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, sameprefixcarponly);
  80 #define V_sameprefixcarponly            VNET(sameprefixcarponly)
  81 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, same_prefix_carp_only, CTLFLAG_RW,
  82         &VNET_NAME(sameprefixcarponly), 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 VNET_DECLARE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
  89 #define V_arpstat               VNET(arpstat)
  90
  91 /*
  92  * Return 1 if an internet address is for a ``local'' host
  93  * (one to which we have a connection).
  94  */
  95 int
  96 in_localaddr(struct in_addr in)
  97 {
  98         register u_long i = ntohl(in.s_addr);
  99         register struct in_ifaddr *ia;
 100
 101         IN_IFADDR_RLOCK();
 102         TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
 103                 if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
 104                         IN_IFADDR_RUNLOCK();
 105                         return (1);
 106                 }
 107         }
 108         IN_IFADDR_RUNLOCK();
 109         return (0);
 110 }
 111
 112 /*
 113  * Return 1 if an internet address is for the local host and configured
 114  * on one of its interfaces.
 115  */
 116 int
 117 in_localip(struct in_addr in)
 118 {
 119         struct in_ifaddr *ia;
 120
 121         IN_IFADDR_RLOCK();
 122         LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
 123                 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
 124                         IN_IFADDR_RUNLOCK();
 125                         return (1);
 126                 }
 127         }
 128         IN_IFADDR_RUNLOCK();
 129         return (0);
 130 }
 131
 132 /*
 133  * Determine whether an IP address is in a reserved set of addresses
 134  * that may not be forwarded, or whether datagrams to that destination
 135  * may be forwarded.
 136  */
 137 int
 138 in_canforward(struct in_addr in)
 139 {
 140         register u_long i = ntohl(in.s_addr);
 141         register u_long net;
 142
 143         if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
 144                 return (0);
 145         if (IN_CLASSA(i)) {
 146                 net = i & IN_CLASSA_NET;
 147                 if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
 148                         return (0);
 149         }
 150         return (1);
 151 }
 152
 153 /*
 154  * Trim a mask in a sockaddr
 155  */
 156 static void
 157 in_socktrim(struct sockaddr_in *ap)
 158 {
 159     register char *cplim = (char *) &ap->sin_addr;
 160     register char *cp = (char *) (&ap->sin_addr + 1);
 161
 162     ap->sin_len = 0;
 163     while (--cp >= cplim)
 164         if (*cp) {
 165             (ap)->sin_len = cp - (char *) (ap) + 1;
 166             break;
 167         }
 168 }
 169
 170 static int
 171 in_mask2len(mask)
 172         struct in_addr *mask;
 173 {
 174         int x, y;
 175         u_char *p;
 176
 177         p = (u_char *)mask;
 178         for (x = 0; x < sizeof(*mask); x++) {
 179                 if (p[x] != 0xff)
 180                         break;
 181         }
 182         y = 0;
 183         if (x < sizeof(*mask)) {
 184                 for (y = 0; y < 8; y++) {
 185                         if ((p[x] & (0x80 >> y)) == 0)
 186                                 break;
 187                 }
 188         }
 189         return (x * 8 + y);
 190 }
 191
 192 static void
 193 in_len2mask(struct in_addr *mask, int len)
 194 {
 195         int i;
 196         u_char *p;
 197
 198         p = (u_char *)mask;
 199         bzero(mask, sizeof(*mask));
 200         for (i = 0; i < len / 8; i++)
 201                 p[i] = 0xff;
 202         if (len % 8)
 203                 p[i] = (0xff00 >> (len % 8)) & 0xff;
 204 }
 205
 206 /*
 207  * Generic internet control operations (ioctl's).
 208  *
 209  * ifp is NULL if not an interface-specific ioctl.
 210  */
 211 /* ARGSUSED */
 212 int
 213 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
 214     struct thread *td)
 215 {
 216         register struct ifreq *ifr = (struct ifreq *)data;
 217         register struct in_ifaddr *ia, *iap;
 218         register struct ifaddr *ifa;
 219         struct in_addr allhosts_addr;
 220         struct in_addr dst;
 221         struct in_ifinfo *ii;
 222         struct in_aliasreq *ifra = (struct in_aliasreq *)data;
 223         struct sockaddr_in oldaddr;
 224         int error, hostIsNew, iaIsNew, maskIsNew;
 225         int iaIsFirst;
 226
 227         ia = NULL;
 228         iaIsFirst = 0;
 229         iaIsNew = 0;
 230         allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
 231
 232         /*
 233          * Filter out ioctls we implement directly; forward the rest on to
 234          * in_lifaddr_ioctl() and ifp->if_ioctl().
 235          */
 236         switch (cmd) {
 237         case SIOCAIFADDR:
 238         case SIOCDIFADDR:
 239         case SIOCGIFADDR:
 240         case SIOCGIFBRDADDR:
 241         case SIOCGIFDSTADDR:
 242         case SIOCGIFNETMASK:
 243         case SIOCSIFADDR:
 244         case SIOCSIFBRDADDR:
 245         case SIOCSIFDSTADDR:
 246         case SIOCSIFNETMASK:
 247                 break;
 248
 249         case SIOCALIFADDR:
 250                 if (td != NULL) {
 251                         error = priv_check(td, PRIV_NET_ADDIFADDR);
 252                         if (error)
 253                                 return (error);
 254                 }
 255                 if (ifp == NULL)
 256                         return (EINVAL);
 257                 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
 258
 259         case SIOCDLIFADDR:
 260                 if (td != NULL) {
 261                         error = priv_check(td, PRIV_NET_DELIFADDR);
 262                         if (error)
 263                                 return (error);
 264                 }
 265                 if (ifp == NULL)
 266                         return (EINVAL);
 267                 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
 268
 269         case SIOCGLIFADDR:
 270                 if (ifp == NULL)
 271                         return (EINVAL);
 272                 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
 273
 274         default:
 275                 if (ifp == NULL || ifp->if_ioctl == NULL)
 276                         return (EOPNOTSUPP);
 277                 return ((*ifp->if_ioctl)(ifp, cmd, data));
 278         }
 279
 280         if (ifp == NULL)
 281                 return (EADDRNOTAVAIL);
 282
 283         /*
 284          * Security checks before we get involved in any work.
 285          */
 286         switch (cmd) {
 287         case SIOCAIFADDR:
 288         case SIOCSIFADDR:
 289         case SIOCSIFBRDADDR:
 290         case SIOCSIFNETMASK:
 291         case SIOCSIFDSTADDR:
 292                 if (td != NULL) {
 293                         error = priv_check(td, PRIV_NET_ADDIFADDR);
 294                         if (error)
 295                                 return (error);
 296                 }
 297                 break;
 298
 299         case SIOCDIFADDR:
 300                 if (td != NULL) {
 301                         error = priv_check(td, PRIV_NET_DELIFADDR);
 302                         if (error)
 303                                 return (error);
 304                 }
 305                 break;
 306         }
 307
 308         /*
 309          * Find address for this interface, if it exists.
 310          *
 311          * If an alias address was specified, find that one instead of the
 312          * first one on the interface, if possible.
 313          */
 314         dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
 315         IN_IFADDR_RLOCK();
 316         LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) {
 317                 if (iap->ia_ifp == ifp &&
 318                     iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
 319                         if (td == NULL || prison_check_ip4(td->td_ucred,
 320                             &dst) == 0)
 321                                 ia = iap;
 322                         break;
 323                 }
 324         }
 325         if (ia != NULL)
 326                 ifa_ref(&ia->ia_ifa);
 327         IN_IFADDR_RUNLOCK();
 328         if (ia == NULL) {
 329                 IF_ADDR_LOCK(ifp);
 330                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 331                         iap = ifatoia(ifa);
 332                         if (iap->ia_addr.sin_family == AF_INET) {
 333                                 if (td != NULL &&
 334                                     prison_check_ip4(td->td_ucred,
 335                                     &iap->ia_addr.sin_addr) != 0)
 336                                         continue;
 337                                 ia = iap;
 338                                 break;
 339                         }
 340                 }
 341                 if (ia != NULL)
 342                         ifa_ref(&ia->ia_ifa);
 343                 IF_ADDR_UNLOCK(ifp);
 344         }
 345         if (ia == NULL)
 346                 iaIsFirst = 1;
 347
 348         error = 0;
 349         switch (cmd) {
 350         case SIOCAIFADDR:
 351         case SIOCDIFADDR:
 352                 if (ifra->ifra_addr.sin_family == AF_INET) {
 353                         struct in_ifaddr *oia;
 354
 355                         IN_IFADDR_RLOCK();
 356                         for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
 357                                 if (ia->ia_ifp == ifp  &&
 358                                     ia->ia_addr.sin_addr.s_addr ==
 359                                     ifra->ifra_addr.sin_addr.s_addr)
 360                                         break;
 361                         }
 362                         if (ia != NULL && ia != oia)
 363                                 ifa_ref(&ia->ia_ifa);
 364                         if (oia != NULL && ia != oia)
 365                                 ifa_free(&oia->ia_ifa);
 366                         IN_IFADDR_RUNLOCK();
 367                         if ((ifp->if_flags & IFF_POINTOPOINT)
 368                             && (cmd == SIOCAIFADDR)
 369                             && (ifra->ifra_dstaddr.sin_addr.s_addr
 370                                 == INADDR_ANY)) {
 371                                 error = EDESTADDRREQ;
 372                                 goto out;
 373                         }
 374                 }
 375                 if (cmd == SIOCDIFADDR && ia == NULL) {
 376                         error = EADDRNOTAVAIL;
 377                         goto out;
 378                 }
 379                 /* FALLTHROUGH */
 380         case SIOCSIFADDR:
 381         case SIOCSIFNETMASK:
 382         case SIOCSIFDSTADDR:
 383                 if (ia == NULL) {
 384                         ia = (struct in_ifaddr *)
 385                                 malloc(sizeof *ia, M_IFADDR, M_NOWAIT |
 386                                     M_ZERO);
 387                         if (ia == NULL) {
 388                                 error = ENOBUFS;
 389                                 goto out;
 390                         }
 391
 392                         ifa = &ia->ia_ifa;
 393                         ifa_init(ifa);
 394                         ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
 395                         ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
 396                         ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
 397
 398                         ia->ia_sockmask.sin_len = 8;
 399                         ia->ia_sockmask.sin_family = AF_INET;
 400                         if (ifp->if_flags & IFF_BROADCAST) {
 401                                 ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
 402                                 ia->ia_broadaddr.sin_family = AF_INET;
 403                         }
 404                         ia->ia_ifp = ifp;
 405
 406                         ifa_ref(ifa);                   /* if_addrhead */
 407                         IF_ADDR_LOCK(ifp);
 408                         TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
 409                         IF_ADDR_UNLOCK(ifp);
 410                         ifa_ref(ifa);                   /* in_ifaddrhead */
 411                         IN_IFADDR_WLOCK();
 412                         TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
 413                         IN_IFADDR_WUNLOCK();
 414                         iaIsNew = 1;
 415                 }
 416                 break;
 417
 418         case SIOCSIFBRDADDR:
 419         case SIOCGIFADDR:
 420         case SIOCGIFNETMASK:
 421         case SIOCGIFDSTADDR:
 422         case SIOCGIFBRDADDR:
 423                 if (ia == NULL) {
 424                         error = EADDRNOTAVAIL;
 425                         goto out;
 426                 }
 427                 break;
 428         }
 429
 430         /*
 431          * Most paths in this switch return directly or via out.  Only paths
 432          * that remove the address break in order to hit common removal code.
 433          */
 434         switch (cmd) {
 435         case SIOCGIFADDR:
 436                 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
 437                 goto out;
 438
 439         case SIOCGIFBRDADDR:
 440                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 441                         error = EINVAL;
 442                         goto out;
 443                 }
 444                 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
 445                 goto out;
 446
 447         case SIOCGIFDSTADDR:
 448                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
 449                         error = EINVAL;
 450                         goto out;
 451                 }
 452                 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
 453                 goto out;
 454
 455         case SIOCGIFNETMASK:
 456                 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
 457                 goto out;
 458
 459         case SIOCSIFDSTADDR:
 460                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
 461                         error = EINVAL;
 462                         goto out;
 463                 }
 464                 oldaddr = ia->ia_dstaddr;
 465                 ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
 466                 if (ifp->if_ioctl != NULL) {
 467                         error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
 468                             (caddr_t)ia);
 469                         if (error) {
 470                                 ia->ia_dstaddr = oldaddr;
 471                                 goto out;
 472                         }
 473                 }
 474                 if (ia->ia_flags & IFA_ROUTE) {
 475                         ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
 476                         rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
 477                         ia->ia_ifa.ifa_dstaddr =
 478                                         (struct sockaddr *)&ia->ia_dstaddr;
 479                         rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
 480                 }
 481                 goto out;
 482
 483         case SIOCSIFBRDADDR:
 484                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 485                         error = EINVAL;
 486                         goto out;
 487                 }
 488                 ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
 489                 goto out;
 490
 491         case SIOCSIFADDR:
 492                 error = in_ifinit(ifp, ia,
 493                     (struct sockaddr_in *) &ifr->ifr_addr, 1);
 494                 if (error != 0 && iaIsNew)
 495                         break;
 496                 if (error == 0) {
 497                         ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
 498                         if (iaIsFirst &&
 499                             (ifp->if_flags & IFF_MULTICAST) != 0) {
 500                                 error = in_joingroup(ifp, &allhosts_addr,
 501                                     NULL, &ii->ii_allhosts);
 502                         }
 503                         EVENTHANDLER_INVOKE(ifaddr_event, ifp);
 504                 }
 505                 error = 0;
 506                 goto out;
 507
 508         case SIOCSIFNETMASK:
 509                 ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
 510                 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
 511                 goto out;
 512
 513         case SIOCAIFADDR:
 514                 maskIsNew = 0;
 515                 hostIsNew = 1;
 516                 error = 0;
 517                 if (ia->ia_addr.sin_family == AF_INET) {
 518                         if (ifra->ifra_addr.sin_len == 0) {
 519                                 ifra->ifra_addr = ia->ia_addr;
 520                                 hostIsNew = 0;
 521                         } else if (ifra->ifra_addr.sin_addr.s_addr ==
 522                                                ia->ia_addr.sin_addr.s_addr)
 523                                 hostIsNew = 0;
 524                 }
 525                 if (ifra->ifra_mask.sin_len) {
 526                         /*
 527                          * QL: XXX
 528                          * Need to scrub the prefix here in case
 529                          * the issued command is SIOCAIFADDR with
 530                          * the same address, but with a different
 531                          * prefix length. And if the prefix length
 532                          * is the same as before, then the call is
 533                          * un-necessarily executed here.
 534                          */
 535                         in_ifscrub(ifp, ia, LLE_STATIC);
 536                         ia->ia_sockmask = ifra->ifra_mask;
 537                         ia->ia_sockmask.sin_family = AF_INET;
 538                         ia->ia_subnetmask =
 539                              ntohl(ia->ia_sockmask.sin_addr.s_addr);
 540                         maskIsNew = 1;
 541                 }
 542                 if ((ifp->if_flags & IFF_POINTOPOINT) &&
 543                     (ifra->ifra_dstaddr.sin_family == AF_INET)) {
 544                         in_ifscrub(ifp, ia, LLE_STATIC);
 545                         ia->ia_dstaddr = ifra->ifra_dstaddr;
 546                         maskIsNew  = 1; /* We lie; but the effect's the same */
 547                 }
 548                 if (ifra->ifra_addr.sin_family == AF_INET &&
 549                     (hostIsNew || maskIsNew))
 550                         error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
 551                 if (error != 0 && iaIsNew)
 552                         break;
 553
 554                 if ((ifp->if_flags & IFF_BROADCAST) &&
 555                     (ifra->ifra_broadaddr.sin_family == AF_INET))
 556                         ia->ia_broadaddr = ifra->ifra_broadaddr;
 557                 if (error == 0) {
 558                         ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
 559                         if (iaIsFirst &&
 560                             (ifp->if_flags & IFF_MULTICAST) != 0) {
 561                                 error = in_joingroup(ifp, &allhosts_addr,
 562                                     NULL, &ii->ii_allhosts);
 563                         }
 564                         EVENTHANDLER_INVOKE(ifaddr_event, ifp);
 565                 }
 566                 goto out;
 567
 568         case SIOCDIFADDR:
 569                 /*
 570                  * in_ifscrub kills the interface route.
 571                  */
 572                 in_ifscrub(ifp, ia, LLE_STATIC);
 573
 574                 /*
 575                  * in_ifadown gets rid of all the rest of
 576                  * the routes.  This is not quite the right
 577                  * thing to do, but at least if we are running
 578                  * a routing process they will come back.
 579                  */
 580                 in_ifadown(&ia->ia_ifa, 1);
 581                 EVENTHANDLER_INVOKE(ifaddr_event, ifp);
 582                 error = 0;
 583                 break;
 584
 585         default:
 586                 panic("in_control: unsupported ioctl");
 587         }
 588
 589         IF_ADDR_LOCK(ifp);
 590         /* Re-check that ia is still part of the list. */
 591         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 592                 if (ifa == &ia->ia_ifa)
 593                         break;
 594         }
 595         if (ifa == NULL) {
 596                 /*
 597                  * If we lost the race with another thread, there is no need to
 598                  * try it again for the next loop as there is no other exit
 599                  * path between here and out.
 600                  */
 601                 IF_ADDR_UNLOCK(ifp);
 602                 error = EADDRNOTAVAIL;
 603                 goto out;
 604         }
 605         TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
 606         IF_ADDR_UNLOCK(ifp);
 607         ifa_free(&ia->ia_ifa);                          /* if_addrhead */
 608
 609         IN_IFADDR_WLOCK();
 610         TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
 611         if (ia->ia_addr.sin_family == AF_INET) {
 612                 struct in_ifaddr *if_ia;
 613
 614                 LIST_REMOVE(ia, ia_hash);
 615                 IN_IFADDR_WUNLOCK();
 616                 /*
 617                  * If this is the last IPv4 address configured on this
 618                  * interface, leave the all-hosts group.
 619                  * No state-change report need be transmitted.
 620                  */
 621                 if_ia = NULL;
 622                 IFP_TO_IA(ifp, if_ia);
 623                 if (if_ia == NULL) {
 624                         ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
 625                         IN_MULTI_LOCK();
 626                         if (ii->ii_allhosts) {
 627                                 (void)in_leavegroup_locked(ii->ii_allhosts,
 628                                     NULL);
 629                                 ii->ii_allhosts = NULL;
 630                         }
 631                         IN_MULTI_UNLOCK();
 632                 } else
 633                         ifa_free(&if_ia->ia_ifa);
 634         } else
 635                 IN_IFADDR_WUNLOCK();
 636         ifa_free(&ia->ia_ifa);                          /* in_ifaddrhead */
 637 out:
 638         if (ia != NULL)
 639                 ifa_free(&ia->ia_ifa);
 640         return (error);
 641 }
 642
 643 /*
 644  * SIOC[GAD]LIFADDR.
 645  *      SIOCGLIFADDR: get first address. (?!?)
 646  *      SIOCGLIFADDR with IFLR_PREFIX:
 647  *              get first address that matches the specified prefix.
 648  *      SIOCALIFADDR: add the specified address.
 649  *      SIOCALIFADDR with IFLR_PREFIX:
 650  *              EINVAL since we can't deduce hostid part of the address.
 651  *      SIOCDLIFADDR: delete the specified address.
 652  *      SIOCDLIFADDR with IFLR_PREFIX:
 653  *              delete the first address that matches the specified prefix.
 654  * return values:
 655  *      EINVAL on invalid parameters
 656  *      EADDRNOTAVAIL on prefix match failed/specified address not found
 657  *      other values may be returned from in_ioctl()
 658  */
 659 static int
 660 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
 661     struct ifnet *ifp, struct thread *td)
 662 {
 663         struct if_laddrreq *iflr = (struct if_laddrreq *)data;
 664         struct ifaddr *ifa;
 665
 666         /* sanity checks */
 667         if (data == NULL || ifp == NULL) {
 668                 panic("invalid argument to in_lifaddr_ioctl");
 669                 /*NOTRECHED*/
 670         }
 671
 672         switch (cmd) {
 673         case SIOCGLIFADDR:
 674                 /* address must be specified on GET with IFLR_PREFIX */
 675                 if ((iflr->flags & IFLR_PREFIX) == 0)
 676                         break;
 677                 /*FALLTHROUGH*/
 678         case SIOCALIFADDR:
 679         case SIOCDLIFADDR:
 680                 /* address must be specified on ADD and DELETE */
 681                 if (iflr->addr.ss_family != AF_INET)
 682                         return (EINVAL);
 683                 if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
 684                         return (EINVAL);
 685                 /* XXX need improvement */
 686                 if (iflr->dstaddr.ss_family
 687                  && iflr->dstaddr.ss_family != AF_INET)
 688                         return (EINVAL);
 689                 if (iflr->dstaddr.ss_family
 690                  && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
 691                         return (EINVAL);
 692                 break;
 693         default: /*shouldn't happen*/
 694                 return (EOPNOTSUPP);
 695         }
 696         if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
 697                 return (EINVAL);
 698
 699         switch (cmd) {
 700         case SIOCALIFADDR:
 701             {
 702                 struct in_aliasreq ifra;
 703
 704                 if (iflr->flags & IFLR_PREFIX)
 705                         return (EINVAL);
 706
 707                 /* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
 708                 bzero(&ifra, sizeof(ifra));
 709                 bcopy(iflr->iflr_name, ifra.ifra_name,
 710                         sizeof(ifra.ifra_name));
 711
 712                 bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
 713
 714                 if (iflr->dstaddr.ss_family) {  /*XXX*/
 715                         bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
 716                                 iflr->dstaddr.ss_len);
 717                 }
 718
 719                 ifra.ifra_mask.sin_family = AF_INET;
 720                 ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
 721                 in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
 722
 723                 return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
 724             }
 725         case SIOCGLIFADDR:
 726         case SIOCDLIFADDR:
 727             {
 728                 struct in_ifaddr *ia;
 729                 struct in_addr mask, candidate, match;
 730                 struct sockaddr_in *sin;
 731
 732                 bzero(&mask, sizeof(mask));
 733                 bzero(&match, sizeof(match));
 734                 if (iflr->flags & IFLR_PREFIX) {
 735                         /* lookup a prefix rather than address. */
 736                         in_len2mask(&mask, iflr->prefixlen);
 737
 738                         sin = (struct sockaddr_in *)&iflr->addr;
 739                         match.s_addr = sin->sin_addr.s_addr;
 740                         match.s_addr &= mask.s_addr;
 741
 742                         /* if you set extra bits, that's wrong */
 743                         if (match.s_addr != sin->sin_addr.s_addr)
 744                                 return (EINVAL);
 745
 746                 } else {
 747                         /* on getting an address, take the 1st match */
 748                         /* on deleting an address, do exact match */
 749                         if (cmd != SIOCGLIFADDR) {
 750                                 in_len2mask(&mask, 32);
 751                                 sin = (struct sockaddr_in *)&iflr->addr;
 752                                 match.s_addr = sin->sin_addr.s_addr;
 753                         }
 754                 }
 755
 756                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 757                         if (ifa->ifa_addr->sa_family != AF_INET6)
 758                                 continue;
 759                         if (match.s_addr == 0)
 760                                 break;
 761                         candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
 762                         candidate.s_addr &= mask.s_addr;
 763                         if (candidate.s_addr == match.s_addr)
 764                                 break;
 765                 }
 766                 if (ifa == NULL)
 767                         return (EADDRNOTAVAIL);
 768                 ia = (struct in_ifaddr *)ifa;
 769
 770                 if (cmd == SIOCGLIFADDR) {
 771                         /* fill in the if_laddrreq structure */
 772                         bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
 773
 774                         if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
 775                                 bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
 776                                         ia->ia_dstaddr.sin_len);
 777                         } else
 778                                 bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
 779
 780                         iflr->prefixlen =
 781                                 in_mask2len(&ia->ia_sockmask.sin_addr);
 782
 783                         iflr->flags = 0;        /*XXX*/
 784
 785                         return (0);
 786                 } else {
 787                         struct in_aliasreq ifra;
 788
 789                         /* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
 790                         bzero(&ifra, sizeof(ifra));
 791                         bcopy(iflr->iflr_name, ifra.ifra_name,
 792                                 sizeof(ifra.ifra_name));
 793
 794                         bcopy(&ia->ia_addr, &ifra.ifra_addr,
 795                                 ia->ia_addr.sin_len);
 796                         if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
 797                                 bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
 798                                         ia->ia_dstaddr.sin_len);
 799                         }
 800                         bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
 801                                 ia->ia_sockmask.sin_len);
 802
 803                         return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
 804                             ifp, td));
 805                 }
 806             }
 807         }
 808
 809         return (EOPNOTSUPP);    /*just for safety*/
 810 }
 811
 812 /*
 813  * Delete any existing route for an interface.
 814  */
 815 void
 816 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia, u_int flags)
 817 {
 818
 819         in_scrubprefix(ia, flags);
 820 }
 821
 822 /*
 823  * Initialize an interface's internet address
 824  * and routing table entry.
 825  */
 826 static int
 827 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
 828     int scrub)
 829 {
 830         register u_long i = ntohl(sin->sin_addr.s_addr);
 831         struct sockaddr_in oldaddr;
 832         int s = splimp(), flags = RTF_UP, error = 0;
 833
 834         oldaddr = ia->ia_addr;
 835         if (oldaddr.sin_family == AF_INET)
 836                 LIST_REMOVE(ia, ia_hash);
 837         ia->ia_addr = *sin;
 838         if (ia->ia_addr.sin_family == AF_INET) {
 839                 IN_IFADDR_WLOCK();
 840                 LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
 841                     ia, ia_hash);
 842                 IN_IFADDR_WUNLOCK();
 843         }
 844         /*
 845          * Give the interface a chance to initialize
 846          * if this is its first address,
 847          * and to validate the address if necessary.
 848          */
 849         if (ifp->if_ioctl != NULL) {
 850                 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
 851                 if (error) {
 852                         splx(s);
 853                         /* LIST_REMOVE(ia, ia_hash) is done in in_control */
 854                         ia->ia_addr = oldaddr;
 855                         IN_IFADDR_WLOCK();
 856                         if (ia->ia_addr.sin_family == AF_INET)
 857                                 LIST_INSERT_HEAD(INADDR_HASH(
 858                                     ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
 859                         else
 860                                 /*
 861                                  * If oldaddr family is not AF_INET (e.g.
 862                                  * interface has been just created) in_control
 863                                  * does not call LIST_REMOVE, and we end up
 864                                  * with bogus ia entries in hash
 865                                  */
 866                                 LIST_REMOVE(ia, ia_hash);
 867                         IN_IFADDR_WUNLOCK();
 868                         return (error);
 869                 }
 870         }
 871         splx(s);
 872         if (scrub) {
 873                 ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
 874                 in_ifscrub(ifp, ia, LLE_STATIC);
 875                 ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
 876         }
 877         /*
 878          * Be compatible with network classes, if netmask isn't supplied,
 879          * guess it based on classes.
 880          */
 881         if (ia->ia_subnetmask == 0) {
 882                 if (IN_CLASSA(i))
 883                         ia->ia_subnetmask = IN_CLASSA_NET;
 884                 else if (IN_CLASSB(i))
 885                         ia->ia_subnetmask = IN_CLASSB_NET;
 886                 else
 887                         ia->ia_subnetmask = IN_CLASSC_NET;
 888                 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
 889         }
 890         ia->ia_subnet = i & ia->ia_subnetmask;
 891         in_socktrim(&ia->ia_sockmask);
 892         /*
 893          * XXX: carp(4) does not have interface route
 894          */
 895         if (ifp->if_type == IFT_CARP)
 896                 return (0);
 897         /*
 898          * Add route for the network.
 899          */
 900         ia->ia_ifa.ifa_metric = ifp->if_metric;
 901         if (ifp->if_flags & IFF_BROADCAST) {
 902                 if (ia->ia_subnetmask == IN_RFC3021_MASK)
 903                         ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
 904                 else
 905                         ia->ia_broadaddr.sin_addr.s_addr =
 906                             htonl(ia->ia_subnet | ~ia->ia_subnetmask);
 907         } else if (ifp->if_flags & IFF_LOOPBACK) {
 908                 ia->ia_dstaddr = ia->ia_addr;
 909                 flags |= RTF_HOST;
 910         } else if (ifp->if_flags & IFF_POINTOPOINT) {
 911                 if (ia->ia_dstaddr.sin_family != AF_INET)
 912                         return (0);
 913                 flags |= RTF_HOST;
 914         }
 915         if ((error = in_addprefix(ia, flags)) != 0)
 916                 return (error);
 917
 918         if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
 919                 return (0);
 920
 921         if (ifp->if_flags & IFF_POINTOPOINT) {
 922                 if (ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
 923                         return (0);
 924         }
 925
 926
 927         /*
 928          * add a loopback route to self
 929          */
 930         if (V_useloopback && !(ifp->if_flags & IFF_LOOPBACK)) {
 931                 struct route ia_ro;
 932
 933                 bzero(&ia_ro, sizeof(ia_ro));
 934                 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr;
 935                 rtalloc_ign_fib(&ia_ro, 0, 0);
 936                 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
 937                     (ia_ro.ro_rt->rt_ifp == V_loif)) {
 938                         RT_LOCK(ia_ro.ro_rt);
 939                         RT_ADDREF(ia_ro.ro_rt);
 940                         RTFREE_LOCKED(ia_ro.ro_rt);
 941                 } else
 942                         error = ifa_add_loopback_route((struct ifaddr *)ia,
 943                                        (struct sockaddr *)&ia->ia_addr);
 944                 if (error == 0)
 945                         ia->ia_flags |= IFA_RTSELF;
 946                 if (ia_ro.ro_rt != NULL)
 947                         RTFREE(ia_ro.ro_rt);
 948         }
 949
 950         return (error);
 951 }
 952
 953 #define rtinitflags(x) \
 954         ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
 955             ? RTF_HOST : 0)
 956
 957 /*
 958  * Generate a routing message when inserting or deleting
 959  * an interface address alias.
 960  */
 961 static void in_addralias_rtmsg(int cmd, struct in_addr *prefix,
 962     struct in_ifaddr *target)
 963 {
 964         struct route pfx_ro;
 965         struct sockaddr_in *pfx_addr;
 966         struct rtentry msg_rt;
 967
 968         /* QL: XXX
 969          * This is a bit questionable because there is no
 970          * additional route entry added/deleted for an address
 971          * alias. Therefore this route report is inaccurate.
 972          */
 973         bzero(&pfx_ro, sizeof(pfx_ro));
 974         pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst);
 975         pfx_addr->sin_len = sizeof(*pfx_addr);
 976         pfx_addr->sin_family = AF_INET;
 977         pfx_addr->sin_addr = *prefix;
 978         rtalloc_ign_fib(&pfx_ro, 0, 0);
 979         if (pfx_ro.ro_rt != NULL) {
 980                 msg_rt = *pfx_ro.ro_rt;
 981
 982                 /* QL: XXX
 983                  * Point the gateway to the new interface
 984                  * address as if a new prefix route entry has
 985                  * been added through the new address alias.
 986                  * All other parts of the rtentry is accurate,
 987                  * e.g., rt_key, rt_mask, rt_ifp etc.
 988                  */
 989                 msg_rt.rt_gateway =
 990                         (struct sockaddr *)&target->ia_addr;
 991                 rt_newaddrmsg(cmd,
 992                               (struct ifaddr *)target,
 993                               0, &msg_rt);
 994                 RTFREE(pfx_ro.ro_rt);
 995         }
 996         return;
 997 }
 998
 999 /*
1000  * Check if we have a route for the given prefix already or add one accordingly.
1001  */
1002 static int
1003 in_addprefix(struct in_ifaddr *target, int flags)
1004 {
1005         struct in_ifaddr *ia;
1006         struct in_addr prefix, mask, p, m;
1007         int error;
1008
1009         if ((flags & RTF_HOST) != 0) {
1010                 prefix = target->ia_dstaddr.sin_addr;
1011                 mask.s_addr = 0;
1012         } else {
1013                 prefix = target->ia_addr.sin_addr;
1014                 mask = target->ia_sockmask.sin_addr;
1015                 prefix.s_addr &= mask.s_addr;
1016         }
1017
1018         IN_IFADDR_RLOCK();
1019         TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1020                 if (rtinitflags(ia)) {
1021                         p = ia->ia_dstaddr.sin_addr;
1022
1023                         if (prefix.s_addr != p.s_addr)
1024                                 continue;
1025                 } else {
1026                         p = ia->ia_addr.sin_addr;
1027                         m = ia->ia_sockmask.sin_addr;
1028                         p.s_addr &= m.s_addr;
1029
1030                         if (prefix.s_addr != p.s_addr ||
1031                             mask.s_addr != m.s_addr)
1032                                 continue;
1033                 }
1034
1035                 /*
1036                  * If we got a matching prefix route inserted by other
1037                  * interface address, we are done here.
1038                  */
1039                 if (ia->ia_flags & IFA_ROUTE) {
1040 #ifdef RADIX_MPATH
1041                         if (ia->ia_addr.sin_addr.s_addr ==
1042                             target->ia_addr.sin_addr.s_addr) {
1043                                 IN_IFADDR_RUNLOCK();
1044                                 return (EEXIST);
1045                         } else
1046                                 break;
1047 #endif
1048                         if (V_sameprefixcarponly &&
1049                             target->ia_ifp->if_type != IFT_CARP &&
1050                             ia->ia_ifp->if_type != IFT_CARP) {
1051                                 IN_IFADDR_RUNLOCK();
1052                                 return (EEXIST);
1053                         } else {
1054                                 in_addralias_rtmsg(RTM_ADD, &prefix, target);
1055                                 IN_IFADDR_RUNLOCK();
1056                                 return (0);
1057                         }
1058                 }
1059         }
1060         IN_IFADDR_RUNLOCK();
1061
1062         /*
1063          * No-one seem to have this prefix route, so we try to insert it.
1064          */
1065         error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
1066         if (!error)
1067                 target->ia_flags |= IFA_ROUTE;
1068         return (error);
1069 }
1070
1071 extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
1072
1073 /*
1074  * If there is no other address in the system that can serve a route to the
1075  * same prefix, remove the route.  Hand over the route to the new address
1076  * otherwise.
1077  */
1078 static int
1079 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1080 {
1081         struct in_ifaddr *ia;
1082         struct in_addr prefix, mask, p;
1083         int error = 0;
1084         struct sockaddr_in prefix0, mask0;
1085
1086         /*
1087          * Remove the loopback route to the interface address.
1088          * The "useloopback" setting is not consulted because if the
1089          * user configures an interface address, turns off this
1090          * setting, and then tries to delete that interface address,
1091          * checking the current setting of "useloopback" would leave
1092          * that interface address loopback route untouched, which
1093          * would be wrong. Therefore the interface address loopback route
1094          * deletion is unconditional.
1095          */
1096         if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1097             !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
1098             (target->ia_flags & IFA_RTSELF)) {
1099                 struct route ia_ro;
1100                 int freeit = 0;
1101
1102                 bzero(&ia_ro, sizeof(ia_ro));
1103                 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr;
1104                 rtalloc_ign_fib(&ia_ro, 0, 0);
1105                 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
1106                     (ia_ro.ro_rt->rt_ifp == V_loif)) {
1107                         RT_LOCK(ia_ro.ro_rt);
1108                         if (ia_ro.ro_rt->rt_refcnt <= 1)
1109                                 freeit = 1;
1110                         else if (flags & LLE_STATIC) {
1111                                 RT_REMREF(ia_ro.ro_rt);
1112                                 target->ia_flags &= ~IFA_RTSELF;
1113                         }
1114                         RTFREE_LOCKED(ia_ro.ro_rt);
1115                 }
1116                 if (freeit && (flags & LLE_STATIC)) {
1117                         error = ifa_del_loopback_route((struct ifaddr *)target,
1118                                        (struct sockaddr *)&target->ia_addr);
1119                         if (error == 0)
1120                                 target->ia_flags &= ~IFA_RTSELF;
1121                 }
1122                 if ((flags & LLE_STATIC) &&
1123                         !(target->ia_ifp->if_flags & IFF_NOARP))
1124                         /* remove arp cache */
1125                         arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1126         }
1127
1128         if (rtinitflags(target))
1129                 prefix = target->ia_dstaddr.sin_addr;
1130         else {
1131                 prefix = target->ia_addr.sin_addr;
1132                 mask = target->ia_sockmask.sin_addr;
1133                 prefix.s_addr &= mask.s_addr;
1134         }
1135
1136         if ((target->ia_flags & IFA_ROUTE) == 0) {
1137                 in_addralias_rtmsg(RTM_DELETE, &prefix, target);
1138                 return (0);
1139         }
1140
1141         IN_IFADDR_RLOCK();
1142         TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1143                 if (rtinitflags(ia))
1144                         p = ia->ia_dstaddr.sin_addr;
1145                 else {
1146                         p = ia->ia_addr.sin_addr;
1147                         p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1148                 }
1149
1150                 if ((prefix.s_addr != p.s_addr) ||
1151                     !(ia->ia_ifp->if_flags & IFF_UP))
1152                         continue;
1153
1154                 /*
1155                  * If we got a matching prefix address, move IFA_ROUTE and
1156                  * the route itself to it.  Make sure that routing daemons
1157                  * get a heads-up.
1158                  *
1159                  * XXX: a special case for carp(4) interface - this should
1160                  *      be more generally specified as an interface that
1161                  *      doesn't support such action.
1162                  */
1163                 if ((ia->ia_flags & IFA_ROUTE) == 0
1164                     && (ia->ia_ifp->if_type != IFT_CARP)) {
1165                         ifa_ref(&ia->ia_ifa);
1166                         IN_IFADDR_RUNLOCK();
1167                         error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1168                             rtinitflags(target));
1169                         if (error == 0)
1170                                 target->ia_flags &= ~IFA_ROUTE;
1171                         else
1172                                 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1173                                         error);
1174                         error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1175                             rtinitflags(ia) | RTF_UP);
1176                         if (error == 0)
1177                                 ia->ia_flags |= IFA_ROUTE;
1178                         else
1179                                 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1180                                         error);
1181                         ifa_free(&ia->ia_ifa);
1182                         return (error);
1183                 }
1184         }
1185         IN_IFADDR_RUNLOCK();
1186
1187         /*
1188          * remove all L2 entries on the given prefix
1189          */
1190         bzero(&prefix0, sizeof(prefix0));
1191         prefix0.sin_len = sizeof(prefix0);
1192         prefix0.sin_family = AF_INET;
1193         prefix0.sin_addr.s_addr = target->ia_subnet;
1194         bzero(&mask0, sizeof(mask0));
1195         mask0.sin_len = sizeof(mask0);
1196         mask0.sin_family = AF_INET;
1197         mask0.sin_addr.s_addr = target->ia_subnetmask;
1198         lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1199                             (struct sockaddr *)&mask0, flags);
1200
1201         /*
1202          * As no-one seem to have this prefix, we can remove the route.
1203          */
1204         error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1205         if (error == 0)
1206                 target->ia_flags &= ~IFA_ROUTE;
1207         else
1208                 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1209         return (error);
1210 }
1211
1212 #undef rtinitflags
1213
1214 /*
1215  * Return 1 if the address might be a local broadcast address.
1216  */
1217 int
1218 in_broadcast(struct in_addr in, struct ifnet *ifp)
1219 {
1220         register struct ifaddr *ifa;
1221         u_long t;
1222
1223         if (in.s_addr == INADDR_BROADCAST ||
1224             in.s_addr == INADDR_ANY)
1225                 return (1);
1226         if ((ifp->if_flags & IFF_BROADCAST) == 0)
1227                 return (0);
1228         t = ntohl(in.s_addr);
1229         /*
1230          * Look through the list of addresses for a match
1231          * with a broadcast address.
1232          */
1233 #define ia ((struct in_ifaddr *)ifa)
1234         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1235                 if (ifa->ifa_addr->sa_family == AF_INET &&
1236                     (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1237                      /*
1238                       * Check for old-style (host 0) broadcast, but
1239                       * taking into account that RFC 3021 obsoletes it.
1240                       */
1241                      (ia->ia_subnetmask != IN_RFC3021_MASK &&
1242                      t == ia->ia_subnet)) &&
1243                      /*
1244                       * Check for an all one subnetmask. These
1245                       * only exist when an interface gets a secondary
1246                       * address.
1247                       */
1248                      ia->ia_subnetmask != (u_long)0xffffffff)
1249                             return (1);
1250         return (0);
1251 #undef ia
1252 }
1253
1254 /*
1255  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1256  */
1257 void
1258 in_ifdetach(struct ifnet *ifp)
1259 {
1260
1261         in_pcbpurgeif0(&V_ripcbinfo, ifp);
1262         in_pcbpurgeif0(&V_udbinfo, ifp);
1263         in_purgemaddrs(ifp);
1264 }
1265
1266 /*
1267  * Delete all IPv4 multicast address records, and associated link-layer
1268  * multicast address records, associated with ifp.
1269  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1270  * XXX This should not race with ifma_protospec being set during
1271  * a new allocation, if it does, we have bigger problems.
1272  */
1273 static void
1274 in_purgemaddrs(struct ifnet *ifp)
1275 {
1276         LIST_HEAD(,in_multi) purgeinms;
1277         struct in_multi         *inm, *tinm;
1278         struct ifmultiaddr      *ifma;
1279
1280         LIST_INIT(&purgeinms);
1281         IN_MULTI_LOCK();
1282
1283         /*
1284          * Extract list of in_multi associated with the detaching ifp
1285          * which the PF_INET layer is about to release.
1286          * We need to do this as IF_ADDR_LOCK() may be re-acquired
1287          * by code further down.
1288          */
1289         IF_ADDR_LOCK(ifp);
1290         TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1291                 if (ifma->ifma_addr->sa_family != AF_INET ||
1292                     ifma->ifma_protospec == NULL)
1293                         continue;
1294 #if 0
1295                 KASSERT(ifma->ifma_protospec != NULL,
1296                     ("%s: ifma_protospec is NULL", __func__));
1297 #endif
1298                 inm = (struct in_multi *)ifma->ifma_protospec;
1299                 LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1300         }
1301         IF_ADDR_UNLOCK(ifp);
1302
1303         LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1304                 LIST_REMOVE(inm, inm_link);
1305                 inm_release_locked(inm);
1306         }
1307         igmp_ifdetach(ifp);
1308
1309         IN_MULTI_UNLOCK();
1310 }
1311
1312 #include <net/if_dl.h>
1313 #include <netinet/if_ether.h>
1314
1315 struct in_llentry {
1316         struct llentry          base;
1317         struct sockaddr_in      l3_addr4;
1318 };
1319
1320 static struct llentry *
1321 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1322 {
1323         struct in_llentry *lle;
1324
1325         lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1326         if (lle == NULL)                /* NB: caller generates msg */
1327                 return NULL;
1328
1329         callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1330         /*
1331          * For IPv4 this will trigger "arpresolve" to generate
1332          * an ARP request.
1333          */
1334         lle->base.la_expire = time_uptime; /* mark expired */
1335         lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1336         lle->base.lle_refcnt = 1;
1337         LLE_LOCK_INIT(&lle->base);
1338         return &lle->base;
1339 }
1340
1341 /*
1342  * Deletes an address from the address table.
1343  * This function is called by the timer functions
1344  * such as arptimer() and nd6_llinfo_timer(), and
1345  * the caller does the locking.
1346  */
1347 static void
1348 in_lltable_free(struct lltable *llt, struct llentry *lle)
1349 {
1350         LLE_WUNLOCK(lle);
1351         LLE_LOCK_DESTROY(lle);
1352         free(lle, M_LLTABLE);
1353 }
1354
1355
1356 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)       (                       \
1357             (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1358
1359 static void
1360 in_lltable_prefix_free(struct lltable *llt,
1361                        const struct sockaddr *prefix,
1362                        const struct sockaddr *mask,
1363                        u_int flags)
1364 {
1365         const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1366         const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1367         struct llentry *lle, *next;
1368         register int i;
1369         size_t pkts_dropped;
1370
1371         for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
1372                 LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1373
1374                         /*
1375                          * (flags & LLE_STATIC) means deleting all entries
1376                          * including static ARP entries
1377                          */
1378                         if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle),
1379                                                      pfx, msk) &&
1380                             ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) {
1381                                 int canceled;
1382
1383                                 canceled = callout_drain(&lle->la_timer);
1384                                 LLE_WLOCK(lle);
1385                                 if (canceled)
1386                                         LLE_REMREF(lle);
1387                                 pkts_dropped = llentry_free(lle);
1388                                 ARPSTAT_ADD(dropped, pkts_dropped);
1389                         }
1390                 }
1391         }
1392 }
1393
1394
1395 static int
1396 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1397 {
1398         struct rtentry *rt;
1399
1400         KASSERT(l3addr->sa_family == AF_INET,
1401             ("sin_family %d", l3addr->sa_family));
1402
1403         /* XXX rtalloc1 should take a const param */
1404         rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1405
1406         if (rt == NULL)
1407                 return (EINVAL);
1408
1409         /*
1410          * If the gateway for an existing host route matches the target L3
1411          * address, which is a special route inserted by some implementation
1412          * such as MANET, and the interface is of the correct type, then
1413          * allow for ARP to proceed.
1414          */
1415         if (rt->rt_flags & RTF_GATEWAY) {
1416                 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
1417                         rt->rt_ifp->if_type != IFT_ETHER ||
1418                           (rt->rt_ifp->if_flags &
1419                            (IFF_NOARP | IFF_STATICARP)) != 0 ||
1420                           memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
1421                                  sizeof(in_addr_t)) != 0) {
1422                         RTFREE_LOCKED(rt);
1423                         return (EINVAL);
1424                 }
1425         }
1426
1427         /*
1428          * Make sure that at least the destination address is covered
1429          * by the route. This is for handling the case where 2 or more
1430          * interfaces have the same prefix. An incoming packet arrives
1431          * on one interface and the corresponding outgoing packet leaves
1432          * another interface.
1433          */
1434         if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
1435                 const char *sa, *mask, *addr, *lim;
1436                 int len;
1437
1438                 mask = (const char *)rt_mask(rt);
1439                 /*
1440                  * Just being extra cautious to avoid some custom
1441                  * code getting into trouble.
1442                  */
1443                 if (mask == NULL) {
1444                         RTFREE_LOCKED(rt);
1445                         return (EINVAL);
1446                 }
1447
1448                 sa = (const char *)rt_key(rt);
1449                 addr = (const char *)l3addr;
1450                 len = ((const struct sockaddr_in *)l3addr)->sin_len;
1451                 lim = addr + len;
1452
1453                 for ( ; addr < lim; sa++, mask++, addr++) {
1454                         if ((*sa ^ *addr) & *mask) {
1455 #ifdef DIAGNOSTIC
1456                                 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1457                                     inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1458 #endif
1459                                 RTFREE_LOCKED(rt);
1460                                 return (EINVAL);
1461                         }
1462                 }
1463         }
1464
1465         RTFREE_LOCKED(rt);
1466         return (0);
1467 }
1468
1469 /*
1470  * Return NULL if not found or marked for deletion.
1471  * If found return lle read locked.
1472  */
1473 static struct llentry *
1474 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1475 {
1476         const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1477         struct ifnet *ifp = llt->llt_ifp;
1478         struct llentry *lle;
1479         struct llentries *lleh;
1480         u_int hashkey;
1481
1482         IF_AFDATA_LOCK_ASSERT(ifp);
1483         KASSERT(l3addr->sa_family == AF_INET,
1484             ("sin_family %d", l3addr->sa_family));
1485
1486         hashkey = sin->sin_addr.s_addr;
1487         lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1488         LIST_FOREACH(lle, lleh, lle_next) {
1489                 struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1490                 if (lle->la_flags & LLE_DELETED)
1491                         continue;
1492                 if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1493                         break;
1494         }
1495         if (lle == NULL) {
1496 #ifdef DIAGNOSTIC
1497                 if (flags & LLE_DELETE)
1498                         log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1499 #endif
1500                 if (!(flags & LLE_CREATE))
1501                         return (NULL);
1502                 /*
1503                  * A route that covers the given address must have
1504                  * been installed 1st because we are doing a resolution,
1505                  * verify this.
1506                  */
1507                 if (!(flags & LLE_IFADDR) &&
1508                     in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1509                         goto done;
1510
1511                 lle = in_lltable_new(l3addr, flags);
1512                 if (lle == NULL) {
1513                         log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1514                         goto done;
1515                 }
1516                 lle->la_flags = flags & ~LLE_CREATE;
1517                 if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1518                         bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1519                         lle->la_flags |= (LLE_VALID | LLE_STATIC);
1520                 }
1521
1522                 lle->lle_tbl  = llt;
1523                 lle->lle_head = lleh;
1524                 LIST_INSERT_HEAD(lleh, lle, lle_next);
1525         } else if (flags & LLE_DELETE) {
1526                 if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1527                         LLE_WLOCK(lle);
1528                         lle->la_flags = LLE_DELETED;
1529                         EVENTHANDLER_INVOKE(arp_update_event, lle);
1530                         LLE_WUNLOCK(lle);
1531 #ifdef DIAGNOSTIC
1532                         log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1533 #endif
1534                 }
1535                 lle = (void *)-1;
1536
1537         }
1538         if (LLE_IS_VALID(lle)) {
1539                 if (flags & LLE_EXCLUSIVE)
1540                         LLE_WLOCK(lle);
1541                 else
1542                         LLE_RLOCK(lle);
1543         }
1544 done:
1545         return (lle);
1546 }
1547
1548 static int
1549 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1550 {
1551 #define SIN(lle)        ((struct sockaddr_in *) L3_ADDR(lle))
1552         struct ifnet *ifp = llt->llt_ifp;
1553         struct llentry *lle;
1554         /* XXX stack use */
1555         struct {
1556                 struct rt_msghdr        rtm;
1557                 struct sockaddr_inarp   sin;
1558                 struct sockaddr_dl      sdl;
1559         } arpc;
1560         int error, i;
1561
1562         LLTABLE_LOCK_ASSERT();
1563
1564         error = 0;
1565         for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1566                 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1567                         struct sockaddr_dl *sdl;
1568
1569                         /* skip deleted entries */
1570                         if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1571                                 continue;
1572                         /* Skip if jailed and not a valid IP of the prison. */
1573                         if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1574                                 continue;
1575                         /*
1576                          * produce a msg made of:
1577                          *  struct rt_msghdr;
1578                          *  struct sockaddr_inarp; (IPv4)
1579                          *  struct sockaddr_dl;
1580                          */
1581                         bzero(&arpc, sizeof(arpc));
1582                         arpc.rtm.rtm_msglen = sizeof(arpc);
1583                         arpc.rtm.rtm_version = RTM_VERSION;
1584                         arpc.rtm.rtm_type = RTM_GET;
1585                         arpc.rtm.rtm_flags = RTF_UP;
1586                         arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1587                         arpc.sin.sin_family = AF_INET;
1588                         arpc.sin.sin_len = sizeof(arpc.sin);
1589                         arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1590
1591                         /* publish */
1592                         if (lle->la_flags & LLE_PUB) {
1593                                 arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1594                                 /* proxy only */
1595                                 if (lle->la_flags & LLE_PROXY)
1596                                         arpc.sin.sin_other = SIN_PROXY;
1597                         }
1598
1599                         sdl = &arpc.sdl;
1600                         sdl->sdl_family = AF_LINK;
1601                         sdl->sdl_len = sizeof(*sdl);
1602                         sdl->sdl_index = ifp->if_index;
1603                         sdl->sdl_type = ifp->if_type;
1604                         if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1605                                 sdl->sdl_alen = ifp->if_addrlen;
1606                                 bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1607                         } else {
1608                                 sdl->sdl_alen = 0;
1609                                 bzero(LLADDR(sdl), ifp->if_addrlen);
1610                         }
1611
1612                         arpc.rtm.rtm_rmx.rmx_expire =
1613                             lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1614                         arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1615                         if (lle->la_flags & LLE_STATIC)
1616                                 arpc.rtm.rtm_flags |= RTF_STATIC;
1617                         arpc.rtm.rtm_index = ifp->if_index;
1618                         error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1619                         if (error)
1620                                 break;
1621                 }
1622         }
1623         return error;
1624 #undef SIN
1625 }
1626
1627 void *
1628 in_domifattach(struct ifnet *ifp)
1629 {
1630         struct in_ifinfo *ii;
1631         struct lltable *llt;
1632
1633         ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1634
1635         llt = lltable_init(ifp, AF_INET);
1636         if (llt != NULL) {
1637                 llt->llt_free = in_lltable_free;
1638                 llt->llt_prefix_free = in_lltable_prefix_free;
1639                 llt->llt_lookup = in_lltable_lookup;
1640                 llt->llt_dump = in_lltable_dump;
1641         }
1642         ii->ii_llt = llt;
1643
1644         ii->ii_igmp = igmp_domifattach(ifp);
1645
1646         return ii;
1647 }
1648
1649 void
1650 in_domifdetach(struct ifnet *ifp, void *aux)
1651 {
1652         struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1653
1654         igmp_domifdetach(ifp);
1655         lltable_free(ii->ii_llt);
1656         free(ii, M_IFADDR);
1657 }