if_tap.c revision 167713
1/*- 2 * Copyright (C) 1999-2000 by Maksim Yevmenkin <m_evmenkin@yahoo.com> 3 * 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 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 * 26 * BASED ON: 27 * ------------------------------------------------------------------------- 28 * 29 * Copyright (c) 1988, Julian Onions <jpo@cs.nott.ac.uk> 30 * Nottingham University 1987. 31 */ 32 33/* 34 * $FreeBSD: head/sys/net/if_tap.c 167713 2007-03-19 18:17:31Z bms $ 35 * $Id: if_tap.c,v 0.21 2000/07/23 21:46:02 max Exp $ 36 */ 37 38#include "opt_compat.h" 39#include "opt_inet.h" 40 41#include <sys/param.h> 42#include <sys/conf.h> 43#include <sys/fcntl.h> 44#include <sys/filio.h> 45#include <sys/kernel.h> 46#include <sys/malloc.h> 47#include <sys/mbuf.h> 48#include <sys/module.h> 49#include <sys/poll.h> 50#include <sys/priv.h> 51#include <sys/proc.h> 52#include <sys/selinfo.h> 53#include <sys/signalvar.h> 54#include <sys/socket.h> 55#include <sys/sockio.h> 56#include <sys/sysctl.h> 57#include <sys/systm.h> 58#include <sys/ttycom.h> 59#include <sys/uio.h> 60#include <sys/queue.h> 61 62#include <net/bpf.h> 63#include <net/ethernet.h> 64#include <net/if.h> 65#include <net/if_clone.h> 66#include <net/if_dl.h> 67#include <net/route.h> 68#include <net/if_types.h> 69 70#include <netinet/in.h> 71 72#include <net/if_tapvar.h> 73#include <net/if_tap.h> 74 75 76#define CDEV_NAME "tap" 77#define TAPDEBUG if (tapdebug) printf 78 79#define TAP "tap" 80#define VMNET "vmnet" 81#define TAPMAXUNIT 0x7fff 82#define VMNET_DEV_MASK CLONE_FLAG0 83 84/* module */ 85static int tapmodevent(module_t, int, void *); 86 87/* device */ 88static void tapclone(void *, struct ucred *, char *, int, 89 struct cdev **); 90static void tapcreate(struct cdev *); 91 92/* network interface */ 93static void tapifstart(struct ifnet *); 94static int tapifioctl(struct ifnet *, u_long, caddr_t); 95static void tapifinit(void *); 96 97static int tap_clone_create(struct if_clone *, int, caddr_t); 98static void tap_clone_destroy(struct ifnet *); 99static int vmnet_clone_create(struct if_clone *, int, caddr_t); 100static void vmnet_clone_destroy(struct ifnet *); 101 102IFC_SIMPLE_DECLARE(tap, 0); 103IFC_SIMPLE_DECLARE(vmnet, 0); 104 105/* character device */ 106static d_open_t tapopen; 107static d_close_t tapclose; 108static d_read_t tapread; 109static d_write_t tapwrite; 110static d_ioctl_t tapioctl; 111static d_poll_t tappoll; 112static d_kqfilter_t tapkqfilter; 113 114/* kqueue(2) */ 115static int tapkqread(struct knote *, long); 116static int tapkqwrite(struct knote *, long); 117static void tapkqdetach(struct knote *); 118 119static struct filterops tap_read_filterops = { 120 .f_isfd = 1, 121 .f_attach = NULL, 122 .f_detach = tapkqdetach, 123 .f_event = tapkqread, 124}; 125 126static struct filterops tap_write_filterops = { 127 .f_isfd = 1, 128 .f_attach = NULL, 129 .f_detach = tapkqdetach, 130 .f_event = tapkqwrite, 131}; 132 133static struct cdevsw tap_cdevsw = { 134 .d_version = D_VERSION, 135 .d_flags = D_PSEUDO | D_NEEDGIANT, 136 .d_open = tapopen, 137 .d_close = tapclose, 138 .d_read = tapread, 139 .d_write = tapwrite, 140 .d_ioctl = tapioctl, 141 .d_poll = tappoll, 142 .d_name = CDEV_NAME, 143 .d_kqfilter = tapkqfilter, 144}; 145 146/* 147 * All global variables in if_tap.c are locked with tapmtx, with the 148 * exception of tapdebug, which is accessed unlocked; tapclones is 149 * static at runtime. 150 */ 151static struct mtx tapmtx; 152static int tapdebug = 0; /* debug flag */ 153static int tapuopen = 0; /* allow user open() */ 154static int tapuponopen = 0; /* IFF_UP on open() */ 155static int tapdclone = 1; /* enable devfs cloning */ 156static SLIST_HEAD(, tap_softc) taphead; /* first device */ 157static struct clonedevs *tapclones; 158 159MALLOC_DECLARE(M_TAP); 160MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface"); 161SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, ""); 162 163SYSCTL_DECL(_net_link); 164SYSCTL_NODE(_net_link, OID_AUTO, tap, CTLFLAG_RW, 0, 165 "Ethernet tunnel software network interface"); 166SYSCTL_INT(_net_link_tap, OID_AUTO, user_open, CTLFLAG_RW, &tapuopen, 0, 167 "Allow user to open /dev/tap (based on node permissions)"); 168SYSCTL_INT(_net_link_tap, OID_AUTO, up_on_open, CTLFLAG_RW, &tapuponopen, 0, 169 "Bring interface up when /dev/tap is opened"); 170SYSCTL_INT(_net_link_tap, OID_AUTO, devfs_cloning, CTLFLAG_RW, &tapdclone, 0, 171 "Enably legacy devfs interface creation"); 172SYSCTL_INT(_net_link_tap, OID_AUTO, debug, CTLFLAG_RW, &tapdebug, 0, ""); 173 174TUNABLE_INT("net.link.tap.devfs_cloning", &tapdclone); 175 176DEV_MODULE(if_tap, tapmodevent, NULL); 177 178static int 179tap_clone_create(struct if_clone *ifc, int unit, caddr_t params) 180{ 181 struct cdev *dev; 182 int i; 183 int extra; 184 185 if (strcmp(ifc->ifc_name, VMNET) == 0) 186 extra = VMNET_DEV_MASK; 187 else 188 extra = 0; 189 190 /* find any existing device, or allocate new unit number */ 191 i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, extra); 192 if (i) { 193 dev = make_dev(&tap_cdevsw, unit2minor(unit | extra), 194 UID_ROOT, GID_WHEEL, 0600, "%s%d", ifc->ifc_name, unit); 195 if (dev != NULL) { 196 dev_ref(dev); 197 dev->si_flags |= SI_CHEAPCLONE; 198 } 199 } 200 201 tapcreate(dev); 202 return (0); 203} 204 205/* vmnet devices are tap devices in disguise */ 206static int 207vmnet_clone_create(struct if_clone *ifc, int unit, caddr_t params) 208{ 209 return tap_clone_create(ifc, unit, params); 210} 211 212static void 213tap_destroy(struct tap_softc *tp) 214{ 215 struct ifnet *ifp = tp->tap_ifp; 216 int s; 217 218 /* Unlocked read. */ 219 KASSERT(!(tp->tap_flags & TAP_OPEN), 220 ("%s flags is out of sync", ifp->if_xname)); 221 222 knlist_destroy(&tp->tap_rsel.si_note); 223 destroy_dev(tp->tap_dev); 224 s = splimp(); 225 ether_ifdetach(ifp); 226 if_free_type(ifp, IFT_ETHER); 227 splx(s); 228 229 mtx_destroy(&tp->tap_mtx); 230 free(tp, M_TAP); 231} 232 233static void 234tap_clone_destroy(struct ifnet *ifp) 235{ 236 struct tap_softc *tp = ifp->if_softc; 237 238 mtx_lock(&tapmtx); 239 SLIST_REMOVE(&taphead, tp, tap_softc, tap_next); 240 mtx_unlock(&tapmtx); 241 tap_destroy(tp); 242} 243 244/* vmnet devices are tap devices in disguise */ 245static void 246vmnet_clone_destroy(struct ifnet *ifp) 247{ 248 tap_clone_destroy(ifp); 249} 250 251/* 252 * tapmodevent 253 * 254 * module event handler 255 */ 256static int 257tapmodevent(module_t mod, int type, void *data) 258{ 259 static eventhandler_tag eh_tag = NULL; 260 struct tap_softc *tp = NULL; 261 struct ifnet *ifp = NULL; 262 263 switch (type) { 264 case MOD_LOAD: 265 266 /* intitialize device */ 267 268 mtx_init(&tapmtx, "tapmtx", NULL, MTX_DEF); 269 SLIST_INIT(&taphead); 270 271 clone_setup(&tapclones); 272 eh_tag = EVENTHANDLER_REGISTER(dev_clone, tapclone, 0, 1000); 273 if (eh_tag == NULL) { 274 clone_cleanup(&tapclones); 275 mtx_destroy(&tapmtx); 276 return (ENOMEM); 277 } 278 if_clone_attach(&tap_cloner); 279 if_clone_attach(&vmnet_cloner); 280 return (0); 281 282 case MOD_UNLOAD: 283 /* 284 * The EBUSY algorithm here can't quite atomically 285 * guarantee that this is race-free since we have to 286 * release the tap mtx to deregister the clone handler. 287 */ 288 mtx_lock(&tapmtx); 289 SLIST_FOREACH(tp, &taphead, tap_next) { 290 mtx_lock(&tp->tap_mtx); 291 if (tp->tap_flags & TAP_OPEN) { 292 mtx_unlock(&tp->tap_mtx); 293 mtx_unlock(&tapmtx); 294 return (EBUSY); 295 } 296 mtx_unlock(&tp->tap_mtx); 297 } 298 mtx_unlock(&tapmtx); 299 300 EVENTHANDLER_DEREGISTER(dev_clone, eh_tag); 301 if_clone_detach(&tap_cloner); 302 if_clone_detach(&vmnet_cloner); 303 304 mtx_lock(&tapmtx); 305 while ((tp = SLIST_FIRST(&taphead)) != NULL) { 306 SLIST_REMOVE_HEAD(&taphead, tap_next); 307 mtx_unlock(&tapmtx); 308 309 ifp = tp->tap_ifp; 310 311 TAPDEBUG("detaching %s\n", ifp->if_xname); 312 313 tap_destroy(tp); 314 mtx_lock(&tapmtx); 315 } 316 mtx_unlock(&tapmtx); 317 clone_cleanup(&tapclones); 318 319 mtx_destroy(&tapmtx); 320 321 break; 322 323 default: 324 return (EOPNOTSUPP); 325 } 326 327 return (0); 328} /* tapmodevent */ 329 330 331/* 332 * DEVFS handler 333 * 334 * We need to support two kind of devices - tap and vmnet 335 */ 336static void 337tapclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev) 338{ 339 char devname[SPECNAMELEN + 1]; 340 int i, unit, append_unit; 341 int extra; 342 343 if (*dev != NULL) 344 return; 345 346 if (!tapdclone || 347 (!tapuopen && priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0)) 348 return; 349 350 unit = 0; 351 append_unit = 0; 352 extra = 0; 353 354 /* We're interested in only tap/vmnet devices. */ 355 if (strcmp(name, TAP) == 0) { 356 unit = -1; 357 } else if (strcmp(name, VMNET) == 0) { 358 unit = -1; 359 extra = VMNET_DEV_MASK; 360 } else if (dev_stdclone(name, NULL, TAP, &unit) != 1) { 361 if (dev_stdclone(name, NULL, VMNET, &unit) != 1) { 362 return; 363 } else { 364 extra = VMNET_DEV_MASK; 365 } 366 } 367 368 if (unit == -1) 369 append_unit = 1; 370 371 /* find any existing device, or allocate new unit number */ 372 i = clone_create(&tapclones, &tap_cdevsw, &unit, dev, extra); 373 if (i) { 374 if (append_unit) { 375 /* 376 * We were passed 'tun' or 'tap', with no unit specified 377 * so we'll need to append it now. 378 */ 379 namelen = snprintf(devname, sizeof(devname), "%s%d", name, 380 unit); 381 name = devname; 382 } 383 384 *dev = make_dev(&tap_cdevsw, unit2minor(unit | extra), 385 UID_ROOT, GID_WHEEL, 0600, "%s", name); 386 if (*dev != NULL) { 387 dev_ref(*dev); 388 (*dev)->si_flags |= SI_CHEAPCLONE; 389 } 390 } 391 392 if_clone_create(name, namelen, NULL); 393} /* tapclone */ 394 395 396/* 397 * tapcreate 398 * 399 * to create interface 400 */ 401static void 402tapcreate(struct cdev *dev) 403{ 404 struct ifnet *ifp = NULL; 405 struct tap_softc *tp = NULL; 406 unsigned short macaddr_hi; 407 int unit, s; 408 char *name = NULL; 409 u_char eaddr[6]; 410 411 dev->si_flags &= ~SI_CHEAPCLONE; 412 413 /* allocate driver storage and create device */ 414 MALLOC(tp, struct tap_softc *, sizeof(*tp), M_TAP, M_WAITOK | M_ZERO); 415 mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF); 416 mtx_lock(&tapmtx); 417 SLIST_INSERT_HEAD(&taphead, tp, tap_next); 418 mtx_unlock(&tapmtx); 419 420 unit = dev2unit(dev); 421 422 /* select device: tap or vmnet */ 423 if (unit & VMNET_DEV_MASK) { 424 name = VMNET; 425 tp->tap_flags |= TAP_VMNET; 426 } else 427 name = TAP; 428 429 unit &= TAPMAXUNIT; 430 431 TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, minor(dev)); 432 433 /* generate fake MAC address: 00 bd xx xx xx unit_no */ 434 macaddr_hi = htons(0x00bd); 435 bcopy(&macaddr_hi, eaddr, sizeof(short)); 436 bcopy(&ticks, &eaddr[2], sizeof(long)); 437 eaddr[5] = (u_char)unit; 438 439 /* fill the rest and attach interface */ 440 ifp = tp->tap_ifp = if_alloc(IFT_ETHER); 441 if (ifp == NULL) 442 panic("%s%d: can not if_alloc()", name, unit); 443 ifp->if_softc = tp; 444 if_initname(ifp, name, unit); 445 ifp->if_init = tapifinit; 446 ifp->if_start = tapifstart; 447 ifp->if_ioctl = tapifioctl; 448 ifp->if_mtu = ETHERMTU; 449 ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST); 450 ifp->if_snd.ifq_maxlen = ifqmaxlen; 451 452 dev->si_drv1 = tp; 453 tp->tap_dev = dev; 454 455 s = splimp(); 456 ether_ifattach(ifp, eaddr); 457 splx(s); 458 459 mtx_lock(&tp->tap_mtx); 460 tp->tap_flags |= TAP_INITED; 461 mtx_unlock(&tp->tap_mtx); 462 463 knlist_init(&tp->tap_rsel.si_note, NULL, NULL, NULL, NULL); 464 465 TAPDEBUG("interface %s is created. minor = %#x\n", 466 ifp->if_xname, minor(dev)); 467} /* tapcreate */ 468 469 470/* 471 * tapopen 472 * 473 * to open tunnel. must be superuser 474 */ 475static int 476tapopen(struct cdev *dev, int flag, int mode, struct thread *td) 477{ 478 struct tap_softc *tp = NULL; 479 struct ifnet *ifp = NULL; 480 int error, s; 481 482 if (tapuopen == 0) { 483 error = priv_check(td, PRIV_NET_TAP); 484 if (error) 485 return (error); 486 } 487 488 if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT) 489 return (ENXIO); 490 491 tp = dev->si_drv1; 492 493 mtx_lock(&tp->tap_mtx); 494 if (tp->tap_flags & TAP_OPEN) { 495 mtx_unlock(&tp->tap_mtx); 496 return (EBUSY); 497 } 498 499 bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr)); 500 tp->tap_pid = td->td_proc->p_pid; 501 tp->tap_flags |= TAP_OPEN; 502 ifp = tp->tap_ifp; 503 mtx_unlock(&tp->tap_mtx); 504 505 s = splimp(); 506 ifp->if_drv_flags |= IFF_DRV_RUNNING; 507 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 508 if (tapuponopen) 509 ifp->if_flags |= IFF_UP; 510 splx(s); 511 512 TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, minor(dev)); 513 514 return (0); 515} /* tapopen */ 516 517 518/* 519 * tapclose 520 * 521 * close the device - mark i/f down & delete routing info 522 */ 523static int 524tapclose(struct cdev *dev, int foo, int bar, struct thread *td) 525{ 526 struct ifaddr *ifa; 527 struct tap_softc *tp = dev->si_drv1; 528 struct ifnet *ifp = tp->tap_ifp; 529 int s; 530 531 /* junk all pending output */ 532 IF_DRAIN(&ifp->if_snd); 533 534 /* 535 * do not bring the interface down, and do not anything with 536 * interface, if we are in VMnet mode. just close the device. 537 */ 538 539 mtx_lock(&tp->tap_mtx); 540 if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) { 541 mtx_unlock(&tp->tap_mtx); 542 s = splimp(); 543 if_down(ifp); 544 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 545 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 546 rtinit(ifa, (int)RTM_DELETE, 0); 547 } 548 if_purgeaddrs(ifp); 549 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 550 } 551 splx(s); 552 } else 553 mtx_unlock(&tp->tap_mtx); 554 555 funsetown(&tp->tap_sigio); 556 selwakeuppri(&tp->tap_rsel, PZERO+1); 557 KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0); 558 559 mtx_lock(&tp->tap_mtx); 560 tp->tap_flags &= ~TAP_OPEN; 561 tp->tap_pid = 0; 562 mtx_unlock(&tp->tap_mtx); 563 564 TAPDEBUG("%s is closed. minor = %#x\n", 565 ifp->if_xname, minor(dev)); 566 567 return (0); 568} /* tapclose */ 569 570 571/* 572 * tapifinit 573 * 574 * network interface initialization function 575 */ 576static void 577tapifinit(void *xtp) 578{ 579 struct tap_softc *tp = (struct tap_softc *)xtp; 580 struct ifnet *ifp = tp->tap_ifp; 581 582 TAPDEBUG("initializing %s\n", ifp->if_xname); 583 584 ifp->if_drv_flags |= IFF_DRV_RUNNING; 585 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 586 587 /* attempt to start output */ 588 tapifstart(ifp); 589} /* tapifinit */ 590 591 592/* 593 * tapifioctl 594 * 595 * Process an ioctl request on network interface 596 */ 597static int 598tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 599{ 600 struct tap_softc *tp = ifp->if_softc; 601 struct ifstat *ifs = NULL; 602 int s, dummy; 603 604 switch (cmd) { 605 case SIOCSIFFLAGS: /* XXX -- just like vmnet does */ 606 case SIOCADDMULTI: 607 case SIOCDELMULTI: 608 break; 609 610 case SIOCGIFSTATUS: 611 s = splimp(); 612 ifs = (struct ifstat *)data; 613 dummy = strlen(ifs->ascii); 614 mtx_lock(&tp->tap_mtx); 615 if (tp->tap_pid != 0 && dummy < sizeof(ifs->ascii)) 616 snprintf(ifs->ascii + dummy, 617 sizeof(ifs->ascii) - dummy, 618 "\tOpened by PID %d\n", tp->tap_pid); 619 mtx_unlock(&tp->tap_mtx); 620 splx(s); 621 break; 622 623 default: 624 s = splimp(); 625 dummy = ether_ioctl(ifp, cmd, data); 626 splx(s); 627 return (dummy); 628 /* NOT REACHED */ 629 } 630 631 return (0); 632} /* tapifioctl */ 633 634 635/* 636 * tapifstart 637 * 638 * queue packets from higher level ready to put out 639 */ 640static void 641tapifstart(struct ifnet *ifp) 642{ 643 struct tap_softc *tp = ifp->if_softc; 644 int s; 645 646 TAPDEBUG("%s starting\n", ifp->if_xname); 647 648 /* 649 * do not junk pending output if we are in VMnet mode. 650 * XXX: can this do any harm because of queue overflow? 651 */ 652 653 mtx_lock(&tp->tap_mtx); 654 if (((tp->tap_flags & TAP_VMNET) == 0) && 655 ((tp->tap_flags & TAP_READY) != TAP_READY)) { 656 struct mbuf *m = NULL; 657 658 mtx_unlock(&tp->tap_mtx); 659 660 /* Unlocked read. */ 661 TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname, 662 tp->tap_flags); 663 664 s = splimp(); 665 do { 666 IF_DEQUEUE(&ifp->if_snd, m); 667 if (m != NULL) 668 m_freem(m); 669 ifp->if_oerrors ++; 670 } while (m != NULL); 671 splx(s); 672 673 return; 674 } 675 mtx_unlock(&tp->tap_mtx); 676 677 s = splimp(); 678 ifp->if_drv_flags |= IFF_DRV_OACTIVE; 679 680 if (ifp->if_snd.ifq_len != 0) { 681 mtx_lock(&tp->tap_mtx); 682 if (tp->tap_flags & TAP_RWAIT) { 683 tp->tap_flags &= ~TAP_RWAIT; 684 wakeup(tp); 685 } 686 687 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) { 688 mtx_unlock(&tp->tap_mtx); 689 pgsigio(&tp->tap_sigio, SIGIO, 0); 690 } else 691 mtx_unlock(&tp->tap_mtx); 692 693 selwakeuppri(&tp->tap_rsel, PZERO+1); 694 KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0); 695 ifp->if_opackets ++; /* obytes are counted in ether_output */ 696 } 697 698 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 699 splx(s); 700} /* tapifstart */ 701 702 703/* 704 * tapioctl 705 * 706 * the cdevsw interface is now pretty minimal 707 */ 708static int 709tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 710{ 711 struct tap_softc *tp = dev->si_drv1; 712 struct ifnet *ifp = tp->tap_ifp; 713 struct tapinfo *tapp = NULL; 714 int s; 715 int f; 716#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \ 717 defined(COMPAT_FREEBSD4) 718 int ival; 719#endif 720 721 switch (cmd) { 722 case TAPSIFINFO: 723 s = splimp(); 724 tapp = (struct tapinfo *)data; 725 ifp->if_mtu = tapp->mtu; 726 ifp->if_type = tapp->type; 727 ifp->if_baudrate = tapp->baudrate; 728 splx(s); 729 break; 730 731 case TAPGIFINFO: 732 tapp = (struct tapinfo *)data; 733 tapp->mtu = ifp->if_mtu; 734 tapp->type = ifp->if_type; 735 tapp->baudrate = ifp->if_baudrate; 736 break; 737 738 case TAPSDEBUG: 739 tapdebug = *(int *)data; 740 break; 741 742 case TAPGDEBUG: 743 *(int *)data = tapdebug; 744 break; 745 746 case FIONBIO: 747 break; 748 749 case FIOASYNC: 750 s = splimp(); 751 mtx_lock(&tp->tap_mtx); 752 if (*(int *)data) 753 tp->tap_flags |= TAP_ASYNC; 754 else 755 tp->tap_flags &= ~TAP_ASYNC; 756 mtx_unlock(&tp->tap_mtx); 757 splx(s); 758 break; 759 760 case FIONREAD: 761 s = splimp(); 762 if (ifp->if_snd.ifq_head) { 763 struct mbuf *mb = ifp->if_snd.ifq_head; 764 765 for(*(int *)data = 0;mb != NULL;mb = mb->m_next) 766 *(int *)data += mb->m_len; 767 } else 768 *(int *)data = 0; 769 splx(s); 770 break; 771 772 case FIOSETOWN: 773 return (fsetown(*(int *)data, &tp->tap_sigio)); 774 775 case FIOGETOWN: 776 *(int *)data = fgetown(&tp->tap_sigio); 777 return (0); 778 779 /* this is deprecated, FIOSETOWN should be used instead */ 780 case TIOCSPGRP: 781 return (fsetown(-(*(int *)data), &tp->tap_sigio)); 782 783 /* this is deprecated, FIOGETOWN should be used instead */ 784 case TIOCGPGRP: 785 *(int *)data = -fgetown(&tp->tap_sigio); 786 return (0); 787 788 /* VMware/VMnet port ioctl's */ 789 790 case SIOCGIFFLAGS: /* get ifnet flags */ 791 bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags)); 792 break; 793 794#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \ 795 defined(COMPAT_FREEBSD4) 796 case _IO('V', 0): 797 ival = IOCPARM_IVAL(data); 798 data = (caddr_t)&ival; 799 /* FALLTHROUGH */ 800#endif 801 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */ 802 f = *(int *)data; 803 f &= 0x0fff; 804 f &= ~IFF_CANTCHANGE; 805 f |= IFF_UP; 806 807 s = splimp(); 808 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE); 809 splx(s); 810 break; 811 812 case OSIOCGIFADDR: /* get MAC address of the remote side */ 813 case SIOCGIFADDR: 814 mtx_lock(&tp->tap_mtx); 815 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr)); 816 mtx_unlock(&tp->tap_mtx); 817 break; 818 819 case SIOCSIFADDR: /* set MAC address of the remote side */ 820 mtx_lock(&tp->tap_mtx); 821 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr)); 822 mtx_unlock(&tp->tap_mtx); 823 break; 824 825 default: 826 return (ENOTTY); 827 } 828 return (0); 829} /* tapioctl */ 830 831 832/* 833 * tapread 834 * 835 * the cdevsw read interface - reads a packet at a time, or at 836 * least as much of a packet as can be read 837 */ 838static int 839tapread(struct cdev *dev, struct uio *uio, int flag) 840{ 841 struct tap_softc *tp = dev->si_drv1; 842 struct ifnet *ifp = tp->tap_ifp; 843 struct mbuf *m = NULL; 844 int error = 0, len, s; 845 846 TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, minor(dev)); 847 848 mtx_lock(&tp->tap_mtx); 849 if ((tp->tap_flags & TAP_READY) != TAP_READY) { 850 mtx_unlock(&tp->tap_mtx); 851 852 /* Unlocked read. */ 853 TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n", 854 ifp->if_xname, minor(dev), tp->tap_flags); 855 856 return (EHOSTDOWN); 857 } 858 859 tp->tap_flags &= ~TAP_RWAIT; 860 mtx_unlock(&tp->tap_mtx); 861 862 /* sleep until we get a packet */ 863 do { 864 s = splimp(); 865 IF_DEQUEUE(&ifp->if_snd, m); 866 splx(s); 867 868 if (m == NULL) { 869 if (flag & O_NONBLOCK) 870 return (EWOULDBLOCK); 871 872 mtx_lock(&tp->tap_mtx); 873 tp->tap_flags |= TAP_RWAIT; 874 mtx_unlock(&tp->tap_mtx); 875 error = tsleep(tp,PCATCH|(PZERO+1),"taprd",0); 876 if (error) 877 return (error); 878 } 879 } while (m == NULL); 880 881 /* feed packet to bpf */ 882 BPF_MTAP(ifp, m); 883 884 /* xfer packet to user space */ 885 while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) { 886 len = min(uio->uio_resid, m->m_len); 887 if (len == 0) 888 break; 889 890 error = uiomove(mtod(m, void *), len, uio); 891 m = m_free(m); 892 } 893 894 if (m != NULL) { 895 TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname, 896 minor(dev)); 897 m_freem(m); 898 } 899 900 return (error); 901} /* tapread */ 902 903 904/* 905 * tapwrite 906 * 907 * the cdevsw write interface - an atomic write is a packet - or else! 908 */ 909static int 910tapwrite(struct cdev *dev, struct uio *uio, int flag) 911{ 912 struct ether_header *eh; 913 struct tap_softc *tp = dev->si_drv1; 914 struct ifnet *ifp = tp->tap_ifp; 915 struct mbuf *m; 916 917 TAPDEBUG("%s writting, minor = %#x\n", 918 ifp->if_xname, minor(dev)); 919 920 if (uio->uio_resid == 0) 921 return (0); 922 923 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) { 924 TAPDEBUG("%s invalid packet len = %d, minor = %#x\n", 925 ifp->if_xname, uio->uio_resid, minor(dev)); 926 927 return (EIO); 928 } 929 930 if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, ETHER_ALIGN, 931 M_PKTHDR)) == NULL) { 932 ifp->if_ierrors ++; 933 return (ENOBUFS); 934 } 935 936 m->m_pkthdr.rcvif = ifp; 937 938 /* 939 * Only pass a unicast frame to ether_input(), if it would actually 940 * have been received by non-virtual hardware. 941 */ 942 if (m->m_len < sizeof(struct ether_header)) { 943 m_freem(m); 944 return (0); 945 } 946 eh = mtod(m, struct ether_header *); 947 948 if (eh && (ifp->if_flags & IFF_PROMISC) == 0 && 949 !ETHER_IS_MULTICAST(eh->ether_dhost) && 950 bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) { 951 m_freem(m); 952 return (0); 953 } 954 955 /* Pass packet up to parent. */ 956 (*ifp->if_input)(ifp, m); 957 ifp->if_ipackets ++; /* ibytes are counted in parent */ 958 959 return (0); 960} /* tapwrite */ 961 962 963/* 964 * tappoll 965 * 966 * the poll interface, this is only useful on reads 967 * really. the write detect always returns true, write never blocks 968 * anyway, it either accepts the packet or drops it 969 */ 970static int 971tappoll(struct cdev *dev, int events, struct thread *td) 972{ 973 struct tap_softc *tp = dev->si_drv1; 974 struct ifnet *ifp = tp->tap_ifp; 975 int s, revents = 0; 976 977 TAPDEBUG("%s polling, minor = %#x\n", 978 ifp->if_xname, minor(dev)); 979 980 s = splimp(); 981 if (events & (POLLIN | POLLRDNORM)) { 982 if (ifp->if_snd.ifq_len > 0) { 983 TAPDEBUG("%s have data in queue. len = %d, " \ 984 "minor = %#x\n", ifp->if_xname, 985 ifp->if_snd.ifq_len, minor(dev)); 986 987 revents |= (events & (POLLIN | POLLRDNORM)); 988 } else { 989 TAPDEBUG("%s waiting for data, minor = %#x\n", 990 ifp->if_xname, minor(dev)); 991 992 selrecord(td, &tp->tap_rsel); 993 } 994 } 995 996 if (events & (POLLOUT | POLLWRNORM)) 997 revents |= (events & (POLLOUT | POLLWRNORM)); 998 999 splx(s); 1000 return (revents); 1001} /* tappoll */ 1002 1003 1004/* 1005 * tap_kqfilter 1006 * 1007 * support for kevent() system call 1008 */ 1009static int 1010tapkqfilter(struct cdev *dev, struct knote *kn) 1011{ 1012 int s; 1013 struct tap_softc *tp = dev->si_drv1; 1014 struct ifnet *ifp = tp->tap_ifp; 1015 1016 s = splimp(); 1017 switch (kn->kn_filter) { 1018 case EVFILT_READ: 1019 TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n", 1020 ifp->if_xname, minor(dev)); 1021 kn->kn_fop = &tap_read_filterops; 1022 break; 1023 1024 case EVFILT_WRITE: 1025 TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n", 1026 ifp->if_xname, minor(dev)); 1027 kn->kn_fop = &tap_write_filterops; 1028 break; 1029 1030 default: 1031 TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n", 1032 ifp->if_xname, minor(dev)); 1033 splx(s); 1034 return (EINVAL); 1035 /* NOT REACHED */ 1036 } 1037 splx(s); 1038 1039 kn->kn_hook = (caddr_t) dev; 1040 knlist_add(&tp->tap_rsel.si_note, kn, 0); 1041 1042 return (0); 1043} /* tapkqfilter */ 1044 1045 1046/* 1047 * tap_kqread 1048 * 1049 * Return true if there is data in the interface queue 1050 */ 1051static int 1052tapkqread(struct knote *kn, long hint) 1053{ 1054 int ret, s; 1055 struct cdev *dev = (struct cdev *)(kn->kn_hook); 1056 struct tap_softc *tp = dev->si_drv1; 1057 struct ifnet *ifp = tp->tap_ifp; 1058 1059 s = splimp(); 1060 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1061 TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n", 1062 ifp->if_xname, ifp->if_snd.ifq_len, minor(dev)); 1063 ret = 1; 1064 } else { 1065 TAPDEBUG("%s waiting for data, minor = %#x\n", 1066 ifp->if_xname, minor(dev)); 1067 ret = 0; 1068 } 1069 splx(s); 1070 1071 return (ret); 1072} /* tapkqread */ 1073 1074 1075/* 1076 * tap_kqwrite 1077 * 1078 * Always can write. Return the MTU in kn->data 1079 */ 1080static int 1081tapkqwrite(struct knote *kn, long hint) 1082{ 1083 int s; 1084 struct tap_softc *tp = ((struct cdev *) kn->kn_hook)->si_drv1; 1085 struct ifnet *ifp = tp->tap_ifp; 1086 1087 s = splimp(); 1088 kn->kn_data = ifp->if_mtu; 1089 splx(s); 1090 1091 return (1); 1092} /* tapkqwrite */ 1093 1094 1095static void 1096tapkqdetach(struct knote *kn) 1097{ 1098 struct tap_softc *tp = ((struct cdev *) kn->kn_hook)->si_drv1; 1099 1100 knlist_remove(&tp->tap_rsel.si_note, kn, 0); 1101} /* tapkqdetach */ 1102 1103