sys/netinet/in.c

   1 /*-
   2  * Copyright (c) 1982, 1986, 1991, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  * Copyright (C) 2001 WIDE Project.  All rights reserved.
   5  *
   6  * Redistribution and use in source and binary forms, with or without
   7  * modification, are permitted provided that the following conditions
   8  * are met:
   9  * 1. Redistributions of source code must retain the above copyright
  10  *    notice, this list of conditions and the following disclaimer.
  11  * 2. Redistributions in binary form must reproduce the above copyright
  12  *    notice, this list of conditions and the following disclaimer in the
  13  *    documentation and/or other materials provided with the distribution.
  14  * 4. Neither the name of the University nor the names of its contributors
  15  *    may be used to endorse or promote products derived from this software
  16  *    without specific prior written permission.
  17  *
  18  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  28  * SUCH DAMAGE.
  29  *
  30  *      @(#)in.c        8.4 (Berkeley) 1/9/95
  31  */
  32
  33 #include <sys/cdefs.h>
  34 __FBSDID("$FreeBSD$");
  35
  36 #include "opt_mpath.h"
  37
  38 #include <sys/param.h>
  39 #include <sys/systm.h>
  40 #include <sys/sockio.h>
  41 #include <sys/malloc.h>
  42 #include <sys/priv.h>
  43 #include <sys/socket.h>
  44 #include <sys/jail.h>
  45 #include <sys/kernel.h>
  46 #include <sys/proc.h>
  47 #include <sys/sysctl.h>
  48 #include <sys/syslog.h>
  49
  50 #include <net/if.h>
  51 #include <net/if_var.h>
  52 #include <net/if_arp.h>
  53 #include <net/if_dl.h>
  54 #include <net/if_llatbl.h>
  55 #include <net/if_types.h>
  56 #include <net/route.h>
  57 #include <net/vnet.h>
  58
  59 #include <netinet/if_ether.h>
  60 #include <netinet/in.h>
  61 #include <netinet/in_var.h>
  62 #include <netinet/in_pcb.h>
  63 #include <netinet/ip_var.h>
  64 #include <netinet/ip_carp.h>
  65 #include <netinet/igmp_var.h>
  66 #include <netinet/udp.h>
  67 #include <netinet/udp_var.h>
  68
  69 static int in_mask2len(struct in_addr *);
  70 static void in_len2mask(struct in_addr *, int);
  71 static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t,
  72         struct ifnet *, struct thread *);
  73
  74 static void     in_socktrim(struct sockaddr_in *);
  75 static int      in_ifinit(struct ifnet *, struct in_ifaddr *,
  76                     struct sockaddr_in *, int, int, int);
  77 static void     in_purgemaddrs(struct ifnet *);
  78
  79 static VNET_DEFINE(int, nosameprefix);
  80 #define V_nosameprefix                  VNET(nosameprefix)
  81 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_RW,
  82         &VNET_NAME(nosameprefix), 0,
  83         "Refuse to create same prefixes on different interfaces");
  84
  85 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
  86 #define V_ripcbinfo                     VNET(ripcbinfo)
  87
  88 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 SIOCGIFADDR:
 238         case SIOCGIFBRDADDR:
 239         case SIOCGIFDSTADDR:
 240         case SIOCGIFNETMASK:
 241         case SIOCDIFADDR:
 242                 break;
 243         case SIOCAIFADDR:
 244                 /*
 245                  * ifra_addr must be present and be of INET family.
 246                  * ifra_broadaddr and ifra_mask are optional.
 247                  */
 248                 if (ifra->ifra_addr.sin_len != sizeof(struct sockaddr_in) ||
 249                     ifra->ifra_addr.sin_family != AF_INET)
 250                         return (EINVAL);
 251                 if (ifra->ifra_broadaddr.sin_len != 0 &&
 252                     (ifra->ifra_broadaddr.sin_len !=
 253                     sizeof(struct sockaddr_in) ||
 254                     ifra->ifra_broadaddr.sin_family != AF_INET))
 255                         return (EINVAL);
 256 #if 0
 257                 /*
 258                  * ifconfig(8) historically doesn't set af_family for mask
 259                  * for unknown reason.
 260                  */
 261                 if (ifra->ifra_mask.sin_len != 0 &&
 262                     (ifra->ifra_mask.sin_len != sizeof(struct sockaddr_in) ||
 263                     ifra->ifra_mask.sin_family != AF_INET))
 264                         return (EINVAL);
 265 #endif
 266                 break;
 267         case SIOCSIFADDR:
 268         case SIOCSIFBRDADDR:
 269         case SIOCSIFDSTADDR:
 270         case SIOCSIFNETMASK:
 271                 if (ifr->ifr_addr.sa_family != AF_INET ||
 272                     ifr->ifr_addr.sa_len != sizeof(struct sockaddr_in))
 273                         return (EINVAL);
 274                 break;
 275
 276         case SIOCALIFADDR:
 277                 if (td != NULL) {
 278                         error = priv_check(td, PRIV_NET_ADDIFADDR);
 279                         if (error)
 280                                 return (error);
 281                 }
 282                 if (ifp == NULL)
 283                         return (EINVAL);
 284                 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
 285
 286         case SIOCDLIFADDR:
 287                 if (td != NULL) {
 288                         error = priv_check(td, PRIV_NET_DELIFADDR);
 289                         if (error)
 290                                 return (error);
 291                 }
 292                 if (ifp == NULL)
 293                         return (EINVAL);
 294                 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
 295
 296         case SIOCGLIFADDR:
 297                 if (ifp == NULL)
 298                         return (EINVAL);
 299                 return in_lifaddr_ioctl(so, cmd, data, ifp, td);
 300
 301         default:
 302                 if (ifp == NULL || ifp->if_ioctl == NULL)
 303                         return (EOPNOTSUPP);
 304                 return ((*ifp->if_ioctl)(ifp, cmd, data));
 305         }
 306
 307         if (ifp == NULL)
 308                 return (EADDRNOTAVAIL);
 309
 310         /*
 311          * Security checks before we get involved in any work.
 312          */
 313         switch (cmd) {
 314         case SIOCAIFADDR:
 315         case SIOCSIFADDR:
 316         case SIOCSIFBRDADDR:
 317         case SIOCSIFNETMASK:
 318         case SIOCSIFDSTADDR:
 319                 if (td != NULL) {
 320                         error = priv_check(td, PRIV_NET_ADDIFADDR);
 321                         if (error)
 322                                 return (error);
 323                 }
 324                 break;
 325
 326         case SIOCDIFADDR:
 327                 if (td != NULL) {
 328                         error = priv_check(td, PRIV_NET_DELIFADDR);
 329                         if (error)
 330                                 return (error);
 331                 }
 332                 break;
 333         }
 334
 335         /*
 336          * Find address for this interface, if it exists.
 337          *
 338          * If an alias address was specified, find that one instead of the
 339          * first one on the interface, if possible.
 340          */
 341         dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
 342         IN_IFADDR_RLOCK();
 343         LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) {
 344                 if (iap->ia_ifp == ifp &&
 345                     iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
 346                         if (td == NULL || prison_check_ip4(td->td_ucred,
 347                             &dst) == 0)
 348                                 ia = iap;
 349                         break;
 350                 }
 351         }
 352         if (ia != NULL)
 353                 ifa_ref(&ia->ia_ifa);
 354         IN_IFADDR_RUNLOCK();
 355         if (ia == NULL) {
 356                 IF_ADDR_LOCK(ifp);
 357                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 358                         iap = ifatoia(ifa);
 359                         if (iap->ia_addr.sin_family == AF_INET) {
 360                                 if (td != NULL &&
 361                                     prison_check_ip4(td->td_ucred,
 362                                     &iap->ia_addr.sin_addr) != 0)
 363                                         continue;
 364                                 ia = iap;
 365                                 break;
 366                         }
 367                 }
 368                 if (ia != NULL)
 369                         ifa_ref(&ia->ia_ifa);
 370                 IF_ADDR_UNLOCK(ifp);
 371         }
 372         if (ia == NULL)
 373                 iaIsFirst = 1;
 374
 375         error = 0;
 376         switch (cmd) {
 377         case SIOCAIFADDR:
 378         case SIOCDIFADDR:
 379                 if (ifra->ifra_addr.sin_family == AF_INET) {
 380                         struct in_ifaddr *oia;
 381
 382                         IN_IFADDR_RLOCK();
 383                         for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
 384                                 if (ia->ia_ifp == ifp  &&
 385                                     ia->ia_addr.sin_addr.s_addr ==
 386                                     ifra->ifra_addr.sin_addr.s_addr)
 387                                         break;
 388                         }
 389                         if (ia != NULL && ia != oia)
 390                                 ifa_ref(&ia->ia_ifa);
 391                         if (oia != NULL && ia != oia)
 392                                 ifa_free(&oia->ia_ifa);
 393                         IN_IFADDR_RUNLOCK();
 394                         if ((ifp->if_flags & IFF_POINTOPOINT)
 395                             && (cmd == SIOCAIFADDR)
 396                             && (ifra->ifra_dstaddr.sin_addr.s_addr
 397                                 == INADDR_ANY)) {
 398                                 error = EDESTADDRREQ;
 399                                 goto out;
 400                         }
 401                 }
 402                 if (cmd == SIOCDIFADDR && ia == NULL) {
 403                         error = EADDRNOTAVAIL;
 404                         goto out;
 405                 }
 406                 /* FALLTHROUGH */
 407         case SIOCSIFADDR:
 408         case SIOCSIFNETMASK:
 409         case SIOCSIFDSTADDR:
 410                 if (ia == NULL) {
 411                         ia = (struct in_ifaddr *)
 412                                 malloc(sizeof *ia, M_IFADDR, M_NOWAIT |
 413                                     M_ZERO);
 414                         if (ia == NULL) {
 415                                 error = ENOBUFS;
 416                                 goto out;
 417                         }
 418
 419                         ifa = &ia->ia_ifa;
 420                         ifa_init(ifa);
 421                         ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
 422                         ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
 423                         ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
 424
 425                         ia->ia_sockmask.sin_len = 8;
 426                         ia->ia_sockmask.sin_family = AF_INET;
 427                         if (ifp->if_flags & IFF_BROADCAST) {
 428                                 ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
 429                                 ia->ia_broadaddr.sin_family = AF_INET;
 430                         }
 431                         ia->ia_ifp = ifp;
 432
 433                         ifa_ref(ifa);                   /* if_addrhead */
 434                         IF_ADDR_LOCK(ifp);
 435                         TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
 436                         IF_ADDR_UNLOCK(ifp);
 437                         ifa_ref(ifa);                   /* in_ifaddrhead */
 438                         IN_IFADDR_WLOCK();
 439                         TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
 440                         IN_IFADDR_WUNLOCK();
 441                         iaIsNew = 1;
 442                 }
 443                 break;
 444
 445         case SIOCSIFBRDADDR:
 446         case SIOCGIFADDR:
 447         case SIOCGIFNETMASK:
 448         case SIOCGIFDSTADDR:
 449         case SIOCGIFBRDADDR:
 450                 if (ia == NULL) {
 451                         error = EADDRNOTAVAIL;
 452                         goto out;
 453                 }
 454                 break;
 455         }
 456
 457         /*
 458          * Most paths in this switch return directly or via out.  Only paths
 459          * that remove the address break in order to hit common removal code.
 460          */
 461         switch (cmd) {
 462         case SIOCGIFADDR:
 463                 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
 464                 goto out;
 465
 466         case SIOCGIFBRDADDR:
 467                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 468                         error = EINVAL;
 469                         goto out;
 470                 }
 471                 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
 472                 goto out;
 473
 474         case SIOCGIFDSTADDR:
 475                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
 476                         error = EINVAL;
 477                         goto out;
 478                 }
 479                 *((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
 480                 goto out;
 481
 482         case SIOCGIFNETMASK:
 483                 *((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
 484                 goto out;
 485
 486         case SIOCSIFDSTADDR:
 487                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
 488                         error = EINVAL;
 489                         goto out;
 490                 }
 491                 oldaddr = ia->ia_dstaddr;
 492                 ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
 493                 if (ifp->if_ioctl != NULL) {
 494                         error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
 495                             (caddr_t)ia);
 496                         if (error) {
 497                                 ia->ia_dstaddr = oldaddr;
 498                                 goto out;
 499                         }
 500                 }
 501                 if (ia->ia_flags & IFA_ROUTE) {
 502                         ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
 503                         rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
 504                         ia->ia_ifa.ifa_dstaddr =
 505                                         (struct sockaddr *)&ia->ia_dstaddr;
 506                         rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
 507                 }
 508                 goto out;
 509
 510         case SIOCSIFBRDADDR:
 511                 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
 512                         error = EINVAL;
 513                         goto out;
 514                 }
 515                 ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
 516                 goto out;
 517
 518         case SIOCSIFADDR:
 519                 error = in_ifinit(ifp, ia,
 520                     (struct sockaddr_in *) &ifr->ifr_addr, 1, 0, 0);
 521                 if (error != 0 && iaIsNew)
 522                         break;
 523                 if (error == 0) {
 524                         ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
 525                         if (iaIsFirst &&
 526                             (ifp->if_flags & IFF_MULTICAST) != 0) {
 527                                 error = in_joingroup(ifp, &allhosts_addr,
 528                                     NULL, &ii->ii_allhosts);
 529                         }
 530                         EVENTHANDLER_INVOKE(ifaddr_event, ifp);
 531                 }
 532                 error = 0;
 533                 goto out;
 534
 535         case SIOCSIFNETMASK:
 536                 ia->ia_sockmask.sin_addr = ((struct sockaddr_in *)
 537                     &ifr->ifr_addr)->sin_addr;
 538                 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
 539                 goto out;
 540
 541         case SIOCAIFADDR:
 542                 maskIsNew = 0;
 543                 hostIsNew = 1;
 544                 error = 0;
 545                 if (ifra->ifra_addr.sin_addr.s_addr ==
 546                             ia->ia_addr.sin_addr.s_addr)
 547                         hostIsNew = 0;
 548                 if (ifra->ifra_mask.sin_len) {
 549                         /*
 550                          * QL: XXX
 551                          * Need to scrub the prefix here in case
 552                          * the issued command is SIOCAIFADDR with
 553                          * the same address, but with a different
 554                          * prefix length. And if the prefix length
 555                          * is the same as before, then the call is
 556                          * un-necessarily executed here.
 557                          */
 558                         in_ifscrub(ifp, ia, LLE_STATIC);
 559                         ia->ia_sockmask = ifra->ifra_mask;
 560                         ia->ia_sockmask.sin_family = AF_INET;
 561                         ia->ia_subnetmask =
 562                              ntohl(ia->ia_sockmask.sin_addr.s_addr);
 563                         maskIsNew = 1;
 564                 }
 565                 if ((ifp->if_flags & IFF_POINTOPOINT) &&
 566                     (ifra->ifra_dstaddr.sin_family == AF_INET)) {
 567                         in_ifscrub(ifp, ia, LLE_STATIC);
 568                         ia->ia_dstaddr = ifra->ifra_dstaddr;
 569                         maskIsNew  = 1; /* We lie; but the effect's the same */
 570                 }
 571                 if (hostIsNew || maskIsNew)
 572                         error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0,
 573                             maskIsNew, ifra->ifra_vhid);
 574                 if (error != 0 && iaIsNew)
 575                         break;
 576
 577                 if ((ifp->if_flags & IFF_BROADCAST) &&
 578                     ifra->ifra_broadaddr.sin_len)
 579                         ia->ia_broadaddr = ifra->ifra_broadaddr;
 580                 if (error == 0) {
 581                         ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
 582                         if (iaIsFirst &&
 583                             (ifp->if_flags & IFF_MULTICAST) != 0) {
 584                                 error = in_joingroup(ifp, &allhosts_addr,
 585                                     NULL, &ii->ii_allhosts);
 586                         }
 587                         EVENTHANDLER_INVOKE(ifaddr_event, ifp);
 588                 }
 589                 goto out;
 590
 591         case SIOCDIFADDR:
 592                 /*
 593                  * in_ifscrub kills the interface route.
 594                  */
 595                 in_ifscrub(ifp, ia, LLE_STATIC);
 596
 597                 /*
 598                  * in_ifadown gets rid of all the rest of
 599                  * the routes.  This is not quite the right
 600                  * thing to do, but at least if we are running
 601                  * a routing process they will come back.
 602                  */
 603                 in_ifadown(&ia->ia_ifa, 1);
 604                 EVENTHANDLER_INVOKE(ifaddr_event, ifp);
 605                 error = 0;
 606                 break;
 607
 608         default:
 609                 panic("in_control: unsupported ioctl");
 610         }
 611
 612         if (ia->ia_ifa.ifa_carp)
 613                 (*carp_detach_p)(&ia->ia_ifa);
 614
 615         IF_ADDR_LOCK(ifp);
 616         /* Re-check that ia is still part of the list. */
 617         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 618                 if (ifa == &ia->ia_ifa)
 619                         break;
 620         }
 621         if (ifa == NULL) {
 622                 /*
 623                  * If we lost the race with another thread, there is no need to
 624                  * try it again for the next loop as there is no other exit
 625                  * path between here and out.
 626                  */
 627                 IF_ADDR_UNLOCK(ifp);
 628                 error = EADDRNOTAVAIL;
 629                 goto out;
 630         }
 631         TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
 632         IF_ADDR_UNLOCK(ifp);
 633         ifa_free(&ia->ia_ifa);                          /* if_addrhead */
 634
 635         IN_IFADDR_WLOCK();
 636         TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
 637
 638         LIST_REMOVE(ia, ia_hash);
 639         IN_IFADDR_WUNLOCK();
 640         /*
 641          * If this is the last IPv4 address configured on this
 642          * interface, leave the all-hosts group.
 643          * No state-change report need be transmitted.
 644          */
 645         IFP_TO_IA(ifp, iap);
 646         if (iap == NULL) {
 647                 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
 648                 IN_MULTI_LOCK();
 649                 if (ii->ii_allhosts) {
 650                         (void)in_leavegroup_locked(ii->ii_allhosts, NULL);
 651                         ii->ii_allhosts = NULL;
 652                 }
 653                 IN_MULTI_UNLOCK();
 654         } else
 655                 ifa_free(&iap->ia_ifa);
 656
 657         ifa_free(&ia->ia_ifa);                          /* in_ifaddrhead */
 658 out:
 659         if (ia != NULL)
 660                 ifa_free(&ia->ia_ifa);
 661         return (error);
 662 }
 663
 664 /*
 665  * SIOC[GAD]LIFADDR.
 666  *      SIOCGLIFADDR: get first address. (?!?)
 667  *      SIOCGLIFADDR with IFLR_PREFIX:
 668  *              get first address that matches the specified prefix.
 669  *      SIOCALIFADDR: add the specified address.
 670  *      SIOCALIFADDR with IFLR_PREFIX:
 671  *              EINVAL since we can't deduce hostid part of the address.
 672  *      SIOCDLIFADDR: delete the specified address.
 673  *      SIOCDLIFADDR with IFLR_PREFIX:
 674  *              delete the first address that matches the specified prefix.
 675  * return values:
 676  *      EINVAL on invalid parameters
 677  *      EADDRNOTAVAIL on prefix match failed/specified address not found
 678  *      other values may be returned from in_ioctl()
 679  */
 680 static int
 681 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
 682     struct ifnet *ifp, struct thread *td)
 683 {
 684         struct if_laddrreq *iflr = (struct if_laddrreq *)data;
 685         struct ifaddr *ifa;
 686
 687         /* sanity checks */
 688         if (data == NULL || ifp == NULL) {
 689                 panic("invalid argument to in_lifaddr_ioctl");
 690                 /*NOTRECHED*/
 691         }
 692
 693         switch (cmd) {
 694         case SIOCGLIFADDR:
 695                 /* address must be specified on GET with IFLR_PREFIX */
 696                 if ((iflr->flags & IFLR_PREFIX) == 0)
 697                         break;
 698                 /*FALLTHROUGH*/
 699         case SIOCALIFADDR:
 700         case SIOCDLIFADDR:
 701                 /* address must be specified on ADD and DELETE */
 702                 if (iflr->addr.ss_family != AF_INET)
 703                         return (EINVAL);
 704                 if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
 705                         return (EINVAL);
 706                 /* XXX need improvement */
 707                 if (iflr->dstaddr.ss_family
 708                  && iflr->dstaddr.ss_family != AF_INET)
 709                         return (EINVAL);
 710                 if (iflr->dstaddr.ss_family
 711                  && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
 712                         return (EINVAL);
 713                 break;
 714         default: /*shouldn't happen*/
 715                 return (EOPNOTSUPP);
 716         }
 717         if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
 718                 return (EINVAL);
 719
 720         switch (cmd) {
 721         case SIOCALIFADDR:
 722             {
 723                 struct in_aliasreq ifra;
 724
 725                 if (iflr->flags & IFLR_PREFIX)
 726                         return (EINVAL);
 727
 728                 /* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
 729                 bzero(&ifra, sizeof(ifra));
 730                 bcopy(iflr->iflr_name, ifra.ifra_name,
 731                         sizeof(ifra.ifra_name));
 732
 733                 bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
 734
 735                 if (iflr->dstaddr.ss_family) {  /*XXX*/
 736                         bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
 737                                 iflr->dstaddr.ss_len);
 738                 }
 739
 740                 ifra.ifra_mask.sin_family = AF_INET;
 741                 ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
 742                 in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
 743
 744                 return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
 745             }
 746         case SIOCGLIFADDR:
 747         case SIOCDLIFADDR:
 748             {
 749                 struct in_ifaddr *ia;
 750                 struct in_addr mask, candidate, match;
 751                 struct sockaddr_in *sin;
 752
 753                 bzero(&mask, sizeof(mask));
 754                 bzero(&match, sizeof(match));
 755                 if (iflr->flags & IFLR_PREFIX) {
 756                         /* lookup a prefix rather than address. */
 757                         in_len2mask(&mask, iflr->prefixlen);
 758
 759                         sin = (struct sockaddr_in *)&iflr->addr;
 760                         match.s_addr = sin->sin_addr.s_addr;
 761                         match.s_addr &= mask.s_addr;
 762
 763                         /* if you set extra bits, that's wrong */
 764                         if (match.s_addr != sin->sin_addr.s_addr)
 765                                 return (EINVAL);
 766
 767                 } else {
 768                         /* on getting an address, take the 1st match */
 769                         /* on deleting an address, do exact match */
 770                         if (cmd != SIOCGLIFADDR) {
 771                                 in_len2mask(&mask, 32);
 772                                 sin = (struct sockaddr_in *)&iflr->addr;
 773                                 match.s_addr = sin->sin_addr.s_addr;
 774                         }
 775                 }
 776
 777                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 778                         if (ifa->ifa_addr->sa_family != AF_INET6)
 779                                 continue;
 780                         if (match.s_addr == 0)
 781                                 break;
 782                         candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
 783                         candidate.s_addr &= mask.s_addr;
 784                         if (candidate.s_addr == match.s_addr)
 785                                 break;
 786                 }
 787                 if (ifa == NULL)
 788                         return (EADDRNOTAVAIL);
 789                 ia = (struct in_ifaddr *)ifa;
 790
 791                 if (cmd == SIOCGLIFADDR) {
 792                         /* fill in the if_laddrreq structure */
 793                         bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
 794
 795                         if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
 796                                 bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
 797                                         ia->ia_dstaddr.sin_len);
 798                         } else
 799                                 bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
 800
 801                         iflr->prefixlen =
 802                                 in_mask2len(&ia->ia_sockmask.sin_addr);
 803
 804                         iflr->flags = 0;        /*XXX*/
 805
 806                         return (0);
 807                 } else {
 808                         struct in_aliasreq ifra;
 809
 810                         /* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
 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
 824                         return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
 825                             ifp, td));
 826                 }
 827             }
 828         }
 829
 830         return (EOPNOTSUPP);    /*just for safety*/
 831 }
 832
 833 /*
 834  * Delete any existing route for an interface.
 835  */
 836 void
 837 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia, u_int flags)
 838 {
 839
 840         in_scrubprefix(ia, flags);
 841 }
 842
 843 /*
 844  * Initialize an interface's internet address
 845  * and routing table entry.
 846  */
 847 static int
 848 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
 849     int scrub, int masksupplied, int vhid)
 850 {
 851         register u_long i = ntohl(sin->sin_addr.s_addr);
 852         int flags = RTF_UP, error = 0;
 853
 854         if (scrub)
 855                 in_scrubprefix(ia, LLE_STATIC);
 856
 857         IN_IFADDR_WLOCK();
 858         if (ia->ia_addr.sin_family == AF_INET)
 859                 LIST_REMOVE(ia, ia_hash);
 860         ia->ia_addr = *sin;
 861         LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
 862             ia, ia_hash);
 863         IN_IFADDR_WUNLOCK();
 864
 865         if (vhid > 0) {
 866                 if (carp_attach_p != NULL)
 867                         error = (*carp_attach_p)(&ia->ia_ifa, vhid);
 868                 else
 869                         error = EPROTONOSUPPORT;
 870         }
 871         if (error)
 872                 return (error);
 873
 874         /*
 875          * Give the interface a chance to initialize
 876          * if this is its first address,
 877          * and to validate the address if necessary.
 878          */
 879         if (ifp->if_ioctl != NULL &&
 880             (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia)) != 0)
 881                         /* LIST_REMOVE(ia, ia_hash) is done in in_control */
 882                         return (error);
 883
 884         /*
 885          * Be compatible with network classes, if netmask isn't supplied,
 886          * guess it based on classes.
 887          */
 888         if (!masksupplied) {
 889                 if (IN_CLASSA(i))
 890                         ia->ia_subnetmask = IN_CLASSA_NET;
 891                 else if (IN_CLASSB(i))
 892                         ia->ia_subnetmask = IN_CLASSB_NET;
 893                 else
 894                         ia->ia_subnetmask = IN_CLASSC_NET;
 895                 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
 896         }
 897         ia->ia_subnet = i & ia->ia_subnetmask;
 898         in_socktrim(&ia->ia_sockmask);
 899         /*
 900          * Add route for the network.
 901          */
 902         ia->ia_ifa.ifa_metric = ifp->if_metric;
 903         if (ifp->if_flags & IFF_BROADCAST) {
 904                 if (ia->ia_subnetmask == IN_RFC3021_MASK)
 905                         ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
 906                 else
 907                         ia->ia_broadaddr.sin_addr.s_addr =
 908                             htonl(ia->ia_subnet | ~ia->ia_subnetmask);
 909         } else if (ifp->if_flags & IFF_LOOPBACK) {
 910                 ia->ia_dstaddr = ia->ia_addr;
 911                 flags |= RTF_HOST;
 912         } else if (ifp->if_flags & IFF_POINTOPOINT) {
 913                 if (ia->ia_dstaddr.sin_family != AF_INET)
 914                         return (0);
 915                 flags |= RTF_HOST;
 916         }
 917         if (!vhid && (error = in_addprefix(ia, flags)) != 0)
 918                 return (error);
 919
 920         if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
 921                 return (0);
 922
 923         if (ifp->if_flags & IFF_POINTOPOINT &&
 924             ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
 925                         return (0);
 926
 927         /*
 928          * add a loopback route to self
 929          */
 930         if (V_useloopback && !vhid && !(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 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_nosameprefix) {
1049                                 IN_IFADDR_RUNLOCK();
1050                                 return (EEXIST);
1051                         } else {
1052                                 in_addralias_rtmsg(RTM_ADD, &prefix, target);
1053                                 IN_IFADDR_RUNLOCK();
1054                                 return (0);
1055                         }
1056                 }
1057         }
1058         IN_IFADDR_RUNLOCK();
1059
1060         /*
1061          * No-one seem to have this prefix route, so we try to insert it.
1062          */
1063         error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
1064         if (!error)
1065                 target->ia_flags |= IFA_ROUTE;
1066         return (error);
1067 }
1068
1069 /*
1070  * If there is no other address in the system that can serve a route to the
1071  * same prefix, remove the route.  Hand over the route to the new address
1072  * otherwise.
1073  */
1074 int
1075 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1076 {
1077         struct in_ifaddr *ia;
1078         struct in_addr prefix, mask, p, m;
1079         int error = 0;
1080         struct sockaddr_in prefix0, mask0;
1081
1082         /*
1083          * Remove the loopback route to the interface address.
1084          * The "useloopback" setting is not consulted because if the
1085          * user configures an interface address, turns off this
1086          * setting, and then tries to delete that interface address,
1087          * checking the current setting of "useloopback" would leave
1088          * that interface address loopback route untouched, which
1089          * would be wrong. Therefore the interface address loopback route
1090          * deletion is unconditional.
1091          */
1092         if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1093             !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
1094             (target->ia_flags & IFA_RTSELF)) {
1095                 struct route ia_ro;
1096                 int freeit = 0;
1097
1098                 bzero(&ia_ro, sizeof(ia_ro));
1099                 *((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr;
1100                 rtalloc_ign_fib(&ia_ro, 0, 0);
1101                 if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
1102                     (ia_ro.ro_rt->rt_ifp == V_loif)) {
1103                         RT_LOCK(ia_ro.ro_rt);
1104                         if (ia_ro.ro_rt->rt_refcnt <= 1)
1105                                 freeit = 1;
1106                         else if (flags & LLE_STATIC) {
1107                                 RT_REMREF(ia_ro.ro_rt);
1108                                 target->ia_flags &= ~IFA_RTSELF;
1109                         }
1110                         RTFREE_LOCKED(ia_ro.ro_rt);
1111                 }
1112                 if (freeit && (flags & LLE_STATIC)) {
1113                         error = ifa_del_loopback_route((struct ifaddr *)target,
1114                                        (struct sockaddr *)&target->ia_addr);
1115                         if (error == 0)
1116                                 target->ia_flags &= ~IFA_RTSELF;
1117                 }
1118                 if ((flags & LLE_STATIC) &&
1119                         !(target->ia_ifp->if_flags & IFF_NOARP))
1120                         /* remove arp cache */
1121                         arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1122         }
1123
1124         if (rtinitflags(target)) {
1125                 prefix = target->ia_dstaddr.sin_addr;
1126                 mask.s_addr = 0;
1127         } else {
1128                 prefix = target->ia_addr.sin_addr;
1129                 mask = target->ia_sockmask.sin_addr;
1130                 prefix.s_addr &= mask.s_addr;
1131         }
1132
1133         if ((target->ia_flags & IFA_ROUTE) == 0) {
1134                 in_addralias_rtmsg(RTM_DELETE, &prefix, target);
1135                 return (0);
1136         }
1137
1138         IN_IFADDR_RLOCK();
1139         TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1140                 if (rtinitflags(ia)) {
1141                         p = ia->ia_dstaddr.sin_addr;
1142
1143                         if (prefix.s_addr != p.s_addr)
1144                                 continue;
1145                 } else {
1146                         p = ia->ia_addr.sin_addr;
1147                         m = ia->ia_sockmask.sin_addr;
1148                         p.s_addr &= m.s_addr;
1149
1150                         if (prefix.s_addr != p.s_addr ||
1151                             mask.s_addr != m.s_addr)
1152                                 continue;
1153                 }
1154
1155                 if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1156                         continue;
1157
1158                 /*
1159                  * If we got a matching prefix address, move IFA_ROUTE and
1160                  * the route itself to it.  Make sure that routing daemons
1161                  * get a heads-up.
1162                  */
1163                 if ((ia->ia_flags & IFA_ROUTE) == 0) {
1164                         ifa_ref(&ia->ia_ifa);
1165                         IN_IFADDR_RUNLOCK();
1166                         error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1167                             rtinitflags(target));
1168                         if (error == 0)
1169                                 target->ia_flags &= ~IFA_ROUTE;
1170                         else
1171                                 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1172                                         error);
1173                         error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1174                             rtinitflags(ia) | RTF_UP);
1175                         if (error == 0)
1176                                 ia->ia_flags |= IFA_ROUTE;
1177                         else
1178                                 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1179                                         error);
1180                         ifa_free(&ia->ia_ifa);
1181                         return (error);
1182                 }
1183         }
1184         IN_IFADDR_RUNLOCK();
1185
1186         /*
1187          * remove all L2 entries on the given prefix
1188          */
1189         bzero(&prefix0, sizeof(prefix0));
1190         prefix0.sin_len = sizeof(prefix0);
1191         prefix0.sin_family = AF_INET;
1192         prefix0.sin_addr.s_addr = target->ia_subnet;
1193         bzero(&mask0, sizeof(mask0));
1194         mask0.sin_len = sizeof(mask0);
1195         mask0.sin_family = AF_INET;
1196         mask0.sin_addr.s_addr = target->ia_subnetmask;
1197         lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1198                             (struct sockaddr *)&mask0, flags);
1199
1200         /*
1201          * As no-one seem to have this prefix, we can remove the route.
1202          */
1203         error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1204         if (error == 0)
1205                 target->ia_flags &= ~IFA_ROUTE;
1206         else
1207                 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1208         return (error);
1209 }
1210
1211 #undef rtinitflags
1212
1213 /*
1214  * Return 1 if the address might be a local broadcast address.
1215  */
1216 int
1217 in_broadcast(struct in_addr in, struct ifnet *ifp)
1218 {
1219         register struct ifaddr *ifa;
1220         u_long t;
1221
1222         if (in.s_addr == INADDR_BROADCAST ||
1223             in.s_addr == INADDR_ANY)
1224                 return (1);
1225         if ((ifp->if_flags & IFF_BROADCAST) == 0)
1226                 return (0);
1227         t = ntohl(in.s_addr);
1228         /*
1229          * Look through the list of addresses for a match
1230          * with a broadcast address.
1231          */
1232 #define ia ((struct in_ifaddr *)ifa)
1233         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1234                 if (ifa->ifa_addr->sa_family == AF_INET &&
1235                     (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1236                      /*
1237                       * Check for old-style (host 0) broadcast, but
1238                       * taking into account that RFC 3021 obsoletes it.
1239                       */
1240                      (ia->ia_subnetmask != IN_RFC3021_MASK &&
1241                      t == ia->ia_subnet)) &&
1242                      /*
1243                       * Check for an all one subnetmask. These
1244                       * only exist when an interface gets a secondary
1245                       * address.
1246                       */
1247                      ia->ia_subnetmask != (u_long)0xffffffff)
1248                             return (1);
1249         return (0);
1250 #undef ia
1251 }
1252
1253 /*
1254  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1255  */
1256 void
1257 in_ifdetach(struct ifnet *ifp)
1258 {
1259
1260         in_pcbpurgeif0(&V_ripcbinfo, ifp);
1261         in_pcbpurgeif0(&V_udbinfo, ifp);
1262         in_purgemaddrs(ifp);
1263 }
1264
1265 /*
1266  * Delete all IPv4 multicast address records, and associated link-layer
1267  * multicast address records, associated with ifp.
1268  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1269  * XXX This should not race with ifma_protospec being set during
1270  * a new allocation, if it does, we have bigger problems.
1271  */
1272 static void
1273 in_purgemaddrs(struct ifnet *ifp)
1274 {
1275         LIST_HEAD(,in_multi) purgeinms;
1276         struct in_multi         *inm, *tinm;
1277         struct ifmultiaddr      *ifma;
1278
1279         LIST_INIT(&purgeinms);
1280         IN_MULTI_LOCK();
1281
1282         /*
1283          * Extract list of in_multi associated with the detaching ifp
1284          * which the PF_INET layer is about to release.
1285          * We need to do this as IF_ADDR_LOCK() may be re-acquired
1286          * by code further down.
1287          */
1288         IF_ADDR_LOCK(ifp);
1289         TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1290                 if (ifma->ifma_addr->sa_family != AF_INET ||
1291                     ifma->ifma_protospec == NULL)
1292                         continue;
1293 #if 0
1294                 KASSERT(ifma->ifma_protospec != NULL,
1295                     ("%s: ifma_protospec is NULL", __func__));
1296 #endif
1297                 inm = (struct in_multi *)ifma->ifma_protospec;
1298                 LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1299         }
1300         IF_ADDR_UNLOCK(ifp);
1301
1302         LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1303                 LIST_REMOVE(inm, inm_link);
1304                 inm_release_locked(inm);
1305         }
1306         igmp_ifdetach(ifp);
1307
1308         IN_MULTI_UNLOCK();
1309 }
1310
1311 struct in_llentry {
1312         struct llentry          base;
1313         struct sockaddr_in      l3_addr4;
1314 };
1315
1316 static struct llentry *
1317 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1318 {
1319         struct in_llentry *lle;
1320
1321         lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1322         if (lle == NULL)                /* NB: caller generates msg */
1323                 return NULL;
1324
1325         callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1326         /*
1327          * For IPv4 this will trigger "arpresolve" to generate
1328          * an ARP request.
1329          */
1330         lle->base.la_expire = time_uptime; /* mark expired */
1331         lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1332         lle->base.lle_refcnt = 1;
1333         LLE_LOCK_INIT(&lle->base);
1334         return &lle->base;
1335 }
1336
1337 /*
1338  * Deletes an address from the address table.
1339  * This function is called by the timer functions
1340  * such as arptimer() and nd6_llinfo_timer(), and
1341  * the caller does the locking.
1342  */
1343 static void
1344 in_lltable_free(struct lltable *llt, struct llentry *lle)
1345 {
1346         LLE_WUNLOCK(lle);
1347         LLE_LOCK_DESTROY(lle);
1348         free(lle, M_LLTABLE);
1349 }
1350
1351
1352 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)       (                       \
1353             (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1354
1355 static void
1356 in_lltable_prefix_free(struct lltable *llt,
1357                        const struct sockaddr *prefix,
1358                        const struct sockaddr *mask,
1359                        u_int flags)
1360 {
1361         const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1362         const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1363         struct llentry *lle, *next;
1364         register int i;
1365         size_t pkts_dropped;
1366
1367         for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
1368                 LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1369
1370                         /*
1371                          * (flags & LLE_STATIC) means deleting all entries
1372                          * including static ARP entries
1373                          */
1374                         if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle),
1375                                                      pfx, msk) &&
1376                             ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) {
1377                                 int canceled;
1378
1379                                 canceled = callout_drain(&lle->la_timer);
1380                                 LLE_WLOCK(lle);
1381                                 if (canceled)
1382                                         LLE_REMREF(lle);
1383                                 pkts_dropped = llentry_free(lle);
1384                                 ARPSTAT_ADD(dropped, pkts_dropped);
1385                         }
1386                 }
1387         }
1388 }
1389
1390
1391 static int
1392 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1393 {
1394         struct rtentry *rt;
1395
1396         KASSERT(l3addr->sa_family == AF_INET,
1397             ("sin_family %d", l3addr->sa_family));
1398
1399         /* XXX rtalloc1 should take a const param */
1400         rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1401
1402         if (rt == NULL)
1403                 return (EINVAL);
1404
1405         /*
1406          * If the gateway for an existing host route matches the target L3
1407          * address, which is a special route inserted by some implementation
1408          * such as MANET, and the interface is of the correct type, then
1409          * allow for ARP to proceed.
1410          */
1411         if (rt->rt_flags & RTF_GATEWAY) {
1412                 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
1413                         rt->rt_ifp->if_type != IFT_ETHER ||
1414                           (rt->rt_ifp->if_flags &
1415                            (IFF_NOARP | IFF_STATICARP)) != 0 ||
1416                           memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
1417                                  sizeof(in_addr_t)) != 0) {
1418                         RTFREE_LOCKED(rt);
1419                         return (EINVAL);
1420                 }
1421         }
1422
1423         /*
1424          * Make sure that at least the destination address is covered
1425          * by the route. This is for handling the case where 2 or more
1426          * interfaces have the same prefix. An incoming packet arrives
1427          * on one interface and the corresponding outgoing packet leaves
1428          * another interface.
1429          */
1430         if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
1431                 const char *sa, *mask, *addr, *lim;
1432                 int len;
1433
1434                 mask = (const char *)rt_mask(rt);
1435                 /*
1436                  * Just being extra cautious to avoid some custom
1437                  * code getting into trouble.
1438                  */
1439                 if (mask == NULL) {
1440                         RTFREE_LOCKED(rt);
1441                         return (EINVAL);
1442                 }
1443
1444                 sa = (const char *)rt_key(rt);
1445                 addr = (const char *)l3addr;
1446                 len = ((const struct sockaddr_in *)l3addr)->sin_len;
1447                 lim = addr + len;
1448
1449                 for ( ; addr < lim; sa++, mask++, addr++) {
1450                         if ((*sa ^ *addr) & *mask) {
1451 #ifdef DIAGNOSTIC
1452                                 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1453                                     inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1454 #endif
1455                                 RTFREE_LOCKED(rt);
1456                                 return (EINVAL);
1457                         }
1458                 }
1459         }
1460
1461         RTFREE_LOCKED(rt);
1462         return (0);
1463 }
1464
1465 /*
1466  * Return NULL if not found or marked for deletion.
1467  * If found return lle read locked.
1468  */
1469 static struct llentry *
1470 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1471 {
1472         const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1473         struct ifnet *ifp = llt->llt_ifp;
1474         struct llentry *lle;
1475         struct llentries *lleh;
1476         u_int hashkey;
1477
1478         IF_AFDATA_LOCK_ASSERT(ifp);
1479         KASSERT(l3addr->sa_family == AF_INET,
1480             ("sin_family %d", l3addr->sa_family));
1481
1482         hashkey = sin->sin_addr.s_addr;
1483         lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1484         LIST_FOREACH(lle, lleh, lle_next) {
1485                 struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1486                 if (lle->la_flags & LLE_DELETED)
1487                         continue;
1488                 if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1489                         break;
1490         }
1491         if (lle == NULL) {
1492 #ifdef DIAGNOSTIC
1493                 if (flags & LLE_DELETE)
1494                         log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1495 #endif
1496                 if (!(flags & LLE_CREATE))
1497                         return (NULL);
1498                 /*
1499                  * A route that covers the given address must have
1500                  * been installed 1st because we are doing a resolution,
1501                  * verify this.
1502                  */
1503                 if (!(flags & LLE_IFADDR) &&
1504                     in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1505                         goto done;
1506
1507                 lle = in_lltable_new(l3addr, flags);
1508                 if (lle == NULL) {
1509                         log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1510                         goto done;
1511                 }
1512                 lle->la_flags = flags & ~LLE_CREATE;
1513                 if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1514                         bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1515                         lle->la_flags |= (LLE_VALID | LLE_STATIC);
1516                 }
1517
1518                 lle->lle_tbl  = llt;
1519                 lle->lle_head = lleh;
1520                 LIST_INSERT_HEAD(lleh, lle, lle_next);
1521         } else if (flags & LLE_DELETE) {
1522                 if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1523                         LLE_WLOCK(lle);
1524                         lle->la_flags = LLE_DELETED;
1525                         EVENTHANDLER_INVOKE(arp_update_event, lle);
1526                         LLE_WUNLOCK(lle);
1527 #ifdef DIAGNOSTIC
1528                         log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1529 #endif
1530                 }
1531                 lle = (void *)-1;
1532
1533         }
1534         if (LLE_IS_VALID(lle)) {
1535                 if (flags & LLE_EXCLUSIVE)
1536                         LLE_WLOCK(lle);
1537                 else
1538                         LLE_RLOCK(lle);
1539         }
1540 done:
1541         return (lle);
1542 }
1543
1544 static int
1545 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1546 {
1547 #define SIN(lle)        ((struct sockaddr_in *) L3_ADDR(lle))
1548         struct ifnet *ifp = llt->llt_ifp;
1549         struct llentry *lle;
1550         /* XXX stack use */
1551         struct {
1552                 struct rt_msghdr        rtm;
1553                 struct sockaddr_inarp   sin;
1554                 struct sockaddr_dl      sdl;
1555         } arpc;
1556         int error, i;
1557
1558         LLTABLE_LOCK_ASSERT();
1559
1560         error = 0;
1561         for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1562                 LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1563                         struct sockaddr_dl *sdl;
1564
1565                         /* skip deleted entries */
1566                         if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1567                                 continue;
1568                         /* Skip if jailed and not a valid IP of the prison. */
1569                         if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1570                                 continue;
1571                         /*
1572                          * produce a msg made of:
1573                          *  struct rt_msghdr;
1574                          *  struct sockaddr_inarp; (IPv4)
1575                          *  struct sockaddr_dl;
1576                          */
1577                         bzero(&arpc, sizeof(arpc));
1578                         arpc.rtm.rtm_msglen = sizeof(arpc);
1579                         arpc.rtm.rtm_version = RTM_VERSION;
1580                         arpc.rtm.rtm_type = RTM_GET;
1581                         arpc.rtm.rtm_flags = RTF_UP;
1582                         arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1583                         arpc.sin.sin_family = AF_INET;
1584                         arpc.sin.sin_len = sizeof(arpc.sin);
1585                         arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1586
1587                         /* publish */
1588                         if (lle->la_flags & LLE_PUB) {
1589                                 arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1590                                 /* proxy only */
1591                                 if (lle->la_flags & LLE_PROXY)
1592                                         arpc.sin.sin_other = SIN_PROXY;
1593                         }
1594
1595                         sdl = &arpc.sdl;
1596                         sdl->sdl_family = AF_LINK;
1597                         sdl->sdl_len = sizeof(*sdl);
1598                         sdl->sdl_index = ifp->if_index;
1599                         sdl->sdl_type = ifp->if_type;
1600                         if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1601                                 sdl->sdl_alen = ifp->if_addrlen;
1602                                 bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1603                         } else {
1604                                 sdl->sdl_alen = 0;
1605                                 bzero(LLADDR(sdl), ifp->if_addrlen);
1606                         }
1607
1608                         arpc.rtm.rtm_rmx.rmx_expire =
1609                             lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1610                         arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1611                         if (lle->la_flags & LLE_STATIC)
1612                                 arpc.rtm.rtm_flags |= RTF_STATIC;
1613                         arpc.rtm.rtm_index = ifp->if_index;
1614                         error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1615                         if (error)
1616                                 break;
1617                 }
1618         }
1619         return error;
1620 #undef SIN
1621 }
1622
1623 void *
1624 in_domifattach(struct ifnet *ifp)
1625 {
1626         struct in_ifinfo *ii;
1627         struct lltable *llt;
1628
1629         ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1630
1631         llt = lltable_init(ifp, AF_INET);
1632         if (llt != NULL) {
1633                 llt->llt_free = in_lltable_free;
1634                 llt->llt_prefix_free = in_lltable_prefix_free;
1635                 llt->llt_lookup = in_lltable_lookup;
1636                 llt->llt_dump = in_lltable_dump;
1637         }
1638         ii->ii_llt = llt;
1639
1640         ii->ii_igmp = igmp_domifattach(ifp);
1641
1642         return ii;
1643 }
1644
1645 void
1646 in_domifdetach(struct ifnet *ifp, void *aux)
1647 {
1648         struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1649
1650         igmp_domifdetach(ifp);
1651         lltable_free(ii->ii_llt);
1652         free(ii, M_IFADDR);
1653 }