if_tap.c revision 166514
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 166514 2007-02-05 11:29:08Z 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 tapdclone = 1; /* enable devfs cloning */ 155static SLIST_HEAD(, tap_softc) taphead; /* first device */ 156static struct clonedevs *tapclones; 157 158MALLOC_DECLARE(M_TAP); 159MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface"); 160SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, ""); 161 162SYSCTL_DECL(_net_link); 163SYSCTL_NODE(_net_link, OID_AUTO, tap, CTLFLAG_RW, 0, 164 "Ethernet tunnel software network interface"); 165SYSCTL_INT(_net_link_tap, OID_AUTO, user_open, CTLFLAG_RW, &tapuopen, 0, 166 "Allow user to open /dev/tap (based on node permissions)"); 167SYSCTL_INT(_net_link_tap, OID_AUTO, devfs_cloning, CTLFLAG_RW, &tapdclone, 0, 168 "Enably legacy devfs interface creation"); 169SYSCTL_INT(_net_link_tap, OID_AUTO, debug, CTLFLAG_RW, &tapdebug, 0, ""); 170 171TUNABLE_INT("net.link.tap.devfs_cloning", &tapdclone); 172 173DEV_MODULE(if_tap, tapmodevent, NULL); 174 175static int 176tap_clone_create(struct if_clone *ifc, int unit, caddr_t params) 177{ 178 struct cdev *dev; 179 int i; 180 int extra; 181 182 if (strcmp(ifc->ifc_name, VMNET) == 0) 183 extra = VMNET_DEV_MASK; 184 else 185 extra = 0; 186 187 /* find any existing device, or allocate new unit number */ 188 i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, extra); 189 if (i) { 190 dev = make_dev(&tap_cdevsw, unit2minor(unit | extra), 191 UID_ROOT, GID_WHEEL, 0600, "%s%d", ifc->ifc_name, unit); 192 if (dev != NULL) { 193 dev_ref(dev); 194 dev->si_flags |= SI_CHEAPCLONE; 195 } 196 } 197 198 tapcreate(dev); 199 return (0); 200} 201 202/* vmnet devices are tap devices in disguise */ 203static int 204vmnet_clone_create(struct if_clone *ifc, int unit, caddr_t params) 205{ 206 return tap_clone_create(ifc, unit, params); 207} 208 209static void 210tap_destroy(struct tap_softc *tp) 211{ 212 struct ifnet *ifp = tp->tap_ifp; 213 int s; 214 215 /* Unlocked read. */ 216 KASSERT(!(tp->tap_flags & TAP_OPEN), 217 ("%s flags is out of sync", ifp->if_xname)); 218 219 knlist_destroy(&tp->tap_rsel.si_note); 220 destroy_dev(tp->tap_dev); 221 s = splimp(); 222 ether_ifdetach(ifp); 223 if_free_type(ifp, IFT_ETHER); 224 splx(s); 225 226 mtx_destroy(&tp->tap_mtx); 227 free(tp, M_TAP); 228} 229 230static void 231tap_clone_destroy(struct ifnet *ifp) 232{ 233 struct tap_softc *tp = ifp->if_softc; 234 235 mtx_lock(&tapmtx); 236 SLIST_REMOVE(&taphead, tp, tap_softc, tap_next); 237 mtx_unlock(&tapmtx); 238 tap_destroy(tp); 239} 240 241/* vmnet devices are tap devices in disguise */ 242static void 243vmnet_clone_destroy(struct ifnet *ifp) 244{ 245 tap_clone_destroy(ifp); 246} 247 248/* 249 * tapmodevent 250 * 251 * module event handler 252 */ 253static int 254tapmodevent(module_t mod, int type, void *data) 255{ 256 static eventhandler_tag eh_tag = NULL; 257 struct tap_softc *tp = NULL; 258 struct ifnet *ifp = NULL; 259 260 switch (type) { 261 case MOD_LOAD: 262 263 /* intitialize device */ 264 265 mtx_init(&tapmtx, "tapmtx", NULL, MTX_DEF); 266 SLIST_INIT(&taphead); 267 268 clone_setup(&tapclones); 269 eh_tag = EVENTHANDLER_REGISTER(dev_clone, tapclone, 0, 1000); 270 if (eh_tag == NULL) { 271 clone_cleanup(&tapclones); 272 mtx_destroy(&tapmtx); 273 return (ENOMEM); 274 } 275 if_clone_attach(&tap_cloner); 276 if_clone_attach(&vmnet_cloner); 277 return (0); 278 279 case MOD_UNLOAD: 280 /* 281 * The EBUSY algorithm here can't quite atomically 282 * guarantee that this is race-free since we have to 283 * release the tap mtx to deregister the clone handler. 284 */ 285 mtx_lock(&tapmtx); 286 SLIST_FOREACH(tp, &taphead, tap_next) { 287 mtx_lock(&tp->tap_mtx); 288 if (tp->tap_flags & TAP_OPEN) { 289 mtx_unlock(&tp->tap_mtx); 290 mtx_unlock(&tapmtx); 291 return (EBUSY); 292 } 293 mtx_unlock(&tp->tap_mtx); 294 } 295 mtx_unlock(&tapmtx); 296 297 EVENTHANDLER_DEREGISTER(dev_clone, eh_tag); 298 if_clone_detach(&tap_cloner); 299 if_clone_detach(&vmnet_cloner); 300 301 mtx_lock(&tapmtx); 302 while ((tp = SLIST_FIRST(&taphead)) != NULL) { 303 SLIST_REMOVE_HEAD(&taphead, tap_next); 304 mtx_unlock(&tapmtx); 305 306 ifp = tp->tap_ifp; 307 308 TAPDEBUG("detaching %s\n", ifp->if_xname); 309 310 tap_destroy(tp); 311 mtx_lock(&tapmtx); 312 } 313 mtx_unlock(&tapmtx); 314 clone_cleanup(&tapclones); 315 316 mtx_destroy(&tapmtx); 317 318 break; 319 320 default: 321 return (EOPNOTSUPP); 322 } 323 324 return (0); 325} /* tapmodevent */ 326 327 328/* 329 * DEVFS handler 330 * 331 * We need to support two kind of devices - tap and vmnet 332 */ 333static void 334tapclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev) 335{ 336 char devname[SPECNAMELEN + 1]; 337 int i, unit, append_unit; 338 int extra; 339 340 if (*dev != NULL) 341 return; 342 343 if (!tapdclone || 344 (!tapuopen && priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0)) 345 return; 346 347 unit = 0; 348 append_unit = 0; 349 extra = 0; 350 351 /* We're interested in only tap/vmnet devices. */ 352 if (strcmp(name, TAP) == 0) { 353 unit = -1; 354 } else if (strcmp(name, VMNET) == 0) { 355 unit = -1; 356 extra = VMNET_DEV_MASK; 357 } else if (dev_stdclone(name, NULL, TAP, &unit) != 1) { 358 if (dev_stdclone(name, NULL, VMNET, &unit) != 1) { 359 return; 360 } else { 361 extra = VMNET_DEV_MASK; 362 } 363 } 364 365 if (unit == -1) 366 append_unit = 1; 367 368 /* find any existing device, or allocate new unit number */ 369 i = clone_create(&tapclones, &tap_cdevsw, &unit, dev, extra); 370 if (i) { 371 if (append_unit) { 372 /* 373 * We were passed 'tun' or 'tap', with no unit specified 374 * so we'll need to append it now. 375 */ 376 namelen = snprintf(devname, sizeof(devname), "%s%d", name, 377 unit); 378 name = devname; 379 } 380 381 *dev = make_dev(&tap_cdevsw, unit2minor(unit | extra), 382 UID_ROOT, GID_WHEEL, 0600, "%s", name); 383 if (*dev != NULL) { 384 dev_ref(*dev); 385 (*dev)->si_flags |= SI_CHEAPCLONE; 386 } 387 } 388 389 if_clone_create(name, namelen, NULL); 390} /* tapclone */ 391 392 393/* 394 * tapcreate 395 * 396 * to create interface 397 */ 398static void 399tapcreate(struct cdev *dev) 400{ 401 struct ifnet *ifp = NULL; 402 struct tap_softc *tp = NULL; 403 unsigned short macaddr_hi; 404 int unit, s; 405 char *name = NULL; 406 u_char eaddr[6]; 407 408 dev->si_flags &= ~SI_CHEAPCLONE; 409 410 /* allocate driver storage and create device */ 411 MALLOC(tp, struct tap_softc *, sizeof(*tp), M_TAP, M_WAITOK | M_ZERO); 412 mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF); 413 mtx_lock(&tapmtx); 414 SLIST_INSERT_HEAD(&taphead, tp, tap_next); 415 mtx_unlock(&tapmtx); 416 417 unit = dev2unit(dev); 418 419 /* select device: tap or vmnet */ 420 if (unit & VMNET_DEV_MASK) { 421 name = VMNET; 422 tp->tap_flags |= TAP_VMNET; 423 } else 424 name = TAP; 425 426 unit &= TAPMAXUNIT; 427 428 TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, minor(dev)); 429 430 /* generate fake MAC address: 00 bd xx xx xx unit_no */ 431 macaddr_hi = htons(0x00bd); 432 bcopy(&macaddr_hi, eaddr, sizeof(short)); 433 bcopy(&ticks, &eaddr[2], sizeof(long)); 434 eaddr[5] = (u_char)unit; 435 436 /* fill the rest and attach interface */ 437 ifp = tp->tap_ifp = if_alloc(IFT_ETHER); 438 if (ifp == NULL) 439 panic("%s%d: can not if_alloc()", name, unit); 440 ifp->if_softc = tp; 441 if_initname(ifp, name, unit); 442 ifp->if_init = tapifinit; 443 ifp->if_start = tapifstart; 444 ifp->if_ioctl = tapifioctl; 445 ifp->if_mtu = ETHERMTU; 446 ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST); 447 ifp->if_snd.ifq_maxlen = ifqmaxlen; 448 449 dev->si_drv1 = tp; 450 tp->tap_dev = dev; 451 452 s = splimp(); 453 ether_ifattach(ifp, eaddr); 454 splx(s); 455 456 mtx_lock(&tp->tap_mtx); 457 tp->tap_flags |= TAP_INITED; 458 mtx_unlock(&tp->tap_mtx); 459 460 knlist_init(&tp->tap_rsel.si_note, NULL, NULL, NULL, NULL); 461 462 TAPDEBUG("interface %s is created. minor = %#x\n", 463 ifp->if_xname, minor(dev)); 464} /* tapcreate */ 465 466 467/* 468 * tapopen 469 * 470 * to open tunnel. must be superuser 471 */ 472static int 473tapopen(struct cdev *dev, int flag, int mode, struct thread *td) 474{ 475 struct tap_softc *tp = NULL; 476 struct ifnet *ifp = NULL; 477 int error, s; 478 479 if (tapuopen == 0) { 480 error = priv_check(td, PRIV_NET_TAP); 481 if (error) 482 return (error); 483 } 484 485 if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT) 486 return (ENXIO); 487 488 tp = dev->si_drv1; 489 490 mtx_lock(&tp->tap_mtx); 491 if (tp->tap_flags & TAP_OPEN) { 492 mtx_unlock(&tp->tap_mtx); 493 return (EBUSY); 494 } 495 496 bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr)); 497 tp->tap_pid = td->td_proc->p_pid; 498 tp->tap_flags |= TAP_OPEN; 499 ifp = tp->tap_ifp; 500 mtx_unlock(&tp->tap_mtx); 501 502 s = splimp(); 503 ifp->if_drv_flags |= IFF_DRV_RUNNING; 504 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 505 splx(s); 506 507 TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, minor(dev)); 508 509 return (0); 510} /* tapopen */ 511 512 513/* 514 * tapclose 515 * 516 * close the device - mark i/f down & delete routing info 517 */ 518static int 519tapclose(struct cdev *dev, int foo, int bar, struct thread *td) 520{ 521 struct ifaddr *ifa; 522 struct tap_softc *tp = dev->si_drv1; 523 struct ifnet *ifp = tp->tap_ifp; 524 int s; 525 526 /* junk all pending output */ 527 IF_DRAIN(&ifp->if_snd); 528 529 /* 530 * do not bring the interface down, and do not anything with 531 * interface, if we are in VMnet mode. just close the device. 532 */ 533 534 mtx_lock(&tp->tap_mtx); 535 if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) { 536 mtx_unlock(&tp->tap_mtx); 537 s = splimp(); 538 if_down(ifp); 539 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 540 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 541 rtinit(ifa, (int)RTM_DELETE, 0); 542 } 543 if_purgeaddrs(ifp); 544 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 545 } 546 splx(s); 547 } else 548 mtx_unlock(&tp->tap_mtx); 549 550 funsetown(&tp->tap_sigio); 551 selwakeuppri(&tp->tap_rsel, PZERO+1); 552 KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0); 553 554 mtx_lock(&tp->tap_mtx); 555 tp->tap_flags &= ~TAP_OPEN; 556 tp->tap_pid = 0; 557 mtx_unlock(&tp->tap_mtx); 558 559 TAPDEBUG("%s is closed. minor = %#x\n", 560 ifp->if_xname, minor(dev)); 561 562 return (0); 563} /* tapclose */ 564 565 566/* 567 * tapifinit 568 * 569 * network interface initialization function 570 */ 571static void 572tapifinit(void *xtp) 573{ 574 struct tap_softc *tp = (struct tap_softc *)xtp; 575 struct ifnet *ifp = tp->tap_ifp; 576 577 TAPDEBUG("initializing %s\n", ifp->if_xname); 578 579 ifp->if_drv_flags |= IFF_DRV_RUNNING; 580 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 581 582 /* attempt to start output */ 583 tapifstart(ifp); 584} /* tapifinit */ 585 586 587/* 588 * tapifioctl 589 * 590 * Process an ioctl request on network interface 591 */ 592static int 593tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 594{ 595 struct tap_softc *tp = ifp->if_softc; 596 struct ifstat *ifs = NULL; 597 int s, dummy; 598 599 switch (cmd) { 600 case SIOCSIFFLAGS: /* XXX -- just like vmnet does */ 601 case SIOCADDMULTI: 602 case SIOCDELMULTI: 603 break; 604 605 case SIOCGIFSTATUS: 606 s = splimp(); 607 ifs = (struct ifstat *)data; 608 dummy = strlen(ifs->ascii); 609 mtx_lock(&tp->tap_mtx); 610 if (tp->tap_pid != 0 && dummy < sizeof(ifs->ascii)) 611 snprintf(ifs->ascii + dummy, 612 sizeof(ifs->ascii) - dummy, 613 "\tOpened by PID %d\n", tp->tap_pid); 614 mtx_unlock(&tp->tap_mtx); 615 splx(s); 616 break; 617 618 default: 619 s = splimp(); 620 dummy = ether_ioctl(ifp, cmd, data); 621 splx(s); 622 return (dummy); 623 /* NOT REACHED */ 624 } 625 626 return (0); 627} /* tapifioctl */ 628 629 630/* 631 * tapifstart 632 * 633 * queue packets from higher level ready to put out 634 */ 635static void 636tapifstart(struct ifnet *ifp) 637{ 638 struct tap_softc *tp = ifp->if_softc; 639 int s; 640 641 TAPDEBUG("%s starting\n", ifp->if_xname); 642 643 /* 644 * do not junk pending output if we are in VMnet mode. 645 * XXX: can this do any harm because of queue overflow? 646 */ 647 648 mtx_lock(&tp->tap_mtx); 649 if (((tp->tap_flags & TAP_VMNET) == 0) && 650 ((tp->tap_flags & TAP_READY) != TAP_READY)) { 651 struct mbuf *m = NULL; 652 653 mtx_unlock(&tp->tap_mtx); 654 655 /* Unlocked read. */ 656 TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname, 657 tp->tap_flags); 658 659 s = splimp(); 660 do { 661 IF_DEQUEUE(&ifp->if_snd, m); 662 if (m != NULL) 663 m_freem(m); 664 ifp->if_oerrors ++; 665 } while (m != NULL); 666 splx(s); 667 668 return; 669 } 670 mtx_unlock(&tp->tap_mtx); 671 672 s = splimp(); 673 ifp->if_drv_flags |= IFF_DRV_OACTIVE; 674 675 if (ifp->if_snd.ifq_len != 0) { 676 mtx_lock(&tp->tap_mtx); 677 if (tp->tap_flags & TAP_RWAIT) { 678 tp->tap_flags &= ~TAP_RWAIT; 679 wakeup(tp); 680 } 681 682 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) { 683 mtx_unlock(&tp->tap_mtx); 684 pgsigio(&tp->tap_sigio, SIGIO, 0); 685 } else 686 mtx_unlock(&tp->tap_mtx); 687 688 selwakeuppri(&tp->tap_rsel, PZERO+1); 689 KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0); 690 ifp->if_opackets ++; /* obytes are counted in ether_output */ 691 } 692 693 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 694 splx(s); 695} /* tapifstart */ 696 697 698/* 699 * tapioctl 700 * 701 * the cdevsw interface is now pretty minimal 702 */ 703static int 704tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 705{ 706 struct tap_softc *tp = dev->si_drv1; 707 struct ifnet *ifp = tp->tap_ifp; 708 struct tapinfo *tapp = NULL; 709 int s; 710 int f; 711#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \ 712 defined(COMPAT_FREEBSD4) 713 int ival; 714#endif 715 716 switch (cmd) { 717 case TAPSIFINFO: 718 s = splimp(); 719 tapp = (struct tapinfo *)data; 720 ifp->if_mtu = tapp->mtu; 721 ifp->if_type = tapp->type; 722 ifp->if_baudrate = tapp->baudrate; 723 splx(s); 724 break; 725 726 case TAPGIFINFO: 727 tapp = (struct tapinfo *)data; 728 tapp->mtu = ifp->if_mtu; 729 tapp->type = ifp->if_type; 730 tapp->baudrate = ifp->if_baudrate; 731 break; 732 733 case TAPSDEBUG: 734 tapdebug = *(int *)data; 735 break; 736 737 case TAPGDEBUG: 738 *(int *)data = tapdebug; 739 break; 740 741 case FIONBIO: 742 break; 743 744 case FIOASYNC: 745 s = splimp(); 746 mtx_lock(&tp->tap_mtx); 747 if (*(int *)data) 748 tp->tap_flags |= TAP_ASYNC; 749 else 750 tp->tap_flags &= ~TAP_ASYNC; 751 mtx_unlock(&tp->tap_mtx); 752 splx(s); 753 break; 754 755 case FIONREAD: 756 s = splimp(); 757 if (ifp->if_snd.ifq_head) { 758 struct mbuf *mb = ifp->if_snd.ifq_head; 759 760 for(*(int *)data = 0;mb != NULL;mb = mb->m_next) 761 *(int *)data += mb->m_len; 762 } else 763 *(int *)data = 0; 764 splx(s); 765 break; 766 767 case FIOSETOWN: 768 return (fsetown(*(int *)data, &tp->tap_sigio)); 769 770 case FIOGETOWN: 771 *(int *)data = fgetown(&tp->tap_sigio); 772 return (0); 773 774 /* this is deprecated, FIOSETOWN should be used instead */ 775 case TIOCSPGRP: 776 return (fsetown(-(*(int *)data), &tp->tap_sigio)); 777 778 /* this is deprecated, FIOGETOWN should be used instead */ 779 case TIOCGPGRP: 780 *(int *)data = -fgetown(&tp->tap_sigio); 781 return (0); 782 783 /* VMware/VMnet port ioctl's */ 784 785 case SIOCGIFFLAGS: /* get ifnet flags */ 786 bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags)); 787 break; 788 789#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \ 790 defined(COMPAT_FREEBSD4) 791 case _IO('V', 0): 792 ival = IOCPARM_IVAL(data); 793 data = (caddr_t)&ival; 794 /* FALLTHROUGH */ 795#endif 796 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */ 797 f = *(int *)data; 798 f &= 0x0fff; 799 f &= ~IFF_CANTCHANGE; 800 f |= IFF_UP; 801 802 s = splimp(); 803 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE); 804 splx(s); 805 break; 806 807 case OSIOCGIFADDR: /* get MAC address of the remote side */ 808 case SIOCGIFADDR: 809 mtx_lock(&tp->tap_mtx); 810 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr)); 811 mtx_unlock(&tp->tap_mtx); 812 break; 813 814 case SIOCSIFADDR: /* set MAC address of the remote side */ 815 mtx_lock(&tp->tap_mtx); 816 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr)); 817 mtx_unlock(&tp->tap_mtx); 818 break; 819 820 default: 821 return (ENOTTY); 822 } 823 return (0); 824} /* tapioctl */ 825 826 827/* 828 * tapread 829 * 830 * the cdevsw read interface - reads a packet at a time, or at 831 * least as much of a packet as can be read 832 */ 833static int 834tapread(struct cdev *dev, struct uio *uio, int flag) 835{ 836 struct tap_softc *tp = dev->si_drv1; 837 struct ifnet *ifp = tp->tap_ifp; 838 struct mbuf *m = NULL; 839 int error = 0, len, s; 840 841 TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, minor(dev)); 842 843 mtx_lock(&tp->tap_mtx); 844 if ((tp->tap_flags & TAP_READY) != TAP_READY) { 845 mtx_unlock(&tp->tap_mtx); 846 847 /* Unlocked read. */ 848 TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n", 849 ifp->if_xname, minor(dev), tp->tap_flags); 850 851 return (EHOSTDOWN); 852 } 853 854 tp->tap_flags &= ~TAP_RWAIT; 855 mtx_unlock(&tp->tap_mtx); 856 857 /* sleep until we get a packet */ 858 do { 859 s = splimp(); 860 IF_DEQUEUE(&ifp->if_snd, m); 861 splx(s); 862 863 if (m == NULL) { 864 if (flag & O_NONBLOCK) 865 return (EWOULDBLOCK); 866 867 mtx_lock(&tp->tap_mtx); 868 tp->tap_flags |= TAP_RWAIT; 869 mtx_unlock(&tp->tap_mtx); 870 error = tsleep(tp,PCATCH|(PZERO+1),"taprd",0); 871 if (error) 872 return (error); 873 } 874 } while (m == NULL); 875 876 /* feed packet to bpf */ 877 BPF_MTAP(ifp, m); 878 879 /* xfer packet to user space */ 880 while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) { 881 len = min(uio->uio_resid, m->m_len); 882 if (len == 0) 883 break; 884 885 error = uiomove(mtod(m, void *), len, uio); 886 m = m_free(m); 887 } 888 889 if (m != NULL) { 890 TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname, 891 minor(dev)); 892 m_freem(m); 893 } 894 895 return (error); 896} /* tapread */ 897 898 899/* 900 * tapwrite 901 * 902 * the cdevsw write interface - an atomic write is a packet - or else! 903 */ 904static int 905tapwrite(struct cdev *dev, struct uio *uio, int flag) 906{ 907 struct ether_header *eh; 908 struct tap_softc *tp = dev->si_drv1; 909 struct ifnet *ifp = tp->tap_ifp; 910 struct mbuf *m; 911 912 TAPDEBUG("%s writting, minor = %#x\n", 913 ifp->if_xname, minor(dev)); 914 915 if (uio->uio_resid == 0) 916 return (0); 917 918 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) { 919 TAPDEBUG("%s invalid packet len = %d, minor = %#x\n", 920 ifp->if_xname, uio->uio_resid, minor(dev)); 921 922 return (EIO); 923 } 924 925 if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, ETHER_ALIGN, 926 M_PKTHDR)) == NULL) { 927 ifp->if_ierrors ++; 928 return (ENOBUFS); 929 } 930 931 m->m_pkthdr.rcvif = ifp; 932 933 /* 934 * Only pass a unicast frame to ether_input(), if it would actually 935 * have been received by non-virtual hardware. 936 */ 937 if (m->m_len < sizeof(struct ether_header)) { 938 m_freem(m); 939 return (0); 940 } 941 eh = mtod(m, struct ether_header *); 942 943 if (eh && (ifp->if_flags & IFF_PROMISC) == 0 && 944 !ETHER_IS_MULTICAST(eh->ether_dhost) && 945 bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) { 946 m_freem(m); 947 return (0); 948 } 949 950 /* Pass packet up to parent. */ 951 (*ifp->if_input)(ifp, m); 952 ifp->if_ipackets ++; /* ibytes are counted in parent */ 953 954 return (0); 955} /* tapwrite */ 956 957 958/* 959 * tappoll 960 * 961 * the poll interface, this is only useful on reads 962 * really. the write detect always returns true, write never blocks 963 * anyway, it either accepts the packet or drops it 964 */ 965static int 966tappoll(struct cdev *dev, int events, struct thread *td) 967{ 968 struct tap_softc *tp = dev->si_drv1; 969 struct ifnet *ifp = tp->tap_ifp; 970 int s, revents = 0; 971 972 TAPDEBUG("%s polling, minor = %#x\n", 973 ifp->if_xname, minor(dev)); 974 975 s = splimp(); 976 if (events & (POLLIN | POLLRDNORM)) { 977 if (ifp->if_snd.ifq_len > 0) { 978 TAPDEBUG("%s have data in queue. len = %d, " \ 979 "minor = %#x\n", ifp->if_xname, 980 ifp->if_snd.ifq_len, minor(dev)); 981 982 revents |= (events & (POLLIN | POLLRDNORM)); 983 } else { 984 TAPDEBUG("%s waiting for data, minor = %#x\n", 985 ifp->if_xname, minor(dev)); 986 987 selrecord(td, &tp->tap_rsel); 988 } 989 } 990 991 if (events & (POLLOUT | POLLWRNORM)) 992 revents |= (events & (POLLOUT | POLLWRNORM)); 993 994 splx(s); 995 return (revents); 996} /* tappoll */ 997 998 999/* 1000 * tap_kqfilter 1001 * 1002 * support for kevent() system call 1003 */ 1004static int 1005tapkqfilter(struct cdev *dev, struct knote *kn) 1006{ 1007 int s; 1008 struct tap_softc *tp = dev->si_drv1; 1009 struct ifnet *ifp = tp->tap_ifp; 1010 1011 s = splimp(); 1012 switch (kn->kn_filter) { 1013 case EVFILT_READ: 1014 TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n", 1015 ifp->if_xname, minor(dev)); 1016 kn->kn_fop = &tap_read_filterops; 1017 break; 1018 1019 case EVFILT_WRITE: 1020 TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n", 1021 ifp->if_xname, minor(dev)); 1022 kn->kn_fop = &tap_write_filterops; 1023 break; 1024 1025 default: 1026 TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n", 1027 ifp->if_xname, minor(dev)); 1028 splx(s); 1029 return (EINVAL); 1030 /* NOT REACHED */ 1031 } 1032 splx(s); 1033 1034 kn->kn_hook = (caddr_t) dev; 1035 knlist_add(&tp->tap_rsel.si_note, kn, 0); 1036 1037 return (0); 1038} /* tapkqfilter */ 1039 1040 1041/* 1042 * tap_kqread 1043 * 1044 * Return true if there is data in the interface queue 1045 */ 1046static int 1047tapkqread(struct knote *kn, long hint) 1048{ 1049 int ret, s; 1050 struct cdev *dev = (struct cdev *)(kn->kn_hook); 1051 struct tap_softc *tp = dev->si_drv1; 1052 struct ifnet *ifp = tp->tap_ifp; 1053 1054 s = splimp(); 1055 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1056 TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n", 1057 ifp->if_xname, ifp->if_snd.ifq_len, minor(dev)); 1058 ret = 1; 1059 } else { 1060 TAPDEBUG("%s waiting for data, minor = %#x\n", 1061 ifp->if_xname, minor(dev)); 1062 ret = 0; 1063 } 1064 splx(s); 1065 1066 return (ret); 1067} /* tapkqread */ 1068 1069 1070/* 1071 * tap_kqwrite 1072 * 1073 * Always can write. Return the MTU in kn->data 1074 */ 1075static int 1076tapkqwrite(struct knote *kn, long hint) 1077{ 1078 int s; 1079 struct tap_softc *tp = ((struct cdev *) kn->kn_hook)->si_drv1; 1080 struct ifnet *ifp = tp->tap_ifp; 1081 1082 s = splimp(); 1083 kn->kn_data = ifp->if_mtu; 1084 splx(s); 1085 1086 return (1); 1087} /* tapkqwrite */ 1088 1089 1090static void 1091tapkqdetach(struct knote *kn) 1092{ 1093 struct tap_softc *tp = ((struct cdev *) kn->kn_hook)->si_drv1; 1094 1095 knlist_remove(&tp->tap_rsel.si_note, kn, 0); 1096} /* tapkqdetach */ 1097 1098