 if (broadaddr->sin_len != 0) {
599 ia->ia_broadaddr = *broadaddr;
600 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
601 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
602 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
603 ia->ia_broadaddr.sin_family = AF_INET;
604 } else {
605 ia->ia_broadaddr.sin_addr.s_addr =
606 htonl(ia->ia_subnet | ~ia->ia_subnetmask);
607 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
608 ia->ia_broadaddr.sin_family = AF_INET;
609 }
610 }
611
612 if (ifp->if_flags & IFF_POINTOPOINT)
613 ia->ia_dstaddr = *dstaddr;
614
615 if (vhid != 0) {
616 error = (*carp_attach_p)(&ia->ia_ifa, vhid);
617 if (error)
618 return (error);
619 }
620
621 /* if_addrhead is already referenced by ifa_alloc() */
622 IF_ADDR_WLOCK(ifp);
623 CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
624 IF_ADDR_WUNLOCK(ifp);
625
626 ifa_ref(ifa); /* in_ifaddrhead */
627 sx_assert(&in_control_sx, SA_XLOCKED);
628 CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
629 CK_LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia,
630 ia_hash);
631
632 /*
633 * Give the interface a chance to initialize
634 * if this is its first address,
635 * and to validate the address if necessary.
636 */
637 if (ifp->if_ioctl != NULL) {
638 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
639 if (error)
640 goto fail1;
641 }
642
643 /*
644 * Add route for the network.
645 */
646 if (vhid == 0) {
647 error = in_addprefix(ia);
648 if (error)
649 goto fail1;
650 }
651
652 /*
653 * Add a loopback route to self.
654 */
655 if (vhid == 0 && ia_need_loopback_route(ia)) {
656 struct in_ifaddr *eia;
657
658 eia = in_localip_more(ia);
659
660 if (eia == NULL) {
661 error = ifa_add_loopback_route((struct ifaddr *)ia,
662 (struct sockaddr *)&ia->ia_addr);
663 if (error)
664 goto fail2;
665 } else
666 ifa_free(&eia->ia_ifa);
667 }
668
669 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
670 struct in_addr allhosts_addr;
671 struct in_ifinfo *ii;
672
673 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
674 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
675
676 error = in_joingroup(ifp, &allhosts_addr, NULL,
677 &ii->ii_allhosts);
678 }
679
680 /*
681 * Note: we don't need extra reference for ifa, since we called
682 * with sx lock held, and ifaddr can not be deleted in concurrent
683 * thread.
684 */
685 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
686
687 return (error);
688
689 fail2:
690 if (vhid == 0)
691 (void )in_scrubprefix(ia, LLE_STATIC);
692
693 fail1:
694 if (ia->ia_ifa.ifa_carp)
695 (*carp_detach_p)(&ia->ia_ifa, false);
696
697 IF_ADDR_WLOCK(ifp);
698 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
699 IF_ADDR_WUNLOCK(ifp);
700 ifa_free(&ia->ia_ifa); /* if_addrhead */
701
702 sx_assert(&in_control_sx, SA_XLOCKED);
703 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
704 CK_LIST_REMOVE(ia, ia_hash);
705 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */
706
707 return (error);
708 }
709
710 static int
in_difaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)711 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
712 {
713 const struct ifreq *ifr = (struct ifreq *)data;
714 const struct sockaddr_in *addr = (const struct sockaddr_in *)
715 &ifr->ifr_addr;
716 struct ifaddr *ifa;
717 struct in_ifaddr *ia;
718 bool deleteAny, iaIsLast;
719 int error;
720
721 if (cred != NULL) {
722 error = priv_check_cred(cred, PRIV_NET_DELIFADDR);
723 if (error)
724 return (error);
725 }
726
727 if (addr->sin_len != sizeof(struct sockaddr_in) ||
728 addr->sin_family != AF_INET)
729 deleteAny = true;
730 else
731 deleteAny = false;
732
733 iaIsLast = true;
734 ia = NULL;
735 IF_ADDR_WLOCK(ifp);
736 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
737 struct in_ifaddr *it;
738
739 if (ifa->ifa_addr->sa_family != AF_INET)
740 continue;
741
742 it = (struct in_ifaddr *)ifa;
743 if (deleteAny && ia == NULL && (cred == NULL ||
744 prison_check_ip4(cred, &it->ia_addr.sin_addr) == 0))
745 ia = it;
746
747 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
748 (cred == NULL || prison_check_ip4(cred,
749 &addr->sin_addr) == 0))
750 ia = it;
751
752 if (it != ia)
753 iaIsLast = false;
754 }
755
756 if (ia == NULL) {
757 IF_ADDR_WUNLOCK(ifp);
758 return (EADDRNOTAVAIL);
759 }
760
761 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
762 IF_ADDR_WUNLOCK(ifp);
763 ifa_free(&ia->ia_ifa); /* if_addrhead */
764
765 sx_assert(&in_control_sx, SA_XLOCKED);
766 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
767 CK_LIST_REMOVE(ia, ia_hash);
768
769 /*
770 * in_scrubprefix() kills the interface route.
771 */
772 in_scrubprefix(ia, LLE_STATIC);
773
774 /*
775 * in_ifadown gets rid of all the rest of
776 * the routes. This is not quite the right
777 * thing to do, but at least if we are running
778 * a routing process they will come back.
779 */
780 in_ifadown(&ia->ia_ifa, 1);
781
782 if (ia->ia_ifa.ifa_carp)
783 (*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
784
785 /*
786 * If this is the last IPv4 address configured on this
787 * interface, leave the all-hosts group.
788 * No state-change report need be transmitted.
789 */
790 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
791 struct in_ifinfo *ii;
792
793 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
794 if (ii->ii_allhosts) {
795 (void)in_leavegroup(ii->ii_allhosts, NULL);
796 ii->ii_allhosts = NULL;
797 }
798 }
799
800 IF_ADDR_WLOCK(ifp);
801 if (callout_stop(&ia->ia_garp_timer) == 1) {
802 ifa_free(&ia->ia_ifa);
803 }
804 IF_ADDR_WUNLOCK(ifp);
805
806 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
807 IFADDR_EVENT_DEL);
808 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */
809
810 return (0);
811 }
812
813 static int
in_gifaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)814 in_gifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
815 {
816 struct in_aliasreq *ifra = (struct in_aliasreq *)data;
817 const struct sockaddr_in *addr = &ifra->ifra_addr;
818 struct epoch_tracker et;
819 struct ifaddr *ifa;
820 struct in_ifaddr *ia;
821
822 /*
823 * ifra_addr must be present and be of INET family.
824 */
825 if (addr->sin_len != sizeof(struct sockaddr_in) ||
826 addr->sin_family != AF_INET)
827 return (EINVAL);
828
829 /*
830 * See whether address exist.
831 */
832 ia = NULL;
833 NET_EPOCH_ENTER(et);
834 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
835 struct in_ifaddr *it;
836
837 if (ifa->ifa_addr->sa_family != AF_INET)
838 continue;
839
840 it = (struct in_ifaddr *)ifa;
841 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
842 prison_check_ip4(cred, &addr->sin_addr) == 0) {
843 ia = it;
844 break;
845 }
846 }
847 if (ia == NULL) {
848 NET_EPOCH_EXIT(et);
849 return (EADDRNOTAVAIL);
850 }
851
852 ifra->ifra_mask = ia->ia_sockmask;
853 if ((ifp->if_flags & IFF_POINTOPOINT) &&
854 ia->ia_dstaddr.sin_family == AF_INET)
855 ifra->ifra_dstaddr = ia->ia_dstaddr;
856 else if ((ifp->if_flags & IFF_BROADCAST) &&
857 ia->ia_broadaddr.sin_family == AF_INET)
858 ifra->ifra_broadaddr = ia->ia_broadaddr;
859 else
860 memset(&ifra->ifra_broadaddr, 0,
861 sizeof(ifra->ifra_broadaddr));
862
863 NET_EPOCH_EXIT(et);
864 return (0);
865 }
866
867 static int
in_match_ifaddr(const struct rtentry * rt,const struct nhop_object * nh,void * arg)868 in_match_ifaddr(const struct rtentry *rt, const struct nhop_object *nh, void *arg)
869 {
870
871 if (nh->nh_ifa == (struct ifaddr *)arg)
872 return (1);
873
874 return (0);
875 }
876
877 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)878 in_handle_prefix_route(uint32_t fibnum, int cmd,
879 struct sockaddr_in *dst, struct sockaddr_in *netmask, struct ifaddr *ifa,
880 struct ifnet *ifp)
881 {
882
883 NET_EPOCH_ASSERT();
884
885 /* Prepare gateway */
886 struct sockaddr_dl_short sdl = {
887 .sdl_family = AF_LINK,
888 .sdl_len = sizeof(struct sockaddr_dl_short),
889 .sdl_type = ifa->ifa_ifp->if_type,
890 .sdl_index = ifa->ifa_ifp->if_index,
891 };
892
893 struct rt_addrinfo info = {
894 .rti_ifa = ifa,
895 .rti_ifp = ifp,
896 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
897 .rti_info = {
898 [RTAX_DST] = (struct sockaddr *)dst,
899 [RTAX_NETMASK] = (struct sockaddr *)netmask,
900 [RTAX_GATEWAY] = (struct sockaddr *)&sdl,
901 },
902 /* Ensure we delete the prefix IFF prefix ifa matches */
903 .rti_filter = in_match_ifaddr,
904 .rti_filterdata = ifa,
905 };
906
907 return (rib_handle_ifaddr_info(fibnum, cmd, &info));
908 }
909
910 /*
911 * Routing table interaction with interface addresses.
912 *
913 * In general, two types of routes needs to be installed:
914 * a) "interface" or "prefix" route, telling user that the addresses
915 * behind the ifa prefix are reached directly.
916 * b) "loopback" route installed for the ifa address, telling user that
917 * the address belongs to local system.
918 *
919 * Handling for (a) and (b) differs in multi-fib aspects, hence they
920 * are implemented in different functions below.
921 *
922 * The cases above may intersect - /32 interface aliases results in
923 * the same prefix produced by (a) and (b). This blurs the definition
924 * of the "loopback" route and complicate interactions. The interaction
925 * table is defined below. The case numbers are used in the multiple
926 * functions below to refer to the particular test case.
927 *
928 * There can be multiple options:
929 * 1) Adding address with prefix on non-p2p/non-loopback interface.
930 * Example: 192.0.2.1/24. Action:
931 * * add "prefix" route towards 192.0.2.0/24 via @ia interface,
932 * using @ia as an address source.
933 * * add "loopback" route towards 192.0.2.1 via V_loif, saving
934 * @ia ifp in the gateway and using @ia as an address source.
935 *
936 * 2) Adding address with /32 mask to non-p2p/non-loopback interface.
937 * Example: 192.0.2.2/32. Action:
938 * * add "prefix" host route via V_loif, using @ia as an address source.
939 *
940 * 3) Adding address with or without prefix to p2p interface.
941 * Example: 10.0.0.1/24->10.0.0.2. Action:
942 * * add "prefix" host route towards 10.0.0.2 via this interface, using @ia
943 * as an address source. Note: no sense in installing full /24 as the interface
944 * is point-to-point.
945 * * add "loopback" route towards 10.0.9.1 via V_loif, saving
946 * @ia ifp in the gateway and using @ia as an address source.
947 *
948 * 4) Adding address with or without prefix to loopback interface.
949 * Example: 192.0.2.1/24. Action:
950 * * add "prefix" host route via @ia interface, using @ia as an address source.
951 * Note: Skip installing /24 prefix as it would introduce TTL loop
952 * for the traffic destined to these addresses.
953 */
954
955 /*
956 * Checks if @ia needs to install loopback route to @ia address via
957 * ifa_maintain_loopback_route().
958 *
959 * Return true on success.
960 */
961 static bool
ia_need_loopback_route(const struct in_ifaddr * ia)962 ia_need_loopback_route(const struct in_ifaddr *ia)
963 {
964 struct ifnet *ifp = ia->ia_ifp;
965
966 /* Case 4: Skip loopback interfaces */
967 if ((ifp->if_flags & IFF_LOOPBACK) ||
968 (ia->ia_addr.sin_addr.s_addr == INADDR_ANY))
969 return (false);
970
971 /* Clash avoidance: Skip p2p interfaces with both addresses are equal */
972 if ((ifp->if_flags & IFF_POINTOPOINT) &&
973 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
974 return (false);
975
976 /* Case 2: skip /32 prefixes */
977 if (!(ifp->if_flags & IFF_POINTOPOINT) &&
978 (ia->ia_sockmask.sin_addr.s_addr == INADDR_BROADCAST))
979 return (false);
980
981 return (true);
982 }
983
984 /*
985 * Calculate "prefix" route corresponding to @ia.
986 */
987 static void
ia_getrtprefix(const struct in_ifaddr * ia,struct in_addr * prefix,struct in_addr * mask)988 ia_getrtprefix(const struct in_ifaddr *ia, struct in_addr *prefix, struct in_addr *mask)
989 {
990
991 if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) {
992 /* Case 3: return host route for dstaddr */
993 *prefix = ia->ia_dstaddr.sin_addr;
994 mask->s_addr = INADDR_BROADCAST;
995 } else if (ia->ia_ifp->if_flags & IFF_LOOPBACK) {
996 /* Case 4: return host route for ifaddr */
997 *prefix = ia->ia_addr.sin_addr;
998 mask->s_addr = INADDR_BROADCAST;
999 } else {
1000 /* Cases 1,2: return actual ia prefix */
1001 *prefix = ia->ia_addr.sin_addr;
1002 *mask = ia->ia_sockmask.sin_addr;
1003 prefix->s_addr &= mask->s_addr;
1004 }
1005 }
1006
1007 /*
1008 * Adds or delete interface "prefix" route corresponding to @ifa.
1009 * Returns 0 on success or errno.
1010 */
1011 static int
in_handle_ifaddr_route(int cmd,struct in_ifaddr * ia)1012 in_handle_ifaddr_route(int cmd, struct in_ifaddr *ia)
1013 {
1014 struct ifaddr *ifa = &ia->ia_ifa;
1015 struct in_addr daddr, maddr;
1016 struct sockaddr_in *pmask;
1017 struct epoch_tracker et;
1018 int error;
1019
1020 ia_getrtprefix(ia, &daddr, &maddr);
1021
1022 struct sockaddr_in mask = {
1023 .sin_family = AF_INET,
1024 .sin_len = sizeof(struct sockaddr_in),
1025 .sin_addr = maddr,
1026 };
1027
1028 pmask = (maddr.s_addr != INADDR_BROADCAST) ? &mask : NULL;
1029
1030 struct sockaddr_in dst = {
1031 .sin_family = AF_INET,
1032 .sin_len = sizeof(struct sockaddr_in),
1033 .sin_addr.s_addr = daddr.s_addr & maddr.s_addr,
1034 };
1035
1036 struct ifnet *ifp = ia->ia_ifp;
1037
1038 if ((maddr.s_addr == INADDR_BROADCAST) &&
1039 (!(ia->ia_ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)))) {
1040 /* Case 2: host route on broadcast interface */
1041 ifp = V_loif;
1042 }
1043
1044 uint32_t fibnum = ifa->ifa_ifp->if_fib;
1045 NET_EPOCH_ENTER(et);
1046 error = in_handle_prefix_route(fibnum, cmd, &dst, pmask, ifa, ifp);
1047 NET_EPOCH_EXIT(et);
1048
1049 return (error);
1050 }
1051
1052 /*
1053 * Check if we have a route for the given prefix already.
1054 */
1055 static bool
in_hasrtprefix(struct in_ifaddr * target)1056 in_hasrtprefix(struct in_ifaddr *target)
1057 {
1058 struct epoch_tracker et;
1059 struct in_ifaddr *ia;
1060 struct in_addr prefix, mask, p, m;
1061 bool result = false;
1062
1063 ia_getrtprefix(target, &prefix, &mask);
1064
1065 /* Look for an existing address with the same prefix, mask, and fib */
1066 NET_EPOCH_ENTER(et);
1067 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1068 ia_getrtprefix(ia, &p, &m);
1069
1070 if (prefix.s_addr != p.s_addr ||
1071 mask.s_addr != m.s_addr)
1072 continue;
1073
1074 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
1075 continue;
1076
1077 /*
1078 * If we got a matching prefix route inserted by other
1079 * interface address, we are done here.
1080 */
1081 if (ia->ia_flags & IFA_ROUTE) {
1082 result = true;
1083 break;
1084 }
1085 }
1086 NET_EPOCH_EXIT(et);
1087
1088 return (result);
1089 }
1090
1091 int
in_addprefix(struct in_ifaddr * target)1092 in_addprefix(struct in_ifaddr *target)
1093 {
1094 int error;
1095
1096 if (in_hasrtprefix(target)) {
1097 if (V_nosameprefix)
1098 return (EEXIST);
1099 else {
1100 rt_addrmsg(RTM_ADD, &target->ia_ifa,
1101 target->ia_ifp->if_fib);
1102 return (0);
1103 }
1104 }
1105
1106 /*
1107 * No-one seem to have this prefix route, so we try to insert it.
1108 */
1109 rt_addrmsg(RTM_ADD, &target->ia_ifa, target->ia_ifp->if_fib);
1110 error = in_handle_ifaddr_route(RTM_ADD, target);
1111 if (!error)
1112 target->ia_flags |= IFA_ROUTE;
1113 return (error);
1114 }
1115
1116 /*
1117 * Removes either all lle entries for given @ia, or lle
1118 * corresponding to @ia address.
1119 */
1120 static void
in_scrubprefixlle(struct in_ifaddr * ia,int all,u_int flags)1121 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
1122 {
1123 struct sockaddr_in addr, mask;
1124 struct sockaddr *saddr, *smask;
1125 struct ifnet *ifp;
1126
1127 saddr = (struct sockaddr *)&addr;
1128 bzero(&addr, sizeof(addr));
1129 addr.sin_len = sizeof(addr);
1130 addr.sin_family = AF_INET;
1131 smask = (struct sockaddr *)&mask;
1132 bzero(&mask, sizeof(mask));
1133 mask.sin_len = sizeof(mask);
1134 mask.sin_family = AF_INET;
1135 mask.sin_addr.s_addr = ia->ia_subnetmask;
1136 ifp = ia->ia_ifp;
1137
1138 if (all) {
1139 /*
1140 * Remove all L2 entries matching given prefix.
1141 * Convert address to host representation to avoid
1142 * doing this on every callback. ia_subnetmask is already
1143 * stored in host representation.
1144 */
1145 addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
1146 lltable_prefix_free(AF_INET, saddr, smask, flags);
1147 } else {
1148 /* Remove interface address only */
1149 addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
1150 lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
1151 }
1152 }
1153
1154 /*
1155 * If there is no other address in the system that can serve a route to the
1156 * same prefix, remove the route. Hand over the route to the new address
1157 * otherwise.
1158 */
1159 int
in_scrubprefix(struct in_ifaddr * target,u_int flags)1160 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1161 {
1162 struct epoch_tracker et;
1163 struct in_ifaddr *ia;
1164 struct in_addr prefix, mask, p, m;
1165 int error = 0;
1166
1167 /*
1168 * Remove the loopback route to the interface address.
1169 */
1170 if (ia_need_loopback_route(target) && (flags & LLE_STATIC)) {
1171 struct in_ifaddr *eia;
1172
1173 eia = in_localip_more(target);
1174
1175 if (eia != NULL) {
1176 error = ifa_switch_loopback_route((struct ifaddr *)eia,
1177 (struct sockaddr *)&target->ia_addr);
1178 ifa_free(&eia->ia_ifa);
1179 } else {
1180 error = ifa_del_loopback_route((struct ifaddr *)target,
1181 (struct sockaddr *)&target->ia_addr);
1182 }
1183 }
1184
1185 ia_getrtprefix(target, &prefix, &mask);
1186
1187 if ((target->ia_flags & IFA_ROUTE) == 0) {
1188 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1189
1190 /*
1191 * Removing address from !IFF_UP interface or
1192 * prefix which exists on other interface (along with route).
1193 * No entries should exist here except target addr.
1194 * Given that, delete this entry only.
1195 */
1196 in_scrubprefixlle(target, 0, flags);
1197 return (0);
1198 }
1199
1200 NET_EPOCH_ENTER(et);
1201 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1202 ia_getrtprefix(ia, &p, &m);
1203
1204 if (prefix.s_addr != p.s_addr ||
1205 mask.s_addr != m.s_addr)
1206 continue;
1207
1208 if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1209 continue;
1210
1211 /*
1212 * If we got a matching prefix address, move IFA_ROUTE and
1213 * the route itself to it. Make sure that routing daemons
1214 * get a heads-up.
1215 */
1216 if ((ia->ia_flags & IFA_ROUTE) == 0) {
1217 ifa_ref(&ia->ia_ifa);
1218 NET_EPOCH_EXIT(et);
1219 error = in_handle_ifaddr_route(RTM_DELETE, target);
1220 if (error == 0)
1221 target->ia_flags &= ~IFA_ROUTE;
1222 else
1223 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1224 error);
1225 /* Scrub all entries IFF interface is different */
1226 in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
1227 flags);
1228 error = in_handle_ifaddr_route(RTM_ADD, ia);
1229 if (error == 0)
1230 ia->ia_flags |= IFA_ROUTE;
1231 else
1232 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1233 error);
1234 ifa_free(&ia->ia_ifa);
1235 return (error);
1236 }
1237 }
1238 NET_EPOCH_EXIT(et);
1239
1240 /*
1241 * remove all L2 entries on the given prefix
1242 */
1243 in_scrubprefixlle(target, 1, flags);
1244
1245 /*
1246 * As no-one seem to have this prefix, we can remove the route.
1247 */
1248 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1249 error = in_handle_ifaddr_route(RTM_DELETE, target);
1250 if (error == 0)
1251 target->ia_flags &= ~IFA_ROUTE;
1252 else
1253 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1254 return (error);
1255 }
1256
1257 void
in_ifscrub_all(void)1258 in_ifscrub_all(void)
1259 {
1260 struct ifnet *ifp;
1261 struct ifaddr *ifa, *nifa;
1262 struct ifreq ifr;
1263
1264 IFNET_RLOCK();
1265 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1266 /* Cannot lock here - lock recursion. */
1267 /* NET_EPOCH_ENTER(et); */
1268 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1269 if (ifa->ifa_addr->sa_family != AF_INET)
1270 continue;
1271
1272 /*
1273 * This is ugly but the only way for legacy IP to
1274 * cleanly remove addresses and everything attached.
1275 */
1276 bzero(&ifr, sizeof(ifr));
1277 ifr.ifr_addr = *ifa->ifa_addr;
1278 (void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
1279 ifp, NULL);
1280 }
1281 /* NET_EPOCH_EXIT(et); */
1282 in_purgemaddrs(ifp);
1283 igmp_domifdetach(ifp);
1284 }
1285 IFNET_RUNLOCK();
1286 }
1287
1288 bool
in_ifaddr_broadcast(struct in_addr in,struct in_ifaddr * ia)1289 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
1290 {
1291
1292 return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1293 /*
1294 * Optionally check for old-style (host 0) broadcast, but
1295 * taking into account that RFC 3021 obsoletes it.
1296 */
1297 __predict_false(V_broadcast_lowest &&
1298 ia->ia_subnetmask != IN_RFC3021_MASK &&
1299 ntohl(in.s_addr) == ia->ia_subnet)) &&
1300 /*
1301 * Check for an all one subnetmask. These
1302 * only exist when an interface gets a secondary
1303 * address.
1304 */
1305 ia->ia_subnetmask != (u_long)0xffffffff);
1306 }
1307
1308 /*
1309 * Return true if the address might be a local broadcast address.
1310 */
1311 bool
in_ifnet_broadcast(struct in_addr in,struct ifnet * ifp)1312 in_ifnet_broadcast(struct in_addr in, struct ifnet *ifp)
1313 {
1314 struct ifaddr *ifa;
1315
1316 NET_EPOCH_ASSERT();
1317
1318 if (in_broadcast(in))
1319 return (true);
1320 if ((ifp->if_flags & IFF_BROADCAST) == 0)
1321 return (false);
1322 /*
1323 * Look through the list of addresses for a match
1324 * with a broadcast address.
1325 */
1326 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1327 if (ifa->ifa_addr->sa_family == AF_INET &&
1328 in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa))
1329 return (true);
1330 return (false);
1331 }
1332
1333 /*
1334 * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1335 */
1336 void
in_ifdetach(struct ifnet * ifp)1337 in_ifdetach(struct ifnet *ifp)
1338 {
1339 IN_MULTI_LOCK();
1340 in_pcbpurgeif0(&V_ripcbinfo, ifp);
1341 in_pcbpurgeif0(&V_udbinfo, ifp);
1342 in_pcbpurgeif0(&V_ulitecbinfo, ifp);
1343 in_purgemaddrs(ifp);
1344 IN_MULTI_UNLOCK();
1345
1346 /*
1347 * Make sure all multicast deletions invoking if_ioctl() are
1348 * completed before returning. Else we risk accessing a freed
1349 * ifnet structure pointer.
1350 */
1351 inm_release_wait(NULL);
1352 }
1353
1354 static void
in_ifnet_event(void * arg __unused,struct ifnet * ifp,int event)1355 in_ifnet_event(void *arg __unused, struct ifnet *ifp, int event)
1356 {
1357 struct epoch_tracker et;
1358 struct ifaddr *ifa;
1359 struct in_ifaddr *ia;
1360 int error;
1361
1362 NET_EPOCH_ENTER(et);
1363 switch (event) {
1364 case IFNET_EVENT_DOWN:
1365 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1366 if (ifa->ifa_addr->sa_family != AF_INET)
1367 continue;
1368 ia = (struct in_ifaddr *)ifa;
1369 if ((ia->ia_flags & IFA_ROUTE) == 0)
1370 continue;
1371 ifa_ref(ifa);
1372 /*
1373 * in_scrubprefix() kills the interface route.
1374 */
1375 in_scrubprefix(ia, 0);
1376 /*
1377 * in_ifadown gets rid of all the rest of the
1378 * routes. This is not quite the right thing
1379 * to do, but at least if we are running a
1380 * routing process they will come back.
1381 */
1382 in_ifadown(ifa, 0);
1383 ifa_free(ifa);
1384 }
1385 break;
1386
1387 case IFNET_EVENT_UP:
1388 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1389 if (ifa->ifa_addr->sa_family != AF_INET)
1390 continue;
1391 ia = (struct in_ifaddr *)ifa;
1392 if (ia->ia_flags & IFA_ROUTE)
1393 continue;
1394 ifa_ref(ifa);
1395 error = ifa_del_loopback_route(ifa, ifa->ifa_addr);
1396 rt_addrmsg(RTM_ADD, ifa, ifa->ifa_ifp->if_fib);
1397 error = in_handle_ifaddr_route(RTM_ADD, ia);
1398 if (error == 0)
1399 ia->ia_flags |= IFA_ROUTE;
1400 error = ifa_add_loopback_route(ifa, ifa->ifa_addr);
1401 ifa_free(ifa);
1402 }
1403 break;
1404 }
1405 NET_EPOCH_EXIT(et);
1406 }
1407 EVENTHANDLER_DEFINE(ifnet_event, in_ifnet_event, NULL, EVENTHANDLER_PRI_ANY);
1408
1409 /*
1410 * Delete all IPv4 multicast address records, and associated link-layer
1411 * multicast address records, associated with ifp.
1412 * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1413 * XXX This should not race with ifma_protospec being set during
1414 * a new allocation, if it does, we have bigger problems.
1415 */
1416 static void
in_purgemaddrs(struct ifnet * ifp)1417 in_purgemaddrs(struct ifnet *ifp)
1418 {
1419 struct epoch_tracker et;
1420 struct in_multi_head purgeinms;
1421 struct in_multi *inm;
1422 struct ifmultiaddr *ifma;
1423
1424 SLIST_INIT(&purgeinms);
1425 IN_MULTI_LIST_LOCK();
1426
1427 /*
1428 * Extract list of in_multi associated with the detaching ifp
1429 * which the PF_INET layer is about to release.
1430 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1431 * by code further down.
1432 */
1433 IF_ADDR_WLOCK(ifp);
1434 NET_EPOCH_ENTER(et);
1435 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1436 inm = inm_ifmultiaddr_get_inm(ifma);
1437 if (inm == NULL)
1438 continue;
1439 inm_rele_locked(&purgeinms, inm);
1440 }
1441 NET_EPOCH_EXIT(et);
1442 IF_ADDR_WUNLOCK(ifp);
1443
1444 inm_release_list_deferred(&purgeinms);
1445 igmp_ifdetach(ifp);
1446 IN_MULTI_LIST_UNLOCK();
1447 }
1448
1449 struct in_llentry {
1450 struct llentry base;
1451 };
1452
1453 #define IN_LLTBL_DEFAULT_HSIZE 32
1454 #define IN_LLTBL_HASH(k, h) \
1455 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1456
1457 /*
1458 * Do actual deallocation of @lle.
1459 */
1460 static void
in_lltable_destroy_lle_unlocked(epoch_context_t ctx)1461 in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1462 {
1463 struct llentry *lle;
1464
1465 lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1466 LLE_LOCK_DESTROY(lle);
1467 LLE_REQ_DESTROY(lle);
1468 free(lle, M_LLTABLE);
1469 }
1470
1471 /*
1472 * Called by LLE_FREE_LOCKED when number of references
1473 * drops to zero.
1474 */
1475 static void
in_lltable_destroy_lle(struct llentry * lle)1476 in_lltable_destroy_lle(struct llentry *lle)
1477 {
1478
1479 LLE_WUNLOCK(lle);
1480 NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1481 }
1482
1483 static struct llentry *
in_lltable_new(struct in_addr addr4,u_int flags)1484 in_lltable_new(struct in_addr addr4, u_int flags)
1485 {
1486 struct in_llentry *lle;
1487
1488 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1489 if (lle == NULL) /* NB: caller generates msg */
1490 return NULL;
1491
1492 /*
1493 * For IPv4 this will trigger "arpresolve" to generate
1494 * an ARP request.
1495 */
1496 lle->base.la_expire = time_uptime; /* mark expired */
1497 lle->base.r_l3addr.addr4 = addr4;
1498 lle->base.lle_refcnt = 1;
1499 lle->base.lle_free = in_lltable_destroy_lle;
1500 LLE_LOCK_INIT(&lle->base);
1501 LLE_REQ_INIT(&lle->base);
1502 callout_init(&lle->base.lle_timer, 1);
1503
1504 return (&lle->base);
1505 }
1506
1507 static int
in_lltable_match_prefix(const struct sockaddr * saddr,const struct sockaddr * smask,u_int flags,struct llentry * lle)1508 in_lltable_match_prefix(const struct sockaddr *saddr,
1509 const struct sockaddr *smask, u_int flags, struct llentry *lle)
1510 {
1511 struct in_addr addr, mask, lle_addr;
1512
1513 addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1514 mask = ((const struct sockaddr_in *)smask)->sin_addr;
1515 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1516
1517 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1518 return (0);
1519
1520 if (lle->la_flags & LLE_IFADDR) {
1521 /*
1522 * Delete LLE_IFADDR records IFF address & flag matches.
1523 * Note that addr is the interface address within prefix
1524 * being matched.
1525 * Note also we should handle 'ifdown' cases without removing
1526 * ifaddr macs.
1527 */
1528 if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1529 return (1);
1530 return (0);
1531 }
1532
1533 /* flags & LLE_STATIC means deleting both dynamic and static entries */
1534 if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1535 return (1);
1536
1537 return (0);
1538 }
1539
1540 static void
in_lltable_free_entry(struct lltable * llt,struct llentry * lle)1541 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1542 {
1543 size_t pkts_dropped;
1544
1545 LLE_WLOCK_ASSERT(lle);
1546 KASSERT(llt != NULL, ("lltable is NULL"));
1547
1548 /* Unlink entry from table if not already */
1549 if ((lle->la_flags & LLE_LINKED) != 0) {
1550 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1551 lltable_unlink_entry(llt, lle);
1552 }
1553
1554 /* Drop hold queue */
1555 pkts_dropped = llentry_free(lle);
1556 ARPSTAT_ADD(dropped, pkts_dropped);
1557 }
1558
1559 static int
in_lltable_rtcheck(struct ifnet * ifp,u_int flags,const struct sockaddr * l3addr)1560 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1561 {
1562 struct nhop_object *nh;
1563 struct in_addr addr;
1564
1565 KASSERT(l3addr->sa_family == AF_INET,
1566 ("sin_family %d", l3addr->sa_family));
1567
1568 addr = ((const struct sockaddr_in *)l3addr)->sin_addr;
1569
1570 nh = fib4_lookup(ifp->if_fib, addr, 0, NHR_NONE, 0);
1571 if (nh == NULL)
1572 return (EINVAL);
1573
1574 /*
1575 * If the gateway for an existing host route matches the target L3
1576 * address, which is a special route inserted by some implementation
1577 * such as MANET, and the interface is of the correct type, then
1578 * allow for ARP to proceed.
1579 */
1580 if (nh->nh_flags & NHF_GATEWAY) {
1581 if (!(nh->nh_flags & NHF_HOST) || nh->nh_ifp->if_type != IFT_ETHER ||
1582 (nh->nh_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1583 memcmp(nh->gw_sa.sa_data, l3addr->sa_data,
1584 sizeof(in_addr_t)) != 0) {
1585 return (EINVAL);
1586 }
1587 }
1588
1589 /*
1590 * Make sure that at least the destination address is covered
1591 * by the route. This is for handling the case where 2 or more
1592 * interfaces have the same prefix. An incoming packet arrives
1593 * on one interface and the corresponding outgoing packet leaves
1594 * another interface.
1595 */
1596 if ((nh->nh_ifp != ifp) && (nh->nh_flags & NHF_HOST) == 0) {
1597 struct in_ifaddr *ia = (struct in_ifaddr *)ifaof_ifpforaddr(l3addr, ifp);
1598 struct in_addr dst_addr, mask_addr;
1599
1600 if (ia == NULL)
1601 return (EINVAL);
1602
1603 /*
1604 * ifaof_ifpforaddr() returns _best matching_ IFA.
1605 * It is possible that ifa prefix does not cover our address.
1606 * Explicitly verify and fail if that's the case.
1607 */
1608 dst_addr = IA_SIN(ia)->sin_addr;
1609 mask_addr.s_addr = htonl(ia->ia_subnetmask);
1610
1611 if (!IN_ARE_MASKED_ADDR_EQUAL(dst_addr, addr, mask_addr))
1612 return (EINVAL);
1613 }
1614
1615 return (0);
1616 }
1617
1618 static inline uint32_t
in_lltable_hash_dst(const struct in_addr dst,uint32_t hsize)1619 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1620 {
1621
1622 return (IN_LLTBL_HASH(dst.s_addr, hsize));
1623 }
1624
1625 static uint32_t
in_lltable_hash(const struct llentry * lle,uint32_t hsize)1626 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1627 {
1628
1629 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1630 }
1631
1632 static void
in_lltable_fill_sa_entry(const struct llentry * lle,struct sockaddr * sa)1633 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1634 {
1635 struct sockaddr_in *sin;
1636
1637 sin = (struct sockaddr_in *)sa;
1638 bzero(sin, sizeof(*sin));
1639 sin->sin_family = AF_INET;
1640 sin->sin_len = sizeof(*sin);
1641 sin->sin_addr = lle->r_l3addr.addr4;
1642 }
1643
1644 static inline struct llentry *
in_lltable_find_dst(struct lltable * llt,struct in_addr dst)1645 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1646 {
1647 struct llentry *lle;
1648 struct llentries *lleh;
1649 u_int hashidx;
1650
1651 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1652 lleh = &llt->lle_head[hashidx];
1653 CK_LIST_FOREACH(lle, lleh, lle_next) {
1654 if (lle->la_flags & LLE_DELETED)
1655 continue;
1656 if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1657 break;
1658 }
1659
1660 return (lle);
1661 }
1662
1663 static void
in_lltable_delete_entry(struct lltable * llt,struct llentry * lle)1664 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1665 {
1666
1667 lle->la_flags |= LLE_DELETED;
1668 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1669 #ifdef DIAGNOSTIC
1670 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1671 #endif
1672 llentry_free(lle);
1673 }
1674
1675 static struct llentry *
in_lltable_alloc(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)1676 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1677 {
1678 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1679 struct ifnet *ifp = llt->llt_ifp;
1680 struct llentry *lle;
1681 char linkhdr[LLE_MAX_LINKHDR];
1682 size_t linkhdrsize;
1683 int lladdr_off;
1684
1685 KASSERT(l3addr->sa_family == AF_INET,
1686 ("sin_family %d", l3addr->sa_family));
1687
1688 /*
1689 * A route that covers the given address must have
1690 * been installed 1st because we are doing a resolution,
1691 * verify this.
1692 */
1693 if (!(flags & LLE_IFADDR) &&
1694 in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1695 return (NULL);
1696
1697 lle = in_lltable_new(sin->sin_addr, flags);
1698 if (lle == NULL) {
1699 log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1700 return (NULL);
1701 }
1702 lle->la_flags = flags;
1703 if (flags & LLE_STATIC)
1704 lle->r_flags |= RLLE_VALID;
1705 if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1706 linkhdrsize = LLE_MAX_LINKHDR;
1707 if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1708 linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1709 in_lltable_free_entry(llt, lle);
1710 return (NULL);
1711 }
1712 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1713 lladdr_off);
1714 lle->la_flags |= LLE_STATIC;
1715 lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1716 lle->la_expire = 0;
1717 }
1718
1719 return (lle);
1720 }
1721
1722 /*
1723 * Return NULL if not found or marked for deletion.
1724 * If found return lle read locked.
1725 */
1726 static struct llentry *
in_lltable_lookup(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)1727 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1728 {
1729 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1730 struct llentry *lle;
1731
1732 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1733 KASSERT(l3addr->sa_family == AF_INET,
1734 ("sin_family %d", l3addr->sa_family));
1735 KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1736 (LLE_UNLOCKED | LLE_EXCLUSIVE),
1737 ("wrong lle request flags: %#x", flags));
1738
1739 lle = in_lltable_find_dst(llt, sin->sin_addr);
1740 if (lle == NULL)
1741 return (NULL);
1742 if (flags & LLE_UNLOCKED)
1743 return (lle);
1744
1745 if (flags & LLE_EXCLUSIVE)
1746 LLE_WLOCK(lle);
1747 else
1748 LLE_RLOCK(lle);
1749
1750 /*
1751 * If the afdata lock is not held, the LLE may have been unlinked while
1752 * we were blocked on the LLE lock. Check for this case.
1753 */
1754 if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1755 if (flags & LLE_EXCLUSIVE)
1756 LLE_WUNLOCK(lle);
1757 else
1758 LLE_RUNLOCK(lle);
1759 return (NULL);
1760 }
1761 return (lle);
1762 }
1763
1764 static int
in_lltable_dump_entry(struct lltable * llt,struct llentry * lle,struct sysctl_req * wr)1765 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1766 struct sysctl_req *wr)
1767 {
1768 struct ifnet *ifp = llt->llt_ifp;
1769 /* XXX stack use */
1770 struct {
1771 struct rt_msghdr rtm;
1772 struct sockaddr_in sin;
1773 struct sockaddr_dl sdl;
1774 } arpc;
1775 struct sockaddr_dl *sdl;
1776 int error;
1777
1778 bzero(&arpc, sizeof(arpc));
1779 /* skip deleted entries */
1780 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1781 return (0);
1782 /* Skip if jailed and not a valid IP of the prison. */
1783 lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1784 if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1785 return (0);
1786 /*
1787 * produce a msg made of:
1788 * struct rt_msghdr;
1789 * struct sockaddr_in; (IPv4)
1790 * struct sockaddr_dl;
1791 */
1792 arpc.rtm.rtm_msglen = sizeof(arpc);
1793 arpc.rtm.rtm_version = RTM_VERSION;
1794 arpc.rtm.rtm_type = RTM_GET;
1795 arpc.rtm.rtm_flags = RTF_UP;
1796 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1797
1798 /* publish */
1799 if (lle->la_flags & LLE_PUB)
1800 arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1801
1802 sdl = &arpc.sdl;
1803 sdl->sdl_family = AF_LINK;
1804 sdl->sdl_len = sizeof(*sdl);
1805 sdl->sdl_index = ifp->if_index;
1806 sdl->sdl_type = ifp->if_type;
1807 if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1808 sdl->sdl_alen = ifp->if_addrlen;
1809 bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1810 } else {
1811 sdl->sdl_alen = 0;
1812 bzero(LLADDR(sdl), ifp->if_addrlen);
1813 }
1814
1815 arpc.rtm.rtm_rmx.rmx_expire =
1816 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1817 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1818 if (lle->la_flags & LLE_STATIC)
1819 arpc.rtm.rtm_flags |= RTF_STATIC;
1820 if (lle->la_flags & LLE_IFADDR)
1821 arpc.rtm.rtm_flags |= RTF_PINNED;
1822 arpc.rtm.rtm_index = ifp->if_index;
1823 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1824
1825 return (error);
1826 }
1827
1828 static void
in_lltable_post_resolved(struct lltable * llt,struct llentry * lle)1829 in_lltable_post_resolved(struct lltable *llt, struct llentry *lle)
1830 {
1831 struct ifnet *ifp = llt->llt_ifp;
1832
1833 /* gratuitous ARP */
1834 if ((lle->la_flags & LLE_PUB) != 0)
1835 arprequest(ifp, &lle->r_l3addr.addr4, &lle->r_l3addr.addr4,
1836 lle->ll_addr);
1837 }
1838
1839 static struct lltable *
in_lltattach(struct ifnet * ifp)1840 in_lltattach(struct ifnet *ifp)
1841 {
1842 struct lltable *llt;
1843
1844 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1845 llt->llt_af = AF_INET;
1846 llt->llt_ifp = ifp;
1847
1848 llt->llt_lookup = in_lltable_lookup;
1849 llt->llt_alloc_entry = in_lltable_alloc;
1850 llt->llt_delete_entry = in_lltable_delete_entry;
1851 llt->llt_dump_entry = in_lltable_dump_entry;
1852 llt->llt_hash = in_lltable_hash;
1853 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1854 llt->llt_free_entry = in_lltable_free_entry;
1855 llt->llt_match_prefix = in_lltable_match_prefix;
1856 llt->llt_mark_used = llentry_mark_used;
1857 llt->llt_post_resolved = in_lltable_post_resolved;
1858 lltable_link(llt);
1859
1860 return (llt);
1861 }
1862
1863 struct lltable *
in_lltable_get(struct ifnet * ifp)1864 in_lltable_get(struct ifnet *ifp)
1865 {
1866 struct lltable *llt = NULL;
1867
1868 void *afdata_ptr = ifp->if_afdata[AF_INET];
1869 if (afdata_ptr != NULL)
1870 llt = ((struct in_ifinfo *)afdata_ptr)->ii_llt;
1871 return (llt);
1872 }
1873
1874 void *
in_domifattach(struct ifnet * ifp)1875 in_domifattach(struct ifnet *ifp)
1876 {
1877 struct in_ifinfo *ii;
1878
1879 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1880
1881 ii->ii_llt = in_lltattach(ifp);
1882 ii->ii_igmp = igmp_domifattach(ifp);
1883
1884 return (ii);
1885 }
1886
1887 void
in_domifdetach(struct ifnet * ifp,void * aux)1888 in_domifdetach(struct ifnet *ifp, void *aux)
1889 {
1890 struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1891
1892 MPASS(ifp->if_afdata[AF_INET] == NULL);
1893
1894 igmp_domifdetach(ifp);
1895 lltable_free(ii->ii_llt);
1896 free(ii, M_IFADDR);
1897 }
1898