1 /* 2 * Copyright (c) 1980, 1986, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)if.c 8.5 (Berkeley) 1/9/95 30 * $FreeBSD$ 31 */ 32 33 #include "opt_compat.h" 34 #include "opt_inet6.h" 35 #include "opt_inet.h" 36 #include "opt_mac.h" 37 38 #include <sys/param.h> 39 #include <sys/conf.h> 40 #include <sys/mac.h> 41 #include <sys/malloc.h> 42 #include <sys/bus.h> 43 #include <sys/mbuf.h> 44 #include <sys/systm.h> 45 #include <sys/proc.h> 46 #include <sys/socket.h> 47 #include <sys/socketvar.h> 48 #include <sys/protosw.h> 49 #include <sys/kernel.h> 50 #include <sys/sockio.h> 51 #include <sys/syslog.h> 52 #include <sys/sysctl.h> 53 #include <sys/taskqueue.h> 54 #include <sys/domain.h> 55 #include <sys/jail.h> 56 #include <machine/stdarg.h> 57 58 #include <net/if.h> 59 #include <net/if_arp.h> 60 #include <net/if_clone.h> 61 #include <net/if_dl.h> 62 #include <net/if_types.h> 63 #include <net/if_var.h> 64 #include <net/radix.h> 65 #include <net/route.h> 66 67 #if defined(INET) || defined(INET6) 68 /*XXX*/ 69 #include <netinet/in.h> 70 #include <netinet/in_var.h> 71 #ifdef INET6 72 #include <netinet6/in6_var.h> 73 #include <netinet6/in6_ifattach.h> 74 #endif 75 #endif 76 #ifdef INET 77 #include <netinet/if_ether.h> 78 #endif 79 80 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 81 82 static void if_attachdomain(void *); 83 static void if_attachdomain1(struct ifnet *); 84 static int ifconf(u_long, caddr_t); 85 static void if_grow(void); 86 static void if_init(void *); 87 static void if_check(void *); 88 static int if_findindex(struct ifnet *); 89 static void if_qflush(struct ifaltq *); 90 static void if_route(struct ifnet *, int flag, int fam); 91 static void if_slowtimo(void *); 92 static void if_unroute(struct ifnet *, int flag, int fam); 93 static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 94 static int if_rtdel(struct radix_node *, void *); 95 static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 96 static void if_start_deferred(void *context, int pending); 97 #ifdef INET6 98 /* 99 * XXX: declare here to avoid to include many inet6 related files.. 100 * should be more generalized? 101 */ 102 extern void nd6_setmtu(struct ifnet *); 103 #endif 104 105 int if_index = 0; 106 struct ifindex_entry *ifindex_table = NULL; 107 int ifqmaxlen = IFQ_MAXLEN; 108 struct ifnethead ifnet; /* depend on static init XXX */ 109 struct mtx ifnet_lock; 110 111 static int if_indexlim = 8; 112 static struct knlist ifklist; 113 114 static void filt_netdetach(struct knote *kn); 115 static int filt_netdev(struct knote *kn, long hint); 116 117 static struct filterops netdev_filtops = 118 { 1, NULL, filt_netdetach, filt_netdev }; 119 120 /* 121 * System initialization 122 */ 123 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL) 124 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL) 125 126 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 127 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 128 129 static d_open_t netopen; 130 static d_close_t netclose; 131 static d_ioctl_t netioctl; 132 static d_kqfilter_t netkqfilter; 133 134 static struct cdevsw net_cdevsw = { 135 .d_version = D_VERSION, 136 .d_flags = D_NEEDGIANT, 137 .d_open = netopen, 138 .d_close = netclose, 139 .d_ioctl = netioctl, 140 .d_name = "net", 141 .d_kqfilter = netkqfilter, 142 }; 143 144 static int 145 netopen(struct cdev *dev, int flag, int mode, struct thread *td) 146 { 147 return (0); 148 } 149 150 static int 151 netclose(struct cdev *dev, int flags, int fmt, struct thread *td) 152 { 153 return (0); 154 } 155 156 static int 157 netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 158 { 159 struct ifnet *ifp; 160 int error, idx; 161 162 /* only support interface specific ioctls */ 163 if (IOCGROUP(cmd) != 'i') 164 return (EOPNOTSUPP); 165 idx = minor(dev); 166 if (idx == 0) { 167 /* 168 * special network device, not interface. 169 */ 170 if (cmd == SIOCGIFCONF) 171 return (ifconf(cmd, data)); /* XXX remove cmd */ 172 return (EOPNOTSUPP); 173 } 174 175 ifp = ifnet_byindex(idx); 176 if (ifp == NULL) 177 return (ENXIO); 178 179 error = ifhwioctl(cmd, ifp, data, td); 180 if (error == ENOIOCTL) 181 error = EOPNOTSUPP; 182 return (error); 183 } 184 185 static int 186 netkqfilter(struct cdev *dev, struct knote *kn) 187 { 188 struct knlist *klist; 189 struct ifnet *ifp; 190 int idx; 191 192 switch (kn->kn_filter) { 193 case EVFILT_NETDEV: 194 kn->kn_fop = &netdev_filtops; 195 break; 196 default: 197 return (1); 198 } 199 200 idx = minor(dev); 201 if (idx == 0) { 202 klist = &ifklist; 203 } else { 204 ifp = ifnet_byindex(idx); 205 if (ifp == NULL) 206 return (1); 207 klist = &ifp->if_klist; 208 } 209 210 kn->kn_hook = (caddr_t)klist; 211 212 knlist_add(klist, kn, 0); 213 214 return (0); 215 } 216 217 static void 218 filt_netdetach(struct knote *kn) 219 { 220 struct knlist *klist = (struct knlist *)kn->kn_hook; 221 222 knlist_remove(klist, kn, 0); 223 } 224 225 static int 226 filt_netdev(struct knote *kn, long hint) 227 { 228 struct knlist *klist = (struct knlist *)kn->kn_hook; 229 230 /* 231 * Currently NOTE_EXIT is abused to indicate device detach. 232 */ 233 if (hint == NOTE_EXIT) { 234 kn->kn_data = NOTE_LINKINV; 235 kn->kn_flags |= (EV_EOF | EV_ONESHOT); 236 knlist_remove_inevent(klist, kn); 237 return (1); 238 } 239 if (hint != 0) 240 kn->kn_data = hint; /* current status */ 241 if (kn->kn_sfflags & hint) 242 kn->kn_fflags |= hint; 243 return (kn->kn_fflags != 0); 244 } 245 246 /* 247 * Network interface utility routines. 248 * 249 * Routines with ifa_ifwith* names take sockaddr *'s as 250 * parameters. 251 */ 252 /* ARGSUSED*/ 253 static void 254 if_init(void *dummy __unused) 255 { 256 257 IFNET_LOCK_INIT(); 258 TAILQ_INIT(&ifnet); 259 knlist_init(&ifklist, NULL); 260 if_grow(); /* create initial table */ 261 ifdev_byindex(0) = make_dev(&net_cdevsw, 0, 262 UID_ROOT, GID_WHEEL, 0600, "network"); 263 if_clone_init(); 264 } 265 266 static void 267 if_grow(void) 268 { 269 u_int n; 270 struct ifindex_entry *e; 271 272 if_indexlim <<= 1; 273 n = if_indexlim * sizeof(*e); 274 e = malloc(n, M_IFADDR, M_WAITOK | M_ZERO); 275 if (ifindex_table != NULL) { 276 memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2); 277 free((caddr_t)ifindex_table, M_IFADDR); 278 } 279 ifindex_table = e; 280 } 281 282 /* ARGSUSED*/ 283 static void 284 if_check(void *dummy __unused) 285 { 286 struct ifnet *ifp; 287 int s; 288 289 s = splimp(); 290 IFNET_RLOCK(); /* could sleep on rare error; mostly okay XXX */ 291 TAILQ_FOREACH(ifp, &ifnet, if_link) { 292 if (ifp->if_snd.ifq_maxlen == 0) { 293 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n"); 294 ifp->if_snd.ifq_maxlen = ifqmaxlen; 295 } 296 if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) { 297 if_printf(ifp, 298 "XXX: driver didn't initialize queue mtx\n"); 299 mtx_init(&ifp->if_snd.ifq_mtx, "unknown", 300 MTX_NETWORK_LOCK, MTX_DEF); 301 } 302 } 303 IFNET_RUNLOCK(); 304 splx(s); 305 if_slowtimo(0); 306 } 307 308 static int 309 if_findindex(struct ifnet *ifp) 310 { 311 int i, unit; 312 char eaddr[18], devname[32]; 313 const char *name, *p; 314 315 switch (ifp->if_type) { 316 case IFT_ETHER: /* these types use struct arpcom */ 317 case IFT_FDDI: 318 case IFT_XETHER: 319 case IFT_ISO88025: 320 case IFT_L2VLAN: 321 snprintf(eaddr, 18, "%6D", IFP2AC(ifp)->ac_enaddr, ":"); 322 break; 323 default: 324 eaddr[0] = '\0'; 325 break; 326 } 327 strlcpy(devname, ifp->if_xname, sizeof(devname)); 328 name = net_cdevsw.d_name; 329 i = 0; 330 while ((resource_find_dev(&i, name, &unit, NULL, NULL)) == 0) { 331 if (resource_string_value(name, unit, "ether", &p) == 0) 332 if (strcmp(p, eaddr) == 0) 333 goto found; 334 if (resource_string_value(name, unit, "dev", &p) == 0) 335 if (strcmp(p, devname) == 0) 336 goto found; 337 } 338 unit = 0; 339 found: 340 if (unit != 0) { 341 if (ifaddr_byindex(unit) == NULL) 342 return (unit); 343 printf("%s%d in use, cannot hardwire it to %s.\n", 344 name, unit, devname); 345 } 346 for (unit = 1; ; unit++) { 347 if (unit <= if_index && ifaddr_byindex(unit) != NULL) 348 continue; 349 if (resource_string_value(name, unit, "ether", &p) == 0 || 350 resource_string_value(name, unit, "dev", &p) == 0) 351 continue; 352 break; 353 } 354 return (unit); 355 } 356 357 /* 358 * Attach an interface to the 359 * list of "active" interfaces. 360 */ 361 void 362 if_attach(struct ifnet *ifp) 363 { 364 unsigned socksize, ifasize; 365 int namelen, masklen; 366 struct sockaddr_dl *sdl; 367 struct ifaddr *ifa; 368 369 TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp); 370 IF_AFDATA_LOCK_INIT(ifp); 371 ifp->if_afdata_initialized = 0; 372 IFNET_WLOCK(); 373 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link); 374 IFNET_WUNLOCK(); 375 /* 376 * XXX - 377 * The old code would work if the interface passed a pre-existing 378 * chain of ifaddrs to this code. We don't trust our callers to 379 * properly initialize the tailq, however, so we no longer allow 380 * this unlikely case. 381 */ 382 TAILQ_INIT(&ifp->if_addrhead); 383 TAILQ_INIT(&ifp->if_prefixhead); 384 TAILQ_INIT(&ifp->if_multiaddrs); 385 knlist_init(&ifp->if_klist, NULL); 386 getmicrotime(&ifp->if_lastchange); 387 ifp->if_data.ifi_epoch = time_second; 388 389 #ifdef MAC 390 mac_init_ifnet(ifp); 391 mac_create_ifnet(ifp); 392 #endif 393 394 ifp->if_index = if_findindex(ifp); 395 if (ifp->if_index > if_index) 396 if_index = ifp->if_index; 397 if (if_index >= if_indexlim) 398 if_grow(); 399 ifp->if_data.ifi_datalen = sizeof(struct if_data); 400 401 ifnet_byindex(ifp->if_index) = ifp; 402 ifdev_byindex(ifp->if_index) = make_dev(&net_cdevsw, 403 unit2minor(ifp->if_index), 404 UID_ROOT, GID_WHEEL, 0600, "%s/%s", 405 net_cdevsw.d_name, ifp->if_xname); 406 make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d", 407 net_cdevsw.d_name, ifp->if_index); 408 409 mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 410 411 /* 412 * create a Link Level name for this device 413 */ 414 namelen = strlen(ifp->if_xname); 415 /* 416 * Always save enough space for any possiable name so we can do 417 * a rename in place later. 418 */ 419 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 420 socksize = masklen + ifp->if_addrlen; 421 if (socksize < sizeof(*sdl)) 422 socksize = sizeof(*sdl); 423 socksize = roundup2(socksize, sizeof(long)); 424 ifasize = sizeof(*ifa) + 2 * socksize; 425 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO); 426 IFA_LOCK_INIT(ifa); 427 sdl = (struct sockaddr_dl *)(ifa + 1); 428 sdl->sdl_len = socksize; 429 sdl->sdl_family = AF_LINK; 430 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 431 sdl->sdl_nlen = namelen; 432 sdl->sdl_index = ifp->if_index; 433 sdl->sdl_type = ifp->if_type; 434 ifaddr_byindex(ifp->if_index) = ifa; 435 ifa->ifa_ifp = ifp; 436 ifa->ifa_rtrequest = link_rtrequest; 437 ifa->ifa_addr = (struct sockaddr *)sdl; 438 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 439 ifa->ifa_netmask = (struct sockaddr *)sdl; 440 sdl->sdl_len = masklen; 441 while (namelen != 0) 442 sdl->sdl_data[--namelen] = 0xff; 443 ifa->ifa_refcnt = 1; 444 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 445 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 446 ifp->if_snd.altq_type = 0; 447 ifp->if_snd.altq_disc = NULL; 448 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 449 ifp->if_snd.altq_tbr = NULL; 450 ifp->if_snd.altq_ifp = ifp; 451 452 if (domains) 453 if_attachdomain1(ifp); 454 455 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 456 457 /* Announce the interface. */ 458 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 459 } 460 461 static void 462 if_attachdomain(void *dummy) 463 { 464 struct ifnet *ifp; 465 int s; 466 467 s = splnet(); 468 TAILQ_FOREACH(ifp, &ifnet, if_link) 469 if_attachdomain1(ifp); 470 splx(s); 471 } 472 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST, 473 if_attachdomain, NULL); 474 475 static void 476 if_attachdomain1(struct ifnet *ifp) 477 { 478 struct domain *dp; 479 int s; 480 481 s = splnet(); 482 483 /* 484 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 485 * cannot lock ifp->if_afdata initialization, entirely. 486 */ 487 if (IF_AFDATA_TRYLOCK(ifp) == 0) { 488 splx(s); 489 return; 490 } 491 if (ifp->if_afdata_initialized) { 492 IF_AFDATA_UNLOCK(ifp); 493 splx(s); 494 return; 495 } 496 ifp->if_afdata_initialized = 1; 497 IF_AFDATA_UNLOCK(ifp); 498 499 /* address family dependent data region */ 500 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 501 for (dp = domains; dp; dp = dp->dom_next) { 502 if (dp->dom_ifattach) 503 ifp->if_afdata[dp->dom_family] = 504 (*dp->dom_ifattach)(ifp); 505 } 506 507 splx(s); 508 } 509 510 /* 511 * Detach an interface, removing it from the 512 * list of "active" interfaces. 513 */ 514 void 515 if_detach(struct ifnet *ifp) 516 { 517 struct ifaddr *ifa, *next; 518 struct radix_node_head *rnh; 519 int s; 520 int i; 521 struct domain *dp; 522 struct ifnet *iter; 523 int found; 524 525 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 526 /* 527 * Remove routes and flush queues. 528 */ 529 s = splnet(); 530 if_down(ifp); 531 #ifdef ALTQ 532 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 533 altq_disable(&ifp->if_snd); 534 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 535 altq_detach(&ifp->if_snd); 536 #endif 537 538 for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa; ifa = next) { 539 next = TAILQ_NEXT(ifa, ifa_link); 540 541 if (ifa->ifa_addr->sa_family == AF_LINK) 542 continue; 543 #ifdef INET 544 /* XXX: Ugly!! ad hoc just for INET */ 545 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) { 546 struct ifaliasreq ifr; 547 548 bzero(&ifr, sizeof(ifr)); 549 ifr.ifra_addr = *ifa->ifa_addr; 550 if (ifa->ifa_dstaddr) 551 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 552 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 553 NULL) == 0) 554 continue; 555 } 556 #endif /* INET */ 557 #ifdef INET6 558 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) { 559 in6_purgeaddr(ifa); 560 /* ifp_addrhead is already updated */ 561 continue; 562 } 563 #endif /* INET6 */ 564 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 565 IFAFREE(ifa); 566 } 567 568 #ifdef INET6 569 /* 570 * Remove all IPv6 kernel structs related to ifp. This should be done 571 * before removing routing entries below, since IPv6 interface direct 572 * routes are expected to be removed by the IPv6-specific kernel API. 573 * Otherwise, the kernel will detect some inconsistency and bark it. 574 */ 575 in6_ifdetach(ifp); 576 #endif 577 /* 578 * Remove address from ifindex_table[] and maybe decrement if_index. 579 * Clean up all addresses. 580 */ 581 ifnet_byindex(ifp->if_index) = NULL; 582 ifaddr_byindex(ifp->if_index) = NULL; 583 destroy_dev(ifdev_byindex(ifp->if_index)); 584 ifdev_byindex(ifp->if_index) = NULL; 585 586 while (if_index > 0 && ifaddr_byindex(if_index) == NULL) 587 if_index--; 588 589 590 /* We can now free link ifaddr. */ 591 if (!TAILQ_EMPTY(&ifp->if_addrhead)) { 592 ifa = TAILQ_FIRST(&ifp->if_addrhead); 593 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 594 IFAFREE(ifa); 595 } 596 597 /* 598 * Delete all remaining routes using this interface 599 * Unfortuneatly the only way to do this is to slog through 600 * the entire routing table looking for routes which point 601 * to this interface...oh well... 602 */ 603 for (i = 1; i <= AF_MAX; i++) { 604 if ((rnh = rt_tables[i]) == NULL) 605 continue; 606 RADIX_NODE_HEAD_LOCK(rnh); 607 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp); 608 RADIX_NODE_HEAD_UNLOCK(rnh); 609 } 610 611 /* Announce that the interface is gone. */ 612 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 613 614 IF_AFDATA_LOCK(ifp); 615 for (dp = domains; dp; dp = dp->dom_next) { 616 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) 617 (*dp->dom_ifdetach)(ifp, 618 ifp->if_afdata[dp->dom_family]); 619 } 620 IF_AFDATA_UNLOCK(ifp); 621 622 #ifdef MAC 623 mac_destroy_ifnet(ifp); 624 #endif /* MAC */ 625 KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT); 626 knlist_clear(&ifp->if_klist, 0); 627 knlist_destroy(&ifp->if_klist); 628 IFNET_WLOCK(); 629 found = 0; 630 TAILQ_FOREACH(iter, &ifnet, if_link) 631 if (iter == ifp) { 632 found = 1; 633 break; 634 } 635 if (found) 636 TAILQ_REMOVE(&ifnet, ifp, if_link); 637 IFNET_WUNLOCK(); 638 mtx_destroy(&ifp->if_snd.ifq_mtx); 639 IF_AFDATA_DESTROY(ifp); 640 splx(s); 641 } 642 643 /* 644 * Delete Routes for a Network Interface 645 * 646 * Called for each routing entry via the rnh->rnh_walktree() call above 647 * to delete all route entries referencing a detaching network interface. 648 * 649 * Arguments: 650 * rn pointer to node in the routing table 651 * arg argument passed to rnh->rnh_walktree() - detaching interface 652 * 653 * Returns: 654 * 0 successful 655 * errno failed - reason indicated 656 * 657 */ 658 static int 659 if_rtdel(struct radix_node *rn, void *arg) 660 { 661 struct rtentry *rt = (struct rtentry *)rn; 662 struct ifnet *ifp = arg; 663 int err; 664 665 if (rt->rt_ifp == ifp) { 666 667 /* 668 * Protect (sorta) against walktree recursion problems 669 * with cloned routes 670 */ 671 if ((rt->rt_flags & RTF_UP) == 0) 672 return (0); 673 674 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway, 675 rt_mask(rt), rt->rt_flags, 676 (struct rtentry **) NULL); 677 if (err) { 678 log(LOG_WARNING, "if_rtdel: error %d\n", err); 679 } 680 } 681 682 return (0); 683 } 684 685 #define equal(a1, a2) (bcmp((a1), (a2), ((a1))->sa_len) == 0) 686 687 /* 688 * Locate an interface based on a complete address. 689 */ 690 /*ARGSUSED*/ 691 struct ifaddr * 692 ifa_ifwithaddr(struct sockaddr *addr) 693 { 694 struct ifnet *ifp; 695 struct ifaddr *ifa; 696 697 IFNET_RLOCK(); 698 TAILQ_FOREACH(ifp, &ifnet, if_link) 699 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 700 if (ifa->ifa_addr->sa_family != addr->sa_family) 701 continue; 702 if (equal(addr, ifa->ifa_addr)) 703 goto done; 704 /* IP6 doesn't have broadcast */ 705 if ((ifp->if_flags & IFF_BROADCAST) && 706 ifa->ifa_broadaddr && 707 ifa->ifa_broadaddr->sa_len != 0 && 708 equal(ifa->ifa_broadaddr, addr)) 709 goto done; 710 } 711 ifa = NULL; 712 done: 713 IFNET_RUNLOCK(); 714 return (ifa); 715 } 716 717 /* 718 * Locate the point to point interface with a given destination address. 719 */ 720 /*ARGSUSED*/ 721 struct ifaddr * 722 ifa_ifwithdstaddr(struct sockaddr *addr) 723 { 724 struct ifnet *ifp; 725 struct ifaddr *ifa; 726 727 IFNET_RLOCK(); 728 TAILQ_FOREACH(ifp, &ifnet, if_link) { 729 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 730 continue; 731 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 732 if (ifa->ifa_addr->sa_family != addr->sa_family) 733 continue; 734 if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)) 735 goto done; 736 } 737 } 738 ifa = NULL; 739 done: 740 IFNET_RUNLOCK(); 741 return (ifa); 742 } 743 744 /* 745 * Find an interface on a specific network. If many, choice 746 * is most specific found. 747 */ 748 struct ifaddr * 749 ifa_ifwithnet(struct sockaddr *addr) 750 { 751 struct ifnet *ifp; 752 struct ifaddr *ifa; 753 struct ifaddr *ifa_maybe = (struct ifaddr *) 0; 754 u_int af = addr->sa_family; 755 char *addr_data = addr->sa_data, *cplim; 756 757 /* 758 * AF_LINK addresses can be looked up directly by their index number, 759 * so do that if we can. 760 */ 761 if (af == AF_LINK) { 762 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr; 763 if (sdl->sdl_index && sdl->sdl_index <= if_index) 764 return (ifaddr_byindex(sdl->sdl_index)); 765 } 766 767 /* 768 * Scan though each interface, looking for ones that have 769 * addresses in this address family. 770 */ 771 IFNET_RLOCK(); 772 TAILQ_FOREACH(ifp, &ifnet, if_link) { 773 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 774 char *cp, *cp2, *cp3; 775 776 if (ifa->ifa_addr->sa_family != af) 777 next: continue; 778 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) { 779 /* 780 * This is a bit broken as it doesn't 781 * take into account that the remote end may 782 * be a single node in the network we are 783 * looking for. 784 * The trouble is that we don't know the 785 * netmask for the remote end. 786 */ 787 if (ifa->ifa_dstaddr != 0 788 && equal(addr, ifa->ifa_dstaddr)) 789 goto done; 790 } else { 791 /* 792 * if we have a special address handler, 793 * then use it instead of the generic one. 794 */ 795 if (ifa->ifa_claim_addr) { 796 if ((*ifa->ifa_claim_addr)(ifa, addr)) 797 goto done; 798 continue; 799 } 800 801 /* 802 * Scan all the bits in the ifa's address. 803 * If a bit dissagrees with what we are 804 * looking for, mask it with the netmask 805 * to see if it really matters. 806 * (A byte at a time) 807 */ 808 if (ifa->ifa_netmask == 0) 809 continue; 810 cp = addr_data; 811 cp2 = ifa->ifa_addr->sa_data; 812 cp3 = ifa->ifa_netmask->sa_data; 813 cplim = ifa->ifa_netmask->sa_len 814 + (char *)ifa->ifa_netmask; 815 while (cp3 < cplim) 816 if ((*cp++ ^ *cp2++) & *cp3++) 817 goto next; /* next address! */ 818 /* 819 * If the netmask of what we just found 820 * is more specific than what we had before 821 * (if we had one) then remember the new one 822 * before continuing to search 823 * for an even better one. 824 */ 825 if (ifa_maybe == 0 || 826 rn_refines((caddr_t)ifa->ifa_netmask, 827 (caddr_t)ifa_maybe->ifa_netmask)) 828 ifa_maybe = ifa; 829 } 830 } 831 } 832 ifa = ifa_maybe; 833 done: 834 IFNET_RUNLOCK(); 835 return (ifa); 836 } 837 838 /* 839 * Find an interface address specific to an interface best matching 840 * a given address. 841 */ 842 struct ifaddr * 843 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp) 844 { 845 struct ifaddr *ifa; 846 char *cp, *cp2, *cp3; 847 char *cplim; 848 struct ifaddr *ifa_maybe = 0; 849 u_int af = addr->sa_family; 850 851 if (af >= AF_MAX) 852 return (0); 853 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 854 if (ifa->ifa_addr->sa_family != af) 855 continue; 856 if (ifa_maybe == 0) 857 ifa_maybe = ifa; 858 if (ifa->ifa_netmask == 0) { 859 if (equal(addr, ifa->ifa_addr) || 860 (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))) 861 goto done; 862 continue; 863 } 864 if (ifp->if_flags & IFF_POINTOPOINT) { 865 if (equal(addr, ifa->ifa_dstaddr)) 866 goto done; 867 } else { 868 cp = addr->sa_data; 869 cp2 = ifa->ifa_addr->sa_data; 870 cp3 = ifa->ifa_netmask->sa_data; 871 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 872 for (; cp3 < cplim; cp3++) 873 if ((*cp++ ^ *cp2++) & *cp3) 874 break; 875 if (cp3 == cplim) 876 goto done; 877 } 878 } 879 ifa = ifa_maybe; 880 done: 881 return (ifa); 882 } 883 884 #include <net/route.h> 885 886 /* 887 * Default action when installing a route with a Link Level gateway. 888 * Lookup an appropriate real ifa to point to. 889 * This should be moved to /sys/net/link.c eventually. 890 */ 891 static void 892 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 893 { 894 struct ifaddr *ifa, *oifa; 895 struct sockaddr *dst; 896 struct ifnet *ifp; 897 898 RT_LOCK_ASSERT(rt); 899 900 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) || 901 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0)) 902 return; 903 ifa = ifaof_ifpforaddr(dst, ifp); 904 if (ifa) { 905 IFAREF(ifa); /* XXX */ 906 oifa = rt->rt_ifa; 907 rt->rt_ifa = ifa; 908 IFAFREE(oifa); 909 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 910 ifa->ifa_rtrequest(cmd, rt, info); 911 } 912 } 913 914 /* 915 * Mark an interface down and notify protocols of 916 * the transition. 917 * NOTE: must be called at splnet or eqivalent. 918 */ 919 static void 920 if_unroute(struct ifnet *ifp, int flag, int fam) 921 { 922 struct ifaddr *ifa; 923 924 ifp->if_flags &= ~flag; 925 getmicrotime(&ifp->if_lastchange); 926 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 927 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 928 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 929 if_qflush(&ifp->if_snd); 930 rt_ifmsg(ifp); 931 } 932 933 /* 934 * Mark an interface up and notify protocols of 935 * the transition. 936 * NOTE: must be called at splnet or eqivalent. 937 */ 938 static void 939 if_route(struct ifnet *ifp, int flag, int fam) 940 { 941 struct ifaddr *ifa; 942 943 ifp->if_flags |= flag; 944 getmicrotime(&ifp->if_lastchange); 945 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 946 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 947 pfctlinput(PRC_IFUP, ifa->ifa_addr); 948 rt_ifmsg(ifp); 949 #ifdef INET6 950 in6_if_up(ifp); 951 #endif 952 } 953 954 /* 955 * Mark an interface down and notify protocols of 956 * the transition. 957 * NOTE: must be called at splnet or eqivalent. 958 */ 959 void 960 if_down(struct ifnet *ifp) 961 { 962 963 if_unroute(ifp, IFF_UP, AF_UNSPEC); 964 } 965 966 /* 967 * Mark an interface up and notify protocols of 968 * the transition. 969 * NOTE: must be called at splnet or eqivalent. 970 */ 971 void 972 if_up(struct ifnet *ifp) 973 { 974 975 if_route(ifp, IFF_UP, AF_UNSPEC); 976 } 977 978 /* 979 * Flush an interface queue. 980 */ 981 static void 982 if_qflush(struct ifaltq *ifq) 983 { 984 struct mbuf *m, *n; 985 986 IFQ_LOCK(ifq); 987 #ifdef ALTQ 988 if (ALTQ_IS_ENABLED(ifq)) 989 ALTQ_PURGE(ifq); 990 #endif 991 n = ifq->ifq_head; 992 while ((m = n) != 0) { 993 n = m->m_act; 994 m_freem(m); 995 } 996 ifq->ifq_head = 0; 997 ifq->ifq_tail = 0; 998 ifq->ifq_len = 0; 999 IFQ_UNLOCK(ifq); 1000 } 1001 1002 /* 1003 * Handle interface watchdog timer routines. Called 1004 * from softclock, we decrement timers (if set) and 1005 * call the appropriate interface routine on expiration. 1006 * 1007 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called 1008 * holding Giant. If we switch to an MPSAFE callout, we likely need to grab 1009 * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface. 1010 */ 1011 static void 1012 if_slowtimo(void *arg) 1013 { 1014 struct ifnet *ifp; 1015 int s = splimp(); 1016 1017 IFNET_RLOCK(); 1018 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1019 if (ifp->if_timer == 0 || --ifp->if_timer) 1020 continue; 1021 if (ifp->if_watchdog) 1022 (*ifp->if_watchdog)(ifp); 1023 } 1024 IFNET_RUNLOCK(); 1025 splx(s); 1026 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ); 1027 } 1028 1029 /* 1030 * Map interface name to 1031 * interface structure pointer. 1032 */ 1033 struct ifnet * 1034 ifunit(const char *name) 1035 { 1036 struct ifnet *ifp; 1037 1038 IFNET_RLOCK(); 1039 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1040 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 1041 break; 1042 } 1043 IFNET_RUNLOCK(); 1044 return (ifp); 1045 } 1046 1047 /* 1048 * Hardware specific interface ioctls. 1049 */ 1050 static int 1051 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 1052 { 1053 struct ifreq *ifr; 1054 struct ifstat *ifs; 1055 int error = 0; 1056 int new_flags; 1057 size_t namelen, onamelen; 1058 char new_name[IFNAMSIZ]; 1059 struct ifaddr *ifa; 1060 struct sockaddr_dl *sdl; 1061 1062 ifr = (struct ifreq *)data; 1063 switch (cmd) { 1064 case SIOCGIFINDEX: 1065 ifr->ifr_index = ifp->if_index; 1066 break; 1067 1068 case SIOCGIFFLAGS: 1069 ifr->ifr_flags = ifp->if_flags & 0xffff; 1070 ifr->ifr_flagshigh = ifp->if_flags >> 16; 1071 break; 1072 1073 case SIOCGIFCAP: 1074 ifr->ifr_reqcap = ifp->if_capabilities; 1075 ifr->ifr_curcap = ifp->if_capenable; 1076 break; 1077 1078 #ifdef MAC 1079 case SIOCGIFMAC: 1080 error = mac_ioctl_ifnet_get(td->td_ucred, ifr, ifp); 1081 break; 1082 #endif 1083 1084 case SIOCGIFMETRIC: 1085 ifr->ifr_metric = ifp->if_metric; 1086 break; 1087 1088 case SIOCGIFMTU: 1089 ifr->ifr_mtu = ifp->if_mtu; 1090 break; 1091 1092 case SIOCGIFPHYS: 1093 ifr->ifr_phys = ifp->if_physical; 1094 break; 1095 1096 case SIOCSIFFLAGS: 1097 error = suser(td); 1098 if (error) 1099 return (error); 1100 new_flags = (ifr->ifr_flags & 0xffff) | 1101 (ifr->ifr_flagshigh << 16); 1102 if (ifp->if_flags & IFF_SMART) { 1103 /* Smart drivers twiddle their own routes */ 1104 } else if (ifp->if_flags & IFF_UP && 1105 (new_flags & IFF_UP) == 0) { 1106 int s = splimp(); 1107 if_down(ifp); 1108 splx(s); 1109 } else if (new_flags & IFF_UP && 1110 (ifp->if_flags & IFF_UP) == 0) { 1111 int s = splimp(); 1112 if_up(ifp); 1113 splx(s); 1114 } 1115 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 1116 (new_flags &~ IFF_CANTCHANGE); 1117 if (new_flags & IFF_PPROMISC) { 1118 /* Permanently promiscuous mode requested */ 1119 ifp->if_flags |= IFF_PROMISC; 1120 } else if (ifp->if_pcount == 0) { 1121 ifp->if_flags &= ~IFF_PROMISC; 1122 } 1123 if (ifp->if_ioctl) 1124 (void) (*ifp->if_ioctl)(ifp, cmd, data); 1125 getmicrotime(&ifp->if_lastchange); 1126 break; 1127 1128 case SIOCSIFCAP: 1129 error = suser(td); 1130 if (error) 1131 return (error); 1132 if (ifp->if_ioctl == NULL) 1133 return (EOPNOTSUPP); 1134 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 1135 return (EINVAL); 1136 error = (*ifp->if_ioctl)(ifp, cmd, data); 1137 if (error == 0) 1138 getmicrotime(&ifp->if_lastchange); 1139 break; 1140 1141 #ifdef MAC 1142 case SIOCSIFMAC: 1143 error = mac_ioctl_ifnet_set(td->td_ucred, ifr, ifp); 1144 break; 1145 #endif 1146 1147 case SIOCSIFNAME: 1148 error = suser(td); 1149 if (error != 0) 1150 return (error); 1151 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); 1152 if (error != 0) 1153 return (error); 1154 if (new_name[0] == '\0') 1155 return (EINVAL); 1156 if (ifunit(new_name) != NULL) 1157 return (EEXIST); 1158 1159 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1160 /* Announce the departure of the interface. */ 1161 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1162 1163 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 1164 ifa = ifaddr_byindex(ifp->if_index); 1165 IFA_LOCK(ifa); 1166 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 1167 namelen = strlen(new_name); 1168 onamelen = sdl->sdl_nlen; 1169 /* 1170 * Move the address if needed. This is safe because we 1171 * allocate space for a name of length IFNAMSIZ when we 1172 * create this in if_attach(). 1173 */ 1174 if (namelen != onamelen) { 1175 bcopy(sdl->sdl_data + onamelen, 1176 sdl->sdl_data + namelen, sdl->sdl_alen); 1177 } 1178 bcopy(new_name, sdl->sdl_data, namelen); 1179 sdl->sdl_nlen = namelen; 1180 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 1181 bzero(sdl->sdl_data, onamelen); 1182 while (namelen != 0) 1183 sdl->sdl_data[--namelen] = 0xff; 1184 IFA_UNLOCK(ifa); 1185 1186 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 1187 /* Announce the return of the interface. */ 1188 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 1189 break; 1190 1191 case SIOCSIFMETRIC: 1192 error = suser(td); 1193 if (error) 1194 return (error); 1195 ifp->if_metric = ifr->ifr_metric; 1196 getmicrotime(&ifp->if_lastchange); 1197 break; 1198 1199 case SIOCSIFPHYS: 1200 error = suser(td); 1201 if (error) 1202 return (error); 1203 if (ifp->if_ioctl == NULL) 1204 return (EOPNOTSUPP); 1205 error = (*ifp->if_ioctl)(ifp, cmd, data); 1206 if (error == 0) 1207 getmicrotime(&ifp->if_lastchange); 1208 break; 1209 1210 case SIOCSIFMTU: 1211 { 1212 u_long oldmtu = ifp->if_mtu; 1213 1214 error = suser(td); 1215 if (error) 1216 return (error); 1217 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 1218 return (EINVAL); 1219 if (ifp->if_ioctl == NULL) 1220 return (EOPNOTSUPP); 1221 error = (*ifp->if_ioctl)(ifp, cmd, data); 1222 if (error == 0) { 1223 getmicrotime(&ifp->if_lastchange); 1224 rt_ifmsg(ifp); 1225 } 1226 /* 1227 * If the link MTU changed, do network layer specific procedure. 1228 */ 1229 if (ifp->if_mtu != oldmtu) { 1230 #ifdef INET6 1231 nd6_setmtu(ifp); 1232 #endif 1233 } 1234 break; 1235 } 1236 1237 case SIOCADDMULTI: 1238 case SIOCDELMULTI: 1239 error = suser(td); 1240 if (error) 1241 return (error); 1242 1243 /* Don't allow group membership on non-multicast interfaces. */ 1244 if ((ifp->if_flags & IFF_MULTICAST) == 0) 1245 return (EOPNOTSUPP); 1246 1247 /* Don't let users screw up protocols' entries. */ 1248 if (ifr->ifr_addr.sa_family != AF_LINK) 1249 return (EINVAL); 1250 1251 if (cmd == SIOCADDMULTI) { 1252 struct ifmultiaddr *ifma; 1253 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 1254 } else { 1255 error = if_delmulti(ifp, &ifr->ifr_addr); 1256 } 1257 if (error == 0) 1258 getmicrotime(&ifp->if_lastchange); 1259 break; 1260 1261 case SIOCSIFPHYADDR: 1262 case SIOCDIFPHYADDR: 1263 #ifdef INET6 1264 case SIOCSIFPHYADDR_IN6: 1265 #endif 1266 case SIOCSLIFPHYADDR: 1267 case SIOCSIFMEDIA: 1268 case SIOCSIFGENERIC: 1269 error = suser(td); 1270 if (error) 1271 return (error); 1272 if (ifp->if_ioctl == NULL) 1273 return (EOPNOTSUPP); 1274 error = (*ifp->if_ioctl)(ifp, cmd, data); 1275 if (error == 0) 1276 getmicrotime(&ifp->if_lastchange); 1277 break; 1278 1279 case SIOCGIFSTATUS: 1280 ifs = (struct ifstat *)data; 1281 ifs->ascii[0] = '\0'; 1282 1283 case SIOCGIFPSRCADDR: 1284 case SIOCGIFPDSTADDR: 1285 case SIOCGLIFPHYADDR: 1286 case SIOCGIFMEDIA: 1287 case SIOCGIFGENERIC: 1288 if (ifp->if_ioctl == NULL) 1289 return (EOPNOTSUPP); 1290 error = (*ifp->if_ioctl)(ifp, cmd, data); 1291 break; 1292 1293 case SIOCSIFLLADDR: 1294 error = suser(td); 1295 if (error) 1296 return (error); 1297 error = if_setlladdr(ifp, 1298 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 1299 break; 1300 1301 default: 1302 error = ENOIOCTL; 1303 break; 1304 } 1305 return (error); 1306 } 1307 1308 /* 1309 * Interface ioctls. 1310 */ 1311 int 1312 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 1313 { 1314 struct ifnet *ifp; 1315 struct ifreq *ifr; 1316 int error; 1317 int oif_flags; 1318 1319 switch (cmd) { 1320 case SIOCGIFCONF: 1321 case OSIOCGIFCONF: 1322 return (ifconf(cmd, data)); 1323 } 1324 ifr = (struct ifreq *)data; 1325 1326 switch (cmd) { 1327 case SIOCIFCREATE: 1328 case SIOCIFDESTROY: 1329 if ((error = suser(td)) != 0) 1330 return (error); 1331 return ((cmd == SIOCIFCREATE) ? 1332 if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) : 1333 if_clone_destroy(ifr->ifr_name)); 1334 1335 case SIOCIFGCLONERS: 1336 return (if_clone_list((struct if_clonereq *)data)); 1337 } 1338 1339 ifp = ifunit(ifr->ifr_name); 1340 if (ifp == 0) 1341 return (ENXIO); 1342 1343 error = ifhwioctl(cmd, ifp, data, td); 1344 if (error != ENOIOCTL) 1345 return (error); 1346 1347 oif_flags = ifp->if_flags; 1348 if (so->so_proto == 0) 1349 return (EOPNOTSUPP); 1350 #ifndef COMPAT_43 1351 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, 1352 data, 1353 ifp, td)); 1354 #else 1355 { 1356 int ocmd = cmd; 1357 1358 switch (cmd) { 1359 1360 case SIOCSIFDSTADDR: 1361 case SIOCSIFADDR: 1362 case SIOCSIFBRDADDR: 1363 case SIOCSIFNETMASK: 1364 #if BYTE_ORDER != BIG_ENDIAN 1365 if (ifr->ifr_addr.sa_family == 0 && 1366 ifr->ifr_addr.sa_len < 16) { 1367 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len; 1368 ifr->ifr_addr.sa_len = 16; 1369 } 1370 #else 1371 if (ifr->ifr_addr.sa_len == 0) 1372 ifr->ifr_addr.sa_len = 16; 1373 #endif 1374 break; 1375 1376 case OSIOCGIFADDR: 1377 cmd = SIOCGIFADDR; 1378 break; 1379 1380 case OSIOCGIFDSTADDR: 1381 cmd = SIOCGIFDSTADDR; 1382 break; 1383 1384 case OSIOCGIFBRDADDR: 1385 cmd = SIOCGIFBRDADDR; 1386 break; 1387 1388 case OSIOCGIFNETMASK: 1389 cmd = SIOCGIFNETMASK; 1390 } 1391 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, 1392 cmd, 1393 data, 1394 ifp, td)); 1395 switch (ocmd) { 1396 1397 case OSIOCGIFADDR: 1398 case OSIOCGIFDSTADDR: 1399 case OSIOCGIFBRDADDR: 1400 case OSIOCGIFNETMASK: 1401 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family; 1402 1403 } 1404 } 1405 #endif /* COMPAT_43 */ 1406 1407 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 1408 #ifdef INET6 1409 DELAY(100);/* XXX: temporary workaround for fxp issue*/ 1410 if (ifp->if_flags & IFF_UP) { 1411 int s = splimp(); 1412 in6_if_up(ifp); 1413 splx(s); 1414 } 1415 #endif 1416 } 1417 return (error); 1418 } 1419 1420 /* 1421 * Set/clear promiscuous mode on interface ifp based on the truth value 1422 * of pswitch. The calls are reference counted so that only the first 1423 * "on" request actually has an effect, as does the final "off" request. 1424 * Results are undefined if the "off" and "on" requests are not matched. 1425 */ 1426 int 1427 ifpromisc(struct ifnet *ifp, int pswitch) 1428 { 1429 struct ifreq ifr; 1430 int error; 1431 int oldflags, oldpcount; 1432 1433 oldpcount = ifp->if_pcount; 1434 oldflags = ifp->if_flags; 1435 if (ifp->if_flags & IFF_PPROMISC) { 1436 /* Do nothing if device is in permanently promiscuous mode */ 1437 ifp->if_pcount += pswitch ? 1 : -1; 1438 return (0); 1439 } 1440 if (pswitch) { 1441 /* 1442 * If the device is not configured up, we cannot put it in 1443 * promiscuous mode. 1444 */ 1445 if ((ifp->if_flags & IFF_UP) == 0) 1446 return (ENETDOWN); 1447 if (ifp->if_pcount++ != 0) 1448 return (0); 1449 ifp->if_flags |= IFF_PROMISC; 1450 } else { 1451 if (--ifp->if_pcount > 0) 1452 return (0); 1453 ifp->if_flags &= ~IFF_PROMISC; 1454 } 1455 ifr.ifr_flags = ifp->if_flags & 0xffff; 1456 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1457 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1458 if (error == 0) { 1459 log(LOG_INFO, "%s: promiscuous mode %s\n", 1460 ifp->if_xname, 1461 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 1462 rt_ifmsg(ifp); 1463 } else { 1464 ifp->if_pcount = oldpcount; 1465 ifp->if_flags = oldflags; 1466 } 1467 return error; 1468 } 1469 1470 /* 1471 * Return interface configuration 1472 * of system. List may be used 1473 * in later ioctl's (above) to get 1474 * other information. 1475 */ 1476 /*ARGSUSED*/ 1477 static int 1478 ifconf(u_long cmd, caddr_t data) 1479 { 1480 struct ifconf *ifc = (struct ifconf *)data; 1481 struct ifnet *ifp; 1482 struct ifaddr *ifa; 1483 struct ifreq ifr, *ifrp; 1484 int space = ifc->ifc_len, error = 0; 1485 1486 ifrp = ifc->ifc_req; 1487 IFNET_RLOCK(); /* could sleep XXX */ 1488 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1489 int addrs; 1490 1491 if (space < sizeof(ifr)) 1492 break; 1493 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 1494 >= sizeof(ifr.ifr_name)) { 1495 error = ENAMETOOLONG; 1496 break; 1497 } 1498 1499 addrs = 0; 1500 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1501 struct sockaddr *sa = ifa->ifa_addr; 1502 1503 if (space < sizeof(ifr)) 1504 break; 1505 if (jailed(curthread->td_ucred) && 1506 prison_if(curthread->td_ucred, sa)) 1507 continue; 1508 addrs++; 1509 #ifdef COMPAT_43 1510 if (cmd == OSIOCGIFCONF) { 1511 struct osockaddr *osa = 1512 (struct osockaddr *)&ifr.ifr_addr; 1513 ifr.ifr_addr = *sa; 1514 osa->sa_family = sa->sa_family; 1515 error = copyout((caddr_t)&ifr, (caddr_t)ifrp, 1516 sizeof (ifr)); 1517 ifrp++; 1518 } else 1519 #endif 1520 if (sa->sa_len <= sizeof(*sa)) { 1521 ifr.ifr_addr = *sa; 1522 error = copyout((caddr_t)&ifr, (caddr_t)ifrp, 1523 sizeof (ifr)); 1524 ifrp++; 1525 } else { 1526 if (space < sizeof (ifr) + sa->sa_len - 1527 sizeof(*sa)) 1528 break; 1529 space -= sa->sa_len - sizeof(*sa); 1530 error = copyout((caddr_t)&ifr, (caddr_t)ifrp, 1531 sizeof (ifr.ifr_name)); 1532 if (error == 0) 1533 error = copyout((caddr_t)sa, 1534 (caddr_t)&ifrp->ifr_addr, sa->sa_len); 1535 ifrp = (struct ifreq *) 1536 (sa->sa_len + (caddr_t)&ifrp->ifr_addr); 1537 } 1538 if (error) 1539 break; 1540 space -= sizeof (ifr); 1541 } 1542 if (error) 1543 break; 1544 if (!addrs) { 1545 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 1546 error = copyout((caddr_t)&ifr, (caddr_t)ifrp, 1547 sizeof (ifr)); 1548 if (error) 1549 break; 1550 space -= sizeof (ifr); 1551 ifrp++; 1552 } 1553 } 1554 IFNET_RUNLOCK(); 1555 ifc->ifc_len -= space; 1556 return (error); 1557 } 1558 1559 /* 1560 * Just like if_promisc(), but for all-multicast-reception mode. 1561 */ 1562 int 1563 if_allmulti(struct ifnet *ifp, int onswitch) 1564 { 1565 int error = 0; 1566 int s = splimp(); 1567 struct ifreq ifr; 1568 1569 if (onswitch) { 1570 if (ifp->if_amcount++ == 0) { 1571 ifp->if_flags |= IFF_ALLMULTI; 1572 ifr.ifr_flags = ifp->if_flags & 0xffff; 1573 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1574 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1575 } 1576 } else { 1577 if (ifp->if_amcount > 1) { 1578 ifp->if_amcount--; 1579 } else { 1580 ifp->if_amcount = 0; 1581 ifp->if_flags &= ~IFF_ALLMULTI; 1582 ifr.ifr_flags = ifp->if_flags & 0xffff;; 1583 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1584 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1585 } 1586 } 1587 splx(s); 1588 1589 if (error == 0) 1590 rt_ifmsg(ifp); 1591 return error; 1592 } 1593 1594 /* 1595 * Add a multicast listenership to the interface in question. 1596 * The link layer provides a routine which converts 1597 */ 1598 int 1599 if_addmulti(struct ifnet *ifp, struct sockaddr *sa, struct ifmultiaddr **retifma) 1600 { 1601 struct sockaddr *llsa, *dupsa; 1602 int error, s; 1603 struct ifmultiaddr *ifma; 1604 1605 /* 1606 * If the matching multicast address already exists 1607 * then don't add a new one, just add a reference 1608 */ 1609 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1610 if (equal(sa, ifma->ifma_addr)) { 1611 ifma->ifma_refcount++; 1612 if (retifma) 1613 *retifma = ifma; 1614 return 0; 1615 } 1616 } 1617 1618 /* 1619 * Give the link layer a chance to accept/reject it, and also 1620 * find out which AF_LINK address this maps to, if it isn't one 1621 * already. 1622 */ 1623 if (ifp->if_resolvemulti) { 1624 error = ifp->if_resolvemulti(ifp, &llsa, sa); 1625 if (error) return error; 1626 } else { 1627 llsa = 0; 1628 } 1629 1630 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK); 1631 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK); 1632 bcopy(sa, dupsa, sa->sa_len); 1633 1634 ifma->ifma_addr = dupsa; 1635 ifma->ifma_lladdr = llsa; 1636 ifma->ifma_ifp = ifp; 1637 ifma->ifma_refcount = 1; 1638 ifma->ifma_protospec = 0; 1639 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 1640 1641 /* 1642 * Some network interfaces can scan the address list at 1643 * interrupt time; lock them out. 1644 */ 1645 s = splimp(); 1646 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 1647 splx(s); 1648 if (retifma != NULL) 1649 *retifma = ifma; 1650 1651 if (llsa != 0) { 1652 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1653 if (equal(ifma->ifma_addr, llsa)) 1654 break; 1655 } 1656 if (ifma) { 1657 ifma->ifma_refcount++; 1658 } else { 1659 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, 1660 M_IFMADDR, M_WAITOK); 1661 MALLOC(dupsa, struct sockaddr *, llsa->sa_len, 1662 M_IFMADDR, M_WAITOK); 1663 bcopy(llsa, dupsa, llsa->sa_len); 1664 ifma->ifma_addr = dupsa; 1665 ifma->ifma_lladdr = NULL; 1666 ifma->ifma_ifp = ifp; 1667 ifma->ifma_refcount = 1; 1668 ifma->ifma_protospec = 0; 1669 s = splimp(); 1670 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 1671 splx(s); 1672 } 1673 } 1674 /* 1675 * We are certain we have added something, so call down to the 1676 * interface to let them know about it. 1677 */ 1678 s = splimp(); 1679 ifp->if_ioctl(ifp, SIOCADDMULTI, 0); 1680 splx(s); 1681 1682 return 0; 1683 } 1684 1685 /* 1686 * Remove a reference to a multicast address on this interface. Yell 1687 * if the request does not match an existing membership. 1688 */ 1689 int 1690 if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 1691 { 1692 struct ifmultiaddr *ifma; 1693 int s; 1694 1695 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 1696 if (equal(sa, ifma->ifma_addr)) 1697 break; 1698 if (ifma == 0) 1699 return ENOENT; 1700 1701 if (ifma->ifma_refcount > 1) { 1702 ifma->ifma_refcount--; 1703 return 0; 1704 } 1705 1706 rt_newmaddrmsg(RTM_DELMADDR, ifma); 1707 sa = ifma->ifma_lladdr; 1708 s = splimp(); 1709 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 1710 /* 1711 * Make sure the interface driver is notified 1712 * in the case of a link layer mcast group being left. 1713 */ 1714 if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0) 1715 ifp->if_ioctl(ifp, SIOCDELMULTI, 0); 1716 splx(s); 1717 free(ifma->ifma_addr, M_IFMADDR); 1718 free(ifma, M_IFMADDR); 1719 if (sa == 0) 1720 return 0; 1721 1722 /* 1723 * Now look for the link-layer address which corresponds to 1724 * this network address. It had been squirreled away in 1725 * ifma->ifma_lladdr for this purpose (so we don't have 1726 * to call ifp->if_resolvemulti() again), and we saved that 1727 * value in sa above. If some nasty deleted the 1728 * link-layer address out from underneath us, we can deal because 1729 * the address we stored was is not the same as the one which was 1730 * in the record for the link-layer address. (So we don't complain 1731 * in that case.) 1732 */ 1733 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 1734 if (equal(sa, ifma->ifma_addr)) 1735 break; 1736 if (ifma == 0) 1737 return 0; 1738 1739 if (ifma->ifma_refcount > 1) { 1740 ifma->ifma_refcount--; 1741 return 0; 1742 } 1743 1744 s = splimp(); 1745 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 1746 ifp->if_ioctl(ifp, SIOCDELMULTI, 0); 1747 splx(s); 1748 free(ifma->ifma_addr, M_IFMADDR); 1749 free(sa, M_IFMADDR); 1750 free(ifma, M_IFMADDR); 1751 1752 return 0; 1753 } 1754 1755 /* 1756 * Set the link layer address on an interface. 1757 * 1758 * At this time we only support certain types of interfaces, 1759 * and we don't allow the length of the address to change. 1760 */ 1761 int 1762 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 1763 { 1764 struct sockaddr_dl *sdl; 1765 struct ifaddr *ifa; 1766 struct ifreq ifr; 1767 1768 ifa = ifaddr_byindex(ifp->if_index); 1769 if (ifa == NULL) 1770 return (EINVAL); 1771 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 1772 if (sdl == NULL) 1773 return (EINVAL); 1774 if (len != sdl->sdl_alen) /* don't allow length to change */ 1775 return (EINVAL); 1776 switch (ifp->if_type) { 1777 case IFT_ETHER: /* these types use struct arpcom */ 1778 case IFT_FDDI: 1779 case IFT_XETHER: 1780 case IFT_ISO88025: 1781 case IFT_L2VLAN: 1782 bcopy(lladdr, IFP2AC(ifp)->ac_enaddr, len); 1783 /* 1784 * XXX We also need to store the lladdr in LLADDR(sdl), 1785 * which is done below. This is a pain because we must 1786 * remember to keep the info in sync. 1787 */ 1788 /* FALLTHROUGH */ 1789 case IFT_ARCNET: 1790 bcopy(lladdr, LLADDR(sdl), len); 1791 break; 1792 default: 1793 return (ENODEV); 1794 } 1795 /* 1796 * If the interface is already up, we need 1797 * to re-init it in order to reprogram its 1798 * address filter. 1799 */ 1800 if ((ifp->if_flags & IFF_UP) != 0) { 1801 ifp->if_flags &= ~IFF_UP; 1802 ifr.ifr_flags = ifp->if_flags & 0xffff; 1803 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1804 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1805 ifp->if_flags |= IFF_UP; 1806 ifr.ifr_flags = ifp->if_flags & 0xffff; 1807 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1808 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1809 #ifdef INET 1810 /* 1811 * Also send gratuitous ARPs to notify other nodes about 1812 * the address change. 1813 */ 1814 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1815 if (ifa->ifa_addr != NULL && 1816 ifa->ifa_addr->sa_family == AF_INET) 1817 arp_ifinit(ifp, ifa); 1818 } 1819 #endif 1820 } 1821 return (0); 1822 } 1823 1824 struct ifmultiaddr * 1825 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp) 1826 { 1827 struct ifmultiaddr *ifma; 1828 1829 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 1830 if (equal(ifma->ifma_addr, sa)) 1831 break; 1832 1833 return ifma; 1834 } 1835 1836 /* 1837 * The name argument must be a pointer to storage which will last as 1838 * long as the interface does. For physical devices, the result of 1839 * device_get_name(dev) is a good choice and for pseudo-devices a 1840 * static string works well. 1841 */ 1842 void 1843 if_initname(struct ifnet *ifp, const char *name, int unit) 1844 { 1845 ifp->if_dname = name; 1846 ifp->if_dunit = unit; 1847 if (unit != IF_DUNIT_NONE) 1848 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 1849 else 1850 strlcpy(ifp->if_xname, name, IFNAMSIZ); 1851 } 1852 1853 int 1854 if_printf(struct ifnet *ifp, const char * fmt, ...) 1855 { 1856 va_list ap; 1857 int retval; 1858 1859 retval = printf("%s: ", ifp->if_xname); 1860 va_start(ap, fmt); 1861 retval += vprintf(fmt, ap); 1862 va_end(ap); 1863 return (retval); 1864 } 1865 1866 /* 1867 * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot 1868 * be called without Giant. However, we often can't acquire the Giant lock 1869 * at those points; instead, we run it via a task queue that holds Giant via 1870 * if_start_deferred. 1871 * 1872 * XXXRW: We need to make sure that the ifnet isn't fully detached until any 1873 * outstanding if_start_deferred() tasks that will run after the free. This 1874 * probably means waiting in if_detach(). 1875 */ 1876 void 1877 if_start(struct ifnet *ifp) 1878 { 1879 1880 NET_ASSERT_GIANT(); 1881 1882 if ((ifp->if_flags & IFF_NEEDSGIANT) != 0 && debug_mpsafenet != 0) { 1883 if (mtx_owned(&Giant)) 1884 (*(ifp)->if_start)(ifp); 1885 else 1886 taskqueue_enqueue(taskqueue_swi_giant, 1887 &ifp->if_starttask); 1888 } else 1889 (*(ifp)->if_start)(ifp); 1890 } 1891 1892 static void 1893 if_start_deferred(void *context, int pending) 1894 { 1895 struct ifnet *ifp; 1896 1897 /* 1898 * This code must be entered with Giant, and should never run if 1899 * we're not running with debug.mpsafenet. 1900 */ 1901 KASSERT(debug_mpsafenet != 0, ("if_start_deferred: debug.mpsafenet")); 1902 GIANT_REQUIRED; 1903 1904 ifp = (struct ifnet *)context; 1905 (ifp->if_start)(ifp); 1906 } 1907 1908 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 1909 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 1910