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 "opt_inet.h"
34
35 #define IN_HISTORICAL_NETS /* include class masks */
36
37 #include <sys/param.h>
38 #include <sys/eventhandler.h>
39 #include <sys/systm.h>
40 #include <sys/sockio.h>
41 #include <sys/malloc.h>
42 #include <sys/priv.h>
43 #include <sys/socket.h>
44 #include <sys/jail.h>
45 #include <sys/kernel.h>
46 #include <sys/lock.h>
47 #include <sys/proc.h>
48 #include <sys/sysctl.h>
49 #include <sys/syslog.h>
50 #include <sys/sx.h>
51
52 #include <net/if.h>
53 #include <net/if_var.h>
54 #include <net/if_arp.h>
55 #include <net/if_dl.h>
56 #include <net/if_llatbl.h>
57 #include <net/if_private.h>
58 #include <net/if_types.h>
59 #include <net/if_bridgevar.h>
60 #include <net/route.h>
61 #include <net/route/nhop.h>
62 #include <net/route/route_ctl.h>
63 #include <net/vnet.h>
64
65 #include <netinet/if_ether.h>
66 #include <netinet/in.h>
67 #include <netinet/in_fib.h>
68 #include <netinet/in_var.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/ip_carp.h>
72 #include <netinet/igmp_var.h>
73 #include <netinet/udp.h>
74 #include <netinet/udp_var.h>
75
76 #ifdef MAC
77 #include <security/mac/mac_framework.h>
78 #endif
79
80 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
81 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
82 static int in_gifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *);
83
84 static void in_socktrim(struct sockaddr_in *);
85 static void in_purgemaddrs(struct ifnet *);
86
87 static bool ia_need_loopback_route(const struct in_ifaddr *);
88
89 VNET_DEFINE_STATIC(int, nosameprefix);
90 #define V_nosameprefix VNET(nosameprefix)
91 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW,
92 &VNET_NAME(nosameprefix), 0,
93 "Refuse to create same prefixes on different interfaces");
94
95 VNET_DEFINE_STATIC(bool, broadcast_lowest);
96 #define V_broadcast_lowest VNET(broadcast_lowest)
97 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, broadcast_lowest, CTLFLAG_VNET | CTLFLAG_RW,
98 &VNET_NAME(broadcast_lowest), 0,
99 "Treat lowest address on a subnet (host 0) as broadcast");
100
101 VNET_DEFINE(bool, ip_allow_net240) = false;
102 #define V_ip_allow_net240 VNET(ip_allow_net240)
103 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net240,
104 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net240), 0,
105 "Allow forwarding of and ICMP response to Experimental addresses, aka Class E (240/4)");
106 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-240 */
107
108 VNET_DEFINE(bool, ip_allow_net0) = false;
109 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net0,
110 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net0), 0,
111 "Allow forwarding of and ICMP response to addresses in network 0/8");
112 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-0 */
113
114 VNET_DEFINE(uint32_t, in_loopback_mask) = IN_LOOPBACK_MASK_DFLT;
115 #define V_in_loopback_mask VNET(in_loopback_mask)
116 static int sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS);
117 SYSCTL_PROC(_net_inet_ip, OID_AUTO, loopback_prefixlen,
118 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
119 NULL, 0, sysctl_loopback_prefixlen, "I",
120 "Prefix length of address space reserved for loopback");
121 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-127 */
122
123 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
124 #define V_ripcbinfo VNET(ripcbinfo)
125
126 static struct sx in_control_sx;
127 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control");
128
129 /*
130 * Return true if an internet address is for a ``local'' host
131 * (one to which we have a connection).
132 */
133 bool
in_localaddr(struct in_addr in)134 in_localaddr(struct in_addr in)
135 {
136 u_long i = ntohl(in.s_addr);
137 struct in_ifaddr *ia;
138
139 NET_EPOCH_ASSERT();
140
141 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
142 if ((i & ia->ia_subnetmask) == ia->ia_subnet)
143 return (true);
144 }
145
146 return (false);
147 }
148
149 /*
150 * Return true if an internet address is for the local host and configured
151 * on one of its interfaces.
152 */
153 bool
in_localip(struct in_addr in)154 in_localip(struct in_addr in)
155 {
156 struct in_ifaddr *ia;
157
158 NET_EPOCH_ASSERT();
159
160 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
161 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
162 return (true);
163
164 return (false);
165 }
166
167 /*
168 * Like in_localip(), but FIB-aware and carp(4)-aware.
169 */
170 bool
in_localip_fib(struct in_addr in,uint16_t fib)171 in_localip_fib(struct in_addr in, uint16_t fib)
172 {
173 struct in_ifaddr *ia;
174
175 NET_EPOCH_ASSERT();
176
177 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash)
178 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr &&
179 (ia->ia_ifa.ifa_carp == NULL ||
180 carp_master_p(&ia->ia_ifa)) &&
181 ia->ia_ifa.ifa_ifp->if_fib == fib)
182 return (true);
183
184 return (false);
185 }
186
187 /*
188 * Return true if an internet address is configured on an interface.
189 */
190 bool
in_ifhasaddr(struct ifnet * ifp,struct in_addr in)191 in_ifhasaddr(struct ifnet *ifp, struct in_addr in)
192 {
193 struct ifaddr *ifa;
194 struct in_ifaddr *ia;
195
196 NET_EPOCH_ASSERT();
197
198 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
199 if (ifa->ifa_addr->sa_family != AF_INET)
200 continue;
201 ia = (struct in_ifaddr *)ifa;
202 if (ia->ia_addr.sin_addr.s_addr == in.s_addr)
203 return (true);
204 }
205
206 return (false);
207 }
208
209 /*
210 * Return a reference to the interface address which is different to
211 * the supplied one but with same IP address value.
212 */
213 static struct in_ifaddr *
in_localip_more(struct in_ifaddr * original_ia)214 in_localip_more(struct in_ifaddr *original_ia)
215 {
216 struct epoch_tracker et;
217 in_addr_t original_addr = IA_SIN(original_ia)->sin_addr.s_addr;
218 uint32_t original_fib = original_ia->ia_ifa.ifa_ifp->if_fib;
219 struct in_ifaddr *ia;
220
221 NET_EPOCH_ENTER(et);
222 CK_LIST_FOREACH(ia, INADDR_HASH(original_addr), ia_hash) {
223 in_addr_t addr = IA_SIN(ia)->sin_addr.s_addr;
224 uint32_t fib = ia->ia_ifa.ifa_ifp->if_fib;
225 if (!V_rt_add_addr_allfibs && (original_fib != fib))
226 continue;
227 if ((original_ia != ia) && (original_addr == addr)) {
228 ifa_ref(&ia->ia_ifa);
229 NET_EPOCH_EXIT(et);
230 return (ia);
231 }
232 }
233 NET_EPOCH_EXIT(et);
234
235 return (NULL);
236 }
237
238 /*
239 * Tries to find first IPv4 address in the provided fib.
240 * Prefers non-loopback addresses and return loopback IFF
241 * @loopback_ok is set.
242 *
243 * Returns ifa or NULL.
244 */
245 struct in_ifaddr *
in_findlocal(uint32_t fibnum,bool loopback_ok)246 in_findlocal(uint32_t fibnum, bool loopback_ok)
247 {
248 struct in_ifaddr *ia = NULL, *ia_lo = NULL;
249
250 NET_EPOCH_ASSERT();
251
252 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
253 uint32_t ia_fib = ia->ia_ifa.ifa_ifp->if_fib;
254 if (!V_rt_add_addr_allfibs && (fibnum != ia_fib))
255 continue;
256
257 if (!IN_LOOPBACK(ntohl(IA_SIN(ia)->sin_addr.s_addr)))
258 break;
259 if (loopback_ok)
260 ia_lo = ia;
261 }
262
263 if (ia == NULL)
264 ia = ia_lo;
265
266 return (ia);
267 }
268
269 /*
270 * Determine whether an IP address is in a reserved set of addresses
271 * that may not be forwarded, or whether datagrams to that destination
272 * may be forwarded.
273 */
274 bool
in_canforward(struct in_addr in)275 in_canforward(struct in_addr in)
276 {
277 u_long i = ntohl(in.s_addr);
278
279 if (IN_MULTICAST(i) || IN_LINKLOCAL(i) || IN_LOOPBACK(i) ||
280 in_nullhost(in))
281 return (false);
282 if (IN_EXPERIMENTAL(i) && !V_ip_allow_net240)
283 return (false);
284 if (IN_ZERONET(i) && !V_ip_allow_net0)
285 return (false);
286 return (true);
287 }
288
289 /*
290 * Sysctl to manage prefix of reserved loopback network; translate
291 * to/from mask. The mask is always contiguous high-order 1 bits
292 * followed by all 0 bits.
293 */
294 static int
sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS)295 sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS)
296 {
297 int error, preflen;
298
299 /* ffs is 1-based; compensate. */
300 preflen = 33 - ffs(V_in_loopback_mask);
301 error = sysctl_handle_int(oidp, &preflen, 0, req);
302 if (error || !req->newptr)
303 return (error);
304 if (preflen < 8 || preflen > 31)
305 return (EINVAL);
306 V_in_loopback_mask = 0xffffffff << (32 - preflen);
307 return (0);
308 }
309
310 /*
311 * Trim a mask in a sockaddr
312 */
313 static void
in_socktrim(struct sockaddr_in * ap)314 in_socktrim(struct sockaddr_in *ap)
315 {
316 char *cplim = (char *) &ap->sin_addr;
317 char *cp = (char *) (&ap->sin_addr + 1);
318
319 ap->sin_len = 0;
320 while (--cp >= cplim)
321 if (*cp) {
322 (ap)->sin_len = cp - (char *) (ap) + 1;
323 break;
324 }
325 }
326
327 /*
328 * Generic internet control operations (ioctl's).
329 */
330 int
in_control_ioctl(u_long cmd,void * data,struct ifnet * ifp,struct ucred * cred)331 in_control_ioctl(u_long cmd, void *data, struct ifnet *ifp,
332 struct ucred *cred)
333 {
334 struct ifreq *ifr = (struct ifreq *)data;
335 struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
336 struct epoch_tracker et;
337 struct ifaddr *ifa;
338 struct in_ifaddr *ia;
339 int error;
340
341 if (ifp == NULL)
342 return (EADDRNOTAVAIL);
343
344 /*
345 * Filter out 4 ioctls we implement directly. Forward the rest
346 * to specific functions and ifp->if_ioctl().
347 */
348 switch (cmd) {
349 case SIOCGIFADDR:
350 case SIOCGIFBRDADDR:
351 case SIOCGIFDSTADDR:
352 case SIOCGIFNETMASK:
353 break;
354 case SIOCGIFALIAS:
355 sx_xlock(&in_control_sx);
356 error = in_gifaddr_ioctl(cmd, data, ifp, cred);
357 sx_xunlock(&in_control_sx);
358 return (error);
359 case SIOCDIFADDR:
360 sx_xlock(&in_control_sx);
361 error = in_difaddr_ioctl(cmd, data, ifp, cred);
362 sx_xunlock(&in_control_sx);
363 return (error);
364 case OSIOCAIFADDR: /* 9.x compat */
365 case SIOCAIFADDR:
366 sx_xlock(&in_control_sx);
367 error = in_aifaddr_ioctl(cmd, data, ifp, cred);
368 sx_xunlock(&in_control_sx);
369 return (error);
370 case SIOCSIFADDR:
371 case SIOCSIFBRDADDR:
372 case SIOCSIFDSTADDR:
373 case SIOCSIFNETMASK:
374 /* We no longer support that old commands. */
375 return (EINVAL);
376 default:
377 if (ifp->if_ioctl == NULL)
378 return (EOPNOTSUPP);
379 return ((*ifp->if_ioctl)(ifp, cmd, data));
380 }
381
382 if (addr->sin_addr.s_addr != INADDR_ANY &&
383 prison_check_ip4(cred, &addr->sin_addr) != 0)
384 return (EADDRNOTAVAIL);
385
386 /*
387 * Find address for this interface, if it exists. If an
388 * address was specified, find that one instead of the
389 * first one on the interface, if possible.
390 */
391 NET_EPOCH_ENTER(et);
392 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
393 if (ifa->ifa_addr->sa_family != AF_INET)
394 continue;
395 ia = (struct in_ifaddr *)ifa;
396 if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
397 break;
398 }
399 if (ifa == NULL)
400 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
401 if (ifa->ifa_addr->sa_family == AF_INET) {
402 ia = (struct in_ifaddr *)ifa;
403 if (prison_check_ip4(cred,
404 &ia->ia_addr.sin_addr) == 0)
405 break;
406 }
407
408 if (ifa == NULL) {
409 NET_EPOCH_EXIT(et);
410 return (EADDRNOTAVAIL);
411 }
412
413 error = 0;
414 switch (cmd) {
415 case SIOCGIFADDR:
416 *addr = ia->ia_addr;
417 break;
418
419 case SIOCGIFBRDADDR:
420 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
421 error = EINVAL;
422 break;
423 }
424 *addr = ia->ia_broadaddr;
425 break;
426
427 case SIOCGIFDSTADDR:
428 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
429 error = EINVAL;
430 break;
431 }
432 *addr = ia->ia_dstaddr;
433 break;
434
435 case SIOCGIFNETMASK:
436 *addr = ia->ia_sockmask;
437 break;
438 }
439
440 NET_EPOCH_EXIT(et);
441
442 return (error);
443 }
444
445 int
in_mask2len(struct in_addr * mask)446 in_mask2len(struct in_addr *mask)
447 {
448 int x, y;
449 u_char *p;
450
451 p = (u_char *)mask;
452 for (x = 0; x < sizeof(*mask); x++) {
453 if (p[x] != 0xff)
454 break;
455 }
456 y = 0;
457 if (x < sizeof(*mask)) {
458 for (y = 0; y < 8; y++) {
459 if ((p[x] & (0x80 >> y)) == 0)
460 break;
461 }
462 }
463 return (x * 8 + y);
464 }
465
466 int
in_control(struct socket * so,u_long cmd,void * data,struct ifnet * ifp,struct thread * td)467 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
468 struct thread *td)
469 {
470 return (in_control_ioctl(cmd, data, ifp, td ? td->td_ucred : NULL));
471 }
472
473 static int
in_aifaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)474 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
475 {
476 const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
477 const struct sockaddr_in *addr = &ifra->ifra_addr;
478 const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
479 const struct sockaddr_in *mask = &ifra->ifra_mask;
480 const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
481 const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
482 struct epoch_tracker et;
483 struct ifaddr *ifa;
484 struct in_ifaddr *ia;
485 bool iaIsFirst;
486 int error = 0;
487
488 error = priv_check_cred(cred, PRIV_NET_ADDIFADDR);
489 if (error)
490 return (error);
491
492 /*
493 * ifra_addr must be present and be of INET family.
494 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
495 */
496 if (addr->sin_len != sizeof(struct sockaddr_in) ||
497 addr->sin_family != AF_INET)
498 return (EINVAL);
499 if (broadaddr->sin_len != 0 &&
500 (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
501 broadaddr->sin_family != AF_INET))
502 return (EINVAL);
503 if (mask->sin_len != 0 &&
504 (mask->sin_len != sizeof(struct sockaddr_in) ||
505 mask->sin_family != AF_INET))
506 return (EINVAL);
507 if ((ifp->if_flags & IFF_POINTOPOINT) &&
508 (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
509 dstaddr->sin_addr.s_addr == INADDR_ANY))
510 return (EDESTADDRREQ);
511 if (vhid != 0 && carp_attach_p == NULL)
512 return (EPROTONOSUPPORT);
513
514 #ifdef MAC
515 /* Check if a MAC policy disallows setting the IPv4 address. */
516 error = mac_inet_check_add_addr(cred, &addr->sin_addr, ifp);
517 if (error != 0)
518 return (error);
519 #endif
520
521 /*
522 * Check if bridge wants to allow adding addrs to member interfaces.
523 */
524 if (ifp->if_bridge != NULL && ifp->if_type != IFT_GIF &&
525 bridge_member_ifaddrs_p != NULL) {
526 if (bridge_member_ifaddrs_p())
527 if_printf(ifp, "WARNING: Assigning an IP address to "
528 "an interface which is also a bridge member is "
529 "deprecated and will be unsupported in a future "
530 "release.\n");
531 else
532 return (EINVAL);
533 }
534
535 /*
536 * See whether address already exist.
537 */
538 iaIsFirst = true;
539 ia = NULL;
540 NET_EPOCH_ENTER(et);
541 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
542 struct in_ifaddr *it;
543
544 if (ifa->ifa_addr->sa_family != AF_INET)
545 continue;
546
547 it = (struct in_ifaddr *)ifa;
548 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
549 prison_check_ip4(cred, &addr->sin_addr) == 0)
550 ia = it;
551 else
552 iaIsFirst = false;
553 }
554 NET_EPOCH_EXIT(et);
555
556 if (ia != NULL)
557 (void )in_difaddr_ioctl(cmd, data, ifp, cred);
558
559 ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
560 ia = (struct in_ifaddr *)ifa;
561 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
562 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
563 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
564 callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
565 CALLOUT_RETURNUNLOCKED);
566
567 ia->ia_ifp = ifp;
568 ia->ia_addr = *addr;
569 if (mask->sin_len != 0) {
570 ia->ia_sockmask = *mask;
571 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
572 } else {
573 in_addr_t i = ntohl(addr->sin_addr.s_addr);
574
575 /*
576 * If netmask isn't supplied, use historical default.
577 * This is deprecated for interfaces other than loopback
578 * or point-to-point; warn in other cases. In the future
579 * we should return an error rather than warning.
580 */
581 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0)
582 printf("%s: set address: WARNING: network mask "
583 "should be specified; using historical default\n",
584 ifp->if_xname);
585 if (IN_CLASSA(i))
586 ia->ia_subnetmask = IN_CLASSA_NET;
587 else if (IN_CLASSB(i))
588 ia->ia_subnetmask = IN_CLASSB_NET;
589 else
590 ia->ia_subnetmask = IN_CLASSC_NET;
591 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
592 }
593 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
594 in_socktrim(&ia->ia_sockmask);
595
596 if (ifp->if_flags & IFF_BROADCAST) {
597 if (broadaddr->sin_len != 0) {
598 ia->ia_broadaddr = *broadaddr;
599 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
600 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
601 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
602 ia->ia_broadaddr.sin_family = AF_INET;
603 } else {
604 ia->ia_broadaddr.sin_addr.s_addr =
605 htonl(ia->ia_subnet | ~ia->ia_subnetmask);
606 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
607 ia->ia_broadaddr.sin_family = AF_INET;
608 }
609 }
610
611 if (ifp->if_flags & IFF_POINTOPOINT)
612 ia->ia_dstaddr = *dstaddr;
613
614 if (vhid != 0) {
615 error = (*carp_attach_p)(&ia->ia_ifa, vhid);
616 if (error)
617 return (error);
618 }
619
620 /* if_addrhead is already referenced by ifa_alloc() */
621 IF_ADDR_WLOCK(ifp);
622 CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
623 IF_ADDR_WUNLOCK(ifp);
624
625 ifa_ref(ifa); /* in_ifaddrhead */
626 sx_assert(&in_control_sx, SA_XLOCKED);
627 CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
628 CK_LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia,
629 ia_hash);
630
631 /*
632 * Give the interface a chance to initialize
633 * if this is its first address,
634 * and to validate the address if necessary.
635 */
636 if (ifp->if_ioctl != NULL) {
637 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
638 if (error)
639 goto fail1;
640 }
641
642 /*
643 * Add route for the network.
644 */
645 if (vhid == 0) {
646 error = in_addprefix(ia);
647 if (error)
648 goto fail1;
649 }
650
651 /*
652 * Add a loopback route to self.
653 */
654 if (vhid == 0 && ia_need_loopback_route(ia)) {
655 struct in_ifaddr *eia;
656
657 eia = in_localip_more(ia);
658
659 if (eia == NULL) {
660 error = ifa_add_loopback_route((struct ifaddr *)ia,
661 (struct sockaddr *)&ia->ia_addr);
662 if (error)
663 goto fail2;
664 } else
665 ifa_free(&eia->ia_ifa);
666 }
667
668 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
669 struct in_addr allhosts_addr;
670 struct in_ifinfo *ii;
671
672 ii = (struct in_ifinfo *)ifp->if_inet;
673 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
674
675 error = in_joingroup(ifp, &allhosts_addr, NULL,
676 &ii->ii_allhosts);
677 }
678
679 /*
680 * Note: we don't need extra reference for ifa, since we called
681 * with sx lock held, and ifaddr can not be deleted in concurrent
682 * thread.
683 */
684 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
685
686 return (error);
687
688 fail2:
689 if (vhid == 0)
690 (void )in_scrubprefix(ia, LLE_STATIC);
691
692 fail1:
693 if (ia->ia_ifa.ifa_carp)
694 (*carp_detach_p)(&ia->ia_ifa, false);
695
696 IF_ADDR_WLOCK(ifp);
697 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
698 IF_ADDR_WUNLOCK(ifp);
699 ifa_free(&ia->ia_ifa); /* if_addrhead */
700
701 sx_assert(&in_control_sx, SA_XLOCKED);
702 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
703 CK_LIST_REMOVE(ia, ia_hash);
704 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */
705
706 return (error);
707 }
708
709 static int
in_difaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)710 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
711 {
712 const struct ifreq *ifr = (struct ifreq *)data;
713 const struct sockaddr_in *addr = (const struct sockaddr_in *)
714 &ifr->ifr_addr;
715 struct ifaddr *ifa;
716 struct in_ifaddr *ia;
717 bool deleteAny, iaIsLast;
718 int error;
719
720 if (cred != NULL) {
721 error = priv_check_cred(cred, PRIV_NET_DELIFADDR);
722 if (error)
723 return (error);
724 }
725
726 if (addr->sin_len != sizeof(struct sockaddr_in) ||
727 addr->sin_family != AF_INET)
728 deleteAny = true;
729 else
730 deleteAny = false;
731
732 iaIsLast = true;
733 ia = NULL;
734 IF_ADDR_WLOCK(ifp);
735 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
736 struct in_ifaddr *it;
737
738 if (ifa->ifa_addr->sa_family != AF_INET)
739 continue;
740
741 it = (struct in_ifaddr *)ifa;
742 if (deleteAny && ia == NULL && (cred == NULL ||
743 prison_check_ip4(cred, &it->ia_addr.sin_addr) == 0))
744 ia = it;
745
746 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
747 (cred == NULL || prison_check_ip4(cred,
748 &addr->sin_addr) == 0))
749 ia = it;
750
751 if (it != ia)
752 iaIsLast = false;
753 }
754
755 if (ia == NULL) {
756 IF_ADDR_WUNLOCK(ifp);
757 return (EADDRNOTAVAIL);
758 }
759
760 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
761 IF_ADDR_WUNLOCK(ifp);
762 ifa_free(&ia->ia_ifa); /* if_addrhead */
763
764 sx_assert(&in_control_sx, SA_XLOCKED);
765 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
766 CK_LIST_REMOVE(ia, ia_hash);
767
768 /*
769 * in_scrubprefix() kills the interface route.
770 */
771 in_scrubprefix(ia, LLE_STATIC);
772
773 /*
774 * in_ifadown gets rid of all the rest of
775 * the routes. This is not quite the right
776 * thing to do, but at least if we are running
777 * a routing process they will come back.
778 */
779 in_ifadown(&ia->ia_ifa, 1);
780
781 if (ia->ia_ifa.ifa_carp)
782 (*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
783
784 /*
785 * If this is the last IPv4 address configured on this
786 * interface, leave the all-hosts group.
787 * No state-change report need be transmitted.
788 */
789 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
790 struct in_ifinfo *ii;
791
792 ii = (struct in_ifinfo *)ifp->if_inet;
793 if (ii->ii_allhosts) {
794 (void)in_leavegroup(ii->ii_allhosts, NULL);
795 ii->ii_allhosts = NULL;
796 }
797 }
798
799 IF_ADDR_WLOCK(ifp);
800 if (callout_stop(&ia->ia_garp_timer) == 1) {
801 ifa_free(&ia->ia_ifa);
802 }
803 IF_ADDR_WUNLOCK(ifp);
804
805 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
806 IFADDR_EVENT_DEL);
807 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */
808
809 return (0);
810 }
811
812 static int
in_gifaddr_ioctl(u_long cmd,caddr_t data,struct ifnet * ifp,struct ucred * cred)813 in_gifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred)
814 {
815 struct in_aliasreq *ifra = (struct in_aliasreq *)data;
816 const struct sockaddr_in *addr = &ifra->ifra_addr;
817 struct epoch_tracker et;
818 struct ifaddr *ifa;
819 struct in_ifaddr *ia;
820
821 /*
822 * ifra_addr must be present and be of INET family.
823 */
824 if (addr->sin_len != sizeof(struct sockaddr_in) ||
825 addr->sin_family != AF_INET)
826 return (EINVAL);
827
828 /*
829 * See whether address exist.
830 */
831 ia = NULL;
832 NET_EPOCH_ENTER(et);
833 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
834 struct in_ifaddr *it;
835
836 if (ifa->ifa_addr->sa_family != AF_INET)
837 continue;
838
839 it = (struct in_ifaddr *)ifa;
840 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
841 prison_check_ip4(cred, &addr->sin_addr) == 0) {
842 ia = it;
843 break;
844 }
845 }
846 if (ia == NULL) {
847 NET_EPOCH_EXIT(et);
848 return (EADDRNOTAVAIL);
849 }
850
851 ifra->ifra_mask = ia->ia_sockmask;
852 if ((ifp->if_flags & IFF_POINTOPOINT) &&
853 ia->ia_dstaddr.sin_family == AF_INET)
854 ifra->ifra_dstaddr = ia->ia_dstaddr;
855 else if ((ifp->if_flags & IFF_BROADCAST) &&
856 ia->ia_broadaddr.sin_family == AF_INET)
857 ifra->ifra_broadaddr = ia->ia_broadaddr;
858 else
859 memset(&ifra->ifra_broadaddr, 0,
860 sizeof(ifra->ifra_broadaddr));
861
862 NET_EPOCH_EXIT(et);
863 return (0);
864 }
865
866 static int
in_match_ifaddr(const struct rtentry * rt,const struct nhop_object * nh,void * arg)867 in_match_ifaddr(const struct rtentry *rt, const struct nhop_object *nh, void *arg)
868 {
869
870 if (nh->nh_ifa == (struct ifaddr *)arg)
871 return (1);
872
873 return (0);
874 }
875
876 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)877 in_handle_prefix_route(uint32_t fibnum, int cmd,
878 struct sockaddr_in *dst, struct sockaddr_in *netmask, struct ifaddr *ifa,
879 struct ifnet *ifp)
880 {
881
882 NET_EPOCH_ASSERT();
883
884 /* Prepare gateway */
885 struct sockaddr_dl_short sdl = {
886 .sdl_family = AF_LINK,
887 .sdl_len = sizeof(struct sockaddr_dl_short),
888 .sdl_type = ifa->ifa_ifp->if_type,
889 .sdl_index = ifa->ifa_ifp->if_index,
890 };
891
892 struct rt_addrinfo info = {
893 .rti_ifa = ifa,
894 .rti_ifp = ifp,
895 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST),
896 .rti_info = {
897 [RTAX_DST] = (struct sockaddr *)dst,
898 [RTAX_NETMASK] = (struct sockaddr *)netmask,
899 [RTAX_GATEWAY] = (struct sockaddr *)&sdl,
900 },
901 /* Ensure we delete the prefix IFF prefix ifa matches */
902 .rti_filter = in_match_ifaddr,
903 .rti_filterdata = ifa,
904 };
905
906 return (rib_handle_ifaddr_info(fibnum, cmd, &info));
907 }
908
909 /*
910 * Routing table interaction with interface addresses.
911 *
912 * In general, two types of routes needs to be installed:
913 * a) "interface" or "prefix" route, telling user that the addresses
914 * behind the ifa prefix are reached directly.
915 * b) "loopback" route installed for the ifa address, telling user that
916 * the address belongs to local system.
917 *
918 * Handling for (a) and (b) differs in multi-fib aspects, hence they
919 * are implemented in different functions below.
920 *
921 * The cases above may intersect - /32 interface aliases results in
922 * the same prefix produced by (a) and (b). This blurs the definition
923 * of the "loopback" route and complicate interactions. The interaction
924 * table is defined below. The case numbers are used in the multiple
925 * functions below to refer to the particular test case.
926 *
927 * There can be multiple options:
928 * 1) Adding address with prefix on non-p2p/non-loopback interface.
929 * Example: 192.0.2.1/24. Action:
930 * * add "prefix" route towards 192.0.2.0/24 via @ia interface,
931 * using @ia as an address source.
932 * * add "loopback" route towards 192.0.2.1 via V_loif, saving
933 * @ia ifp in the gateway and using @ia as an address source.
934 *
935 * 2) Adding address with /32 mask to non-p2p/non-loopback interface.
936 * Example: 192.0.2.2/32. Action:
937 * * add "prefix" host route via V_loif, using @ia as an address source.
938 *
939 * 3) Adding address with or without prefix to p2p interface.
940 * Example: 10.0.0.1/24->10.0.0.2. Action:
941 * * add "prefix" host route towards 10.0.0.2 via this interface, using @ia
942 * as an address source. Note: no sense in installing full /24 as the interface
943 * is point-to-point.
944 * * add "loopback" route towards 10.0.9.1 via V_loif, saving
945 * @ia ifp in the gateway and using @ia as an address source.
946 *
947 * 4) Adding address with or without prefix to loopback interface.
948 * Example: 192.0.2.1/24. Action:
949 * * add "prefix" host route via @ia interface, using @ia as an address source.
950 * Note: Skip installing /24 prefix as it would introduce TTL loop
951 * for the traffic destined to these addresses.
952 */
953
954 /*
955 * Checks if @ia needs to install loopback route to @ia address via
956 * ifa_maintain_loopback_route().
957 *
958 * Return true on success.
959 */
960 static bool
ia_need_loopback_route(const struct in_ifaddr * ia)961 ia_need_loopback_route(const struct in_ifaddr *ia)
962 {
963 struct ifnet *ifp = ia->ia_ifp;
964
965 /* Case 4: Skip loopback interfaces */
966 if ((ifp->if_flags & IFF_LOOPBACK) ||
967 (ia->ia_addr.sin_addr.s_addr == INADDR_ANY))
968 return (false);
969
970 /* Clash avoidance: Skip p2p interfaces with both addresses are equal */
971 if ((ifp->if_flags & IFF_POINTOPOINT) &&
972 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
973 return (false);
974
975 /* Case 2: skip /32 prefixes */
976 if (!(ifp->if_flags & IFF_POINTOPOINT) &&
977 (ia->ia_sockmask.sin_addr.s_addr == INADDR_BROADCAST))
978 return (false);
979
980 return (true);
981 }
982
983 /*
984 * Calculate "prefix" route corresponding to @ia.
985 */
986 static void
ia_getrtprefix(const struct in_ifaddr * ia,struct in_addr * prefix,struct in_addr * mask)987 ia_getrtprefix(const struct in_ifaddr *ia, struct in_addr *prefix, struct in_addr *mask)
988 {
989
990 if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) {
991 /* Case 3: return host route for dstaddr */
992 *prefix = ia->ia_dstaddr.sin_addr;
993 mask->s_addr = INADDR_BROADCAST;
994 } else if (ia->ia_ifp->if_flags & IFF_LOOPBACK) {
995 /* Case 4: return host route for ifaddr */
996 *prefix = ia->ia_addr.sin_addr;
997 mask->s_addr = INADDR_BROADCAST;
998 } else {
999 /* Cases 1,2: return actual ia prefix */
1000 *prefix = ia->ia_addr.sin_addr;
1001 *mask = ia->ia_sockmask.sin_addr;
1002 prefix->s_addr &= mask->s_addr;
1003 }
1004 }
1005
1006 /*
1007 * Adds or delete interface "prefix" route corresponding to @ifa.
1008 * Returns 0 on success or errno.
1009 */
1010 static int
in_handle_ifaddr_route(int cmd,struct in_ifaddr * ia)1011 in_handle_ifaddr_route(int cmd, struct in_ifaddr *ia)
1012 {
1013 struct ifaddr *ifa = &ia->ia_ifa;
1014 struct in_addr daddr, maddr;
1015 struct sockaddr_in *pmask;
1016 struct epoch_tracker et;
1017 int error;
1018
1019 ia_getrtprefix(ia, &daddr, &maddr);
1020
1021 struct sockaddr_in mask = {
1022 .sin_family = AF_INET,
1023 .sin_len = sizeof(struct sockaddr_in),
1024 .sin_addr = maddr,
1025 };
1026
1027 pmask = (maddr.s_addr != INADDR_BROADCAST) ? &mask : NULL;
1028
1029 struct sockaddr_in dst = {
1030 .sin_family = AF_INET,
1031 .sin_len = sizeof(struct sockaddr_in),
1032 .sin_addr.s_addr = daddr.s_addr & maddr.s_addr,
1033 };
1034
1035 struct ifnet *ifp = ia->ia_ifp;
1036
1037 if ((maddr.s_addr == INADDR_BROADCAST) &&
1038 (!(ia->ia_ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)))) {
1039 /* Case 2: host route on broadcast interface */
1040 ifp = V_loif;
1041 }
1042
1043 uint32_t fibnum = ifa->ifa_ifp->if_fib;
1044 NET_EPOCH_ENTER(et);
1045 error = in_handle_prefix_route(fibnum, cmd, &dst, pmask, ifa, ifp);
1046 NET_EPOCH_EXIT(et);
1047
1048 return (error);
1049 }
1050
1051 /*
1052 * Check if we have a route for the given prefix already.
1053 */
1054 static bool
in_hasrtprefix(struct in_ifaddr * target)1055 in_hasrtprefix(struct in_ifaddr *target)
1056 {
1057 struct epoch_tracker et;
1058 struct in_ifaddr *ia;
1059 struct in_addr prefix, mask, p, m;
1060 bool result = false;
1061
1062 ia_getrtprefix(target, &prefix, &mask);
1063
1064 /* Look for an existing address with the same prefix, mask, and fib */
1065 NET_EPOCH_ENTER(et);
1066 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1067 ia_getrtprefix(ia, &p, &m);
1068
1069 if (prefix.s_addr != p.s_addr ||
1070 mask.s_addr != m.s_addr)
1071 continue;
1072
1073 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
1074 continue;
1075
1076 /*
1077 * If we got a matching prefix route inserted by other
1078 * interface address, we are done here.
1079 */
1080 if (ia->ia_flags & IFA_ROUTE) {
1081 result = true;
1082 break;
1083 }
1084 }
1085 NET_EPOCH_EXIT(et);
1086
1087 return (result);
1088 }
1089
1090 int
in_addprefix(struct in_ifaddr * target)1091 in_addprefix(struct in_ifaddr *target)
1092 {
1093 int error;
1094
1095 if (in_hasrtprefix(target)) {
1096 if (V_nosameprefix)
1097 return (EEXIST);
1098 else {
1099 rt_addrmsg(RTM_ADD, &target->ia_ifa,
1100 target->ia_ifp->if_fib);
1101 return (0);
1102 }
1103 }
1104
1105 /*
1106 * No-one seem to have this prefix route, so we try to insert it.
1107 */
1108 rt_addrmsg(RTM_ADD, &target->ia_ifa, target->ia_ifp->if_fib);
1109 error = in_handle_ifaddr_route(RTM_ADD, target);
1110 if (!error)
1111 target->ia_flags |= IFA_ROUTE;
1112 return (error);
1113 }
1114
1115 /*
1116 * Removes either all lle entries for given @ia, or lle
1117 * corresponding to @ia address.
1118 */
1119 static void
in_scrubprefixlle(struct in_ifaddr * ia,int all,u_int flags)1120 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
1121 {
1122 struct sockaddr_in addr, mask;
1123 struct sockaddr *saddr, *smask;
1124 struct ifnet *ifp;
1125
1126 saddr = (struct sockaddr *)&addr;
1127 bzero(&addr, sizeof(addr));
1128 addr.sin_len = sizeof(addr);
1129 addr.sin_family = AF_INET;
1130 smask = (struct sockaddr *)&mask;
1131 bzero(&mask, sizeof(mask));
1132 mask.sin_len = sizeof(mask);
1133 mask.sin_family = AF_INET;
1134 mask.sin_addr.s_addr = ia->ia_subnetmask;
1135 ifp = ia->ia_ifp;
1136
1137 if (all) {
1138 /*
1139 * Remove all L2 entries matching given prefix.
1140 * Convert address to host representation to avoid
1141 * doing this on every callback. ia_subnetmask is already
1142 * stored in host representation.
1143 */
1144 addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
1145 lltable_prefix_free(AF_INET, saddr, smask, flags);
1146 } else {
1147 /* Remove interface address only */
1148 addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
1149 lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
1150 }
1151 }
1152
1153 /*
1154 * If there is no other address in the system that can serve a route to the
1155 * same prefix, remove the route. Hand over the route to the new address
1156 * otherwise.
1157 */
1158 int
in_scrubprefix(struct in_ifaddr * target,u_int flags)1159 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1160 {
1161 struct epoch_tracker et;
1162 struct in_ifaddr *ia;
1163 struct in_addr prefix, mask, p, m;
1164 int error = 0;
1165
1166 /*
1167 * Remove the loopback route to the interface address.
1168 */
1169 if (ia_need_loopback_route(target) && (flags & LLE_STATIC)) {
1170 struct in_ifaddr *eia;
1171
1172 eia = in_localip_more(target);
1173
1174 if (eia != NULL) {
1175 error = ifa_switch_loopback_route((struct ifaddr *)eia,
1176 (struct sockaddr *)&target->ia_addr);
1177 ifa_free(&eia->ia_ifa);
1178 } else {
1179 error = ifa_del_loopback_route((struct ifaddr *)target,
1180 (struct sockaddr *)&target->ia_addr);
1181 }
1182 }
1183
1184 ia_getrtprefix(target, &prefix, &mask);
1185
1186 if ((target->ia_flags & IFA_ROUTE) == 0) {
1187 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1188
1189 /*
1190 * Removing address from !IFF_UP interface or
1191 * prefix which exists on other interface (along with route).
1192 * No entries should exist here except target addr.
1193 * Given that, delete this entry only.
1194 */
1195 in_scrubprefixlle(target, 0, flags);
1196 return (0);
1197 }
1198
1199 NET_EPOCH_ENTER(et);
1200 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1201 ia_getrtprefix(ia, &p, &m);
1202
1203 if (prefix.s_addr != p.s_addr ||
1204 mask.s_addr != m.s_addr)
1205 continue;
1206
1207 if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1208 continue;
1209
1210 /*
1211 * If we got a matching prefix address, move IFA_ROUTE and
1212 * the route itself to it. Make sure that routing daemons
1213 * get a heads-up.
1214 */
1215 if ((ia->ia_flags & IFA_ROUTE) == 0) {
1216 ifa_ref(&ia->ia_ifa);
1217 NET_EPOCH_EXIT(et);
1218 error = in_handle_ifaddr_route(RTM_DELETE, target);
1219 if (error == 0)
1220 target->ia_flags &= ~IFA_ROUTE;
1221 else
1222 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1223 error);
1224 /* Scrub all entries IFF interface is different */
1225 in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
1226 flags);
1227 error = in_handle_ifaddr_route(RTM_ADD, ia);
1228 if (error == 0)
1229 ia->ia_flags |= IFA_ROUTE;
1230 else
1231 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1232 error);
1233 ifa_free(&ia->ia_ifa);
1234 return (error);
1235 }
1236 }
1237 NET_EPOCH_EXIT(et);
1238
1239 /*
1240 * remove all L2 entries on the given prefix
1241 */
1242 in_scrubprefixlle(target, 1, flags);
1243
1244 /*
1245 * As no-one seem to have this prefix, we can remove the route.
1246 */
1247 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib);
1248 error = in_handle_ifaddr_route(RTM_DELETE, target);
1249 if (error == 0)
1250 target->ia_flags &= ~IFA_ROUTE;
1251 else
1252 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1253 return (error);
1254 }
1255
1256 void
in_ifscrub_all(void)1257 in_ifscrub_all(void)
1258 {
1259 struct ifnet *ifp;
1260 struct ifaddr *ifa, *nifa;
1261 struct ifreq ifr;
1262
1263 IFNET_RLOCK();
1264 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1265 /* Cannot lock here - lock recursion. */
1266 /* NET_EPOCH_ENTER(et); */
1267 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1268 if (ifa->ifa_addr->sa_family != AF_INET)
1269 continue;
1270
1271 /*
1272 * This is ugly but the only way for legacy IP to
1273 * cleanly remove addresses and everything attached.
1274 */
1275 bzero(&ifr, sizeof(ifr));
1276 ifr.ifr_addr = *ifa->ifa_addr;
1277 (void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
1278 ifp, NULL);
1279 }
1280 /* NET_EPOCH_EXIT(et); */
1281 in_purgemaddrs(ifp);
1282 igmp_domifdetach(ifp);
1283 }
1284 IFNET_RUNLOCK();
1285 }
1286
1287 bool
in_ifaddr_broadcast(struct in_addr in,struct in_ifaddr * ia)1288 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
1289 {
1290
1291 return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1292 /*
1293 * Optionally check for old-style (host 0) broadcast, but
1294 * taking into account that RFC 3021 obsoletes it.
1295 */
1296 __predict_false(V_broadcast_lowest &&
1297 ia->ia_subnetmask != IN_RFC3021_MASK &&
1298 ntohl(in.s_addr) == ia->ia_subnet)) &&
1299 /*
1300 * Check for an all one subnetmask. These
1301 * only exist when an interface gets a secondary
1302 * address.
1303 */
1304 ia->ia_subnetmask != (u_long)0xffffffff);
1305 }
1306
1307 /*
1308 * Return true if the address might be a local broadcast address.
1309 */
1310 bool
in_ifnet_broadcast(struct in_addr in,struct ifnet * ifp)1311 in_ifnet_broadcast(struct in_addr in, struct ifnet *ifp)
1312 {
1313 struct ifaddr *ifa;
1314
1315 NET_EPOCH_ASSERT();
1316
1317 if (in_broadcast(in))
1318 return (true);
1319 if ((ifp->if_flags & IFF_BROADCAST) == 0)
1320 return (false);
1321 /*
1322 * Look through the list of addresses for a match
1323 * with a broadcast address.
1324 */
1325 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1326 if (ifa->ifa_addr->sa_family == AF_INET &&
1327 in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa))
1328 return (true);
1329 return (false);
1330 }
1331
1332
1333 static struct lltable *in_lltattach(struct ifnet *);
1334 void
in_ifattach(void * arg __unused,struct ifnet * ifp)1335 in_ifattach(void *arg __unused, struct ifnet *ifp)
1336 {
1337 struct in_ifinfo *ii;
1338
1339 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1340 ii->ii_llt = in_lltattach(ifp);
1341 ii->ii_igmp = igmp_domifattach(ifp);
1342 ifp->if_inet = ii;
1343 }
1344 EVENTHANDLER_DEFINE(ifnet_arrival_event, in_ifattach, NULL,
1345 EVENTHANDLER_PRI_ANY);
1346
1347 /*
1348 * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1349 */
1350 static void
in_ifdetach(void * arg __unused,struct ifnet * ifp)1351 in_ifdetach(void *arg __unused, struct ifnet *ifp)
1352 {
1353 struct in_ifinfo *ii = ifp->if_inet;
1354
1355 #ifdef VIMAGE
1356 /*
1357 * On VNET shutdown abort here as the stack teardown will do all
1358 * the work top-down for us. XXXGL: this logic is copied from
1359 * if_detach() before dom_ifattach removal. Ideally we'd like to have
1360 * same logic for VNET shutdown and normal detach. This means that
1361 * interfaces should be detach before protocols destroyed during VNET
1362 * shutdown.
1363 */
1364 if (!VNET_IS_SHUTTING_DOWN(ifp->if_vnet))
1365 #endif
1366 {
1367 IN_MULTI_LOCK();
1368 in_pcbpurgeif0(&V_ripcbinfo, ifp);
1369 in_pcbpurgeif0(&V_udbinfo, ifp);
1370 in_pcbpurgeif0(&V_ulitecbinfo, ifp);
1371 in_purgemaddrs(ifp);
1372 IN_MULTI_UNLOCK();
1373
1374 /*
1375 * Make sure all multicast deletions invoking if_ioctl() are
1376 * completed before returning. Else we risk accessing a freed
1377 * ifnet structure pointer.
1378 */
1379 inm_release_wait(NULL);
1380 }
1381
1382 igmp_domifdetach(ifp);
1383 lltable_free(ii->ii_llt);
1384 free(ii, M_IFADDR);
1385 }
1386 EVENTHANDLER_DEFINE(ifnet_departure_event, in_ifdetach, NULL,
1387 EVENTHANDLER_PRI_ANY);
1388
1389 static void
in_ifnet_event(void * arg __unused,struct ifnet * ifp,int event)1390 in_ifnet_event(void *arg __unused, struct ifnet *ifp, int event)
1391 {
1392 struct epoch_tracker et;
1393 struct ifaddr *ifa;
1394 struct in_ifaddr *ia;
1395 int error;
1396
1397 NET_EPOCH_ENTER(et);
1398 switch (event) {
1399 case IFNET_EVENT_DOWN:
1400 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1401 if (ifa->ifa_addr->sa_family != AF_INET)
1402 continue;
1403 ia = (struct in_ifaddr *)ifa;
1404 if ((ia->ia_flags & IFA_ROUTE) == 0)
1405 continue;
1406 ifa_ref(ifa);
1407 /*
1408 * in_scrubprefix() kills the interface route.
1409 */
1410 in_scrubprefix(ia, 0);
1411 /*
1412 * in_ifadown gets rid of all the rest of the
1413 * routes. This is not quite the right thing
1414 * to do, but at least if we are running a
1415 * routing process they will come back.
1416 */
1417 in_ifadown(ifa, 0);
1418 ifa_free(ifa);
1419 }
1420 break;
1421
1422 case IFNET_EVENT_UP:
1423 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1424 if (ifa->ifa_addr->sa_family != AF_INET)
1425 continue;
1426 ia = (struct in_ifaddr *)ifa;
1427 if (ia->ia_flags & IFA_ROUTE)
1428 continue;
1429 ifa_ref(ifa);
1430 error = ifa_del_loopback_route(ifa, ifa->ifa_addr);
1431 rt_addrmsg(RTM_ADD, ifa, ifa->ifa_ifp->if_fib);
1432 error = in_handle_ifaddr_route(RTM_ADD, ia);
1433 if (error == 0)
1434 ia->ia_flags |= IFA_ROUTE;
1435 error = ifa_add_loopback_route(ifa, ifa->ifa_addr);
1436 ifa_free(ifa);
1437 }
1438 break;
1439 }
1440 NET_EPOCH_EXIT(et);
1441 }
1442 EVENTHANDLER_DEFINE(ifnet_event, in_ifnet_event, NULL, EVENTHANDLER_PRI_ANY);
1443
1444 /*
1445 * Delete all IPv4 multicast address records, and associated link-layer
1446 * multicast address records, associated with ifp.
1447 * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1448 * XXX This should not race with ifma_protospec being set during
1449 * a new allocation, if it does, we have bigger problems.
1450 */
1451 static void
in_purgemaddrs(struct ifnet * ifp)1452 in_purgemaddrs(struct ifnet *ifp)
1453 {
1454 struct epoch_tracker et;
1455 struct in_multi_head purgeinms;
1456 struct in_multi *inm;
1457 struct ifmultiaddr *ifma;
1458
1459 SLIST_INIT(&purgeinms);
1460 IN_MULTI_LIST_LOCK();
1461
1462 /*
1463 * Extract list of in_multi associated with the detaching ifp
1464 * which the PF_INET layer is about to release.
1465 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1466 * by code further down.
1467 */
1468 IF_ADDR_WLOCK(ifp);
1469 NET_EPOCH_ENTER(et);
1470 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1471 inm = inm_ifmultiaddr_get_inm(ifma);
1472 if (inm == NULL)
1473 continue;
1474 inm_rele_locked(&purgeinms, inm);
1475 }
1476 NET_EPOCH_EXIT(et);
1477 IF_ADDR_WUNLOCK(ifp);
1478
1479 inm_release_list_deferred(&purgeinms);
1480 igmp_ifdetach(ifp);
1481 IN_MULTI_LIST_UNLOCK();
1482 }
1483
1484 struct in_llentry {
1485 struct llentry base;
1486 };
1487
1488 #define IN_LLTBL_DEFAULT_HSIZE 32
1489 #define IN_LLTBL_HASH(k, h) \
1490 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1491
1492 /*
1493 * Do actual deallocation of @lle.
1494 */
1495 static void
in_lltable_destroy_lle_unlocked(epoch_context_t ctx)1496 in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1497 {
1498 struct llentry *lle;
1499
1500 lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1501 LLE_LOCK_DESTROY(lle);
1502 LLE_REQ_DESTROY(lle);
1503 free(lle, M_LLTABLE);
1504 }
1505
1506 /*
1507 * Called by LLE_FREE_LOCKED when number of references
1508 * drops to zero.
1509 */
1510 static void
in_lltable_destroy_lle(struct llentry * lle)1511 in_lltable_destroy_lle(struct llentry *lle)
1512 {
1513
1514 LLE_WUNLOCK(lle);
1515 NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1516 }
1517
1518 static struct llentry *
in_lltable_new(struct in_addr addr4,u_int flags)1519 in_lltable_new(struct in_addr addr4, u_int flags)
1520 {
1521 struct in_llentry *lle;
1522
1523 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1524 if (lle == NULL) /* NB: caller generates msg */
1525 return NULL;
1526
1527 /*
1528 * For IPv4 this will trigger "arpresolve" to generate
1529 * an ARP request.
1530 */
1531 lle->base.la_expire = time_uptime; /* mark expired */
1532 lle->base.r_l3addr.addr4 = addr4;
1533 lle->base.lle_refcnt = 1;
1534 lle->base.lle_free = in_lltable_destroy_lle;
1535 LLE_LOCK_INIT(&lle->base);
1536 LLE_REQ_INIT(&lle->base);
1537 callout_init(&lle->base.lle_timer, 1);
1538
1539 return (&lle->base);
1540 }
1541
1542 static int
in_lltable_match_prefix(const struct sockaddr * saddr,const struct sockaddr * smask,u_int flags,struct llentry * lle)1543 in_lltable_match_prefix(const struct sockaddr *saddr,
1544 const struct sockaddr *smask, u_int flags, struct llentry *lle)
1545 {
1546 struct in_addr addr, mask, lle_addr;
1547
1548 addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1549 mask = ((const struct sockaddr_in *)smask)->sin_addr;
1550 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1551
1552 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1553 return (0);
1554
1555 if (lle->la_flags & LLE_IFADDR) {
1556 /*
1557 * Delete LLE_IFADDR records IFF address & flag matches.
1558 * Note that addr is the interface address within prefix
1559 * being matched.
1560 * Note also we should handle 'ifdown' cases without removing
1561 * ifaddr macs.
1562 */
1563 if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1564 return (1);
1565 return (0);
1566 }
1567
1568 /* flags & LLE_STATIC means deleting both dynamic and static entries */
1569 if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1570 return (1);
1571
1572 return (0);
1573 }
1574
1575 static void
in_lltable_free_entry(struct lltable * llt,struct llentry * lle)1576 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1577 {
1578 size_t pkts_dropped;
1579
1580 LLE_WLOCK_ASSERT(lle);
1581 KASSERT(llt != NULL, ("lltable is NULL"));
1582
1583 /* Unlink entry from table if not already */
1584 if ((lle->la_flags & LLE_LINKED) != 0) {
1585 LLTABLE_LOCK_ASSERT(llt);
1586 lltable_unlink_entry(llt, lle);
1587 }
1588
1589 /* Drop hold queue */
1590 pkts_dropped = llentry_free(lle);
1591 ARPSTAT_ADD(dropped, pkts_dropped);
1592 }
1593
1594 static int
in_lltable_rtcheck(struct ifnet * ifp,u_int flags,const struct sockaddr * l3addr)1595 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1596 {
1597 struct nhop_object *nh;
1598 struct in_addr addr;
1599
1600 KASSERT(l3addr->sa_family == AF_INET,
1601 ("sin_family %d", l3addr->sa_family));
1602
1603 addr = ((const struct sockaddr_in *)l3addr)->sin_addr;
1604
1605 nh = fib4_lookup(ifp->if_fib, addr, 0, NHR_NONE, 0);
1606 if (nh == NULL)
1607 return (EINVAL);
1608
1609 /*
1610 * If the gateway for an existing host route matches the target L3
1611 * address, which is a special route inserted by some implementation
1612 * such as MANET, and the interface is of the correct type, then
1613 * allow for ARP to proceed.
1614 */
1615 if (nh->nh_flags & NHF_GATEWAY) {
1616 if (!(nh->nh_flags & NHF_HOST) || nh->nh_ifp->if_type != IFT_ETHER ||
1617 (nh->nh_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1618 memcmp(nh->gw_sa.sa_data, l3addr->sa_data,
1619 sizeof(in_addr_t)) != 0) {
1620 return (EINVAL);
1621 }
1622 }
1623
1624 /*
1625 * Make sure that at least the destination address is covered
1626 * by the route. This is for handling the case where 2 or more
1627 * interfaces have the same prefix. An incoming packet arrives
1628 * on one interface and the corresponding outgoing packet leaves
1629 * another interface.
1630 */
1631 if ((nh->nh_ifp != ifp) && (nh->nh_flags & NHF_HOST) == 0) {
1632 struct in_ifaddr *ia = (struct in_ifaddr *)ifaof_ifpforaddr(l3addr, ifp);
1633 struct in_addr dst_addr, mask_addr;
1634
1635 if (ia == NULL)
1636 return (EINVAL);
1637
1638 /*
1639 * ifaof_ifpforaddr() returns _best matching_ IFA.
1640 * It is possible that ifa prefix does not cover our address.
1641 * Explicitly verify and fail if that's the case.
1642 */
1643 dst_addr = IA_SIN(ia)->sin_addr;
1644 mask_addr.s_addr = htonl(ia->ia_subnetmask);
1645
1646 if (!IN_ARE_MASKED_ADDR_EQUAL(dst_addr, addr, mask_addr))
1647 return (EINVAL);
1648 }
1649
1650 return (0);
1651 }
1652
1653 static inline uint32_t
in_lltable_hash_dst(const struct in_addr dst,uint32_t hsize)1654 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1655 {
1656
1657 return (IN_LLTBL_HASH(dst.s_addr, hsize));
1658 }
1659
1660 static uint32_t
in_lltable_hash(const struct llentry * lle,uint32_t hsize)1661 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1662 {
1663
1664 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1665 }
1666
1667 static void
in_lltable_fill_sa_entry(const struct llentry * lle,struct sockaddr * sa)1668 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1669 {
1670 struct sockaddr_in *sin;
1671
1672 sin = (struct sockaddr_in *)sa;
1673 bzero(sin, sizeof(*sin));
1674 sin->sin_family = AF_INET;
1675 sin->sin_len = sizeof(*sin);
1676 sin->sin_addr = lle->r_l3addr.addr4;
1677 }
1678
1679 static inline struct llentry *
in_lltable_find_dst(struct lltable * llt,struct in_addr dst)1680 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1681 {
1682 struct llentry *lle;
1683 struct llentries *lleh;
1684 u_int hashidx;
1685
1686 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1687 lleh = &llt->lle_head[hashidx];
1688 CK_LIST_FOREACH(lle, lleh, lle_next) {
1689 if (lle->la_flags & LLE_DELETED)
1690 continue;
1691 if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1692 break;
1693 }
1694
1695 return (lle);
1696 }
1697
1698 static void
in_lltable_delete_entry(struct lltable * llt,struct llentry * lle)1699 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1700 {
1701
1702 lle->la_flags |= LLE_DELETED;
1703 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1704 #ifdef DIAGNOSTIC
1705 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1706 #endif
1707 llentry_free(lle);
1708 }
1709
1710 static struct llentry *
in_lltable_alloc(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)1711 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1712 {
1713 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1714 struct ifnet *ifp = llt->llt_ifp;
1715 struct llentry *lle;
1716 char linkhdr[LLE_MAX_LINKHDR];
1717 size_t linkhdrsize;
1718 int lladdr_off;
1719
1720 KASSERT(l3addr->sa_family == AF_INET,
1721 ("sin_family %d", l3addr->sa_family));
1722
1723 /*
1724 * A route that covers the given address must have
1725 * been installed 1st because we are doing a resolution,
1726 * verify this.
1727 */
1728 if (!(flags & LLE_IFADDR) &&
1729 in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1730 return (NULL);
1731
1732 lle = in_lltable_new(sin->sin_addr, flags);
1733 if (lle == NULL) {
1734 log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1735 return (NULL);
1736 }
1737 lle->la_flags = flags;
1738 if (flags & LLE_STATIC)
1739 lle->r_flags |= RLLE_VALID;
1740 if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1741 linkhdrsize = LLE_MAX_LINKHDR;
1742 if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1743 linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1744 in_lltable_free_entry(llt, lle);
1745 return (NULL);
1746 }
1747 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1748 lladdr_off);
1749 lle->la_flags |= LLE_STATIC;
1750 lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1751 lle->la_expire = 0;
1752 }
1753
1754 return (lle);
1755 }
1756
1757 /*
1758 * Return NULL if not found or marked for deletion.
1759 * If found return lle read locked.
1760 */
1761 static struct llentry *
in_lltable_lookup(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)1762 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1763 {
1764 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1765 struct llentry *lle;
1766
1767 LLTABLE_RLOCK_ASSERT(llt);
1768 KASSERT(l3addr->sa_family == AF_INET,
1769 ("sin_family %d", l3addr->sa_family));
1770 KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1771 (LLE_UNLOCKED | LLE_EXCLUSIVE),
1772 ("wrong lle request flags: %#x", flags));
1773
1774 lle = in_lltable_find_dst(llt, sin->sin_addr);
1775 if (lle == NULL)
1776 return (NULL);
1777 if (flags & LLE_UNLOCKED)
1778 return (lle);
1779
1780 if (flags & LLE_EXCLUSIVE)
1781 LLE_WLOCK(lle);
1782 else
1783 LLE_RLOCK(lle);
1784
1785 /*
1786 * If the afdata lock is not held, the LLE may have been unlinked while
1787 * we were blocked on the LLE lock. Check for this case.
1788 */
1789 if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1790 if (flags & LLE_EXCLUSIVE)
1791 LLE_WUNLOCK(lle);
1792 else
1793 LLE_RUNLOCK(lle);
1794 return (NULL);
1795 }
1796 return (lle);
1797 }
1798
1799 static int
in_lltable_dump_entry(struct lltable * llt,struct llentry * lle,struct sysctl_req * wr)1800 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1801 struct sysctl_req *wr)
1802 {
1803 struct ifnet *ifp = llt->llt_ifp;
1804 /* XXX stack use */
1805 struct {
1806 struct rt_msghdr rtm;
1807 struct sockaddr_in sin;
1808 struct sockaddr_dl sdl;
1809 } arpc;
1810 struct sockaddr_dl *sdl;
1811 int error;
1812
1813 bzero(&arpc, sizeof(arpc));
1814 /* skip deleted entries */
1815 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1816 return (0);
1817 /* Skip if jailed and not a valid IP of the prison. */
1818 lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1819 if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1820 return (0);
1821 /*
1822 * produce a msg made of:
1823 * struct rt_msghdr;
1824 * struct sockaddr_in; (IPv4)
1825 * struct sockaddr_dl;
1826 */
1827 arpc.rtm.rtm_msglen = sizeof(arpc);
1828 arpc.rtm.rtm_version = RTM_VERSION;
1829 arpc.rtm.rtm_type = RTM_GET;
1830 arpc.rtm.rtm_flags = RTF_UP;
1831 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1832
1833 /* publish */
1834 if (lle->la_flags & LLE_PUB)
1835 arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1836
1837 sdl = &arpc.sdl;
1838 sdl->sdl_family = AF_LINK;
1839 sdl->sdl_len = sizeof(*sdl);
1840 sdl->sdl_index = ifp->if_index;
1841 sdl->sdl_type = ifp->if_type;
1842 if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1843 sdl->sdl_alen = ifp->if_addrlen;
1844 bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1845 } else {
1846 sdl->sdl_alen = 0;
1847 bzero(LLADDR(sdl), ifp->if_addrlen);
1848 }
1849
1850 arpc.rtm.rtm_rmx.rmx_expire =
1851 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1852 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1853 if (lle->la_flags & LLE_STATIC)
1854 arpc.rtm.rtm_flags |= RTF_STATIC;
1855 if (lle->la_flags & LLE_IFADDR)
1856 arpc.rtm.rtm_flags |= RTF_PINNED;
1857 arpc.rtm.rtm_index = ifp->if_index;
1858 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1859
1860 return (error);
1861 }
1862
1863 static void
in_lltable_post_resolved(struct lltable * llt,struct llentry * lle)1864 in_lltable_post_resolved(struct lltable *llt, struct llentry *lle)
1865 {
1866 struct ifnet *ifp = llt->llt_ifp;
1867
1868 /* gratuitous ARP */
1869 if ((lle->la_flags & LLE_PUB) != 0)
1870 arprequest(ifp, &lle->r_l3addr.addr4, &lle->r_l3addr.addr4,
1871 lle->ll_addr);
1872 }
1873
1874 static struct lltable *
in_lltattach(struct ifnet * ifp)1875 in_lltattach(struct ifnet *ifp)
1876 {
1877 struct lltable *llt;
1878
1879 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1880 llt->llt_af = AF_INET;
1881 llt->llt_ifp = ifp;
1882
1883 llt->llt_lookup = in_lltable_lookup;
1884 llt->llt_alloc_entry = in_lltable_alloc;
1885 llt->llt_delete_entry = in_lltable_delete_entry;
1886 llt->llt_dump_entry = in_lltable_dump_entry;
1887 llt->llt_hash = in_lltable_hash;
1888 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1889 llt->llt_free_entry = in_lltable_free_entry;
1890 llt->llt_match_prefix = in_lltable_match_prefix;
1891 llt->llt_mark_used = llentry_mark_used;
1892 llt->llt_post_resolved = in_lltable_post_resolved;
1893 lltable_link(llt);
1894
1895 return (llt);
1896 }
1897
1898 struct lltable *
in_lltable_get(struct ifnet * ifp)1899 in_lltable_get(struct ifnet *ifp)
1900 {
1901 /* XXXGL: ??? */
1902 if (ifp->if_inet == NULL)
1903 return (NULL);
1904
1905 return (((struct in_ifinfo *)ifp->if_inet)->ii_llt);
1906 }
1907