if.c revision 134399
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: head/sys/net/if.c 134399 2004-08-27 19:42:40Z brooks $ 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 80struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL; 81 82static void if_attachdomain(void *); 83static void if_attachdomain1(struct ifnet *); 84static int ifconf(u_long, caddr_t); 85static void if_grow(void); 86static void if_init(void *); 87static void if_check(void *); 88static int if_findindex(struct ifnet *); 89static void if_qflush(struct ifaltq *); 90static void if_route(struct ifnet *, int flag, int fam); 91static void if_slowtimo(void *); 92static void if_unroute(struct ifnet *, int flag, int fam); 93static void link_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 94static int if_rtdel(struct radix_node *, void *); 95static int ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *); 96static 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 */ 102extern void nd6_setmtu(struct ifnet *); 103#endif 104 105int if_index = 0; 106struct ifindex_entry *ifindex_table = NULL; 107int ifqmaxlen = IFQ_MAXLEN; 108struct ifnethead ifnet; /* depend on static init XXX */ 109struct mtx ifnet_lock; 110 111static int if_indexlim = 8; 112static struct knlist ifklist; 113 114static void filt_netdetach(struct knote *kn); 115static int filt_netdev(struct knote *kn, long hint); 116 117static struct filterops netdev_filtops = 118 { 1, NULL, filt_netdetach, filt_netdev }; 119 120/* 121 * System initialization 122 */ 123SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL) 124SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL) 125 126MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address"); 127MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address"); 128 129static d_open_t netopen; 130static d_close_t netclose; 131static d_ioctl_t netioctl; 132static d_kqfilter_t netkqfilter; 133 134static 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 144static int 145netopen(struct cdev *dev, int flag, int mode, struct thread *td) 146{ 147 return (0); 148} 149 150static int 151netclose(struct cdev *dev, int flags, int fmt, struct thread *td) 152{ 153 return (0); 154} 155 156static int 157netioctl(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 185static int 186netkqfilter(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 217static void 218filt_netdetach(struct knote *kn) 219{ 220 struct knlist *klist = (struct knlist *)kn->kn_hook; 221 222 if (kn->kn_status & KN_DETACHED) 223 return; 224 225 knlist_remove(klist, kn, 0); 226} 227 228static int 229filt_netdev(struct knote *kn, long hint) 230{ 231 struct knlist *klist = (struct knlist *)kn->kn_hook; 232 233 /* 234 * Currently NOTE_EXIT is abused to indicate device detach. 235 */ 236 if (hint == NOTE_EXIT) { 237 kn->kn_data = NOTE_LINKINV; 238 kn->kn_flags |= (EV_EOF | EV_ONESHOT); 239 knlist_remove_inevent(klist, kn); 240 return (1); 241 } 242 if (hint != 0) 243 kn->kn_data = hint; /* current status */ 244 if (kn->kn_sfflags & hint) 245 kn->kn_fflags |= hint; 246 return (kn->kn_fflags != 0); 247} 248 249/* 250 * Network interface utility routines. 251 * 252 * Routines with ifa_ifwith* names take sockaddr *'s as 253 * parameters. 254 */ 255/* ARGSUSED*/ 256static void 257if_init(void *dummy __unused) 258{ 259 260 IFNET_LOCK_INIT(); 261 TAILQ_INIT(&ifnet); 262 knlist_init(&ifklist, NULL); 263 if_grow(); /* create initial table */ 264 ifdev_byindex(0) = make_dev(&net_cdevsw, 0, 265 UID_ROOT, GID_WHEEL, 0600, "network"); 266 if_clone_init(); 267} 268 269static void 270if_grow(void) 271{ 272 u_int n; 273 struct ifindex_entry *e; 274 275 if_indexlim <<= 1; 276 n = if_indexlim * sizeof(*e); 277 e = malloc(n, M_IFADDR, M_WAITOK | M_ZERO); 278 if (ifindex_table != NULL) { 279 memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2); 280 free((caddr_t)ifindex_table, M_IFADDR); 281 } 282 ifindex_table = e; 283} 284 285/* ARGSUSED*/ 286static void 287if_check(void *dummy __unused) 288{ 289 struct ifnet *ifp; 290 int s; 291 292 s = splimp(); 293 IFNET_RLOCK(); /* could sleep on rare error; mostly okay XXX */ 294 TAILQ_FOREACH(ifp, &ifnet, if_link) { 295 if (ifp->if_snd.ifq_maxlen == 0) { 296 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n"); 297 ifp->if_snd.ifq_maxlen = ifqmaxlen; 298 } 299 if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) { 300 if_printf(ifp, 301 "XXX: driver didn't initialize queue mtx\n"); 302 mtx_init(&ifp->if_snd.ifq_mtx, "unknown", 303 MTX_NETWORK_LOCK, MTX_DEF); 304 } 305 } 306 IFNET_RUNLOCK(); 307 splx(s); 308 if_slowtimo(0); 309} 310 311static int 312if_findindex(struct ifnet *ifp) 313{ 314 int i, unit; 315 char eaddr[18], devname[32]; 316 const char *name, *p; 317 318 switch (ifp->if_type) { 319 case IFT_ETHER: /* these types use struct arpcom */ 320 case IFT_FDDI: 321 case IFT_XETHER: 322 case IFT_ISO88025: 323 case IFT_L2VLAN: 324 snprintf(eaddr, 18, "%6D", IFP2AC(ifp)->ac_enaddr, ":"); 325 break; 326 default: 327 eaddr[0] = '\0'; 328 break; 329 } 330 strlcpy(devname, ifp->if_xname, sizeof(devname)); 331 name = net_cdevsw.d_name; 332 i = 0; 333 while ((resource_find_dev(&i, name, &unit, NULL, NULL)) == 0) { 334 if (resource_string_value(name, unit, "ether", &p) == 0) 335 if (strcmp(p, eaddr) == 0) 336 goto found; 337 if (resource_string_value(name, unit, "dev", &p) == 0) 338 if (strcmp(p, devname) == 0) 339 goto found; 340 } 341 unit = 0; 342found: 343 if (unit != 0) { 344 if (ifaddr_byindex(unit) == NULL) 345 return (unit); 346 printf("%s%d in use, cannot hardwire it to %s.\n", 347 name, unit, devname); 348 } 349 for (unit = 1; ; unit++) { 350 if (unit <= if_index && ifaddr_byindex(unit) != NULL) 351 continue; 352 if (resource_string_value(name, unit, "ether", &p) == 0 || 353 resource_string_value(name, unit, "dev", &p) == 0) 354 continue; 355 break; 356 } 357 return (unit); 358} 359 360/* 361 * Attach an interface to the 362 * list of "active" interfaces. 363 */ 364void 365if_attach(struct ifnet *ifp) 366{ 367 unsigned socksize, ifasize; 368 int namelen, masklen; 369 struct sockaddr_dl *sdl; 370 struct ifaddr *ifa; 371 372 TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp); 373 IF_AFDATA_LOCK_INIT(ifp); 374 ifp->if_afdata_initialized = 0; 375 IFNET_WLOCK(); 376 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link); 377 IFNET_WUNLOCK(); 378 /* 379 * XXX - 380 * The old code would work if the interface passed a pre-existing 381 * chain of ifaddrs to this code. We don't trust our callers to 382 * properly initialize the tailq, however, so we no longer allow 383 * this unlikely case. 384 */ 385 TAILQ_INIT(&ifp->if_addrhead); 386 TAILQ_INIT(&ifp->if_prefixhead); 387 TAILQ_INIT(&ifp->if_multiaddrs); 388 knlist_init(&ifp->if_klist, NULL); 389 getmicrotime(&ifp->if_lastchange); 390 391#ifdef MAC 392 mac_init_ifnet(ifp); 393 mac_create_ifnet(ifp); 394#endif 395 396 ifp->if_index = if_findindex(ifp); 397 if (ifp->if_index > if_index) 398 if_index = ifp->if_index; 399 if (if_index >= if_indexlim) 400 if_grow(); 401 402 ifnet_byindex(ifp->if_index) = ifp; 403 ifdev_byindex(ifp->if_index) = make_dev(&net_cdevsw, 404 unit2minor(ifp->if_index), 405 UID_ROOT, GID_WHEEL, 0600, "%s/%s", 406 net_cdevsw.d_name, ifp->if_xname); 407 make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d", 408 net_cdevsw.d_name, ifp->if_index); 409 410 mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF); 411 412 /* 413 * create a Link Level name for this device 414 */ 415 namelen = strlen(ifp->if_xname); 416 /* 417 * Always save enough space for any possiable name so we can do 418 * a rename in place later. 419 */ 420 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ; 421 socksize = masklen + ifp->if_addrlen; 422 if (socksize < sizeof(*sdl)) 423 socksize = sizeof(*sdl); 424 socksize = roundup2(socksize, sizeof(long)); 425 ifasize = sizeof(*ifa) + 2 * socksize; 426 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO); 427 IFA_LOCK_INIT(ifa); 428 sdl = (struct sockaddr_dl *)(ifa + 1); 429 sdl->sdl_len = socksize; 430 sdl->sdl_family = AF_LINK; 431 bcopy(ifp->if_xname, sdl->sdl_data, namelen); 432 sdl->sdl_nlen = namelen; 433 sdl->sdl_index = ifp->if_index; 434 sdl->sdl_type = ifp->if_type; 435 ifaddr_byindex(ifp->if_index) = ifa; 436 ifa->ifa_ifp = ifp; 437 ifa->ifa_rtrequest = link_rtrequest; 438 ifa->ifa_addr = (struct sockaddr *)sdl; 439 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl); 440 ifa->ifa_netmask = (struct sockaddr *)sdl; 441 sdl->sdl_len = masklen; 442 while (namelen != 0) 443 sdl->sdl_data[--namelen] = 0xff; 444 ifa->ifa_refcnt = 1; 445 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link); 446 ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */ 447 ifp->if_snd.altq_type = 0; 448 ifp->if_snd.altq_disc = NULL; 449 ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE; 450 ifp->if_snd.altq_tbr = NULL; 451 ifp->if_snd.altq_ifp = ifp; 452 453 if (domains) 454 if_attachdomain1(ifp); 455 456 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 457 458 /* Announce the interface. */ 459 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 460} 461 462static void 463if_attachdomain(void *dummy) 464{ 465 struct ifnet *ifp; 466 int s; 467 468 s = splnet(); 469 TAILQ_FOREACH(ifp, &ifnet, if_link) 470 if_attachdomain1(ifp); 471 splx(s); 472} 473SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST, 474 if_attachdomain, NULL); 475 476static void 477if_attachdomain1(struct ifnet *ifp) 478{ 479 struct domain *dp; 480 int s; 481 482 s = splnet(); 483 484 /* 485 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we 486 * cannot lock ifp->if_afdata initialization, entirely. 487 */ 488 if (IF_AFDATA_TRYLOCK(ifp) == 0) { 489 splx(s); 490 return; 491 } 492 if (ifp->if_afdata_initialized) { 493 IF_AFDATA_UNLOCK(ifp); 494 splx(s); 495 return; 496 } 497 ifp->if_afdata_initialized = 1; 498 IF_AFDATA_UNLOCK(ifp); 499 500 /* address family dependent data region */ 501 bzero(ifp->if_afdata, sizeof(ifp->if_afdata)); 502 for (dp = domains; dp; dp = dp->dom_next) { 503 if (dp->dom_ifattach) 504 ifp->if_afdata[dp->dom_family] = 505 (*dp->dom_ifattach)(ifp); 506 } 507 508 splx(s); 509} 510 511/* 512 * Detach an interface, removing it from the 513 * list of "active" interfaces. 514 */ 515void 516if_detach(struct ifnet *ifp) 517{ 518 struct ifaddr *ifa, *next; 519 struct radix_node_head *rnh; 520 int s; 521 int i; 522 struct domain *dp; 523 struct ifnet *iter; 524 int found; 525 526 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 527 /* 528 * Remove routes and flush queues. 529 */ 530 s = splnet(); 531 if_down(ifp); 532#ifdef ALTQ 533 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 534 altq_disable(&ifp->if_snd); 535 if (ALTQ_IS_ATTACHED(&ifp->if_snd)) 536 altq_detach(&ifp->if_snd); 537#endif 538 539 for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa; ifa = next) { 540 next = TAILQ_NEXT(ifa, ifa_link); 541 542 if (ifa->ifa_addr->sa_family == AF_LINK) 543 continue; 544#ifdef INET 545 /* XXX: Ugly!! ad hoc just for INET */ 546 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) { 547 struct ifaliasreq ifr; 548 549 bzero(&ifr, sizeof(ifr)); 550 ifr.ifra_addr = *ifa->ifa_addr; 551 if (ifa->ifa_dstaddr) 552 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 553 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp, 554 NULL) == 0) 555 continue; 556 } 557#endif /* INET */ 558#ifdef INET6 559 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) { 560 in6_purgeaddr(ifa); 561 /* ifp_addrhead is already updated */ 562 continue; 563 } 564#endif /* INET6 */ 565 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 566 IFAFREE(ifa); 567 } 568 569#ifdef INET6 570 /* 571 * Remove all IPv6 kernel structs related to ifp. This should be done 572 * before removing routing entries below, since IPv6 interface direct 573 * routes are expected to be removed by the IPv6-specific kernel API. 574 * Otherwise, the kernel will detect some inconsistency and bark it. 575 */ 576 in6_ifdetach(ifp); 577#endif 578 /* 579 * Remove address from ifindex_table[] and maybe decrement if_index. 580 * Clean up all addresses. 581 */ 582 ifnet_byindex(ifp->if_index) = NULL; 583 ifaddr_byindex(ifp->if_index) = NULL; 584 destroy_dev(ifdev_byindex(ifp->if_index)); 585 ifdev_byindex(ifp->if_index) = NULL; 586 587 while (if_index > 0 && ifaddr_byindex(if_index) == NULL) 588 if_index--; 589 590 591 /* We can now free link ifaddr. */ 592 if (!TAILQ_EMPTY(&ifp->if_addrhead)) { 593 ifa = TAILQ_FIRST(&ifp->if_addrhead); 594 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link); 595 IFAFREE(ifa); 596 } 597 598 /* 599 * Delete all remaining routes using this interface 600 * Unfortuneatly the only way to do this is to slog through 601 * the entire routing table looking for routes which point 602 * to this interface...oh well... 603 */ 604 for (i = 1; i <= AF_MAX; i++) { 605 if ((rnh = rt_tables[i]) == NULL) 606 continue; 607 RADIX_NODE_HEAD_LOCK(rnh); 608 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp); 609 RADIX_NODE_HEAD_UNLOCK(rnh); 610 } 611 612 /* Announce that the interface is gone. */ 613 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 614 615 IF_AFDATA_LOCK(ifp); 616 for (dp = domains; dp; dp = dp->dom_next) { 617 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) 618 (*dp->dom_ifdetach)(ifp, 619 ifp->if_afdata[dp->dom_family]); 620 } 621 IF_AFDATA_UNLOCK(ifp); 622 623#ifdef MAC 624 mac_destroy_ifnet(ifp); 625#endif /* MAC */ 626 KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT); 627 knlist_clear(&ifp->if_klist, 0); 628 knlist_destroy(&ifp->if_klist); 629 IFNET_WLOCK(); 630 found = 0; 631 TAILQ_FOREACH(iter, &ifnet, if_link) 632 if (iter == ifp) { 633 found = 1; 634 break; 635 } 636 if (found) 637 TAILQ_REMOVE(&ifnet, ifp, if_link); 638 IFNET_WUNLOCK(); 639 mtx_destroy(&ifp->if_snd.ifq_mtx); 640 IF_AFDATA_DESTROY(ifp); 641 splx(s); 642} 643 644/* 645 * Delete Routes for a Network Interface 646 * 647 * Called for each routing entry via the rnh->rnh_walktree() call above 648 * to delete all route entries referencing a detaching network interface. 649 * 650 * Arguments: 651 * rn pointer to node in the routing table 652 * arg argument passed to rnh->rnh_walktree() - detaching interface 653 * 654 * Returns: 655 * 0 successful 656 * errno failed - reason indicated 657 * 658 */ 659static int 660if_rtdel(struct radix_node *rn, void *arg) 661{ 662 struct rtentry *rt = (struct rtentry *)rn; 663 struct ifnet *ifp = arg; 664 int err; 665 666 if (rt->rt_ifp == ifp) { 667 668 /* 669 * Protect (sorta) against walktree recursion problems 670 * with cloned routes 671 */ 672 if ((rt->rt_flags & RTF_UP) == 0) 673 return (0); 674 675 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway, 676 rt_mask(rt), rt->rt_flags, 677 (struct rtentry **) NULL); 678 if (err) { 679 log(LOG_WARNING, "if_rtdel: error %d\n", err); 680 } 681 } 682 683 return (0); 684} 685 686#define equal(a1, a2) (bcmp((a1), (a2), ((a1))->sa_len) == 0) 687 688/* 689 * Locate an interface based on a complete address. 690 */ 691/*ARGSUSED*/ 692struct ifaddr * 693ifa_ifwithaddr(struct sockaddr *addr) 694{ 695 struct ifnet *ifp; 696 struct ifaddr *ifa; 697 698 IFNET_RLOCK(); 699 TAILQ_FOREACH(ifp, &ifnet, if_link) 700 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 701 if (ifa->ifa_addr->sa_family != addr->sa_family) 702 continue; 703 if (equal(addr, ifa->ifa_addr)) 704 goto done; 705 /* IP6 doesn't have broadcast */ 706 if ((ifp->if_flags & IFF_BROADCAST) && 707 ifa->ifa_broadaddr && 708 ifa->ifa_broadaddr->sa_len != 0 && 709 equal(ifa->ifa_broadaddr, addr)) 710 goto done; 711 } 712 ifa = NULL; 713done: 714 IFNET_RUNLOCK(); 715 return (ifa); 716} 717 718/* 719 * Locate the point to point interface with a given destination address. 720 */ 721/*ARGSUSED*/ 722struct ifaddr * 723ifa_ifwithdstaddr(struct sockaddr *addr) 724{ 725 struct ifnet *ifp; 726 struct ifaddr *ifa; 727 728 IFNET_RLOCK(); 729 TAILQ_FOREACH(ifp, &ifnet, if_link) { 730 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) 731 continue; 732 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 733 if (ifa->ifa_addr->sa_family != addr->sa_family) 734 continue; 735 if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)) 736 goto done; 737 } 738 } 739 ifa = NULL; 740done: 741 IFNET_RUNLOCK(); 742 return (ifa); 743} 744 745/* 746 * Find an interface on a specific network. If many, choice 747 * is most specific found. 748 */ 749struct ifaddr * 750ifa_ifwithnet(struct sockaddr *addr) 751{ 752 struct ifnet *ifp; 753 struct ifaddr *ifa; 754 struct ifaddr *ifa_maybe = (struct ifaddr *) 0; 755 u_int af = addr->sa_family; 756 char *addr_data = addr->sa_data, *cplim; 757 758 /* 759 * AF_LINK addresses can be looked up directly by their index number, 760 * so do that if we can. 761 */ 762 if (af == AF_LINK) { 763 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr; 764 if (sdl->sdl_index && sdl->sdl_index <= if_index) 765 return (ifaddr_byindex(sdl->sdl_index)); 766 } 767 768 /* 769 * Scan though each interface, looking for ones that have 770 * addresses in this address family. 771 */ 772 IFNET_RLOCK(); 773 TAILQ_FOREACH(ifp, &ifnet, if_link) { 774 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 775 char *cp, *cp2, *cp3; 776 777 if (ifa->ifa_addr->sa_family != af) 778next: continue; 779 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) { 780 /* 781 * This is a bit broken as it doesn't 782 * take into account that the remote end may 783 * be a single node in the network we are 784 * looking for. 785 * The trouble is that we don't know the 786 * netmask for the remote end. 787 */ 788 if (ifa->ifa_dstaddr != 0 789 && equal(addr, ifa->ifa_dstaddr)) 790 goto done; 791 } else { 792 /* 793 * if we have a special address handler, 794 * then use it instead of the generic one. 795 */ 796 if (ifa->ifa_claim_addr) { 797 if ((*ifa->ifa_claim_addr)(ifa, addr)) 798 goto done; 799 continue; 800 } 801 802 /* 803 * Scan all the bits in the ifa's address. 804 * If a bit dissagrees with what we are 805 * looking for, mask it with the netmask 806 * to see if it really matters. 807 * (A byte at a time) 808 */ 809 if (ifa->ifa_netmask == 0) 810 continue; 811 cp = addr_data; 812 cp2 = ifa->ifa_addr->sa_data; 813 cp3 = ifa->ifa_netmask->sa_data; 814 cplim = ifa->ifa_netmask->sa_len 815 + (char *)ifa->ifa_netmask; 816 while (cp3 < cplim) 817 if ((*cp++ ^ *cp2++) & *cp3++) 818 goto next; /* next address! */ 819 /* 820 * If the netmask of what we just found 821 * is more specific than what we had before 822 * (if we had one) then remember the new one 823 * before continuing to search 824 * for an even better one. 825 */ 826 if (ifa_maybe == 0 || 827 rn_refines((caddr_t)ifa->ifa_netmask, 828 (caddr_t)ifa_maybe->ifa_netmask)) 829 ifa_maybe = ifa; 830 } 831 } 832 } 833 ifa = ifa_maybe; 834done: 835 IFNET_RUNLOCK(); 836 return (ifa); 837} 838 839/* 840 * Find an interface address specific to an interface best matching 841 * a given address. 842 */ 843struct ifaddr * 844ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp) 845{ 846 struct ifaddr *ifa; 847 char *cp, *cp2, *cp3; 848 char *cplim; 849 struct ifaddr *ifa_maybe = 0; 850 u_int af = addr->sa_family; 851 852 if (af >= AF_MAX) 853 return (0); 854 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 855 if (ifa->ifa_addr->sa_family != af) 856 continue; 857 if (ifa_maybe == 0) 858 ifa_maybe = ifa; 859 if (ifa->ifa_netmask == 0) { 860 if (equal(addr, ifa->ifa_addr) || 861 (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))) 862 goto done; 863 continue; 864 } 865 if (ifp->if_flags & IFF_POINTOPOINT) { 866 if (equal(addr, ifa->ifa_dstaddr)) 867 goto done; 868 } else { 869 cp = addr->sa_data; 870 cp2 = ifa->ifa_addr->sa_data; 871 cp3 = ifa->ifa_netmask->sa_data; 872 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask; 873 for (; cp3 < cplim; cp3++) 874 if ((*cp++ ^ *cp2++) & *cp3) 875 break; 876 if (cp3 == cplim) 877 goto done; 878 } 879 } 880 ifa = ifa_maybe; 881done: 882 return (ifa); 883} 884 885#include <net/route.h> 886 887/* 888 * Default action when installing a route with a Link Level gateway. 889 * Lookup an appropriate real ifa to point to. 890 * This should be moved to /sys/net/link.c eventually. 891 */ 892static void 893link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info) 894{ 895 struct ifaddr *ifa, *oifa; 896 struct sockaddr *dst; 897 struct ifnet *ifp; 898 899 RT_LOCK_ASSERT(rt); 900 901 if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) || 902 ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0)) 903 return; 904 ifa = ifaof_ifpforaddr(dst, ifp); 905 if (ifa) { 906 IFAREF(ifa); /* XXX */ 907 oifa = rt->rt_ifa; 908 rt->rt_ifa = ifa; 909 IFAFREE(oifa); 910 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest) 911 ifa->ifa_rtrequest(cmd, rt, info); 912 } 913} 914 915/* 916 * Mark an interface down and notify protocols of 917 * the transition. 918 * NOTE: must be called at splnet or eqivalent. 919 */ 920static void 921if_unroute(struct ifnet *ifp, int flag, int fam) 922{ 923 struct ifaddr *ifa; 924 925 ifp->if_flags &= ~flag; 926 getmicrotime(&ifp->if_lastchange); 927 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 928 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 929 pfctlinput(PRC_IFDOWN, ifa->ifa_addr); 930 if_qflush(&ifp->if_snd); 931 rt_ifmsg(ifp); 932} 933 934/* 935 * Mark an interface up and notify protocols of 936 * the transition. 937 * NOTE: must be called at splnet or eqivalent. 938 */ 939static void 940if_route(struct ifnet *ifp, int flag, int fam) 941{ 942 struct ifaddr *ifa; 943 944 ifp->if_flags |= flag; 945 getmicrotime(&ifp->if_lastchange); 946 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 947 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family)) 948 pfctlinput(PRC_IFUP, ifa->ifa_addr); 949 rt_ifmsg(ifp); 950#ifdef INET6 951 in6_if_up(ifp); 952#endif 953} 954 955/* 956 * Mark an interface down and notify protocols of 957 * the transition. 958 * NOTE: must be called at splnet or eqivalent. 959 */ 960void 961if_down(struct ifnet *ifp) 962{ 963 964 if_unroute(ifp, IFF_UP, AF_UNSPEC); 965} 966 967/* 968 * Mark an interface up and notify protocols of 969 * the transition. 970 * NOTE: must be called at splnet or eqivalent. 971 */ 972void 973if_up(struct ifnet *ifp) 974{ 975 976 if_route(ifp, IFF_UP, AF_UNSPEC); 977} 978 979/* 980 * Flush an interface queue. 981 */ 982static void 983if_qflush(struct ifaltq *ifq) 984{ 985 struct mbuf *m, *n; 986 987#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} 1000 1001/* 1002 * Handle interface watchdog timer routines. Called 1003 * from softclock, we decrement timers (if set) and 1004 * call the appropriate interface routine on expiration. 1005 * 1006 * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called 1007 * holding Giant. If we switch to an MPSAFE callout, we likely need to grab 1008 * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface. 1009 */ 1010static void 1011if_slowtimo(void *arg) 1012{ 1013 struct ifnet *ifp; 1014 int s = splimp(); 1015 1016 IFNET_RLOCK(); 1017 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1018 if (ifp->if_timer == 0 || --ifp->if_timer) 1019 continue; 1020 if (ifp->if_watchdog) 1021 (*ifp->if_watchdog)(ifp); 1022 } 1023 IFNET_RUNLOCK(); 1024 splx(s); 1025 timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ); 1026} 1027 1028/* 1029 * Map interface name to 1030 * interface structure pointer. 1031 */ 1032struct ifnet * 1033ifunit(const char *name) 1034{ 1035 struct ifnet *ifp; 1036 1037 IFNET_RLOCK(); 1038 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1039 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0) 1040 break; 1041 } 1042 IFNET_RUNLOCK(); 1043 return (ifp); 1044} 1045 1046/* 1047 * Hardware specific interface ioctls. 1048 */ 1049static int 1050ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) 1051{ 1052 struct ifreq *ifr; 1053 struct ifstat *ifs; 1054 int error = 0; 1055 int new_flags; 1056 size_t namelen, onamelen; 1057 char new_name[IFNAMSIZ]; 1058 struct ifaddr *ifa; 1059 struct sockaddr_dl *sdl; 1060 1061 ifr = (struct ifreq *)data; 1062 switch (cmd) { 1063 case SIOCGIFINDEX: 1064 ifr->ifr_index = ifp->if_index; 1065 break; 1066 1067 case SIOCGIFFLAGS: 1068 ifr->ifr_flags = ifp->if_flags & 0xffff; 1069 ifr->ifr_flagshigh = ifp->if_flags >> 16; 1070 break; 1071 1072 case SIOCGIFCAP: 1073 ifr->ifr_reqcap = ifp->if_capabilities; 1074 ifr->ifr_curcap = ifp->if_capenable; 1075 break; 1076 1077#ifdef MAC 1078 case SIOCGIFMAC: 1079 error = mac_ioctl_ifnet_get(td->td_ucred, ifr, ifp); 1080 break; 1081#endif 1082 1083 case SIOCGIFMETRIC: 1084 ifr->ifr_metric = ifp->if_metric; 1085 break; 1086 1087 case SIOCGIFMTU: 1088 ifr->ifr_mtu = ifp->if_mtu; 1089 break; 1090 1091 case SIOCGIFPHYS: 1092 ifr->ifr_phys = ifp->if_physical; 1093 break; 1094 1095 case SIOCSIFFLAGS: 1096 error = suser(td); 1097 if (error) 1098 return (error); 1099 new_flags = (ifr->ifr_flags & 0xffff) | 1100 (ifr->ifr_flagshigh << 16); 1101 if (ifp->if_flags & IFF_SMART) { 1102 /* Smart drivers twiddle their own routes */ 1103 } else if (ifp->if_flags & IFF_UP && 1104 (new_flags & IFF_UP) == 0) { 1105 int s = splimp(); 1106 if_down(ifp); 1107 splx(s); 1108 } else if (new_flags & IFF_UP && 1109 (ifp->if_flags & IFF_UP) == 0) { 1110 int s = splimp(); 1111 if_up(ifp); 1112 splx(s); 1113 } 1114 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) | 1115 (new_flags &~ IFF_CANTCHANGE); 1116 if (new_flags & IFF_PPROMISC) { 1117 /* Permanently promiscuous mode requested */ 1118 ifp->if_flags |= IFF_PROMISC; 1119 } else if (ifp->if_pcount == 0) { 1120 ifp->if_flags &= ~IFF_PROMISC; 1121 } 1122 if (ifp->if_ioctl) 1123 (void) (*ifp->if_ioctl)(ifp, cmd, data); 1124 getmicrotime(&ifp->if_lastchange); 1125 break; 1126 1127 case SIOCSIFCAP: 1128 error = suser(td); 1129 if (error) 1130 return (error); 1131 if (ifp->if_ioctl == NULL) 1132 return (EOPNOTSUPP); 1133 if (ifr->ifr_reqcap & ~ifp->if_capabilities) 1134 return (EINVAL); 1135 error = (*ifp->if_ioctl)(ifp, cmd, data); 1136 if (error == 0) 1137 getmicrotime(&ifp->if_lastchange); 1138 break; 1139 1140#ifdef MAC 1141 case SIOCSIFMAC: 1142 error = mac_ioctl_ifnet_set(td->td_ucred, ifr, ifp); 1143 break; 1144#endif 1145 1146 case SIOCSIFNAME: 1147 error = suser(td); 1148 if (error != 0) 1149 return (error); 1150 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL); 1151 if (error != 0) 1152 return (error); 1153 if (new_name[0] == '\0') 1154 return (EINVAL); 1155 if (ifunit(new_name) != NULL) 1156 return (EEXIST); 1157 1158 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp); 1159 /* Announce the departure of the interface. */ 1160 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 1161 1162 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname)); 1163 ifa = ifaddr_byindex(ifp->if_index); 1164 IFA_LOCK(ifa); 1165 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 1166 namelen = strlen(new_name); 1167 onamelen = sdl->sdl_nlen; 1168 /* 1169 * Move the address if needed. This is safe because we 1170 * allocate space for a name of length IFNAMSIZ when we 1171 * create this in if_attach(). 1172 */ 1173 if (namelen != onamelen) { 1174 bcopy(sdl->sdl_data + onamelen, 1175 sdl->sdl_data + namelen, sdl->sdl_alen); 1176 } 1177 bcopy(new_name, sdl->sdl_data, namelen); 1178 sdl->sdl_nlen = namelen; 1179 sdl = (struct sockaddr_dl *)ifa->ifa_netmask; 1180 bzero(sdl->sdl_data, onamelen); 1181 while (namelen != 0) 1182 sdl->sdl_data[--namelen] = 0xff; 1183 IFA_UNLOCK(ifa); 1184 1185 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp); 1186 /* Announce the return of the interface. */ 1187 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 1188 break; 1189 1190 case SIOCSIFMETRIC: 1191 error = suser(td); 1192 if (error) 1193 return (error); 1194 ifp->if_metric = ifr->ifr_metric; 1195 getmicrotime(&ifp->if_lastchange); 1196 break; 1197 1198 case SIOCSIFPHYS: 1199 error = suser(td); 1200 if (error) 1201 return (error); 1202 if (ifp->if_ioctl == NULL) 1203 return (EOPNOTSUPP); 1204 error = (*ifp->if_ioctl)(ifp, cmd, data); 1205 if (error == 0) 1206 getmicrotime(&ifp->if_lastchange); 1207 break; 1208 1209 case SIOCSIFMTU: 1210 { 1211 u_long oldmtu = ifp->if_mtu; 1212 1213 error = suser(td); 1214 if (error) 1215 return (error); 1216 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU) 1217 return (EINVAL); 1218 if (ifp->if_ioctl == NULL) 1219 return (EOPNOTSUPP); 1220 error = (*ifp->if_ioctl)(ifp, cmd, data); 1221 if (error == 0) { 1222 getmicrotime(&ifp->if_lastchange); 1223 rt_ifmsg(ifp); 1224 } 1225 /* 1226 * If the link MTU changed, do network layer specific procedure. 1227 */ 1228 if (ifp->if_mtu != oldmtu) { 1229#ifdef INET6 1230 nd6_setmtu(ifp); 1231#endif 1232 } 1233 break; 1234 } 1235 1236 case SIOCADDMULTI: 1237 case SIOCDELMULTI: 1238 error = suser(td); 1239 if (error) 1240 return (error); 1241 1242 /* Don't allow group membership on non-multicast interfaces. */ 1243 if ((ifp->if_flags & IFF_MULTICAST) == 0) 1244 return (EOPNOTSUPP); 1245 1246 /* Don't let users screw up protocols' entries. */ 1247 if (ifr->ifr_addr.sa_family != AF_LINK) 1248 return (EINVAL); 1249 1250 if (cmd == SIOCADDMULTI) { 1251 struct ifmultiaddr *ifma; 1252 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma); 1253 } else { 1254 error = if_delmulti(ifp, &ifr->ifr_addr); 1255 } 1256 if (error == 0) 1257 getmicrotime(&ifp->if_lastchange); 1258 break; 1259 1260 case SIOCSIFPHYADDR: 1261 case SIOCDIFPHYADDR: 1262#ifdef INET6 1263 case SIOCSIFPHYADDR_IN6: 1264#endif 1265 case SIOCSLIFPHYADDR: 1266 case SIOCSIFMEDIA: 1267 case SIOCSIFGENERIC: 1268 error = suser(td); 1269 if (error) 1270 return (error); 1271 if (ifp->if_ioctl == NULL) 1272 return (EOPNOTSUPP); 1273 error = (*ifp->if_ioctl)(ifp, cmd, data); 1274 if (error == 0) 1275 getmicrotime(&ifp->if_lastchange); 1276 break; 1277 1278 case SIOCGIFSTATUS: 1279 ifs = (struct ifstat *)data; 1280 ifs->ascii[0] = '\0'; 1281 1282 case SIOCGIFPSRCADDR: 1283 case SIOCGIFPDSTADDR: 1284 case SIOCGLIFPHYADDR: 1285 case SIOCGIFMEDIA: 1286 case SIOCGIFGENERIC: 1287 if (ifp->if_ioctl == NULL) 1288 return (EOPNOTSUPP); 1289 error = (*ifp->if_ioctl)(ifp, cmd, data); 1290 break; 1291 1292 case SIOCSIFLLADDR: 1293 error = suser(td); 1294 if (error) 1295 return (error); 1296 error = if_setlladdr(ifp, 1297 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len); 1298 break; 1299 1300 default: 1301 error = ENOIOCTL; 1302 break; 1303 } 1304 return (error); 1305} 1306 1307/* 1308 * Interface ioctls. 1309 */ 1310int 1311ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td) 1312{ 1313 struct ifnet *ifp; 1314 struct ifreq *ifr; 1315 int error; 1316 int oif_flags; 1317 1318 switch (cmd) { 1319 case SIOCGIFCONF: 1320 case OSIOCGIFCONF: 1321 return (ifconf(cmd, data)); 1322 } 1323 ifr = (struct ifreq *)data; 1324 1325 switch (cmd) { 1326 case SIOCIFCREATE: 1327 case SIOCIFDESTROY: 1328 if ((error = suser(td)) != 0) 1329 return (error); 1330 return ((cmd == SIOCIFCREATE) ? 1331 if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) : 1332 if_clone_destroy(ifr->ifr_name)); 1333 1334 case SIOCIFGCLONERS: 1335 return (if_clone_list((struct if_clonereq *)data)); 1336 } 1337 1338 ifp = ifunit(ifr->ifr_name); 1339 if (ifp == 0) 1340 return (ENXIO); 1341 1342 error = ifhwioctl(cmd, ifp, data, td); 1343 if (error != ENOIOCTL) 1344 return (error); 1345 1346 oif_flags = ifp->if_flags; 1347 if (so->so_proto == 0) 1348 return (EOPNOTSUPP); 1349#ifndef COMPAT_43 1350 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd, 1351 data, 1352 ifp, td)); 1353#else 1354 { 1355 int ocmd = cmd; 1356 1357 switch (cmd) { 1358 1359 case SIOCSIFDSTADDR: 1360 case SIOCSIFADDR: 1361 case SIOCSIFBRDADDR: 1362 case SIOCSIFNETMASK: 1363#if BYTE_ORDER != BIG_ENDIAN 1364 if (ifr->ifr_addr.sa_family == 0 && 1365 ifr->ifr_addr.sa_len < 16) { 1366 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len; 1367 ifr->ifr_addr.sa_len = 16; 1368 } 1369#else 1370 if (ifr->ifr_addr.sa_len == 0) 1371 ifr->ifr_addr.sa_len = 16; 1372#endif 1373 break; 1374 1375 case OSIOCGIFADDR: 1376 cmd = SIOCGIFADDR; 1377 break; 1378 1379 case OSIOCGIFDSTADDR: 1380 cmd = SIOCGIFDSTADDR; 1381 break; 1382 1383 case OSIOCGIFBRDADDR: 1384 cmd = SIOCGIFBRDADDR; 1385 break; 1386 1387 case OSIOCGIFNETMASK: 1388 cmd = SIOCGIFNETMASK; 1389 } 1390 error = ((*so->so_proto->pr_usrreqs->pru_control)(so, 1391 cmd, 1392 data, 1393 ifp, td)); 1394 switch (ocmd) { 1395 1396 case OSIOCGIFADDR: 1397 case OSIOCGIFDSTADDR: 1398 case OSIOCGIFBRDADDR: 1399 case OSIOCGIFNETMASK: 1400 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family; 1401 1402 } 1403 } 1404#endif /* COMPAT_43 */ 1405 1406 if ((oif_flags ^ ifp->if_flags) & IFF_UP) { 1407#ifdef INET6 1408 DELAY(100);/* XXX: temporary workaround for fxp issue*/ 1409 if (ifp->if_flags & IFF_UP) { 1410 int s = splimp(); 1411 in6_if_up(ifp); 1412 splx(s); 1413 } 1414#endif 1415 } 1416 return (error); 1417} 1418 1419/* 1420 * Set/clear promiscuous mode on interface ifp based on the truth value 1421 * of pswitch. The calls are reference counted so that only the first 1422 * "on" request actually has an effect, as does the final "off" request. 1423 * Results are undefined if the "off" and "on" requests are not matched. 1424 */ 1425int 1426ifpromisc(struct ifnet *ifp, int pswitch) 1427{ 1428 struct ifreq ifr; 1429 int error; 1430 int oldflags, oldpcount; 1431 1432 oldpcount = ifp->if_pcount; 1433 oldflags = ifp->if_flags; 1434 if (ifp->if_flags & IFF_PPROMISC) { 1435 /* Do nothing if device is in permanently promiscuous mode */ 1436 ifp->if_pcount += pswitch ? 1 : -1; 1437 return (0); 1438 } 1439 if (pswitch) { 1440 /* 1441 * If the device is not configured up, we cannot put it in 1442 * promiscuous mode. 1443 */ 1444 if ((ifp->if_flags & IFF_UP) == 0) 1445 return (ENETDOWN); 1446 if (ifp->if_pcount++ != 0) 1447 return (0); 1448 ifp->if_flags |= IFF_PROMISC; 1449 } else { 1450 if (--ifp->if_pcount > 0) 1451 return (0); 1452 ifp->if_flags &= ~IFF_PROMISC; 1453 } 1454 ifr.ifr_flags = ifp->if_flags & 0xffff; 1455 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1456 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1457 if (error == 0) { 1458 log(LOG_INFO, "%s: promiscuous mode %s\n", 1459 ifp->if_xname, 1460 (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled"); 1461 rt_ifmsg(ifp); 1462 } else { 1463 ifp->if_pcount = oldpcount; 1464 ifp->if_flags = oldflags; 1465 } 1466 return error; 1467} 1468 1469/* 1470 * Return interface configuration 1471 * of system. List may be used 1472 * in later ioctl's (above) to get 1473 * other information. 1474 */ 1475/*ARGSUSED*/ 1476static int 1477ifconf(u_long cmd, caddr_t data) 1478{ 1479 struct ifconf *ifc = (struct ifconf *)data; 1480 struct ifnet *ifp; 1481 struct ifaddr *ifa; 1482 struct ifreq ifr, *ifrp; 1483 int space = ifc->ifc_len, error = 0; 1484 1485 ifrp = ifc->ifc_req; 1486 IFNET_RLOCK(); /* could sleep XXX */ 1487 TAILQ_FOREACH(ifp, &ifnet, if_link) { 1488 int addrs; 1489 1490 if (space < sizeof(ifr)) 1491 break; 1492 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name)) 1493 >= sizeof(ifr.ifr_name)) { 1494 error = ENAMETOOLONG; 1495 break; 1496 } 1497 1498 addrs = 0; 1499 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1500 struct sockaddr *sa = ifa->ifa_addr; 1501 1502 if (space < sizeof(ifr)) 1503 break; 1504 if (jailed(curthread->td_ucred) && 1505 prison_if(curthread->td_ucred, sa)) 1506 continue; 1507 addrs++; 1508#ifdef COMPAT_43 1509 if (cmd == OSIOCGIFCONF) { 1510 struct osockaddr *osa = 1511 (struct osockaddr *)&ifr.ifr_addr; 1512 ifr.ifr_addr = *sa; 1513 osa->sa_family = sa->sa_family; 1514 error = copyout((caddr_t)&ifr, (caddr_t)ifrp, 1515 sizeof (ifr)); 1516 ifrp++; 1517 } else 1518#endif 1519 if (sa->sa_len <= sizeof(*sa)) { 1520 ifr.ifr_addr = *sa; 1521 error = copyout((caddr_t)&ifr, (caddr_t)ifrp, 1522 sizeof (ifr)); 1523 ifrp++; 1524 } else { 1525 if (space < sizeof (ifr) + sa->sa_len - 1526 sizeof(*sa)) 1527 break; 1528 space -= sa->sa_len - sizeof(*sa); 1529 error = copyout((caddr_t)&ifr, (caddr_t)ifrp, 1530 sizeof (ifr.ifr_name)); 1531 if (error == 0) 1532 error = copyout((caddr_t)sa, 1533 (caddr_t)&ifrp->ifr_addr, sa->sa_len); 1534 ifrp = (struct ifreq *) 1535 (sa->sa_len + (caddr_t)&ifrp->ifr_addr); 1536 } 1537 if (error) 1538 break; 1539 space -= sizeof (ifr); 1540 } 1541 if (error) 1542 break; 1543 if (!addrs) { 1544 bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr)); 1545 error = copyout((caddr_t)&ifr, (caddr_t)ifrp, 1546 sizeof (ifr)); 1547 if (error) 1548 break; 1549 space -= sizeof (ifr); 1550 ifrp++; 1551 } 1552 } 1553 IFNET_RUNLOCK(); 1554 ifc->ifc_len -= space; 1555 return (error); 1556} 1557 1558/* 1559 * Just like if_promisc(), but for all-multicast-reception mode. 1560 */ 1561int 1562if_allmulti(struct ifnet *ifp, int onswitch) 1563{ 1564 int error = 0; 1565 int s = splimp(); 1566 struct ifreq ifr; 1567 1568 if (onswitch) { 1569 if (ifp->if_amcount++ == 0) { 1570 ifp->if_flags |= IFF_ALLMULTI; 1571 ifr.ifr_flags = ifp->if_flags & 0xffff; 1572 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1573 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1574 } 1575 } else { 1576 if (ifp->if_amcount > 1) { 1577 ifp->if_amcount--; 1578 } else { 1579 ifp->if_amcount = 0; 1580 ifp->if_flags &= ~IFF_ALLMULTI; 1581 ifr.ifr_flags = ifp->if_flags & 0xffff;; 1582 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1583 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1584 } 1585 } 1586 splx(s); 1587 1588 if (error == 0) 1589 rt_ifmsg(ifp); 1590 return error; 1591} 1592 1593/* 1594 * Add a multicast listenership to the interface in question. 1595 * The link layer provides a routine which converts 1596 */ 1597int 1598if_addmulti(struct ifnet *ifp, struct sockaddr *sa, struct ifmultiaddr **retifma) 1599{ 1600 struct sockaddr *llsa, *dupsa; 1601 int error, s; 1602 struct ifmultiaddr *ifma; 1603 1604 /* 1605 * If the matching multicast address already exists 1606 * then don't add a new one, just add a reference 1607 */ 1608 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1609 if (equal(sa, ifma->ifma_addr)) { 1610 ifma->ifma_refcount++; 1611 if (retifma) 1612 *retifma = ifma; 1613 return 0; 1614 } 1615 } 1616 1617 /* 1618 * Give the link layer a chance to accept/reject it, and also 1619 * find out which AF_LINK address this maps to, if it isn't one 1620 * already. 1621 */ 1622 if (ifp->if_resolvemulti) { 1623 error = ifp->if_resolvemulti(ifp, &llsa, sa); 1624 if (error) return error; 1625 } else { 1626 llsa = 0; 1627 } 1628 1629 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK); 1630 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK); 1631 bcopy(sa, dupsa, sa->sa_len); 1632 1633 ifma->ifma_addr = dupsa; 1634 ifma->ifma_lladdr = llsa; 1635 ifma->ifma_ifp = ifp; 1636 ifma->ifma_refcount = 1; 1637 ifma->ifma_protospec = 0; 1638 rt_newmaddrmsg(RTM_NEWMADDR, ifma); 1639 1640 /* 1641 * Some network interfaces can scan the address list at 1642 * interrupt time; lock them out. 1643 */ 1644 s = splimp(); 1645 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 1646 splx(s); 1647 if (retifma != NULL) 1648 *retifma = ifma; 1649 1650 if (llsa != 0) { 1651 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1652 if (equal(ifma->ifma_addr, llsa)) 1653 break; 1654 } 1655 if (ifma) { 1656 ifma->ifma_refcount++; 1657 } else { 1658 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, 1659 M_IFMADDR, M_WAITOK); 1660 MALLOC(dupsa, struct sockaddr *, llsa->sa_len, 1661 M_IFMADDR, M_WAITOK); 1662 bcopy(llsa, dupsa, llsa->sa_len); 1663 ifma->ifma_addr = dupsa; 1664 ifma->ifma_lladdr = NULL; 1665 ifma->ifma_ifp = ifp; 1666 ifma->ifma_refcount = 1; 1667 ifma->ifma_protospec = 0; 1668 s = splimp(); 1669 TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link); 1670 splx(s); 1671 } 1672 } 1673 /* 1674 * We are certain we have added something, so call down to the 1675 * interface to let them know about it. 1676 */ 1677 s = splimp(); 1678 ifp->if_ioctl(ifp, SIOCADDMULTI, 0); 1679 splx(s); 1680 1681 return 0; 1682} 1683 1684/* 1685 * Remove a reference to a multicast address on this interface. Yell 1686 * if the request does not match an existing membership. 1687 */ 1688int 1689if_delmulti(struct ifnet *ifp, struct sockaddr *sa) 1690{ 1691 struct ifmultiaddr *ifma; 1692 int s; 1693 1694 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 1695 if (equal(sa, ifma->ifma_addr)) 1696 break; 1697 if (ifma == 0) 1698 return ENOENT; 1699 1700 if (ifma->ifma_refcount > 1) { 1701 ifma->ifma_refcount--; 1702 return 0; 1703 } 1704 1705 rt_newmaddrmsg(RTM_DELMADDR, ifma); 1706 sa = ifma->ifma_lladdr; 1707 s = splimp(); 1708 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 1709 /* 1710 * Make sure the interface driver is notified 1711 * in the case of a link layer mcast group being left. 1712 */ 1713 if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0) 1714 ifp->if_ioctl(ifp, SIOCDELMULTI, 0); 1715 splx(s); 1716 free(ifma->ifma_addr, M_IFMADDR); 1717 free(ifma, M_IFMADDR); 1718 if (sa == 0) 1719 return 0; 1720 1721 /* 1722 * Now look for the link-layer address which corresponds to 1723 * this network address. It had been squirreled away in 1724 * ifma->ifma_lladdr for this purpose (so we don't have 1725 * to call ifp->if_resolvemulti() again), and we saved that 1726 * value in sa above. If some nasty deleted the 1727 * link-layer address out from underneath us, we can deal because 1728 * the address we stored was is not the same as the one which was 1729 * in the record for the link-layer address. (So we don't complain 1730 * in that case.) 1731 */ 1732 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 1733 if (equal(sa, ifma->ifma_addr)) 1734 break; 1735 if (ifma == 0) 1736 return 0; 1737 1738 if (ifma->ifma_refcount > 1) { 1739 ifma->ifma_refcount--; 1740 return 0; 1741 } 1742 1743 s = splimp(); 1744 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link); 1745 ifp->if_ioctl(ifp, SIOCDELMULTI, 0); 1746 splx(s); 1747 free(ifma->ifma_addr, M_IFMADDR); 1748 free(sa, M_IFMADDR); 1749 free(ifma, M_IFMADDR); 1750 1751 return 0; 1752} 1753 1754/* 1755 * Set the link layer address on an interface. 1756 * 1757 * At this time we only support certain types of interfaces, 1758 * and we don't allow the length of the address to change. 1759 */ 1760int 1761if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len) 1762{ 1763 struct sockaddr_dl *sdl; 1764 struct ifaddr *ifa; 1765 struct ifreq ifr; 1766 1767 ifa = ifaddr_byindex(ifp->if_index); 1768 if (ifa == NULL) 1769 return (EINVAL); 1770 sdl = (struct sockaddr_dl *)ifa->ifa_addr; 1771 if (sdl == NULL) 1772 return (EINVAL); 1773 if (len != sdl->sdl_alen) /* don't allow length to change */ 1774 return (EINVAL); 1775 switch (ifp->if_type) { 1776 case IFT_ETHER: /* these types use struct arpcom */ 1777 case IFT_FDDI: 1778 case IFT_XETHER: 1779 case IFT_ISO88025: 1780 case IFT_L2VLAN: 1781 bcopy(lladdr, IFP2AC(ifp)->ac_enaddr, len); 1782 /* 1783 * XXX We also need to store the lladdr in LLADDR(sdl), 1784 * which is done below. This is a pain because we must 1785 * remember to keep the info in sync. 1786 */ 1787 /* FALLTHROUGH */ 1788 case IFT_ARCNET: 1789 bcopy(lladdr, LLADDR(sdl), len); 1790 break; 1791 default: 1792 return (ENODEV); 1793 } 1794 /* 1795 * If the interface is already up, we need 1796 * to re-init it in order to reprogram its 1797 * address filter. 1798 */ 1799 if ((ifp->if_flags & IFF_UP) != 0) { 1800 ifp->if_flags &= ~IFF_UP; 1801 ifr.ifr_flags = ifp->if_flags & 0xffff; 1802 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1803 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1804 ifp->if_flags |= IFF_UP; 1805 ifr.ifr_flags = ifp->if_flags & 0xffff; 1806 ifr.ifr_flagshigh = ifp->if_flags >> 16; 1807 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr); 1808#ifdef INET 1809 /* 1810 * Also send gratuitous ARPs to notify other nodes about 1811 * the address change. 1812 */ 1813 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1814 if (ifa->ifa_addr != NULL && 1815 ifa->ifa_addr->sa_family == AF_INET) 1816 arp_ifinit(ifp, ifa); 1817 } 1818#endif 1819 } 1820 return (0); 1821} 1822 1823struct ifmultiaddr * 1824ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp) 1825{ 1826 struct ifmultiaddr *ifma; 1827 1828 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) 1829 if (equal(ifma->ifma_addr, sa)) 1830 break; 1831 1832 return ifma; 1833} 1834 1835/* 1836 * The name argument must be a pointer to storage which will last as 1837 * long as the interface does. For physical devices, the result of 1838 * device_get_name(dev) is a good choice and for pseudo-devices a 1839 * static string works well. 1840 */ 1841void 1842if_initname(struct ifnet *ifp, const char *name, int unit) 1843{ 1844 ifp->if_dname = name; 1845 ifp->if_dunit = unit; 1846 if (unit != IF_DUNIT_NONE) 1847 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit); 1848 else 1849 strlcpy(ifp->if_xname, name, IFNAMSIZ); 1850} 1851 1852int 1853if_printf(struct ifnet *ifp, const char * fmt, ...) 1854{ 1855 va_list ap; 1856 int retval; 1857 1858 retval = printf("%s: ", ifp->if_xname); 1859 va_start(ap, fmt); 1860 retval += vprintf(fmt, ap); 1861 va_end(ap); 1862 return (retval); 1863} 1864 1865/* 1866 * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot 1867 * be called without Giant. However, we often can't acquire the Giant lock 1868 * at those points; instead, we run it via a task queue that holds Giant via 1869 * if_start_deferred. 1870 * 1871 * XXXRW: We need to make sure that the ifnet isn't fully detached until any 1872 * outstanding if_start_deferred() tasks that will run after the free. This 1873 * probably means waiting in if_detach(). 1874 */ 1875void 1876if_start(struct ifnet *ifp) 1877{ 1878 1879 NET_ASSERT_GIANT(); 1880 1881 if ((ifp->if_flags & IFF_NEEDSGIANT) != 0 && debug_mpsafenet != 0) { 1882 if (mtx_owned(&Giant)) 1883 (*(ifp)->if_start)(ifp); 1884 else 1885 taskqueue_enqueue(taskqueue_swi_giant, 1886 &ifp->if_starttask); 1887 } else 1888 (*(ifp)->if_start)(ifp); 1889} 1890 1891static void 1892if_start_deferred(void *context, int pending) 1893{ 1894 struct ifnet *ifp; 1895 1896 /* 1897 * This code must be entered with Giant, and should never run if 1898 * we're not running with debug.mpsafenet. 1899 */ 1900 KASSERT(debug_mpsafenet != 0, ("if_start_deferred: debug.mpsafenet")); 1901 GIANT_REQUIRED; 1902 1903 ifp = (struct ifnet *)context; 1904 (ifp->if_start)(ifp); 1905} 1906 1907SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers"); 1908SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management"); 1909