FreeBSD/Linux Kernel Cross Reference
sys/netinet/in.c
1 /*-
2 * Copyright (c) 1982, 1986, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
4 * Copyright (C) 2001 WIDE Project. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 4. Neither the name of the University nor the names of its contributors
15 * may be used to endorse or promote products derived from this software
16 * without specific prior written permission.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 * SUCH DAMAGE.
29 *
30 * @(#)in.c 8.4 (Berkeley) 1/9/95
31 */
32
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
35
36 #include "opt_mpath.h"
37
38 #include <sys/param.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/proc.h>
47 #include <sys/sysctl.h>
48 #include <sys/syslog.h>
49
50 #include <net/if.h>
51 #include <net/if_var.h>
52 #include <net/if_arp.h>
53 #include <net/if_dl.h>
54 #include <net/if_llatbl.h>
55 #include <net/if_types.h>
56 #include <net/route.h>
57 #include <net/vnet.h>
58
59 #include <netinet/if_ether.h>
60 #include <netinet/in.h>
61 #include <netinet/in_var.h>
62 #include <netinet/in_pcb.h>
63 #include <netinet/ip_var.h>
64 #include <netinet/ip_carp.h>
65 #include <netinet/igmp_var.h>
66 #include <netinet/udp.h>
67 #include <netinet/udp_var.h>
68
69 static int in_mask2len(struct in_addr *);
70 static void in_len2mask(struct in_addr *, int);
71 static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t,
72 struct ifnet *, struct thread *);
73
74 static void in_socktrim(struct sockaddr_in *);
75 static int in_ifinit(struct ifnet *, struct in_ifaddr *,
76 struct sockaddr_in *, int, int);
77 static void in_purgemaddrs(struct ifnet *);
78
79 static VNET_DEFINE(int, nosameprefix);
80 #define V_nosameprefix VNET(nosameprefix)
81 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_RW,
82 &VNET_NAME(nosameprefix), 0,
83 "Refuse to create same prefixes on different interfaces");
84
85 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
86 #define V_ripcbinfo VNET(ripcbinfo)
87
88 /*
89 * Return 1 if an internet address is for a ``local'' host
90 * (one to which we have a connection).
91 */
92 int
93 in_localaddr(struct in_addr in)
94 {
95 register u_long i = ntohl(in.s_addr);
96 register struct in_ifaddr *ia;
97
98 IN_IFADDR_RLOCK();
99 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
100 if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
101 IN_IFADDR_RUNLOCK();
102 return (1);
103 }
104 }
105 IN_IFADDR_RUNLOCK();
106 return (0);
107 }
108
109 /*
110 * Return 1 if an internet address is for the local host and configured
111 * on one of its interfaces.
112 */
113 int
114 in_localip(struct in_addr in)
115 {
116 struct in_ifaddr *ia;
117
118 IN_IFADDR_RLOCK();
119 LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
120 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
121 IN_IFADDR_RUNLOCK();
122 return (1);
123 }
124 }
125 IN_IFADDR_RUNLOCK();
126 return (0);
127 }
128
129 /*
130 * Determine whether an IP address is in a reserved set of addresses
131 * that may not be forwarded, or whether datagrams to that destination
132 * may be forwarded.
133 */
134 int
135 in_canforward(struct in_addr in)
136 {
137 register u_long i = ntohl(in.s_addr);
138 register u_long net;
139
140 if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
141 return (0);
142 if (IN_CLASSA(i)) {
143 net = i & IN_CLASSA_NET;
144 if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
145 return (0);
146 }
147 return (1);
148 }
149
150 /*
151 * Trim a mask in a sockaddr
152 */
153 static void
154 in_socktrim(struct sockaddr_in *ap)
155 {
156 register char *cplim = (char *) &ap->sin_addr;
157 register char *cp = (char *) (&ap->sin_addr + 1);
158
159 ap->sin_len = 0;
160 while (--cp >= cplim)
161 if (*cp) {
162 (ap)->sin_len = cp - (char *) (ap) + 1;
163 break;
164 }
165 }
166
167 static int
168 in_mask2len(mask)
169 struct in_addr *mask;
170 {
171 int x, y;
172 u_char *p;
173
174 p = (u_char *)mask;
175 for (x = 0; x < sizeof(*mask); x++) {
176 if (p[x] != 0xff)
177 break;
178 }
179 y = 0;
180 if (x < sizeof(*mask)) {
181 for (y = 0; y < 8; y++) {
182 if ((p[x] & (0x80 >> y)) == 0)
183 break;
184 }
185 }
186 return (x * 8 + y);
187 }
188
189 static void
190 in_len2mask(struct in_addr *mask, int len)
191 {
192 int i;
193 u_char *p;
194
195 p = (u_char *)mask;
196 bzero(mask, sizeof(*mask));
197 for (i = 0; i < len / 8; i++)
198 p[i] = 0xff;
199 if (len % 8)
200 p[i] = (0xff00 >> (len % 8)) & 0xff;
201 }
202
203 /*
204 * Generic internet control operations (ioctl's).
205 *
206 * ifp is NULL if not an interface-specific ioctl.
207 */
208 /* ARGSUSED */
209 int
210 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
211 struct thread *td)
212 {
213 register struct ifreq *ifr = (struct ifreq *)data;
214 register struct in_ifaddr *ia, *iap;
215 register struct ifaddr *ifa;
216 struct in_addr allhosts_addr;
217 struct in_addr dst;
218 struct in_ifinfo *ii;
219 struct in_aliasreq *ifra = (struct in_aliasreq *)data;
220 int error, hostIsNew, iaIsNew, maskIsNew;
221 int iaIsFirst;
222 u_long ocmd = cmd;
223
224 /*
225 * Pre-10.x compat: OSIOCAIFADDR passes a shorter
226 * struct in_aliasreq, without ifra_vhid.
227 */
228 if (cmd == OSIOCAIFADDR)
229 cmd = SIOCAIFADDR;
230
231 ia = NULL;
232 iaIsFirst = 0;
233 iaIsNew = 0;
234 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
235
236 /*
237 * Filter out ioctls we implement directly; forward the rest on to
238 * in_lifaddr_ioctl() and ifp->if_ioctl().
239 */
240 switch (cmd) {
241 case SIOCGIFADDR:
242 case SIOCGIFBRDADDR:
243 case SIOCGIFDSTADDR:
244 case SIOCGIFNETMASK:
245 case SIOCDIFADDR:
246 break;
247 case SIOCAIFADDR:
248 /*
249 * ifra_addr must be present and be of INET family.
250 * ifra_broadaddr and ifra_mask are optional.
251 */
252 if (ifra->ifra_addr.sin_len != sizeof(struct sockaddr_in) ||
253 ifra->ifra_addr.sin_family != AF_INET)
254 return (EINVAL);
255 if (ifra->ifra_broadaddr.sin_len != 0 &&
256 (ifra->ifra_broadaddr.sin_len !=
257 sizeof(struct sockaddr_in) ||
258 ifra->ifra_broadaddr.sin_family != AF_INET))
259 return (EINVAL);
260 #if 0
261 /*
262 * ifconfig(8) in pre-10.x doesn't set sin_family for the
263 * mask. The code is disabled for the 10.x timeline, to
264 * make SIOCAIFADDR compatible with 9.x ifconfig(8).
265 * The code should be enabled in 11.x
266 */
267 if (ifra->ifra_mask.sin_len != 0 &&
268 (ifra->ifra_mask.sin_len != sizeof(struct sockaddr_in) ||
269 ifra->ifra_mask.sin_family != AF_INET))
270 return (EINVAL);
271 #endif
272 break;
273 case SIOCSIFADDR:
274 case SIOCSIFBRDADDR:
275 case SIOCSIFDSTADDR:
276 case SIOCSIFNETMASK:
277 /* We no longer support that old commands. */
278 return (EINVAL);
279
280 case SIOCALIFADDR:
281 if (td != NULL) {
282 error = priv_check(td, PRIV_NET_ADDIFADDR);
283 if (error)
284 return (error);
285 }
286 if (ifp == NULL)
287 return (EINVAL);
288 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
289
290 case SIOCDLIFADDR:
291 if (td != NULL) {
292 error = priv_check(td, PRIV_NET_DELIFADDR);
293 if (error)
294 return (error);
295 }
296 if (ifp == NULL)
297 return (EINVAL);
298 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
299
300 case SIOCGLIFADDR:
301 if (ifp == NULL)
302 return (EINVAL);
303 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
304
305 default:
306 if (ifp == NULL || ifp->if_ioctl == NULL)
307 return (EOPNOTSUPP);
308 return ((*ifp->if_ioctl)(ifp, cmd, data));
309 }
310
311 if (ifp == NULL)
312 return (EADDRNOTAVAIL);
313
314 /*
315 * Security checks before we get involved in any work.
316 */
317 switch (cmd) {
318 case SIOCAIFADDR:
319 if (td != NULL) {
320 error = priv_check(td, PRIV_NET_ADDIFADDR);
321 if (error)
322 return (error);
323 }
324 break;
325
326 case SIOCDIFADDR:
327 if (td != NULL) {
328 error = priv_check(td, PRIV_NET_DELIFADDR);
329 if (error)
330 return (error);
331 }
332 break;
333 }
334
335 /*
336 * Find address for this interface, if it exists.
337 *
338 * If an alias address was specified, find that one instead of the
339 * first one on the interface, if possible.
340 */
341 dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
342 IN_IFADDR_RLOCK();
343 LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) {
344 if (iap->ia_ifp == ifp &&
345 iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
346 if (td == NULL || prison_check_ip4(td->td_ucred,
347 &dst) == 0)
348 ia = iap;
349 break;
350 }
351 }
352 if (ia != NULL)
353 ifa_ref(&ia->ia_ifa);
354 IN_IFADDR_RUNLOCK();
355 if (ia == NULL) {
356 IF_ADDR_RLOCK(ifp);
357 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
358 iap = ifatoia(ifa);
359 if (iap->ia_addr.sin_family == AF_INET) {
360 if (td != NULL &&
361 prison_check_ip4(td->td_ucred,
362 &iap->ia_addr.sin_addr) != 0)
363 continue;
364 ia = iap;
365 break;
366 }
367 }
368 if (ia != NULL)
369 ifa_ref(&ia->ia_ifa);
370 IF_ADDR_RUNLOCK(ifp);
371 }
372 if (ia == NULL)
373 iaIsFirst = 1;
374
375 error = 0;
376 switch (cmd) {
377 case SIOCAIFADDR:
378 case SIOCDIFADDR:
379 if (ifra->ifra_addr.sin_family == AF_INET) {
380 struct in_ifaddr *oia;
381
382 IN_IFADDR_RLOCK();
383 for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
384 if (ia->ia_ifp == ifp &&
385 ia->ia_addr.sin_addr.s_addr ==
386 ifra->ifra_addr.sin_addr.s_addr)
387 break;
388 }
389 if (ia != NULL && ia != oia)
390 ifa_ref(&ia->ia_ifa);
391 if (oia != NULL && ia != oia)
392 ifa_free(&oia->ia_ifa);
393 IN_IFADDR_RUNLOCK();
394 if ((ifp->if_flags & IFF_POINTOPOINT)
395 && (cmd == SIOCAIFADDR)
396 && (ifra->ifra_dstaddr.sin_addr.s_addr
397 == INADDR_ANY)) {
398 error = EDESTADDRREQ;
399 goto out;
400 }
401 }
402 if (cmd == SIOCDIFADDR && ia == NULL) {
403 error = EADDRNOTAVAIL;
404 goto out;
405 }
406 if (ia == NULL) {
407 ia = (struct in_ifaddr *)
408 malloc(sizeof *ia, M_IFADDR, M_NOWAIT |
409 M_ZERO);
410 if (ia == NULL) {
411 error = ENOBUFS;
412 goto out;
413 }
414
415 ifa = &ia->ia_ifa;
416 ifa_init(ifa);
417 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
418 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
419 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
420 callout_init_mtx(&ia->ia_garp_timer, &ifa->ifa_mtx,
421 CALLOUT_RETURNUNLOCKED);
422
423 ia->ia_sockmask.sin_len = 8;
424 ia->ia_sockmask.sin_family = AF_INET;
425 if (ifp->if_flags & IFF_BROADCAST) {
426 ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
427 ia->ia_broadaddr.sin_family = AF_INET;
428 }
429 ia->ia_ifp = ifp;
430
431 ifa_ref(ifa); /* if_addrhead */
432 IF_ADDR_WLOCK(ifp);
433 TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
434 IF_ADDR_WUNLOCK(ifp);
435 ifa_ref(ifa); /* in_ifaddrhead */
436 IN_IFADDR_WLOCK();
437 TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
438 IN_IFADDR_WUNLOCK();
439 iaIsNew = 1;
440 }
441 break;
442
443 case SIOCGIFADDR:
444 case SIOCGIFNETMASK:
445 case SIOCGIFDSTADDR:
446 case SIOCGIFBRDADDR:
447 if (ia == NULL) {
448 error = EADDRNOTAVAIL;
449 goto out;
450 }
451 break;
452 }
453
454 /*
455 * Most paths in this switch return directly or via out. Only paths
456 * that remove the address break in order to hit common removal code.
457 */
458 switch (cmd) {
459 case SIOCGIFADDR:
460 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
461 goto out;
462
463 case SIOCGIFBRDADDR:
464 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
465 error = EINVAL;
466 goto out;
467 }
468 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
469 goto out;
470
471 case SIOCGIFDSTADDR:
472 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
473 error = EINVAL;
474 goto out;
475 }
476 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
477 goto out;
478
479 case SIOCGIFNETMASK:
480 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
481 goto out;
482
483 case SIOCAIFADDR:
484 maskIsNew = 0;
485 hostIsNew = 1;
486 error = 0;
487 if (ifra->ifra_addr.sin_addr.s_addr ==
488 ia->ia_addr.sin_addr.s_addr)
489 hostIsNew = 0;
490 if (ifra->ifra_mask.sin_len) {
491 /*
492 * QL: XXX
493 * Need to scrub the prefix here in case
494 * the issued command is SIOCAIFADDR with
495 * the same address, but with a different
496 * prefix length. And if the prefix length
497 * is the same as before, then the call is
498 * un-necessarily executed here.
499 */
500 in_ifscrub(ifp, ia, LLE_STATIC);
501 ia->ia_sockmask = ifra->ifra_mask;
502 ia->ia_sockmask.sin_family = AF_INET;
503 ia->ia_subnetmask =
504 ntohl(ia->ia_sockmask.sin_addr.s_addr);
505 maskIsNew = 1;
506 }
507 if ((ifp->if_flags & IFF_POINTOPOINT) &&
508 (ifra->ifra_dstaddr.sin_family == AF_INET)) {
509 in_ifscrub(ifp, ia, LLE_STATIC);
510 ia->ia_dstaddr = ifra->ifra_dstaddr;
511 maskIsNew = 1; /* We lie; but the effect's the same */
512 }
513 if (hostIsNew || maskIsNew)
514 error = in_ifinit(ifp, ia, &ifra->ifra_addr, maskIsNew,
515 (ocmd == cmd ? ifra->ifra_vhid : 0));
516 if (error != 0 && iaIsNew)
517 break;
518
519 if ((ifp->if_flags & IFF_BROADCAST) &&
520 ifra->ifra_broadaddr.sin_len)
521 ia->ia_broadaddr = ifra->ifra_broadaddr;
522 if (error == 0) {
523 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
524 if (iaIsFirst &&
525 (ifp->if_flags & IFF_MULTICAST) != 0) {
526 error = in_joingroup(ifp, &allhosts_addr,
527 NULL, &ii->ii_allhosts);
528 }
529 EVENTHANDLER_INVOKE(ifaddr_event, ifp);
530 }
531 goto out;
532
533 case SIOCDIFADDR:
534 /*
535 * in_ifscrub kills the interface route.
536 */
537 in_ifscrub(ifp, ia, LLE_STATIC);
538
539 /*
540 * in_ifadown gets rid of all the rest of
541 * the routes. This is not quite the right
542 * thing to do, but at least if we are running
543 * a routing process they will come back.
544 */
545 in_ifadown(&ia->ia_ifa, 1);
546 EVENTHANDLER_INVOKE(ifaddr_event, ifp);
547 error = 0;
548 break;
549
550 default:
551 panic("in_control: unsupported ioctl");
552 }
553
554 if (ia->ia_ifa.ifa_carp)
555 (*carp_detach_p)(&ia->ia_ifa);
556
557 IF_ADDR_WLOCK(ifp);
558 /* Re-check that ia is still part of the list. */
559 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
560 if (ifa == &ia->ia_ifa)
561 break;
562 }
563 if (ifa == NULL) {
564 /*
565 * If we lost the race with another thread, there is no need to
566 * try it again for the next loop as there is no other exit
567 * path between here and out.
568 */
569 IF_ADDR_WUNLOCK(ifp);
570 error = EADDRNOTAVAIL;
571 goto out;
572 }
573 TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
574 IF_ADDR_WUNLOCK(ifp);
575 ifa_free(&ia->ia_ifa); /* if_addrhead */
576
577 IN_IFADDR_WLOCK();
578 TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
579
580 LIST_REMOVE(ia, ia_hash);
581 IN_IFADDR_WUNLOCK();
582 /*
583 * If this is the last IPv4 address configured on this
584 * interface, leave the all-hosts group.
585 * No state-change report need be transmitted.
586 */
587 IFP_TO_IA(ifp, iap);
588 if (iap == NULL) {
589 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
590 IN_MULTI_LOCK();
591 if (ii->ii_allhosts) {
592 (void)in_leavegroup_locked(ii->ii_allhosts, NULL);
593 ii->ii_allhosts = NULL;
594 }
595 IN_MULTI_UNLOCK();
596 } else
597 ifa_free(&iap->ia_ifa);
598
599 IFA_LOCK(&ia->ia_ifa);
600 if (callout_stop(&ia->ia_garp_timer))
601 ifa_free(&ia->ia_ifa);
602 IFA_UNLOCK(&ia->ia_ifa);
603 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */
604 out:
605 if (ia != NULL)
606 ifa_free(&ia->ia_ifa);
607 return (error);
608 }
609
610 /*
611 * SIOC[GAD]LIFADDR.
612 * SIOCGLIFADDR: get first address. (?!?)
613 * SIOCGLIFADDR with IFLR_PREFIX:
614 * get first address that matches the specified prefix.
615 * SIOCALIFADDR: add the specified address.
616 * SIOCALIFADDR with IFLR_PREFIX:
617 * EINVAL since we can't deduce hostid part of the address.
618 * SIOCDLIFADDR: delete the specified address.
619 * SIOCDLIFADDR with IFLR_PREFIX:
620 * delete the first address that matches the specified prefix.
621 * return values:
622 * EINVAL on invalid parameters
623 * EADDRNOTAVAIL on prefix match failed/specified address not found
624 * other values may be returned from in_ioctl()
625 */
626 static int
627 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
628 struct ifnet *ifp, struct thread *td)
629 {
630 struct if_laddrreq *iflr = (struct if_laddrreq *)data;
631 struct ifaddr *ifa;
632
633 /* sanity checks */
634 if (data == NULL || ifp == NULL) {
635 panic("invalid argument to in_lifaddr_ioctl");
636 /*NOTRECHED*/
637 }
638
639 switch (cmd) {
640 case SIOCGLIFADDR:
641 /* address must be specified on GET with IFLR_PREFIX */
642 if ((iflr->flags & IFLR_PREFIX) == 0)
643 break;
644 /*FALLTHROUGH*/
645 case SIOCALIFADDR:
646 case SIOCDLIFADDR:
647 /* address must be specified on ADD and DELETE */
648 if (iflr->addr.ss_family != AF_INET)
649 return (EINVAL);
650 if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
651 return (EINVAL);
652 /* XXX need improvement */
653 if (iflr->dstaddr.ss_family
654 && iflr->dstaddr.ss_family != AF_INET)
655 return (EINVAL);
656 if (iflr->dstaddr.ss_family
657 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
658 return (EINVAL);
659 break;
660 default: /*shouldn't happen*/
661 return (EOPNOTSUPP);
662 }
663 if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
664 return (EINVAL);
665
666 switch (cmd) {
667 case SIOCALIFADDR:
668 {
669 struct in_aliasreq ifra;
670
671 if (iflr->flags & IFLR_PREFIX)
672 return (EINVAL);
673
674 /* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
675 bzero(&ifra, sizeof(ifra));
676 bcopy(iflr->iflr_name, ifra.ifra_name,
677 sizeof(ifra.ifra_name));
678
679 bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
680
681 if (iflr->dstaddr.ss_family) { /*XXX*/
682 bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
683 iflr->dstaddr.ss_len);
684 }
685
686 ifra.ifra_mask.sin_family = AF_INET;
687 ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
688 in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
689
690 return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
691 }
692 case SIOCGLIFADDR:
693 case SIOCDLIFADDR:
694 {
695 struct in_ifaddr *ia;
696 struct in_addr mask, candidate, match;
697 struct sockaddr_in *sin;
698
699 bzero(&mask, sizeof(mask));
700 bzero(&match, sizeof(match));
701 if (iflr->flags & IFLR_PREFIX) {
702 /* lookup a prefix rather than address. */
703 in_len2mask(&mask, iflr->prefixlen);
704
705 sin = (struct sockaddr_in *)&iflr->addr;
706 match.s_addr = sin->sin_addr.s_addr;
707 match.s_addr &= mask.s_addr;
708
709 /* if you set extra bits, that's wrong */
710 if (match.s_addr != sin->sin_addr.s_addr)
711 return (EINVAL);
712
713 } else {
714 /* on getting an address, take the 1st match */
715 /* on deleting an address, do exact match */
716 if (cmd != SIOCGLIFADDR) {
717 in_len2mask(&mask, 32);
718 sin = (struct sockaddr_in *)&iflr->addr;
719 match.s_addr = sin->sin_addr.s_addr;
720 }
721 }
722
723 IF_ADDR_RLOCK(ifp);
724 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
725 if (ifa->ifa_addr->sa_family != AF_INET)
726 continue;
727 if (match.s_addr == 0)
728 break;
729 sin = (struct sockaddr_in *)&ifa->ifa_addr;
730 candidate.s_addr = sin->sin_addr.s_addr;
731 candidate.s_addr &= mask.s_addr;
732 if (candidate.s_addr == match.s_addr)
733 break;
734 }
735 if (ifa != NULL)
736 ifa_ref(ifa);
737 IF_ADDR_RUNLOCK(ifp);
738 if (ifa == NULL)
739 return (EADDRNOTAVAIL);
740 ia = (struct in_ifaddr *)ifa;
741
742 if (cmd == SIOCGLIFADDR) {
743 /* fill in the if_laddrreq structure */
744 bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
745
746 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
747 bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
748 ia->ia_dstaddr.sin_len);
749 } else
750 bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
751
752 iflr->prefixlen =
753 in_mask2len(&ia->ia_sockmask.sin_addr);
754
755 iflr->flags = 0; /*XXX*/
756 ifa_free(ifa);
757
758 return (0);
759 } else {
760 struct in_aliasreq ifra;
761
762 /* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
763 bzero(&ifra, sizeof(ifra));
764 bcopy(iflr->iflr_name, ifra.ifra_name,
765 sizeof(ifra.ifra_name));
766
767 bcopy(&ia->ia_addr, &ifra.ifra_addr,
768 ia->ia_addr.sin_len);
769 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
770 bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
771 ia->ia_dstaddr.sin_len);
772 }
773 bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
774 ia->ia_sockmask.sin_len);
775 ifa_free(ifa);
776
777 return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
778 ifp, td));
779 }
780 }
781 }
782
783 return (EOPNOTSUPP); /*just for safety*/
784 }
785
786 /*
787 * Delete any existing route for an interface.
788 */
789 void
790 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia, u_int flags)
791 {
792
793 in_scrubprefix(ia, flags);
794 }
795
796 /*
797 * Initialize an interface's internet address
798 * and routing table entry.
799 */
800 static int
801 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
802 int masksupplied, int vhid)
803 {
804 register u_long i = ntohl(sin->sin_addr.s_addr);
805 int flags, error = 0;
806
807 IN_IFADDR_WLOCK();
808 if (ia->ia_addr.sin_family == AF_INET)
809 LIST_REMOVE(ia, ia_hash);
810 ia->ia_addr = *sin;
811 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
812 ia, ia_hash);
813 IN_IFADDR_WUNLOCK();
814
815 if (vhid > 0) {
816 if (carp_attach_p != NULL)
817 error = (*carp_attach_p)(&ia->ia_ifa, vhid);
818 else
819 error = EPROTONOSUPPORT;
820 }
821 if (error)
822 return (error);
823
824 /*
825 * Give the interface a chance to initialize
826 * if this is its first address,
827 * and to validate the address if necessary.
828 */
829 if (ifp->if_ioctl != NULL &&
830 (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia)) != 0)
831 /* LIST_REMOVE(ia, ia_hash) is done in in_control */
832 return (error);
833
834 /*
835 * Be compatible with network classes, if netmask isn't supplied,
836 * guess it based on classes.
837 */
838 if (!masksupplied) {
839 if (IN_CLASSA(i))
840 ia->ia_subnetmask = IN_CLASSA_NET;
841 else if (IN_CLASSB(i))
842 ia->ia_subnetmask = IN_CLASSB_NET;
843 else
844 ia->ia_subnetmask = IN_CLASSC_NET;
845 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
846 }
847 ia->ia_subnet = i & ia->ia_subnetmask;
848 in_socktrim(&ia->ia_sockmask);
849
850 /*
851 * Add route for the network.
852 */
853 flags = RTF_UP;
854 ia->ia_ifa.ifa_metric = ifp->if_metric;
855 if (ifp->if_flags & IFF_BROADCAST) {
856 if (ia->ia_subnetmask == IN_RFC3021_MASK)
857 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
858 else
859 ia->ia_broadaddr.sin_addr.s_addr =
860 htonl(ia->ia_subnet | ~ia->ia_subnetmask);
861 } else if (ifp->if_flags & IFF_LOOPBACK) {
862 ia->ia_dstaddr = ia->ia_addr;
863 flags |= RTF_HOST;
864 } else if (ifp->if_flags & IFF_POINTOPOINT) {
865 if (ia->ia_dstaddr.sin_family != AF_INET)
866 return (0);
867 flags |= RTF_HOST;
868 }
869 if (!vhid && (error = in_addprefix(ia, flags)) != 0)
870 return (error);
871
872 if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
873 return (0);
874
875 if (ifp->if_flags & IFF_POINTOPOINT &&
876 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
877 return (0);
878
879 /*
880 * add a loopback route to self
881 */
882 if (V_useloopback && !vhid && !(ifp->if_flags & IFF_LOOPBACK)) {
883 struct route ia_ro;
884
885 bzero(&ia_ro, sizeof(ia_ro));
886 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr;
887 rtalloc_ign_fib(&ia_ro, 0, RT_DEFAULT_FIB);
888 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
889 (ia_ro.ro_rt->rt_ifp == V_loif)) {
890 RT_LOCK(ia_ro.ro_rt);
891 RT_ADDREF(ia_ro.ro_rt);
892 RTFREE_LOCKED(ia_ro.ro_rt);
893 } else
894 error = ifa_add_loopback_route((struct ifaddr *)ia,
895 (struct sockaddr *)&ia->ia_addr);
896 if (error == 0)
897 ia->ia_flags |= IFA_RTSELF;
898 if (ia_ro.ro_rt != NULL)
899 RTFREE(ia_ro.ro_rt);
900 }
901
902 return (error);
903 }
904
905 #define rtinitflags(x) \
906 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
907 ? RTF_HOST : 0)
908
909 /*
910 * Check if we have a route for the given prefix already or add one accordingly.
911 */
912 int
913 in_addprefix(struct in_ifaddr *target, int flags)
914 {
915 struct in_ifaddr *ia;
916 struct in_addr prefix, mask, p, m;
917 int error;
918
919 if ((flags & RTF_HOST) != 0) {
920 prefix = target->ia_dstaddr.sin_addr;
921 mask.s_addr = 0;
922 } else {
923 prefix = target->ia_addr.sin_addr;
924 mask = target->ia_sockmask.sin_addr;
925 prefix.s_addr &= mask.s_addr;
926 }
927
928 IN_IFADDR_RLOCK();
929 /* Look for an existing address with the same prefix, mask, and fib */
930 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
931 if (rtinitflags(ia)) {
932 p = ia->ia_dstaddr.sin_addr;
933
934 if (prefix.s_addr != p.s_addr)
935 continue;
936 } else {
937 p = ia->ia_addr.sin_addr;
938 m = ia->ia_sockmask.sin_addr;
939 p.s_addr &= m.s_addr;
940
941 if (prefix.s_addr != p.s_addr ||
942 mask.s_addr != m.s_addr)
943 continue;
944 }
945 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
946 continue;
947
948 /*
949 * If we got a matching prefix route inserted by other
950 * interface address, we are done here.
951 */
952 if (ia->ia_flags & IFA_ROUTE) {
953 #ifdef RADIX_MPATH
954 if (ia->ia_addr.sin_addr.s_addr ==
955 target->ia_addr.sin_addr.s_addr) {
956 IN_IFADDR_RUNLOCK();
957 return (EEXIST);
958 } else
959 break;
960 #endif
961 if (V_nosameprefix) {
962 IN_IFADDR_RUNLOCK();
963 return (EEXIST);
964 } else {
965 int fibnum;
966
967 fibnum = rt_add_addr_allfibs ? RT_ALL_FIBS :
968 target->ia_ifp->if_fib;
969 rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum);
970 IN_IFADDR_RUNLOCK();
971 return (0);
972 }
973 }
974 }
975 IN_IFADDR_RUNLOCK();
976
977 /*
978 * No-one seem to have this prefix route, so we try to insert it.
979 */
980 error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
981 if (!error)
982 target->ia_flags |= IFA_ROUTE;
983 return (error);
984 }
985
986 /*
987 * If there is no other address in the system that can serve a route to the
988 * same prefix, remove the route. Hand over the route to the new address
989 * otherwise.
990 */
991 int
992 in_scrubprefix(struct in_ifaddr *target, u_int flags)
993 {
994 struct in_ifaddr *ia;
995 struct in_addr prefix, mask, p, m;
996 int error = 0;
997 struct sockaddr_in prefix0, mask0;
998
999 /*
1000 * Remove the loopback route to the interface address.
1001 * The "useloopback" setting is not consulted because if the
1002 * user configures an interface address, turns off this
1003 * setting, and then tries to delete that interface address,
1004 * checking the current setting of "useloopback" would leave
1005 * that interface address loopback route untouched, which
1006 * would be wrong. Therefore the interface address loopback route
1007 * deletion is unconditional.
1008 */
1009 if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1010 !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
1011 (target->ia_flags & IFA_RTSELF)) {
1012 struct route ia_ro;
1013 int freeit = 0;
1014 int fib;
1015
1016 bzero(&ia_ro, sizeof(ia_ro));
1017 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr;
1018 fib = target->ia_ifa.ifa_ifp->if_fib;
1019 rtalloc_ign_fib(&ia_ro, 0, fib);
1020 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
1021 (ia_ro.ro_rt->rt_ifp == V_loif)) {
1022 RT_LOCK(ia_ro.ro_rt);
1023 if (ia_ro.ro_rt->rt_refcnt <= 1)
1024 freeit = 1;
1025 else if (flags & LLE_STATIC) {
1026 RT_REMREF(ia_ro.ro_rt);
1027 target->ia_flags &= ~IFA_RTSELF;
1028 }
1029 RTFREE_LOCKED(ia_ro.ro_rt);
1030 }
1031 if (freeit && (flags & LLE_STATIC)) {
1032 error = ifa_del_loopback_route((struct ifaddr *)target,
1033 (struct sockaddr *)&target->ia_addr);
1034 if (error == 0)
1035 target->ia_flags &= ~IFA_RTSELF;
1036 }
1037 if ((flags & LLE_STATIC) &&
1038 !(target->ia_ifp->if_flags & IFF_NOARP))
1039 /* remove arp cache */
1040 arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1041 }
1042
1043 if (rtinitflags(target)) {
1044 prefix = target->ia_dstaddr.sin_addr;
1045 mask.s_addr = 0;
1046 } else {
1047 prefix = target->ia_addr.sin_addr;
1048 mask = target->ia_sockmask.sin_addr;
1049 prefix.s_addr &= mask.s_addr;
1050 }
1051
1052 if ((target->ia_flags & IFA_ROUTE) == 0) {
1053 int fibnum;
1054
1055 fibnum = rt_add_addr_allfibs ? RT_ALL_FIBS :
1056 target->ia_ifp->if_fib;
1057 rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum);
1058 return (0);
1059 }
1060
1061 IN_IFADDR_RLOCK();
1062 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1063 if (rtinitflags(ia)) {
1064 p = ia->ia_dstaddr.sin_addr;
1065
1066 if (prefix.s_addr != p.s_addr)
1067 continue;
1068 } else {
1069 p = ia->ia_addr.sin_addr;
1070 m = ia->ia_sockmask.sin_addr;
1071 p.s_addr &= m.s_addr;
1072
1073 if (prefix.s_addr != p.s_addr ||
1074 mask.s_addr != m.s_addr)
1075 continue;
1076 }
1077
1078 if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1079 continue;
1080
1081 /*
1082 * If we got a matching prefix address, move IFA_ROUTE and
1083 * the route itself to it. Make sure that routing daemons
1084 * get a heads-up.
1085 */
1086 if ((ia->ia_flags & IFA_ROUTE) == 0) {
1087 ifa_ref(&ia->ia_ifa);
1088 IN_IFADDR_RUNLOCK();
1089 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1090 rtinitflags(target));
1091 if (error == 0)
1092 target->ia_flags &= ~IFA_ROUTE;
1093 else
1094 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1095 error);
1096 error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1097 rtinitflags(ia) | RTF_UP);
1098 if (error == 0)
1099 ia->ia_flags |= IFA_ROUTE;
1100 else
1101 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1102 error);
1103 ifa_free(&ia->ia_ifa);
1104 return (error);
1105 }
1106 }
1107 IN_IFADDR_RUNLOCK();
1108
1109 /*
1110 * remove all L2 entries on the given prefix
1111 */
1112 bzero(&prefix0, sizeof(prefix0));
1113 prefix0.sin_len = sizeof(prefix0);
1114 prefix0.sin_family = AF_INET;
1115 prefix0.sin_addr.s_addr = target->ia_subnet;
1116 bzero(&mask0, sizeof(mask0));
1117 mask0.sin_len = sizeof(mask0);
1118 mask0.sin_family = AF_INET;
1119 mask0.sin_addr.s_addr = target->ia_subnetmask;
1120 lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1121 (struct sockaddr *)&mask0, flags);
1122
1123 /*
1124 * As no-one seem to have this prefix, we can remove the route.
1125 */
1126 error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1127 if (error == 0)
1128 target->ia_flags &= ~IFA_ROUTE;
1129 else
1130 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1131 return (error);
1132 }
1133
1134 #undef rtinitflags
1135
1136 /*
1137 * Return 1 if the address might be a local broadcast address.
1138 */
1139 int
1140 in_broadcast(struct in_addr in, struct ifnet *ifp)
1141 {
1142 register struct ifaddr *ifa;
1143 u_long t;
1144
1145 if (in.s_addr == INADDR_BROADCAST ||
1146 in.s_addr == INADDR_ANY)
1147 return (1);
1148 if ((ifp->if_flags & IFF_BROADCAST) == 0)
1149 return (0);
1150 t = ntohl(in.s_addr);
1151 /*
1152 * Look through the list of addresses for a match
1153 * with a broadcast address.
1154 */
1155 #define ia ((struct in_ifaddr *)ifa)
1156 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1157 if (ifa->ifa_addr->sa_family == AF_INET &&
1158 (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1159 /*
1160 * Check for old-style (host 0) broadcast, but
1161 * taking into account that RFC 3021 obsoletes it.
1162 */
1163 (ia->ia_subnetmask != IN_RFC3021_MASK &&
1164 t == ia->ia_subnet)) &&
1165 /*
1166 * Check for an all one subnetmask. These
1167 * only exist when an interface gets a secondary
1168 * address.
1169 */
1170 ia->ia_subnetmask != (u_long)0xffffffff)
1171 return (1);
1172 return (0);
1173 #undef ia
1174 }
1175
1176 /*
1177 * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1178 */
1179 void
1180 in_ifdetach(struct ifnet *ifp)
1181 {
1182
1183 in_pcbpurgeif0(&V_ripcbinfo, ifp);
1184 in_pcbpurgeif0(&V_udbinfo, ifp);
1185 in_pcbpurgeif0(&V_ulitecbinfo, ifp);
1186 in_purgemaddrs(ifp);
1187 }
1188
1189 /*
1190 * Delete all IPv4 multicast address records, and associated link-layer
1191 * multicast address records, associated with ifp.
1192 * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1193 * XXX This should not race with ifma_protospec being set during
1194 * a new allocation, if it does, we have bigger problems.
1195 */
1196 static void
1197 in_purgemaddrs(struct ifnet *ifp)
1198 {
1199 LIST_HEAD(,in_multi) purgeinms;
1200 struct in_multi *inm, *tinm;
1201 struct ifmultiaddr *ifma;
1202
1203 LIST_INIT(&purgeinms);
1204 IN_MULTI_LOCK();
1205
1206 /*
1207 * Extract list of in_multi associated with the detaching ifp
1208 * which the PF_INET layer is about to release.
1209 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1210 * by code further down.
1211 */
1212 IF_ADDR_RLOCK(ifp);
1213 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1214 if (ifma->ifma_addr->sa_family != AF_INET ||
1215 ifma->ifma_protospec == NULL)
1216 continue;
1217 #if 0
1218 KASSERT(ifma->ifma_protospec != NULL,
1219 ("%s: ifma_protospec is NULL", __func__));
1220 #endif
1221 inm = (struct in_multi *)ifma->ifma_protospec;
1222 LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1223 }
1224 IF_ADDR_RUNLOCK(ifp);
1225
1226 LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1227 LIST_REMOVE(inm, inm_link);
1228 inm_release_locked(inm);
1229 }
1230 igmp_ifdetach(ifp);
1231
1232 IN_MULTI_UNLOCK();
1233 }
1234
1235 struct in_llentry {
1236 struct llentry base;
1237 struct sockaddr_in l3_addr4;
1238 };
1239
1240 /*
1241 * Deletes an address from the address table.
1242 * This function is called by the timer functions
1243 * such as arptimer() and nd6_llinfo_timer(), and
1244 * the caller does the locking.
1245 */
1246 static void
1247 in_lltable_free(struct lltable *llt, struct llentry *lle)
1248 {
1249 LLE_WUNLOCK(lle);
1250 LLE_LOCK_DESTROY(lle);
1251 free(lle, M_LLTABLE);
1252 }
1253
1254 static struct llentry *
1255 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1256 {
1257 struct in_llentry *lle;
1258
1259 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1260 if (lle == NULL) /* NB: caller generates msg */
1261 return NULL;
1262
1263 /*
1264 * For IPv4 this will trigger "arpresolve" to generate
1265 * an ARP request.
1266 */
1267 lle->base.la_expire = time_uptime; /* mark expired */
1268 lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1269 lle->base.lle_refcnt = 1;
1270 lle->base.lle_free = in_lltable_free;
1271 LLE_LOCK_INIT(&lle->base);
1272 callout_init(&lle->base.la_timer, 1);
1273
1274 return (&lle->base);
1275 }
1276
1277 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \
1278 (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1279
1280 static void
1281 in_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix,
1282 const struct sockaddr *mask, u_int flags)
1283 {
1284 const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1285 const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1286 struct llentry *lle, *next;
1287 int i;
1288 size_t pkts_dropped;
1289
1290 IF_AFDATA_WLOCK(llt->llt_ifp);
1291 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1292 LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1293 /*
1294 * (flags & LLE_STATIC) means deleting all entries
1295 * including static ARP entries.
1296 */
1297 if (IN_ARE_MASKED_ADDR_EQUAL(satosin(L3_ADDR(lle)),
1298 pfx, msk) && ((flags & LLE_STATIC) ||
1299 !(lle->la_flags & LLE_STATIC))) {
1300 LLE_WLOCK(lle);
1301 if (callout_stop(&lle->la_timer))
1302 LLE_REMREF(lle);
1303 pkts_dropped = llentry_free(lle);
1304 ARPSTAT_ADD(dropped, pkts_dropped);
1305 }
1306 }
1307 }
1308 IF_AFDATA_WUNLOCK(llt->llt_ifp);
1309 }
1310
1311
1312 static int
1313 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1314 {
1315 struct rtentry *rt;
1316
1317 KASSERT(l3addr->sa_family == AF_INET,
1318 ("sin_family %d", l3addr->sa_family));
1319
1320 /* XXX rtalloc1_fib should take a const param */
1321 rt = rtalloc1_fib(__DECONST(struct sockaddr *, l3addr), 0, 0,
1322 ifp->if_fib);
1323
1324 if (rt == NULL)
1325 return (EINVAL);
1326
1327 /*
1328 * If the gateway for an existing host route matches the target L3
1329 * address, which is a special route inserted by some implementation
1330 * such as MANET, and the interface is of the correct type, then
1331 * allow for ARP to proceed.
1332 */
1333 if (rt->rt_flags & RTF_GATEWAY) {
1334 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
1335 rt->rt_ifp->if_type != IFT_ETHER ||
1336 (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1337 memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
1338 sizeof(in_addr_t)) != 0) {
1339 RTFREE_LOCKED(rt);
1340 return (EINVAL);
1341 }
1342 }
1343
1344 /*
1345 * Make sure that at least the destination address is covered
1346 * by the route. This is for handling the case where 2 or more
1347 * interfaces have the same prefix. An incoming packet arrives
1348 * on one interface and the corresponding outgoing packet leaves
1349 * another interface.
1350 */
1351 if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
1352 const char *sa, *mask, *addr, *lim;
1353 int len;
1354
1355 mask = (const char *)rt_mask(rt);
1356 /*
1357 * Just being extra cautious to avoid some custom
1358 * code getting into trouble.
1359 */
1360 if (mask == NULL) {
1361 RTFREE_LOCKED(rt);
1362 return (EINVAL);
1363 }
1364
1365 sa = (const char *)rt_key(rt);
1366 addr = (const char *)l3addr;
1367 len = ((const struct sockaddr_in *)l3addr)->sin_len;
1368 lim = addr + len;
1369
1370 for ( ; addr < lim; sa++, mask++, addr++) {
1371 if ((*sa ^ *addr) & *mask) {
1372 #ifdef DIAGNOSTIC
1373 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1374 inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1375 #endif
1376 RTFREE_LOCKED(rt);
1377 return (EINVAL);
1378 }
1379 }
1380 }
1381
1382 RTFREE_LOCKED(rt);
1383 return (0);
1384 }
1385
1386 /*
1387 * Return NULL if not found or marked for deletion.
1388 * If found return lle read locked.
1389 */
1390 static struct llentry *
1391 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1392 {
1393 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1394 struct ifnet *ifp = llt->llt_ifp;
1395 struct llentry *lle;
1396 struct llentries *lleh;
1397 u_int hashkey;
1398
1399 IF_AFDATA_LOCK_ASSERT(ifp);
1400 KASSERT(l3addr->sa_family == AF_INET,
1401 ("sin_family %d", l3addr->sa_family));
1402
1403 hashkey = sin->sin_addr.s_addr;
1404 lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1405 LIST_FOREACH(lle, lleh, lle_next) {
1406 struct sockaddr_in *sa2 = satosin(L3_ADDR(lle));
1407 if (lle->la_flags & LLE_DELETED)
1408 continue;
1409 if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1410 break;
1411 }
1412 if (lle == NULL) {
1413 #ifdef DIAGNOSTIC
1414 if (flags & LLE_DELETE)
1415 log(LOG_INFO, "interface address is missing from cache = %p in delete\n", lle);
1416 #endif
1417 if (!(flags & LLE_CREATE))
1418 return (NULL);
1419 IF_AFDATA_WLOCK_ASSERT(ifp);
1420 /*
1421 * A route that covers the given address must have
1422 * been installed 1st because we are doing a resolution,
1423 * verify this.
1424 */
1425 if (!(flags & LLE_IFADDR) &&
1426 in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1427 goto done;
1428
1429 lle = in_lltable_new(l3addr, flags);
1430 if (lle == NULL) {
1431 log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1432 goto done;
1433 }
1434 lle->la_flags = flags & ~LLE_CREATE;
1435 if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1436 bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1437 lle->la_flags |= (LLE_VALID | LLE_STATIC);
1438 }
1439
1440 lle->lle_tbl = llt;
1441 lle->lle_head = lleh;
1442 lle->la_flags |= LLE_LINKED;
1443 LIST_INSERT_HEAD(lleh, lle, lle_next);
1444 } else if (flags & LLE_DELETE) {
1445 if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1446 LLE_WLOCK(lle);
1447 lle->la_flags |= LLE_DELETED;
1448 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1449 #ifdef DIAGNOSTIC
1450 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1451 #endif
1452 if ((lle->la_flags &
1453 (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
1454 llentry_free(lle);
1455 else
1456 LLE_WUNLOCK(lle);
1457 }
1458 lle = (void *)-1;
1459
1460 }
1461 if (LLE_IS_VALID(lle)) {
1462 if (flags & LLE_EXCLUSIVE)
1463 LLE_WLOCK(lle);
1464 else
1465 LLE_RLOCK(lle);
1466 }
1467 done:
1468 return (lle);
1469 }
1470
1471 static int
1472 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1473 {
1474 #define SIN(lle) ((struct sockaddr_in *) L3_ADDR(lle))
1475 struct ifnet *ifp = llt->llt_ifp;
1476 struct llentry *lle;
1477 /* XXX stack use */
1478 struct {
1479 struct rt_msghdr rtm;
1480 struct sockaddr_in sin;
1481 struct sockaddr_dl sdl;
1482 } arpc;
1483 int error, i;
1484
1485 LLTABLE_LOCK_ASSERT();
1486
1487 error = 0;
1488 for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1489 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1490 struct sockaddr_dl *sdl;
1491
1492 /* skip deleted entries */
1493 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1494 continue;
1495 /* Skip if jailed and not a valid IP of the prison. */
1496 if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1497 continue;
1498 /*
1499 * produce a msg made of:
1500 * struct rt_msghdr;
1501 * struct sockaddr_in; (IPv4)
1502 * struct sockaddr_dl;
1503 */
1504 bzero(&arpc, sizeof(arpc));
1505 arpc.rtm.rtm_msglen = sizeof(arpc);
1506 arpc.rtm.rtm_version = RTM_VERSION;
1507 arpc.rtm.rtm_type = RTM_GET;
1508 arpc.rtm.rtm_flags = RTF_UP;
1509 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1510 arpc.sin.sin_family = AF_INET;
1511 arpc.sin.sin_len = sizeof(arpc.sin);
1512 arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1513
1514 /* publish */
1515 if (lle->la_flags & LLE_PUB)
1516 arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1517
1518 sdl = &arpc.sdl;
1519 sdl->sdl_family = AF_LINK;
1520 sdl->sdl_len = sizeof(*sdl);
1521 sdl->sdl_index = ifp->if_index;
1522 sdl->sdl_type = ifp->if_type;
1523 if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1524 sdl->sdl_alen = ifp->if_addrlen;
1525 bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1526 } else {
1527 sdl->sdl_alen = 0;
1528 bzero(LLADDR(sdl), ifp->if_addrlen);
1529 }
1530
1531 arpc.rtm.rtm_rmx.rmx_expire =
1532 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1533 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1534 if (lle->la_flags & LLE_STATIC)
1535 arpc.rtm.rtm_flags |= RTF_STATIC;
1536 arpc.rtm.rtm_index = ifp->if_index;
1537 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1538 if (error)
1539 break;
1540 }
1541 }
1542 return error;
1543 #undef SIN
1544 }
1545
1546 void *
1547 in_domifattach(struct ifnet *ifp)
1548 {
1549 struct in_ifinfo *ii;
1550 struct lltable *llt;
1551
1552 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1553
1554 llt = lltable_init(ifp, AF_INET);
1555 if (llt != NULL) {
1556 llt->llt_prefix_free = in_lltable_prefix_free;
1557 llt->llt_lookup = in_lltable_lookup;
1558 llt->llt_dump = in_lltable_dump;
1559 }
1560 ii->ii_llt = llt;
1561
1562 ii->ii_igmp = igmp_domifattach(ifp);
1563
1564 return ii;
1565 }
1566
1567 void
1568 in_domifdetach(struct ifnet *ifp, void *aux)
1569 {
1570 struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1571
1572 igmp_domifdetach(ifp);
1573 lltable_free(ii->ii_llt);
1574 free(ii, M_IFADDR);
1575 }
Cache object: 02d19fc5d9de6dc90fc9c0d9eeb9f052
|