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/in.h>
  60 #include <netinet/in_var.h>
  61 #include <netinet/in_pcb.h>
  62 #include <netinet/ip_var.h>
  63 #include <netinet/igmp_var.h>
  64 #include <netinet/udp.h>
  65 #include <netinet/udp_var.h>
  66
  67 static int in_mask2len(struct in_addr *);
  68 static void in_len2mask(struct in_addr *, int);
  69 static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t,
  70         struct ifnet *, struct thread *);
  71
  72 static int      in_addprefix(struct in_ifaddr *, int);
  73 static int      in_scrubprefix(struct in_ifaddr *, u_int);
  74 static void     in_socktrim(struct sockaddr_in *);
  75 static int      in_ifinit(struct ifnet *,
  76             struct in_ifaddr *, struct sockaddr_in *, int);
  77 static void     in_purgemaddrs(struct ifnet *);
  78
  79 static VNET_DEFINE(int, subnetsarelocal);
  80 #define V_subnetsarelocal               VNET(subnetsarelocal)
  81 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, subnets_are_local, CTLFLAG_RW,
  82         &VNET_NAME(subnetsarelocal), 0,
  83         "Treat all subnets as directly connected");
  84 static VNET_DEFINE(int, sameprefixcarponly);
  85 #define V_sameprefixcarponly            VNET(sameprefixcarponly)
  86 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, same_prefix_carp_only, CTLFLAG_RW,
  87         &VNET_NAME(sameprefixcarponly), 0,
  88         "Refuse to create same prefixes on different interfaces");
  89
  90 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
  91 #define V_ripcbinfo                     VNET(ripcbinfo)
  92
  93 VNET_DECLARE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
  94 #define V_arpstat               VNET(arpstat)
  95
  96 /*
  97  * Return 1 if an internet address is for a ``local'' host
  98  * (one to which we have a connection).  If subnetsarelocal
  99  * is true, this includes other subnets of the local net.
 100  * Otherwise, it includes only the directly-connected (sub)nets.
 101  */
 102 int
 103 in_localaddr(struct in_addr in)
 104 {
 105         register u_long i = ntohl(in.s_addr);
 106         register struct in_ifaddr *ia;
 107
 108         IN_IFADDR_RLOCK();
 109         if (V_subnetsarelocal) {
 110                 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
 111                         if ((i & ia->ia_netmask) == ia->ia_net) {
 112                                 IN_IFADDR_RUNLOCK();
 113                                 return (1);
 114                         }
 115                 }
 116         } else {
 117                 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
 118                         if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
 119                                 IN_IFADDR_RUNLOCK();
 120                                 return (1);
 121                         }
 122                 }
 123         }
 124         IN_IFADDR_RUNLOCK();
 125         return (0);
 126 }
 127
 128 /*
 129  * Return 1 if an internet address is for the local host and configured
 130  * on one of its interfaces.
 131  */
 132 int
 133 in_localip(struct in_addr in)
 134 {
 135         struct in_ifaddr *ia;
 136
 137         IN_IFADDR_RLOCK();
 138         LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
 139                 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
 140                         IN_IFADDR_RUNLOCK();
 141                         return (1);
 142                 }
 143         }
 144         IN_IFADDR_RUNLOCK();
 145         return (0);
 146 }
 147
 148 /*
 149  * Determine whether an IP address is in a reserved set of addresses
 150  * that may not be forwarded, or whether datagrams to that destination
 151  * may be forwarded.
 152  */
 153 int
 154 in_canforward(struct in_addr in)
 155 {
 156         register u_long i = ntohl(in.s_addr);
 157         register u_long net;
 158
 159         if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
 160                 return (0);
 161         if (IN_CLASSA(i)) {
 162                 net = i & IN_CLASSA_NET;
 163                 if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
 164                         return (0);
 165         }
 166         return (1);
 167 }
 168
 169 /*
 170  * Trim a mask in a sockaddr
 171  */
 172 static void
 173 in_socktrim(struct sockaddr_in *ap)
 174 {
 175     register char *cplim = (char *) &ap->sin_addr;
 176     register char *cp = (char *) (&ap->sin_addr + 1);
 177
 178     ap->sin_len = 0;
 179     while (--cp >= cplim)
 180         if (*cp) {
 181             (ap)->sin_len = cp - (char *) (ap) + 1;
 182             break;
 183         }
 184 }
 185
 186 static int
 187 in_mask2len(mask)
 188         struct in_addr *mask;
 189 {
 190         int x, y;
 191         u_char *p;
 192
 193         p = (u_char *)mask;
 194         for (x = 0; x < sizeof(*mask); x++) {
 195                 if (p[x] != 0xff)
 196                         break;
 197         }
 198         y = 0;
 199         if (x < sizeof(*mask)) {
 200                 for (y = 0; y < 8; y++) {
 201                         if ((p[x] & (0x80 >> y)) == 0)
 202                                 break;
 203                 }
 204         }
 205         return (x * 8 + y);
 206 }
 207
 208 static void
 209 in_len2mask(struct in_addr *mask, int len)
 210 {
 211         int i;
 212         u_char *p;
 213
 214         p = (u_char *)mask;
 215         bzero(mask, sizeof(*mask));
 216         for (i = 0; i < len / 8; i++)
 217                 p[i] = 0xff;
 218         if (len % 8)
 219                 p[i] = (0xff00 >> (len % 8)) & 0xff;
 220 }
 221
 222 /*
 223  * Generic internet control operations (ioctl's).
 224  *
 225  * ifp is NULL if not an interface-specific ioctl.
 226  */
 227 /* ARGSUSED */
 228 int
 229 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
 230     struct thread *td)
 231 {
 232         register struct ifreq *ifr = (struct ifreq *)data;
 233         register struct in_ifaddr *ia, *iap;
 234         register struct ifaddr *ifa;
 235         struct in_addr allhosts_addr;
 236         struct in_addr dst;
 237         struct in_ifinfo *ii;
 238         struct in_aliasreq *ifra = (struct in_aliasreq *)data;
 239         struct sockaddr_in oldaddr;
 240         int error, hostIsNew, iaIsNew, maskIsNew;
 241         int iaIsFirst;
 242
 243         ia = NULL;
 244         iaIsFirst = 0;
 245         iaIsNew = 0;
 246         allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
 247
 248         /*
 249          * Filter out ioctls we implement directly; forward the rest on to
 250          * in_lifaddr_ioctl() and ifp->if_ioctl().
 251          */
 252         switch (cmd) {
 253         case SIOCAIFADDR:
 254         case SIOCDIFADDR:
 255         case SIOCGIFADDR:
 256         case SIOCGIFBRDADDR:
 257         case SIOCGIFDSTADDR:
 258         case SIOCGIFNETMASK:
 259         case SIOCSIFADDR:
 260         case SIOCSIFBRDADDR:
 261         case SIOCSIFDSTADDR:
 262         case SIOCSIFNETMASK:
 263                 break;
 264
 265         case SIOCALIFADDR:
 266                 if (td != NULL) {
 267                         error = priv_check(td, PRIV_NET_ADDIFADDR);
 268                         if (error)
 269                                 return (error);
 270                 }
 271                 if (ifp == NULL)
 272                         return (EINVAL);
 273                 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
 274
 275         case SIOCDLIFADDR:
 276                 if (td != NULL) {
 277                         error = priv_check(td, PRIV_NET_DELIFADDR);
 278                         if (error)
 279                                 return (error);
 280                 }
 281                 if (ifp == NULL)
 282                         return (EINVAL);
 283                 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
 284
 285         case SIOCGLIFADDR:
 286                 if (ifp == NULL)
 287                         return (EINVAL);
 288                 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
 289
 290         default:
 291                 if (ifp == NULL || ifp->if_ioctl == NULL)
 292                         return (EOPNOTSUPP);
 293                 return ((*ifp->if_ioctl)(ifp, cmd, data));
 294         }
 295
 296         if (ifp == NULL)
 297                 return (EADDRNOTAVAIL);
 298
 299         /*
 300          * Security checks before we get involved in any work.
 301          */
 302         switch (cmd) {
 303         case SIOCAIFADDR:
 304         case SIOCSIFADDR:
 305         case SIOCSIFBRDADDR:
 306         case SIOCSIFNETMASK:
 307         case SIOCSIFDSTADDR:
 308                 if (td != NULL) {
 309                         error = priv_check(td, PRIV_NET_ADDIFADDR);
 310                         if (error)
 311                                 return (error);
 312                 }
 313                 break;
 314
 315         case SIOCDIFADDR:
 316                 if (td != NULL) {
 317                         error = priv_check(td, PRIV_NET_DELIFADDR);
 318                         if (error)
 319                                 return (error);
 320                 }
 321                 break;
 322         }
 323
 324         /*
 325          * Find address for this interface, if it exists.
 326          *
 327          * If an alias address was specified, find that one instead of the
 328          * first one on the interface, if possible.
 329          */
 330         dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
 331         IN_IFADDR_RLOCK();
 332         LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) {
 333                 if (iap->ia_ifp == ifp &&
 334                     iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
 335                         if (td == NULL || prison_check_ip4(td->td_ucred,
 336                             &dst) == 0)
 337                                 ia = iap;
 338                         break;
 339                 }
 340         }
 341         if (ia != NULL)
 342                 ifa_ref(&ia->ia_ifa);
 343         IN_IFADDR_RUNLOCK();
 344         if (ia == NULL) {
 345                 IF_ADDR_LOCK(ifp);
 346                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 347                         iap = ifatoia(ifa);
 348                         if (iap->ia_addr.sin_family == AF_INET) {
 349                                 if (td != NULL &&
 350                                     prison_check_ip4(td->td_ucred,
 351                                     &iap->ia_addr.sin_addr) != 0)
 352                                         continue;
 353                                 ia = iap;
 354                                 break;
 355                         }
 356                 }
 357                 if (ia != NULL)
 358                         ifa_ref(&ia->ia_ifa);
 359                 IF_ADDR_UNLOCK(ifp);
 360         }
 361         if (ia == NULL)
 362                 iaIsFirst = 1;
 363
 364         error = 0;
 365         switch (cmd) {
 366         case SIOCAIFADDR:
 367         case SIOCDIFADDR:
 368                 if (ifra->ifra_addr.sin_family == AF_INET) {
 369                         struct in_ifaddr *oia;
 370
 371                         IN_IFADDR_RLOCK();
 372                         for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
 373                                 if (ia->ia_ifp == ifp  &&
 374                                     ia->ia_addr.sin_addr.s_addr ==
 375                                     ifra->ifra_addr.sin_addr.s_addr)
 376                                         break;
 377                         }
 378                         if (ia != NULL && ia != oia)
 379                                 ifa_ref(&ia->ia_ifa);
 380                         if (oia != NULL && ia != oia)
 381                                 ifa_free(&oia->ia_ifa);
 382                         IN_IFADDR_RUNLOCK();
 383                         if ((ifp->if_flags & IFF_POINTOPOINT)
 384                             && (cmd == SIOCAIFADDR)
 385                             && (ifra->ifra_dstaddr.sin_addr.s_addr
 386                                 == INADDR_ANY)) {
 387                                 error = EDESTADDRREQ;
 388                                 goto out;
 389                         }
 390                 }
 391                 if (cmd == SIOCDIFADDR && ia == NULL) {
 392                         error = EADDRNOTAVAIL;
 393                         goto out;
 394                 }
 395                 /* FALLTHROUGH */
 396         case SIOCSIFADDR:
 397         case SIOCSIFNETMASK:
 398         case SIOCSIFDSTADDR:
 399                 if (ia == NULL) {
 400                         ia = (struct in_ifaddr *)
 401                                 malloc(sizeof *ia, M_IFADDR, M_NOWAIT |
 402                                     M_ZERO);
 403                         if (ia == NULL) {
 404                                 error = ENOBUFS;
 405                                 goto out;
 406                         }
 407
 408                         ifa = &ia->ia_ifa;
 409                         ifa_init(ifa);
 410                         ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
 411                         ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
 412                         ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
 413
 414                         ia->ia_sockmask.sin_len = 8;
 415                         ia->ia_sockmask.sin_family = AF_INET;
 416                         if (ifp->if_flags & IFF_BROADCAST) {
 417                                 ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
 418                                 ia->ia_broadaddr.sin_family = AF_INET;
 419                         }
 420                         ia->ia_ifp = ifp;
 421
 422                         ifa_ref(ifa);                   /* if_addrhead */
 423                         IF_ADDR_LOCK(ifp);
 424                         TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
 425                         IF_ADDR_UNLOCK(ifp);
 426                         ifa_ref(ifa);                   /* in_ifaddrhead */
 427                         IN_IFADDR_WLOCK();
 428                         TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
 429                         IN_IFADDR_WUNLOCK();
 430                         iaIsNew = 1;
 431                 }
 432                 break;
 433
 434         case SIOCSIFBRDADDR:
 435         case SIOCGIFADDR:
 436         case SIOCGIFNETMASK:
 437         case SIOCGIFDSTADDR:
 438         case SIOCGIFBRDADDR:
 439                 if (ia == NULL) {
 440                         error = EADDRNOTAVAIL;
 441                         goto out;
 442                 }
 443                 break;
 444         }
 445
 446         /*
 447          * Most paths in this switch return directly or via out.  Only paths
 448          * that remove the address break in order to hit common removal code.
 449          */
 450         switch (cmd) {
 451         case SIOCGIFADDR:
 452                 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
 453                 goto out;
 454
 455         case SIOCGIFBRDADDR:
 456                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 457                         error = EINVAL;
 458                         goto out;
 459                 }
 460                 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
 461                 goto out;
 462
 463         case SIOCGIFDSTADDR:
 464                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
 465                         error = EINVAL;
 466                         goto out;
 467                 }
 468                 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
 469                 goto out;
 470
 471         case SIOCGIFNETMASK:
 472                 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
 473                 goto out;
 474
 475         case SIOCSIFDSTADDR:
 476                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
 477                         error = EINVAL;
 478                         goto out;
 479                 }
 480                 oldaddr = ia->ia_dstaddr;
 481                 ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
 482                 if (ifp->if_ioctl != NULL) {
 483                         error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
 484                             (caddr_t)ia);
 485                         if (error) {
 486                                 ia->ia_dstaddr = oldaddr;
 487                                 goto out;
 488                         }
 489                 }
 490                 if (ia->ia_flags & IFA_ROUTE) {
 491                         ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
 492                         rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
 493                         ia->ia_ifa.ifa_dstaddr =
 494                                         (struct sockaddr *)&ia->ia_dstaddr;
 495                         rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
 496                 }
 497                 goto out;
 498
 499         case SIOCSIFBRDADDR:
 500                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 501                         error = EINVAL;
 502                         goto out;
 503                 }
 504                 ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
 505                 goto out;
 506
 507         case SIOCSIFADDR:
 508                 error = in_ifinit(ifp, ia,
 509                     (struct sockaddr_in *) &ifr->ifr_addr, 1);
 510                 if (error != 0 && iaIsNew)
 511                         break;
 512                 if (error == 0) {
 513                         ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
 514                         if (iaIsFirst &&
 515                             (ifp->if_flags & IFF_MULTICAST) != 0) {
 516                                 error = in_joingroup(ifp, &allhosts_addr,
 517                                     NULL, &ii->ii_allhosts);
 518                         }
 519                         EVENTHANDLER_INVOKE(ifaddr_event, ifp);
 520                 }
 521                 error = 0;
 522                 goto out;
 523
 524         case SIOCSIFNETMASK:
 525                 ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
 526                 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
 527                 goto out;
 528
 529         case SIOCAIFADDR:
 530                 maskIsNew = 0;
 531                 hostIsNew = 1;
 532                 error = 0;
 533                 if (ia->ia_addr.sin_family == AF_INET) {
 534                         if (ifra->ifra_addr.sin_len == 0) {
 535                                 ifra->ifra_addr = ia->ia_addr;
 536                                 hostIsNew = 0;
 537                         } else if (ifra->ifra_addr.sin_addr.s_addr ==
 538                                                ia->ia_addr.sin_addr.s_addr)
 539                                 hostIsNew = 0;
 540                 }
 541                 if (ifra->ifra_mask.sin_len) {
 542                         /*
 543                          * QL: XXX
 544                          * Need to scrub the prefix here in case
 545                          * the issued command is SIOCAIFADDR with
 546                          * the same address, but with a different
 547                          * prefix length. And if the prefix length
 548                          * is the same as before, then the call is
 549                          * un-necessarily executed here.
 550                          */
 551                         in_ifscrub(ifp, ia, LLE_STATIC);
 552                         ia->ia_sockmask = ifra->ifra_mask;
 553                         ia->ia_sockmask.sin_family = AF_INET;
 554                         ia->ia_subnetmask =
 555                              ntohl(ia->ia_sockmask.sin_addr.s_addr);
 556                         maskIsNew = 1;
 557                 }
 558                 if ((ifp->if_flags & IFF_POINTOPOINT) &&
 559                     (ifra->ifra_dstaddr.sin_family == AF_INET)) {
 560                         in_ifscrub(ifp, ia, LLE_STATIC);
 561                         ia->ia_dstaddr = ifra->ifra_dstaddr;
 562                         maskIsNew  = 1; /* We lie; but the effect's the same */
 563                 }
 564                 if (ifra->ifra_addr.sin_family == AF_INET &&
 565                     (hostIsNew || maskIsNew))
 566                         error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
 567                 if (error != 0 && iaIsNew)
 568                         goto out;
 569
 570                 if ((ifp->if_flags & IFF_BROADCAST) &&
 571                     (ifra->ifra_broadaddr.sin_family == AF_INET))
 572                         ia->ia_broadaddr = ifra->ifra_broadaddr;
 573                 if (error == 0) {
 574                         ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
 575                         if (iaIsFirst &&
 576                             (ifp->if_flags & IFF_MULTICAST) != 0) {
 577                                 error = in_joingroup(ifp, &allhosts_addr,
 578                                     NULL, &ii->ii_allhosts);
 579                         }
 580                         EVENTHANDLER_INVOKE(ifaddr_event, ifp);
 581                 }
 582                 goto out;
 583
 584         case SIOCDIFADDR:
 585                 /*
 586                  * in_ifscrub kills the interface route.
 587                  */
 588                 in_ifscrub(ifp, ia, LLE_STATIC);
 589
 590                 /*
 591                  * in_ifadown gets rid of all the rest of
 592                  * the routes.  This is not quite the right
 593                  * thing to do, but at least if we are running
 594                  * a routing process they will come back.
 595                  */
 596                 in_ifadown(&ia->ia_ifa, 1);
 597                 EVENTHANDLER_INVOKE(ifaddr_event, ifp);
 598                 error = 0;
 599                 break;
 600
 601         default:
 602                 panic("in_control: unsupported ioctl");
 603         }
 604
 605         IF_ADDR_LOCK(ifp);
 606         /* Re-check that ia is still part of the list. */
 607         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 608                 if (ifa == &ia->ia_ifa)
 609                         break;
 610         }
 611         if (ifa == NULL) {
 612                 /*
 613                  * If we lost the race with another thread, there is no need to
 614                  * try it again for the next loop as there is no other exit
 615                  * path between here and out.
 616                  */
 617                 IF_ADDR_UNLOCK(ifp);
 618                 error = EADDRNOTAVAIL;
 619                 goto out;
 620         }
 621         TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
 622         IF_ADDR_UNLOCK(ifp);
 623         ifa_free(&ia->ia_ifa);                          /* if_addrhead */
 624
 625         IN_IFADDR_WLOCK();
 626         TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
 627         if (ia->ia_addr.sin_family == AF_INET) {
 628                 struct in_ifaddr *if_ia;
 629
 630                 LIST_REMOVE(ia, ia_hash);
 631                 IN_IFADDR_WUNLOCK();
 632                 /*
 633                  * If this is the last IPv4 address configured on this
 634                  * interface, leave the all-hosts group.
 635                  * No state-change report need be transmitted.
 636                  */
 637                 if_ia = NULL;
 638                 IFP_TO_IA(ifp, if_ia);
 639                 if (if_ia == NULL) {
 640                         ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
 641                         IN_MULTI_LOCK();
 642                         if (ii->ii_allhosts) {
 643                                 (void)in_leavegroup_locked(ii->ii_allhosts,
 644                                     NULL);
 645                                 ii->ii_allhosts = NULL;
 646                         }
 647                         IN_MULTI_UNLOCK();
 648                 } else
 649                         ifa_free(&if_ia->ia_ifa);
 650         } else
 651                 IN_IFADDR_WUNLOCK();
 652         ifa_free(&ia->ia_ifa);                          /* in_ifaddrhead */
 653 out:
 654         if (ia != NULL)
 655                 ifa_free(&ia->ia_ifa);
 656         return (error);
 657 }
 658
 659 /*
 660  * SIOC[GAD]LIFADDR.
 661  *      SIOCGLIFADDR: get first address. (?!?)
 662  *      SIOCGLIFADDR with IFLR_PREFIX:
 663  *              get first address that matches the specified prefix.
 664  *      SIOCALIFADDR: add the specified address.
 665  *      SIOCALIFADDR with IFLR_PREFIX:
 666  *              EINVAL since we can't deduce hostid part of the address.
 667  *      SIOCDLIFADDR: delete the specified address.
 668  *      SIOCDLIFADDR with IFLR_PREFIX:
 669  *              delete the first address that matches the specified prefix.
 670  * return values:
 671  *      EINVAL on invalid parameters
 672  *      EADDRNOTAVAIL on prefix match failed/specified address not found
 673  *      other values may be returned from in_ioctl()
 674  */
 675 static int
 676 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
 677     struct ifnet *ifp, struct thread *td)
 678 {
 679         struct if_laddrreq *iflr = (struct if_laddrreq *)data;
 680         struct ifaddr *ifa;
 681
 682         /* sanity checks */
 683         if (data == NULL || ifp == NULL) {
 684                 panic("invalid argument to in_lifaddr_ioctl");
 685                 /*NOTRECHED*/
 686         }
 687
 688         switch (cmd) {
 689         case SIOCGLIFADDR:
 690                 /* address must be specified on GET with IFLR_PREFIX */
 691                 if ((iflr->flags & IFLR_PREFIX) == 0)
 692                         break;
 693                 /*FALLTHROUGH*/
 694         case SIOCALIFADDR:
 695         case SIOCDLIFADDR:
 696                 /* address must be specified on ADD and DELETE */
 697                 if (iflr->addr.ss_family != AF_INET)
 698                         return (EINVAL);
 699                 if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
 700                         return (EINVAL);
 701                 /* XXX need improvement */
 702                 if (iflr->dstaddr.ss_family
 703                  && iflr->dstaddr.ss_family != AF_INET)
 704                         return (EINVAL);
 705                 if (iflr->dstaddr.ss_family
 706                  && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
 707                         return (EINVAL);
 708                 break;
 709         default: /*shouldn't happen*/
 710                 return (EOPNOTSUPP);
 711         }
 712         if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
 713                 return (EINVAL);
 714
 715         switch (cmd) {
 716         case SIOCALIFADDR:
 717             {
 718                 struct in_aliasreq ifra;
 719
 720                 if (iflr->flags & IFLR_PREFIX)
 721                         return (EINVAL);
 722
 723                 /* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
 724                 bzero(&ifra, sizeof(ifra));
 725                 bcopy(iflr->iflr_name, ifra.ifra_name,
 726                         sizeof(ifra.ifra_name));
 727
 728                 bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
 729
 730                 if (iflr->dstaddr.ss_family) {  /*XXX*/
 731                         bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
 732                                 iflr->dstaddr.ss_len);
 733                 }
 734
 735                 ifra.ifra_mask.sin_family = AF_INET;
 736                 ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
 737                 in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
 738
 739                 return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
 740             }
 741         case SIOCGLIFADDR:
 742         case SIOCDLIFADDR:
 743             {
 744                 struct in_ifaddr *ia;
 745                 struct in_addr mask, candidate, match;
 746                 struct sockaddr_in *sin;
 747
 748                 bzero(&mask, sizeof(mask));
 749                 bzero(&match, sizeof(match));
 750                 if (iflr->flags & IFLR_PREFIX) {
 751                         /* lookup a prefix rather than address. */
 752                         in_len2mask(&mask, iflr->prefixlen);
 753
 754                         sin = (struct sockaddr_in *)&iflr->addr;
 755                         match.s_addr = sin->sin_addr.s_addr;
 756                         match.s_addr &= mask.s_addr;
 757
 758                         /* if you set extra bits, that's wrong */
 759                         if (match.s_addr != sin->sin_addr.s_addr)
 760                                 return (EINVAL);
 761
 762                 } else {
 763                         /* on getting an address, take the 1st match */
 764                         /* on deleting an address, do exact match */
 765                         if (cmd != SIOCGLIFADDR) {
 766                                 in_len2mask(&mask, 32);
 767                                 sin = (struct sockaddr_in *)&iflr->addr;
 768                                 match.s_addr = sin->sin_addr.s_addr;
 769                         }
 770                 }
 771
 772                 IF_ADDR_LOCK(ifp);
 773                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 774                         if (ifa->ifa_addr->sa_family != AF_INET)
 775                                 continue;
 776                         if (match.s_addr == 0)
 777                                 break;
 778                         candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
 779                         candidate.s_addr &= mask.s_addr;
 780                         if (candidate.s_addr == match.s_addr)
 781                                 break;
 782                 }
 783                 if (ifa != NULL)
 784                         ifa_ref(ifa);
 785                 IF_ADDR_UNLOCK(ifp);
 786                 if (ifa == NULL)
 787                         return (EADDRNOTAVAIL);
 788                 ia = (struct in_ifaddr *)ifa;
 789
 790                 if (cmd == SIOCGLIFADDR) {
 791                         /* fill in the if_laddrreq structure */
 792                         bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
 793
 794                         if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
 795                                 bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
 796                                         ia->ia_dstaddr.sin_len);
 797                         } else
 798                                 bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
 799
 800                         iflr->prefixlen =
 801                                 in_mask2len(&ia->ia_sockmask.sin_addr);
 802
 803                         iflr->flags = 0;        /*XXX*/
 804                         ifa_free(ifa);
 805
 806                         return (0);
 807                 } else {
 808                         struct in_aliasreq ifra;
 809
 810                         /* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
 811                         bzero(&ifra, sizeof(ifra));
 812                         bcopy(iflr->iflr_name, ifra.ifra_name,
 813                                 sizeof(ifra.ifra_name));
 814
 815                         bcopy(&ia->ia_addr, &ifra.ifra_addr,
 816                                 ia->ia_addr.sin_len);
 817                         if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
 818                                 bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
 819                                         ia->ia_dstaddr.sin_len);
 820                         }
 821                         bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
 822                                 ia->ia_sockmask.sin_len);
 823                         ifa_free(ifa);
 824
 825                         return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
 826                             ifp, td));
 827                 }
 828             }
 829         }
 830
 831         return (EOPNOTSUPP);    /*just for safety*/
 832 }
 833
 834 /*
 835  * Delete any existing route for an interface.
 836  */
 837 void
 838 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia, u_int flags)
 839 {
 840
 841         in_scrubprefix(ia, flags);
 842 }
 843
 844 /*
 845  * Initialize an interface's internet address
 846  * and routing table entry.
 847  */
 848 static int
 849 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
 850     int scrub)
 851 {
 852         register u_long i = ntohl(sin->sin_addr.s_addr);
 853         struct sockaddr_in oldaddr;
 854         int s = splimp(), flags = RTF_UP, error = 0;
 855
 856         oldaddr = ia->ia_addr;
 857         if (oldaddr.sin_family == AF_INET)
 858                 LIST_REMOVE(ia, ia_hash);
 859         ia->ia_addr = *sin;
 860         if (ia->ia_addr.sin_family == AF_INET) {
 861                 IN_IFADDR_WLOCK();
 862                 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
 863                     ia, ia_hash);
 864                 IN_IFADDR_WUNLOCK();
 865         }
 866         /*
 867          * Give the interface a chance to initialize
 868          * if this is its first address,
 869          * and to validate the address if necessary.
 870          */
 871         if (ifp->if_ioctl != NULL) {
 872                 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
 873                 if (error) {
 874                         splx(s);
 875                         /* LIST_REMOVE(ia, ia_hash) is done in in_control */
 876                         ia->ia_addr = oldaddr;
 877                         IN_IFADDR_WLOCK();
 878                         if (ia->ia_addr.sin_family == AF_INET)
 879                                 LIST_INSERT_HEAD(INADDR_HASH(
 880                                     ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
 881                         else
 882                                 /*
 883                                  * If oldaddr family is not AF_INET (e.g.
 884                                  * interface has been just created) in_control
 885                                  * does not call LIST_REMOVE, and we end up
 886                                  * with bogus ia entries in hash
 887                                  */
 888                                 LIST_REMOVE(ia, ia_hash);
 889                         IN_IFADDR_WUNLOCK();
 890                         return (error);
 891                 }
 892         }
 893         splx(s);
 894         if (scrub) {
 895                 ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
 896                 in_ifscrub(ifp, ia, LLE_STATIC);
 897                 ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
 898         }
 899         if (IN_CLASSA(i))
 900                 ia->ia_netmask = IN_CLASSA_NET;
 901         else if (IN_CLASSB(i))
 902                 ia->ia_netmask = IN_CLASSB_NET;
 903         else
 904                 ia->ia_netmask = IN_CLASSC_NET;
 905         /*
 906          * The subnet mask usually includes at least the standard network part,
 907          * but may may be smaller in the case of supernetting.
 908          * If it is set, we believe it.
 909          */
 910         if (ia->ia_subnetmask == 0) {
 911                 ia->ia_subnetmask = ia->ia_netmask;
 912                 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
 913         } else
 914                 ia->ia_netmask &= ia->ia_subnetmask;
 915         ia->ia_net = i & ia->ia_netmask;
 916         ia->ia_subnet = i & ia->ia_subnetmask;
 917         in_socktrim(&ia->ia_sockmask);
 918         /*
 919          * XXX: carp(4) does not have interface route
 920          */
 921         if (ifp->if_type == IFT_CARP)
 922                 return (0);
 923         /*
 924          * Add route for the network.
 925          */
 926         ia->ia_ifa.ifa_metric = ifp->if_metric;
 927         if (ifp->if_flags & IFF_BROADCAST) {
 928                 ia->ia_broadaddr.sin_addr.s_addr =
 929                         htonl(ia->ia_subnet | ~ia->ia_subnetmask);
 930                 ia->ia_netbroadcast.s_addr =
 931                         htonl(ia->ia_net | ~ ia->ia_netmask);
 932         } else if (ifp->if_flags & IFF_LOOPBACK) {
 933                 ia->ia_dstaddr = ia->ia_addr;
 934                 flags |= RTF_HOST;
 935         } else if (ifp->if_flags & IFF_POINTOPOINT) {
 936                 if (ia->ia_dstaddr.sin_family != AF_INET)
 937                         return (0);
 938                 flags |= RTF_HOST;
 939         }
 940         if ((error = in_addprefix(ia, flags)) != 0)
 941                 return (error);
 942
 943         if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
 944                 return (0);
 945
 946         if (ifp->if_flags & IFF_POINTOPOINT) {
 947                 if (ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
 948                         return (0);
 949         }
 950
 951
 952         /*
 953          * add a loopback route to self
 954          */
 955         if (V_useloopback && !(ifp->if_flags & IFF_LOOPBACK)) {
 956                 struct route ia_ro;
 957
 958                 bzero(&ia_ro, sizeof(ia_ro));
 959                 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr;
 960                 rtalloc_ign_fib(&ia_ro, 0, RT_DEFAULT_FIB);
 961                 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
 962                     (ia_ro.ro_rt->rt_ifp == V_loif)) {
 963                         RT_LOCK(ia_ro.ro_rt);
 964                         RT_ADDREF(ia_ro.ro_rt);
 965                         RTFREE_LOCKED(ia_ro.ro_rt);
 966                 } else
 967                         error = ifa_add_loopback_route((struct ifaddr *)ia,
 968                                        (struct sockaddr *)&ia->ia_addr);
 969                 if (error == 0)
 970                         ia->ia_flags |= IFA_RTSELF;
 971                 if (ia_ro.ro_rt != NULL)
 972                         RTFREE(ia_ro.ro_rt);
 973         }
 974
 975         return (error);
 976 }
 977
 978 #define rtinitflags(x) \
 979         ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
 980             ? RTF_HOST : 0)
 981
 982 /*
 983  * Generate a routing message when inserting or deleting
 984  * an interface address alias.
 985  */
 986 static void in_addralias_rtmsg(int cmd, struct in_addr *prefix,
 987     struct in_ifaddr *target)
 988 {
 989         struct route pfx_ro;
 990         struct sockaddr_in *pfx_addr;
 991         struct rtentry msg_rt;
 992
 993         /* QL: XXX
 994          * This is a bit questionable because there is no
 995          * additional route entry added/deleted for an address
 996          * alias. Therefore this route report is inaccurate.
 997          */
 998         bzero(&pfx_ro, sizeof(pfx_ro));
 999         pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst);
1000         pfx_addr->sin_len = sizeof(*pfx_addr);
1001         pfx_addr->sin_family = AF_INET;
1002         pfx_addr->sin_addr = *prefix;
1003         rtalloc_ign_fib(&pfx_ro, 0, 0);
1004         if (pfx_ro.ro_rt != NULL) {
1005                 msg_rt = *pfx_ro.ro_rt;
1006
1007                 /* QL: XXX
1008                  * Point the gateway to the new interface
1009                  * address as if a new prefix route entry has
1010                  * been added through the new address alias.
1011                  * All other parts of the rtentry is accurate,
1012                  * e.g., rt_key, rt_mask, rt_ifp etc.
1013                  */
1014                 msg_rt.rt_gateway =
1015                         (struct sockaddr *)&target->ia_addr;
1016                 rt_newaddrmsg(cmd,
1017                               (struct ifaddr *)target,
1018                               0, &msg_rt);
1019                 RTFREE(pfx_ro.ro_rt);
1020         }
1021         return;
1022 }
1023
1024 /*
1025  * Check if we have a route for the given prefix already or add one accordingly.
1026  */
1027 static int
1028 in_addprefix(struct in_ifaddr *target, int flags)
1029 {
1030         struct in_ifaddr *ia;
1031         struct in_addr prefix, mask, p, m;
1032         int error;
1033
1034         if ((flags & RTF_HOST) != 0) {
1035                 prefix = target->ia_dstaddr.sin_addr;
1036                 mask.s_addr = 0;
1037         } else {
1038                 prefix = target->ia_addr.sin_addr;
1039                 mask = target->ia_sockmask.sin_addr;
1040                 prefix.s_addr &= mask.s_addr;
1041         }
1042
1043         IN_IFADDR_RLOCK();
1044         TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1045                 if (rtinitflags(ia)) {
1046                         p = ia->ia_dstaddr.sin_addr;
1047
1048                         if (prefix.s_addr != p.s_addr)
1049                                 continue;
1050                 } else {
1051                         p = ia->ia_addr.sin_addr;
1052                         m = ia->ia_sockmask.sin_addr;
1053                         p.s_addr &= m.s_addr;
1054
1055                         if (prefix.s_addr != p.s_addr ||
1056                             mask.s_addr != m.s_addr)
1057                                 continue;
1058                 }
1059
1060                 /*
1061                  * If we got a matching prefix route inserted by other
1062                  * interface address, we are done here.
1063                  */
1064                 if (ia->ia_flags & IFA_ROUTE) {
1065 #ifdef RADIX_MPATH
1066                         if (ia->ia_addr.sin_addr.s_addr ==
1067                             target->ia_addr.sin_addr.s_addr) {
1068                                 IN_IFADDR_RUNLOCK();
1069                                 return (EEXIST);
1070                         } else
1071                                 break;
1072 #endif
1073                         if (V_sameprefixcarponly &&
1074                             target->ia_ifp->if_type != IFT_CARP &&
1075                             ia->ia_ifp->if_type != IFT_CARP) {
1076                                 IN_IFADDR_RUNLOCK();
1077                                 return (EEXIST);
1078                         } else {
1079                                 in_addralias_rtmsg(RTM_ADD, &prefix, target);
1080                                 IN_IFADDR_RUNLOCK();
1081                                 return (0);
1082                         }
1083                 }
1084         }
1085         IN_IFADDR_RUNLOCK();
1086
1087         /*
1088          * No-one seem to have this prefix route, so we try to insert it.
1089          */
1090         error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
1091         if (!error)
1092                 target->ia_flags |= IFA_ROUTE;
1093         return (error);
1094 }
1095
1096 extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
1097
1098 /*
1099  * If there is no other address in the system that can serve a route to the
1100  * same prefix, remove the route.  Hand over the route to the new address
1101  * otherwise.
1102  */
1103 static int
1104 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1105 {
1106         struct in_ifaddr *ia;
1107         struct in_addr prefix, mask, p;
1108         int error = 0;
1109         struct sockaddr_in prefix0, mask0;
1110
1111         /*
1112          * Remove the loopback route to the interface address.
1113          * The "useloopback" setting is not consulted because if the
1114          * user configures an interface address, turns off this
1115          * setting, and then tries to delete that interface address,
1116          * checking the current setting of "useloopback" would leave
1117          * that interface address loopback route untouched, which
1118          * would be wrong. Therefore the interface address loopback route
1119          * deletion is unconditional.
1120          */
1121         if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1122             !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
1123             (target->ia_flags & IFA_RTSELF)) {
1124                 struct route ia_ro;
1125                 int freeit = 0;
1126
1127                 bzero(&ia_ro, sizeof(ia_ro));
1128                 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr;
1129                 rtalloc_ign_fib(&ia_ro, 0, 0);
1130                 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
1131                     (ia_ro.ro_rt->rt_ifp == V_loif)) {
1132                         RT_LOCK(ia_ro.ro_rt);
1133                         if (ia_ro.ro_rt->rt_refcnt <= 1)
1134                                 freeit = 1;
1135                         else if (flags & LLE_STATIC) {
1136                                 RT_REMREF(ia_ro.ro_rt);
1137                                 target->ia_flags &= ~IFA_RTSELF;
1138                         }
1139                         RTFREE_LOCKED(ia_ro.ro_rt);
1140                 }
1141                 if (freeit && (flags & LLE_STATIC)) {
1142                         error = ifa_del_loopback_route((struct ifaddr *)target,
1143                                        (struct sockaddr *)&target->ia_addr);
1144                         if (error == 0)
1145                                 target->ia_flags &= ~IFA_RTSELF;
1146                 }
1147                 if ((flags & LLE_STATIC) &&
1148                         !(target->ia_ifp->if_flags & IFF_NOARP))
1149                         /* remove arp cache */
1150                         arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1151         }
1152
1153         if (rtinitflags(target))
1154                 prefix = target->ia_dstaddr.sin_addr;
1155         else {
1156                 prefix = target->ia_addr.sin_addr;
1157                 mask = target->ia_sockmask.sin_addr;
1158                 prefix.s_addr &= mask.s_addr;
1159         }
1160
1161         if ((target->ia_flags & IFA_ROUTE) == 0) {
1162                 in_addralias_rtmsg(RTM_DELETE, &prefix, target);
1163                 return (0);
1164         }
1165
1166         IN_IFADDR_RLOCK();
1167         TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1168                 if (rtinitflags(ia))
1169                         p = ia->ia_dstaddr.sin_addr;
1170                 else {
1171                         p = ia->ia_addr.sin_addr;
1172                         p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1173                 }
1174
1175                 if ((prefix.s_addr != p.s_addr) ||
1176                     !(ia->ia_ifp->if_flags & IFF_UP))
1177                         continue;
1178
1179                 /*
1180                  * If we got a matching prefix address, move IFA_ROUTE and
1181                  * the route itself to it.  Make sure that routing daemons
1182                  * get a heads-up.
1183                  *
1184                  * XXX: a special case for carp(4) interface - this should
1185                  *      be more generally specified as an interface that
1186                  *      doesn't support such action.
1187                  */
1188                 if ((ia->ia_flags & IFA_ROUTE) == 0
1189                     && (ia->ia_ifp->if_type != IFT_CARP)) {
1190                         ifa_ref(&ia->ia_ifa);
1191                         IN_IFADDR_RUNLOCK();
1192                         error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1193                             rtinitflags(target));
1194                         if (error == 0)
1195                                 target->ia_flags &= ~IFA_ROUTE;
1196                         else
1197                                 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1198                                         error);
1199                         error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1200                             rtinitflags(ia) | RTF_UP);
1201                         if (error == 0)
1202                                 ia->ia_flags |= IFA_ROUTE;
1203                         else
1204                                 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1205                                         error);
1206                         ifa_free(&ia->ia_ifa);
1207                         return (error);
1208                 }
1209         }
1210         IN_IFADDR_RUNLOCK();
1211
1212         /*
1213          * remove all L2 entries on the given prefix
1214          */
1215         bzero(&prefix0, sizeof(prefix0));
1216         prefix0.sin_len = sizeof(prefix0);
1217         prefix0.sin_family = AF_INET;
1218         prefix0.sin_addr.s_addr = target->ia_subnet;
1219         bzero(&mask0, sizeof(mask0));
1220         mask0.sin_len = sizeof(mask0);
1221         mask0.sin_family = AF_INET;
1222         mask0.sin_addr.s_addr = target->ia_subnetmask;
1223         lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1224                             (struct sockaddr *)&mask0, flags);
1225
1226         /*
1227          * As no-one seem to have this prefix, we can remove the route.
1228          */
1229         error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1230         if (error == 0)
1231                 target->ia_flags &= ~IFA_ROUTE;
1232         else
1233                 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1234         return (error);
1235 }
1236
1237 #undef rtinitflags
1238
1239 /*
1240  * Return 1 if the address might be a local broadcast address.
1241  */
1242 int
1243 in_broadcast(struct in_addr in, struct ifnet *ifp)
1244 {
1245         register struct ifaddr *ifa;
1246         u_long t;
1247
1248         if (in.s_addr == INADDR_BROADCAST ||
1249             in.s_addr == INADDR_ANY)
1250                 return (1);
1251         if ((ifp->if_flags & IFF_BROADCAST) == 0)
1252                 return (0);
1253         t = ntohl(in.s_addr);
1254         /*
1255          * Look through the list of addresses for a match
1256          * with a broadcast address.
1257          */
1258 #define ia ((struct in_ifaddr *)ifa)
1259         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1260                 if (ifa->ifa_addr->sa_family == AF_INET &&
1261                     (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1262                      in.s_addr == ia->ia_netbroadcast.s_addr ||
1263                      /*
1264                       * Check for old-style (host 0) broadcast.
1265                       */
1266                      t == ia->ia_subnet || t == ia->ia_net) &&
1267                      /*
1268                       * Check for an all one subnetmask. These
1269                       * only exist when an interface gets a secondary
1270                       * address.
1271                       */
1272                      ia->ia_subnetmask != (u_long)0xffffffff)
1273                             return (1);
1274         return (0);
1275 #undef ia
1276 }
1277
1278 /*
1279  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1280  */
1281 void
1282 in_ifdetach(struct ifnet *ifp)
1283 {
1284
1285         in_pcbpurgeif0(&V_ripcbinfo, ifp);
1286         in_pcbpurgeif0(&V_udbinfo, ifp);
1287         in_purgemaddrs(ifp);
1288 }
1289
1290 /*
1291  * Delete all IPv4 multicast address records, and associated link-layer
1292  * multicast address records, associated with ifp.
1293  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1294  * XXX This should not race with ifma_protospec being set during
1295  * a new allocation, if it does, we have bigger problems.
1296  */
1297 static void
1298 in_purgemaddrs(struct ifnet *ifp)
1299 {
1300         LIST_HEAD(,in_multi) purgeinms;
1301         struct in_multi         *inm, *tinm;
1302         struct ifmultiaddr      *ifma;
1303
1304         LIST_INIT(&purgeinms);
1305         IN_MULTI_LOCK();
1306
1307         /*
1308          * Extract list of in_multi associated with the detaching ifp
1309          * which the PF_INET layer is about to release.
1310          * We need to do this as IF_ADDR_LOCK() may be re-acquired
1311          * by code further down.
1312          */
1313         IF_ADDR_LOCK(ifp);
1314         TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1315                 if (ifma->ifma_addr->sa_family != AF_INET ||
1316                     ifma->ifma_protospec == NULL)
1317                         continue;
1318 #if 0
1319                 KASSERT(ifma->ifma_protospec != NULL,
1320                     ("%s: ifma_protospec is NULL", __func__));
1321 #endif
1322                 inm = (struct in_multi *)ifma->ifma_protospec;
1323                 LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1324         }
1325         IF_ADDR_UNLOCK(ifp);
1326
1327         LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1328                 LIST_REMOVE(inm, inm_link);
1329                 inm_release_locked(inm);
1330         }
1331         igmp_ifdetach(ifp);
1332
1333         IN_MULTI_UNLOCK();
1334 }
1335
1336 #include <net/if_dl.h>
1337 #include <netinet/if_ether.h>
1338
1339 struct in_llentry {
1340         struct llentry          base;
1341         struct sockaddr_in      l3_addr4;
1342 };
1343
1344 static struct llentry *
1345 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1346 {
1347         struct in_llentry *lle;
1348
1349         lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1350         if (lle == NULL)                /* NB: caller generates msg */
1351                 return NULL;
1352
1353         callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1354         /*
1355          * For IPv4 this will trigger "arpresolve" to generate
1356          * an ARP request.
1357          */
1358         lle->base.la_expire = time_second; /* mark expired */
1359         lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1360         lle->base.lle_refcnt = 1;
1361         LLE_LOCK_INIT(&lle->base);
1362         return &lle->base;
1363 }
1364
1365 /*
1366  * Deletes an address from the address table.
1367  * This function is called by the timer functions
1368  * such as arptimer() and nd6_llinfo_timer(), and
1369  * the caller does the locking.
1370  */
1371 static void
1372 in_lltable_free(struct lltable *llt, struct llentry *lle)
1373 {
1374         LLE_WUNLOCK(lle);
1375         LLE_LOCK_DESTROY(lle);
1376         free(lle, M_LLTABLE);
1377 }
1378
1379
1380 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)       (                       \
1381             (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1382
1383 static void
1384 in_lltable_prefix_free(struct lltable *llt,
1385                        const struct sockaddr *prefix,
1386                        const struct sockaddr *mask,
1387                        u_int flags)
1388 {
1389         const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1390         const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1391         struct llentry *lle, *next;
1392         register int i;
1393         size_t pkts_dropped;
1394
1395         for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
1396                 LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1397
1398                         /*
1399                          * (flags & LLE_STATIC) means deleting all entries
1400                          * including static ARP entries
1401                          */
1402                         if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle),
1403                                                      pfx, msk) &&
1404                             ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) {
1405                                 int canceled;
1406
1407                                 canceled = callout_drain(&lle->la_timer);
1408                                 LLE_WLOCK(lle);
1409                                 if (canceled)
1410                                         LLE_REMREF(lle);
1411                                 pkts_dropped = llentry_free(lle);
1412                                 ARPSTAT_ADD(dropped, pkts_dropped);
1413                         }
1414                 }
1415         }
1416 }
1417
1418
1419 static int
1420 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1421 {
1422         struct rtentry *rt;
1423
1424         KASSERT(l3addr->sa_family == AF_INET,
1425             ("sin_family %d", l3addr->sa_family));
1426
1427         /* XXX rtalloc1 should take a const param */
1428         rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1429
1430         if (rt == NULL)
1431                 return (EINVAL);
1432
1433         /*
1434          * If the gateway for an existing host route matches the target L3
1435          * address, which is a special route inserted by some implementation
1436          * such as MANET, and the interface is of the correct type, then
1437          * allow for ARP to proceed.
1438          */
1439         if (rt->rt_flags & RTF_GATEWAY) {
1440                 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
1441                         rt->rt_ifp->if_type != IFT_ETHER ||
1442                           (rt->rt_ifp->if_flags &
1443                            (IFF_NOARP | IFF_STATICARP)) != 0 ||
1444                           memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
1445                                  sizeof(in_addr_t)) != 0) {
1446                         RTFREE_LOCKED(rt);
1447                         return (EINVAL);
1448                 }
1449         }
1450
1451         /*
1452          * Make sure that at least the destination address is covered
1453          * by the route. This is for handling the case where 2 or more
1454          * interfaces have the same prefix. An incoming packet arrives
1455          * on one interface and the corresponding outgoing packet leaves
1456          * another interface.
1457          */
1458         if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
1459                 const char *sa, *mask, *addr, *lim;
1460                 int len;
1461
1462                 mask = (const char *)rt_mask(rt);
1463                 /*
1464                  * Just being extra cautious to avoid some custom
1465                  * code getting into trouble.
1466                  */
1467                 if (mask == NULL) {
1468                         RTFREE_LOCKED(rt);
1469                         return (EINVAL);
1470                 }
1471
1472                 sa = (const char *)rt_key(rt);
1473                 addr = (const char *)l3addr;
1474                 len = ((const struct sockaddr_in *)l3addr)->sin_len;
1475                 lim = addr + len;
1476
1477                 for ( ; addr < lim; sa++, mask++, addr++) {
1478                         if ((*sa ^ *addr) & *mask) {
1479 #ifdef DIAGNOSTIC
1480                                 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1481                                     inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1482 #endif
1483                                 RTFREE_LOCKED(rt);
1484                                 return (EINVAL);
1485                         }
1486                 }
1487         }
1488
1489         RTFREE_LOCKED(rt);
1490         return (0);
1491 }
1492
1493 /*
1494  * Return NULL if not found or marked for deletion.
1495  * If found return lle read locked.
1496  */
1497 static struct llentry *
1498 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1499 {
1500         const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1501         struct ifnet *ifp = llt->llt_ifp;
1502         struct llentry *lle;
1503         struct llentries *lleh;
1504         u_int hashkey;
1505
1506         IF_AFDATA_LOCK_ASSERT(ifp);
1507         KASSERT(l3addr->sa_family == AF_INET,
1508             ("sin_family %d", l3addr->sa_family));
1509
1510         hashkey = sin->sin_addr.s_addr;
1511         lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1512         LIST_FOREACH(lle, lleh, lle_next) {
1513                 struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1514                 if (lle->la_flags & LLE_DELETED)
1515                         continue;
1516                 if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1517                         break;
1518         }
1519         if (lle == NULL) {
1520 #ifdef DIAGNOSTIC
1521                 if (flags & LLE_DELETE)
1522                         log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1523 #endif
1524                 if (!(flags & LLE_CREATE))
1525                         return (NULL);
1526                 /*
1527                  * A route that covers the given address must have
1528                  * been installed 1st because we are doing a resolution,
1529                  * verify this.
1530                  */
1531                 if (!(flags & LLE_IFADDR) &&
1532                     in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1533                         goto done;
1534
1535                 lle = in_lltable_new(l3addr, flags);
1536                 if (lle == NULL) {
1537                         log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1538                         goto done;
1539                 }
1540                 lle->la_flags = flags & ~LLE_CREATE;
1541                 if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1542                         bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1543                         lle->la_flags |= (LLE_VALID | LLE_STATIC);
1544                 }
1545
1546                 lle->lle_tbl  = llt;
1547                 lle->lle_head = lleh;
1548                 LIST_INSERT_HEAD(lleh, lle, lle_next);
1549         } else if (flags & LLE_DELETE) {
1550                 if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1551                         LLE_WLOCK(lle);
1552                         lle->la_flags = LLE_DELETED;
1553                         LLE_WUNLOCK(lle);
1554 #ifdef DIAGNOSTIC
1555                         log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1556 #endif
1557                 }
1558                 lle = (void *)-1;
1559
1560         }
1561         if (LLE_IS_VALID(lle)) {
1562                 if (flags & LLE_EXCLUSIVE)
1563                         LLE_WLOCK(lle);
1564                 else
1565                         LLE_RLOCK(lle);
1566         }
1567 done:
1568         return (lle);
1569 }
1570
1571 static int
1572 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1573 {
1574 #define SIN(lle)        ((struct sockaddr_in *) L3_ADDR(lle))
1575         struct ifnet *ifp = llt->llt_ifp;
1576         struct llentry *lle;
1577         /* XXX stack use */
1578         struct {
1579                 struct rt_msghdr        rtm;
1580                 struct sockaddr_inarp   sin;
1581                 struct sockaddr_dl      sdl;
1582         } arpc;
1583         int error, i;
1584
1585         LLTABLE_LOCK_ASSERT();
1586
1587         error = 0;
1588         for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1589                 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1590                         struct sockaddr_dl *sdl;
1591
1592                         /* skip deleted entries */
1593                         if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1594                                 continue;
1595                         /* Skip if jailed and not a valid IP of the prison. */
1596                         if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1597                                 continue;
1598                         /*
1599                          * produce a msg made of:
1600                          *  struct rt_msghdr;
1601                          *  struct sockaddr_inarp; (IPv4)
1602                          *  struct sockaddr_dl;
1603                          */
1604                         bzero(&arpc, sizeof(arpc));
1605                         arpc.rtm.rtm_msglen = sizeof(arpc);
1606                         arpc.rtm.rtm_version = RTM_VERSION;
1607                         arpc.rtm.rtm_type = RTM_GET;
1608                         arpc.rtm.rtm_flags = RTF_UP;
1609                         arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1610                         arpc.sin.sin_family = AF_INET;
1611                         arpc.sin.sin_len = sizeof(arpc.sin);
1612                         arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1613
1614                         /* publish */
1615                         if (lle->la_flags & LLE_PUB) {
1616                                 arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1617                                 /* proxy only */
1618                                 if (lle->la_flags & LLE_PROXY)
1619                                         arpc.sin.sin_other = SIN_PROXY;
1620                         }
1621
1622                         sdl = &arpc.sdl;
1623                         sdl->sdl_family = AF_LINK;
1624                         sdl->sdl_len = sizeof(*sdl);
1625                         sdl->sdl_index = ifp->if_index;
1626                         sdl->sdl_type = ifp->if_type;
1627                         if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1628                                 sdl->sdl_alen = ifp->if_addrlen;
1629                                 bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1630                         } else {
1631                                 sdl->sdl_alen = 0;
1632                                 bzero(LLADDR(sdl), ifp->if_addrlen);
1633                         }
1634
1635                         arpc.rtm.rtm_rmx.rmx_expire =
1636                             lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1637                         arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1638                         if (lle->la_flags & LLE_STATIC)
1639                                 arpc.rtm.rtm_flags |= RTF_STATIC;
1640                         arpc.rtm.rtm_index = ifp->if_index;
1641                         error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1642                         if (error)
1643                                 break;
1644                 }
1645         }
1646         return error;
1647 #undef SIN
1648 }
1649
1650 void *
1651 in_domifattach(struct ifnet *ifp)
1652 {
1653         struct in_ifinfo *ii;
1654         struct lltable *llt;
1655
1656         ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1657
1658         llt = lltable_init(ifp, AF_INET);
1659         if (llt != NULL) {
1660                 llt->llt_free = in_lltable_free;
1661                 llt->llt_prefix_free = in_lltable_prefix_free;
1662                 llt->llt_lookup = in_lltable_lookup;
1663                 llt->llt_dump = in_lltable_dump;
1664         }
1665         ii->ii_llt = llt;
1666
1667         ii->ii_igmp = igmp_domifattach(ifp);
1668
1669         return ii;
1670 }
1671
1672 void
1673 in_domifdetach(struct ifnet *ifp, void *aux)
1674 {
1675         struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1676
1677         igmp_domifdetach(ifp);
1678         lltable_free(ii->ii_llt);
1679         free(ii, M_IFADDR);
1680 }