if_tap.c revision 281363
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 281363 2015-04-10 09:50:13Z glebius $ 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/jail.h> 46#include <sys/kernel.h> 47#include <sys/malloc.h> 48#include <sys/mbuf.h> 49#include <sys/module.h> 50#include <sys/poll.h> 51#include <sys/priv.h> 52#include <sys/proc.h> 53#include <sys/selinfo.h> 54#include <sys/signalvar.h> 55#include <sys/socket.h> 56#include <sys/sockio.h> 57#include <sys/sysctl.h> 58#include <sys/systm.h> 59#include <sys/ttycom.h> 60#include <sys/uio.h> 61#include <sys/queue.h> 62 63#include <net/bpf.h> 64#include <net/ethernet.h> 65#include <net/if.h> 66#include <net/if_var.h> 67#include <net/if_clone.h> 68#include <net/if_dl.h> 69#include <net/if_media.h> 70#include <net/if_types.h> 71#include <net/route.h> 72#include <net/vnet.h> 73 74#include <netinet/in.h> 75 76#include <net/if_tapvar.h> 77#include <net/if_tap.h> 78 79 80#define CDEV_NAME "tap" 81#define TAPDEBUG if (tapdebug) printf 82 83static const char tapname[] = "tap"; 84static const char vmnetname[] = "vmnet"; 85#define TAPMAXUNIT 0x7fff 86#define VMNET_DEV_MASK CLONE_FLAG0 87 88/* module */ 89static int tapmodevent(module_t, int, void *); 90 91/* device */ 92static void tapclone(void *, struct ucred *, char *, int, 93 struct cdev **); 94static void tapcreate(struct cdev *); 95 96/* network interface */ 97static void tapifstart(struct ifnet *); 98static int tapifioctl(struct ifnet *, u_long, caddr_t); 99static void tapifinit(void *); 100 101static int tap_clone_create(struct if_clone *, int, caddr_t); 102static void tap_clone_destroy(struct ifnet *); 103static struct if_clone *tap_cloner; 104static int vmnet_clone_create(struct if_clone *, int, caddr_t); 105static void vmnet_clone_destroy(struct ifnet *); 106static struct if_clone *vmnet_cloner; 107 108/* character device */ 109static d_open_t tapopen; 110static d_close_t tapclose; 111static d_read_t tapread; 112static d_write_t tapwrite; 113static d_ioctl_t tapioctl; 114static d_poll_t tappoll; 115static d_kqfilter_t tapkqfilter; 116 117/* kqueue(2) */ 118static int tapkqread(struct knote *, long); 119static int tapkqwrite(struct knote *, long); 120static void tapkqdetach(struct knote *); 121 122static struct filterops tap_read_filterops = { 123 .f_isfd = 1, 124 .f_attach = NULL, 125 .f_detach = tapkqdetach, 126 .f_event = tapkqread, 127}; 128 129static struct filterops tap_write_filterops = { 130 .f_isfd = 1, 131 .f_attach = NULL, 132 .f_detach = tapkqdetach, 133 .f_event = tapkqwrite, 134}; 135 136static struct cdevsw tap_cdevsw = { 137 .d_version = D_VERSION, 138 .d_flags = D_NEEDMINOR, 139 .d_open = tapopen, 140 .d_close = tapclose, 141 .d_read = tapread, 142 .d_write = tapwrite, 143 .d_ioctl = tapioctl, 144 .d_poll = tappoll, 145 .d_name = CDEV_NAME, 146 .d_kqfilter = tapkqfilter, 147}; 148 149/* 150 * All global variables in if_tap.c are locked with tapmtx, with the 151 * exception of tapdebug, which is accessed unlocked; tapclones is 152 * static at runtime. 153 */ 154static struct mtx tapmtx; 155static int tapdebug = 0; /* debug flag */ 156static int tapuopen = 0; /* allow user open() */ 157static int tapuponopen = 0; /* IFF_UP on open() */ 158static int tapdclone = 1; /* enable devfs cloning */ 159static SLIST_HEAD(, tap_softc) taphead; /* first device */ 160static struct clonedevs *tapclones; 161 162MALLOC_DECLARE(M_TAP); 163MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface"); 164SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, ""); 165 166SYSCTL_DECL(_net_link); 167static SYSCTL_NODE(_net_link, OID_AUTO, tap, CTLFLAG_RW, 0, 168 "Ethernet tunnel software network interface"); 169SYSCTL_INT(_net_link_tap, OID_AUTO, user_open, CTLFLAG_RW, &tapuopen, 0, 170 "Allow user to open /dev/tap (based on node permissions)"); 171SYSCTL_INT(_net_link_tap, OID_AUTO, up_on_open, CTLFLAG_RW, &tapuponopen, 0, 172 "Bring interface up when /dev/tap is opened"); 173SYSCTL_INT(_net_link_tap, OID_AUTO, devfs_cloning, CTLFLAG_RWTUN, &tapdclone, 0, 174 "Enably legacy devfs interface creation"); 175SYSCTL_INT(_net_link_tap, OID_AUTO, debug, CTLFLAG_RW, &tapdebug, 0, ""); 176 177DEV_MODULE(if_tap, tapmodevent, NULL); 178 179static int 180tap_clone_create(struct if_clone *ifc, int unit, caddr_t params) 181{ 182 struct cdev *dev; 183 int i; 184 185 /* Find any existing device, or allocate new unit number. */ 186 i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, 0); 187 if (i) { 188 dev = make_dev(&tap_cdevsw, unit, UID_ROOT, GID_WHEEL, 0600, 189 "%s%d", tapname, unit); 190 } 191 192 tapcreate(dev); 193 return (0); 194} 195 196/* vmnet devices are tap devices in disguise */ 197static int 198vmnet_clone_create(struct if_clone *ifc, int unit, caddr_t params) 199{ 200 struct cdev *dev; 201 int i; 202 203 /* Find any existing device, or allocate new unit number. */ 204 i = clone_create(&tapclones, &tap_cdevsw, &unit, &dev, VMNET_DEV_MASK); 205 if (i) { 206 dev = make_dev(&tap_cdevsw, unit | VMNET_DEV_MASK, UID_ROOT, 207 GID_WHEEL, 0600, "%s%d", vmnetname, unit); 208 } 209 210 tapcreate(dev); 211 return (0); 212} 213 214static void 215tap_destroy(struct tap_softc *tp) 216{ 217 struct ifnet *ifp = tp->tap_ifp; 218 219 CURVNET_SET(ifp->if_vnet); 220 destroy_dev(tp->tap_dev); 221 seldrain(&tp->tap_rsel); 222 knlist_clear(&tp->tap_rsel.si_note, 0); 223 knlist_destroy(&tp->tap_rsel.si_note); 224 ether_ifdetach(ifp); 225 if_free(ifp); 226 227 mtx_destroy(&tp->tap_mtx); 228 free(tp, M_TAP); 229 CURVNET_RESTORE(); 230} 231 232static void 233tap_clone_destroy(struct ifnet *ifp) 234{ 235 struct tap_softc *tp = ifp->if_softc; 236 237 mtx_lock(&tapmtx); 238 SLIST_REMOVE(&taphead, tp, tap_softc, tap_next); 239 mtx_unlock(&tapmtx); 240 tap_destroy(tp); 241} 242 243/* vmnet devices are tap devices in disguise */ 244static void 245vmnet_clone_destroy(struct ifnet *ifp) 246{ 247 tap_clone_destroy(ifp); 248} 249 250/* 251 * tapmodevent 252 * 253 * module event handler 254 */ 255static int 256tapmodevent(module_t mod, int type, void *data) 257{ 258 static eventhandler_tag eh_tag = NULL; 259 struct tap_softc *tp = NULL; 260 struct ifnet *ifp = NULL; 261 262 switch (type) { 263 case MOD_LOAD: 264 265 /* intitialize device */ 266 267 mtx_init(&tapmtx, "tapmtx", NULL, MTX_DEF); 268 SLIST_INIT(&taphead); 269 270 clone_setup(&tapclones); 271 eh_tag = EVENTHANDLER_REGISTER(dev_clone, tapclone, 0, 1000); 272 if (eh_tag == NULL) { 273 clone_cleanup(&tapclones); 274 mtx_destroy(&tapmtx); 275 return (ENOMEM); 276 } 277 tap_cloner = if_clone_simple(tapname, tap_clone_create, 278 tap_clone_destroy, 0); 279 vmnet_cloner = if_clone_simple(vmnetname, vmnet_clone_create, 280 vmnet_clone_destroy, 0); 281 return (0); 282 283 case MOD_UNLOAD: 284 /* 285 * The EBUSY algorithm here can't quite atomically 286 * guarantee that this is race-free since we have to 287 * release the tap mtx to deregister the clone handler. 288 */ 289 mtx_lock(&tapmtx); 290 SLIST_FOREACH(tp, &taphead, tap_next) { 291 mtx_lock(&tp->tap_mtx); 292 if (tp->tap_flags & TAP_OPEN) { 293 mtx_unlock(&tp->tap_mtx); 294 mtx_unlock(&tapmtx); 295 return (EBUSY); 296 } 297 mtx_unlock(&tp->tap_mtx); 298 } 299 mtx_unlock(&tapmtx); 300 301 EVENTHANDLER_DEREGISTER(dev_clone, eh_tag); 302 if_clone_detach(tap_cloner); 303 if_clone_detach(vmnet_cloner); 304 drain_dev_clone_events(); 305 306 mtx_lock(&tapmtx); 307 while ((tp = SLIST_FIRST(&taphead)) != NULL) { 308 SLIST_REMOVE_HEAD(&taphead, tap_next); 309 mtx_unlock(&tapmtx); 310 311 ifp = tp->tap_ifp; 312 313 TAPDEBUG("detaching %s\n", ifp->if_xname); 314 315 tap_destroy(tp); 316 mtx_lock(&tapmtx); 317 } 318 mtx_unlock(&tapmtx); 319 clone_cleanup(&tapclones); 320 321 mtx_destroy(&tapmtx); 322 323 break; 324 325 default: 326 return (EOPNOTSUPP); 327 } 328 329 return (0); 330} /* tapmodevent */ 331 332 333/* 334 * DEVFS handler 335 * 336 * We need to support two kind of devices - tap and vmnet 337 */ 338static void 339tapclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev) 340{ 341 char devname[SPECNAMELEN + 1]; 342 int i, unit, append_unit; 343 int extra; 344 345 if (*dev != NULL) 346 return; 347 348 if (!tapdclone || 349 (!tapuopen && priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0)) 350 return; 351 352 unit = 0; 353 append_unit = 0; 354 extra = 0; 355 356 /* We're interested in only tap/vmnet devices. */ 357 if (strcmp(name, tapname) == 0) { 358 unit = -1; 359 } else if (strcmp(name, vmnetname) == 0) { 360 unit = -1; 361 extra = VMNET_DEV_MASK; 362 } else if (dev_stdclone(name, NULL, tapname, &unit) != 1) { 363 if (dev_stdclone(name, NULL, vmnetname, &unit) != 1) { 364 return; 365 } else { 366 extra = VMNET_DEV_MASK; 367 } 368 } 369 370 if (unit == -1) 371 append_unit = 1; 372 373 CURVNET_SET(CRED_TO_VNET(cred)); 374 /* find any existing device, or allocate new unit number */ 375 i = clone_create(&tapclones, &tap_cdevsw, &unit, dev, extra); 376 if (i) { 377 if (append_unit) { 378 /* 379 * We were passed 'tun' or 'tap', with no unit specified 380 * so we'll need to append it now. 381 */ 382 namelen = snprintf(devname, sizeof(devname), "%s%d", name, 383 unit); 384 name = devname; 385 } 386 387 *dev = make_dev_credf(MAKEDEV_REF, &tap_cdevsw, unit | extra, 388 cred, UID_ROOT, GID_WHEEL, 0600, "%s", name); 389 } 390 391 if_clone_create(name, namelen, NULL); 392 CURVNET_RESTORE(); 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 uint32_t macaddr_mid; 408 int unit; 409 const char *name = NULL; 410 u_char eaddr[6]; 411 412 /* allocate driver storage and create device */ 413 tp = malloc(sizeof(*tp), M_TAP, M_WAITOK | M_ZERO); 414 mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF); 415 mtx_lock(&tapmtx); 416 SLIST_INSERT_HEAD(&taphead, tp, tap_next); 417 mtx_unlock(&tapmtx); 418 419 unit = dev2unit(dev); 420 421 /* select device: tap or vmnet */ 422 if (unit & VMNET_DEV_MASK) { 423 name = vmnetname; 424 tp->tap_flags |= TAP_VMNET; 425 } else 426 name = tapname; 427 428 unit &= TAPMAXUNIT; 429 430 TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, dev2unit(dev)); 431 432 /* generate fake MAC address: 00 bd xx xx xx unit_no */ 433 macaddr_hi = htons(0x00bd); 434 macaddr_mid = (uint32_t) ticks; 435 bcopy(&macaddr_hi, eaddr, sizeof(short)); 436 bcopy(&macaddr_mid, &eaddr[2], sizeof(uint32_t)); 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 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 451 ifp->if_capabilities |= IFCAP_LINKSTATE; 452 ifp->if_capenable |= IFCAP_LINKSTATE; 453 454 dev->si_drv1 = tp; 455 tp->tap_dev = dev; 456 457 ether_ifattach(ifp, eaddr); 458 459 mtx_lock(&tp->tap_mtx); 460 tp->tap_flags |= TAP_INITED; 461 mtx_unlock(&tp->tap_mtx); 462 463 knlist_init_mtx(&tp->tap_rsel.si_note, &tp->tap_mtx); 464 465 TAPDEBUG("interface %s is created. minor = %#x\n", 466 ifp->if_xname, dev2unit(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; 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; 503 504 ifp->if_drv_flags |= IFF_DRV_RUNNING; 505 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 506 if (tapuponopen) 507 ifp->if_flags |= IFF_UP; 508 if_link_state_change(ifp, LINK_STATE_UP); 509 mtx_unlock(&tp->tap_mtx); 510 511 TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, dev2unit(dev)); 512 513 return (0); 514} /* tapopen */ 515 516 517/* 518 * tapclose 519 * 520 * close the device - mark i/f down & delete routing info 521 */ 522static int 523tapclose(struct cdev *dev, int foo, int bar, struct thread *td) 524{ 525 struct ifaddr *ifa; 526 struct tap_softc *tp = dev->si_drv1; 527 struct ifnet *ifp = tp->tap_ifp; 528 529 /* junk all pending output */ 530 mtx_lock(&tp->tap_mtx); 531 CURVNET_SET(ifp->if_vnet); 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 if (((tp->tap_flags & TAP_VMNET) == 0) && 539 (ifp->if_flags & (IFF_UP | IFF_LINK0)) == IFF_UP) { 540 mtx_unlock(&tp->tap_mtx); 541 if_down(ifp); 542 mtx_lock(&tp->tap_mtx); 543 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 544 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 545 mtx_unlock(&tp->tap_mtx); 546 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 547 rtinit(ifa, (int)RTM_DELETE, 0); 548 } 549 if_purgeaddrs(ifp); 550 mtx_lock(&tp->tap_mtx); 551 } 552 } 553 554 if_link_state_change(ifp, LINK_STATE_DOWN); 555 CURVNET_RESTORE(); 556 557 funsetown(&tp->tap_sigio); 558 selwakeuppri(&tp->tap_rsel, PZERO+1); 559 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0); 560 561 tp->tap_flags &= ~TAP_OPEN; 562 tp->tap_pid = 0; 563 mtx_unlock(&tp->tap_mtx); 564 565 TAPDEBUG("%s is closed. minor = %#x\n", 566 ifp->if_xname, dev2unit(dev)); 567 568 return (0); 569} /* tapclose */ 570 571 572/* 573 * tapifinit 574 * 575 * network interface initialization function 576 */ 577static void 578tapifinit(void *xtp) 579{ 580 struct tap_softc *tp = (struct tap_softc *)xtp; 581 struct ifnet *ifp = tp->tap_ifp; 582 583 TAPDEBUG("initializing %s\n", ifp->if_xname); 584 585 mtx_lock(&tp->tap_mtx); 586 ifp->if_drv_flags |= IFF_DRV_RUNNING; 587 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 588 mtx_unlock(&tp->tap_mtx); 589 590 /* attempt to start output */ 591 tapifstart(ifp); 592} /* tapifinit */ 593 594 595/* 596 * tapifioctl 597 * 598 * Process an ioctl request on network interface 599 */ 600static int 601tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 602{ 603 struct tap_softc *tp = ifp->if_softc; 604 struct ifreq *ifr = (struct ifreq *)data; 605 struct ifstat *ifs = NULL; 606 struct ifmediareq *ifmr = NULL; 607 int dummy, error = 0; 608 609 switch (cmd) { 610 case SIOCSIFFLAGS: /* XXX -- just like vmnet does */ 611 case SIOCADDMULTI: 612 case SIOCDELMULTI: 613 break; 614 615 case SIOCGIFMEDIA: 616 ifmr = (struct ifmediareq *)data; 617 dummy = ifmr->ifm_count; 618 ifmr->ifm_count = 1; 619 ifmr->ifm_status = IFM_AVALID; 620 ifmr->ifm_active = IFM_ETHER; 621 if (tp->tap_flags & TAP_OPEN) 622 ifmr->ifm_status |= IFM_ACTIVE; 623 ifmr->ifm_current = ifmr->ifm_active; 624 if (dummy >= 1) { 625 int media = IFM_ETHER; 626 error = copyout(&media, ifmr->ifm_ulist, 627 sizeof(int)); 628 } 629 break; 630 631 case SIOCSIFMTU: 632 ifp->if_mtu = ifr->ifr_mtu; 633 break; 634 635 case SIOCGIFSTATUS: 636 ifs = (struct ifstat *)data; 637 mtx_lock(&tp->tap_mtx); 638 if (tp->tap_pid != 0) 639 snprintf(ifs->ascii, sizeof(ifs->ascii), 640 "\tOpened by PID %d\n", tp->tap_pid); 641 else 642 ifs->ascii[0] = '\0'; 643 mtx_unlock(&tp->tap_mtx); 644 break; 645 646 default: 647 error = ether_ioctl(ifp, cmd, data); 648 break; 649 } 650 651 return (error); 652} /* tapifioctl */ 653 654 655/* 656 * tapifstart 657 * 658 * queue packets from higher level ready to put out 659 */ 660static void 661tapifstart(struct ifnet *ifp) 662{ 663 struct tap_softc *tp = ifp->if_softc; 664 665 TAPDEBUG("%s starting\n", ifp->if_xname); 666 667 /* 668 * do not junk pending output if we are in VMnet mode. 669 * XXX: can this do any harm because of queue overflow? 670 */ 671 672 mtx_lock(&tp->tap_mtx); 673 if (((tp->tap_flags & TAP_VMNET) == 0) && 674 ((tp->tap_flags & TAP_READY) != TAP_READY)) { 675 struct mbuf *m; 676 677 /* Unlocked read. */ 678 TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname, 679 tp->tap_flags); 680 681 for (;;) { 682 IF_DEQUEUE(&ifp->if_snd, m); 683 if (m != NULL) { 684 m_freem(m); 685 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 686 } else 687 break; 688 } 689 mtx_unlock(&tp->tap_mtx); 690 691 return; 692 } 693 694 ifp->if_drv_flags |= IFF_DRV_OACTIVE; 695 696 if (!IFQ_IS_EMPTY(&ifp->if_snd)) { 697 if (tp->tap_flags & TAP_RWAIT) { 698 tp->tap_flags &= ~TAP_RWAIT; 699 wakeup(tp); 700 } 701 702 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) { 703 mtx_unlock(&tp->tap_mtx); 704 pgsigio(&tp->tap_sigio, SIGIO, 0); 705 mtx_lock(&tp->tap_mtx); 706 } 707 708 selwakeuppri(&tp->tap_rsel, PZERO+1); 709 KNOTE_LOCKED(&tp->tap_rsel.si_note, 0); 710 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); /* obytes are counted in ether_output */ 711 } 712 713 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 714 mtx_unlock(&tp->tap_mtx); 715} /* tapifstart */ 716 717 718/* 719 * tapioctl 720 * 721 * the cdevsw interface is now pretty minimal 722 */ 723static int 724tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 725{ 726 struct tap_softc *tp = dev->si_drv1; 727 struct ifnet *ifp = tp->tap_ifp; 728 struct tapinfo *tapp = NULL; 729 int f; 730#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \ 731 defined(COMPAT_FREEBSD4) 732 int ival; 733#endif 734 735 switch (cmd) { 736 case TAPSIFINFO: 737 tapp = (struct tapinfo *)data; 738 mtx_lock(&tp->tap_mtx); 739 ifp->if_mtu = tapp->mtu; 740 ifp->if_type = tapp->type; 741 ifp->if_baudrate = tapp->baudrate; 742 mtx_unlock(&tp->tap_mtx); 743 break; 744 745 case TAPGIFINFO: 746 tapp = (struct tapinfo *)data; 747 mtx_lock(&tp->tap_mtx); 748 tapp->mtu = ifp->if_mtu; 749 tapp->type = ifp->if_type; 750 tapp->baudrate = ifp->if_baudrate; 751 mtx_unlock(&tp->tap_mtx); 752 break; 753 754 case TAPSDEBUG: 755 tapdebug = *(int *)data; 756 break; 757 758 case TAPGDEBUG: 759 *(int *)data = tapdebug; 760 break; 761 762 case TAPGIFNAME: { 763 struct ifreq *ifr = (struct ifreq *) data; 764 765 strlcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ); 766 } break; 767 768 case FIONBIO: 769 break; 770 771 case FIOASYNC: 772 mtx_lock(&tp->tap_mtx); 773 if (*(int *)data) 774 tp->tap_flags |= TAP_ASYNC; 775 else 776 tp->tap_flags &= ~TAP_ASYNC; 777 mtx_unlock(&tp->tap_mtx); 778 break; 779 780 case FIONREAD: 781 if (!IFQ_IS_EMPTY(&ifp->if_snd)) { 782 struct mbuf *mb; 783 784 IFQ_LOCK(&ifp->if_snd); 785 IFQ_POLL_NOLOCK(&ifp->if_snd, mb); 786 for (*(int *)data = 0; mb != NULL; 787 mb = mb->m_next) 788 *(int *)data += mb->m_len; 789 IFQ_UNLOCK(&ifp->if_snd); 790 } else 791 *(int *)data = 0; 792 break; 793 794 case FIOSETOWN: 795 return (fsetown(*(int *)data, &tp->tap_sigio)); 796 797 case FIOGETOWN: 798 *(int *)data = fgetown(&tp->tap_sigio); 799 return (0); 800 801 /* this is deprecated, FIOSETOWN should be used instead */ 802 case TIOCSPGRP: 803 return (fsetown(-(*(int *)data), &tp->tap_sigio)); 804 805 /* this is deprecated, FIOGETOWN should be used instead */ 806 case TIOCGPGRP: 807 *(int *)data = -fgetown(&tp->tap_sigio); 808 return (0); 809 810 /* VMware/VMnet port ioctl's */ 811 812#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \ 813 defined(COMPAT_FREEBSD4) 814 case _IO('V', 0): 815 ival = IOCPARM_IVAL(data); 816 data = (caddr_t)&ival; 817 /* FALLTHROUGH */ 818#endif 819 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */ 820 f = *(int *)data; 821 f &= 0x0fff; 822 f &= ~IFF_CANTCHANGE; 823 f |= IFF_UP; 824 825 mtx_lock(&tp->tap_mtx); 826 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE); 827 mtx_unlock(&tp->tap_mtx); 828 break; 829 830 case SIOCGIFADDR: /* get MAC address of the remote side */ 831 mtx_lock(&tp->tap_mtx); 832 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr)); 833 mtx_unlock(&tp->tap_mtx); 834 break; 835 836 case SIOCSIFADDR: /* set MAC address of the remote side */ 837 mtx_lock(&tp->tap_mtx); 838 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr)); 839 mtx_unlock(&tp->tap_mtx); 840 break; 841 842 default: 843 return (ENOTTY); 844 } 845 return (0); 846} /* tapioctl */ 847 848 849/* 850 * tapread 851 * 852 * the cdevsw read interface - reads a packet at a time, or at 853 * least as much of a packet as can be read 854 */ 855static int 856tapread(struct cdev *dev, struct uio *uio, int flag) 857{ 858 struct tap_softc *tp = dev->si_drv1; 859 struct ifnet *ifp = tp->tap_ifp; 860 struct mbuf *m = NULL; 861 int error = 0, len; 862 863 TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, dev2unit(dev)); 864 865 mtx_lock(&tp->tap_mtx); 866 if ((tp->tap_flags & TAP_READY) != TAP_READY) { 867 mtx_unlock(&tp->tap_mtx); 868 869 /* Unlocked read. */ 870 TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n", 871 ifp->if_xname, dev2unit(dev), tp->tap_flags); 872 873 return (EHOSTDOWN); 874 } 875 876 tp->tap_flags &= ~TAP_RWAIT; 877 878 /* sleep until we get a packet */ 879 do { 880 IF_DEQUEUE(&ifp->if_snd, m); 881 882 if (m == NULL) { 883 if (flag & O_NONBLOCK) { 884 mtx_unlock(&tp->tap_mtx); 885 return (EWOULDBLOCK); 886 } 887 888 tp->tap_flags |= TAP_RWAIT; 889 error = mtx_sleep(tp, &tp->tap_mtx, PCATCH | (PZERO + 1), 890 "taprd", 0); 891 if (error) { 892 mtx_unlock(&tp->tap_mtx); 893 return (error); 894 } 895 } 896 } while (m == NULL); 897 mtx_unlock(&tp->tap_mtx); 898 899 /* feed packet to bpf */ 900 BPF_MTAP(ifp, m); 901 902 /* xfer packet to user space */ 903 while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) { 904 len = min(uio->uio_resid, m->m_len); 905 if (len == 0) 906 break; 907 908 error = uiomove(mtod(m, void *), len, uio); 909 m = m_free(m); 910 } 911 912 if (m != NULL) { 913 TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname, 914 dev2unit(dev)); 915 m_freem(m); 916 } 917 918 return (error); 919} /* tapread */ 920 921 922/* 923 * tapwrite 924 * 925 * the cdevsw write interface - an atomic write is a packet - or else! 926 */ 927static int 928tapwrite(struct cdev *dev, struct uio *uio, int flag) 929{ 930 struct ether_header *eh; 931 struct tap_softc *tp = dev->si_drv1; 932 struct ifnet *ifp = tp->tap_ifp; 933 struct mbuf *m; 934 935 TAPDEBUG("%s writing, minor = %#x\n", 936 ifp->if_xname, dev2unit(dev)); 937 938 if (uio->uio_resid == 0) 939 return (0); 940 941 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) { 942 TAPDEBUG("%s invalid packet len = %zd, minor = %#x\n", 943 ifp->if_xname, uio->uio_resid, dev2unit(dev)); 944 945 return (EIO); 946 } 947 948 if ((m = m_uiotombuf(uio, M_NOWAIT, 0, ETHER_ALIGN, 949 M_PKTHDR)) == NULL) { 950 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); 951 return (ENOBUFS); 952 } 953 954 m->m_pkthdr.rcvif = ifp; 955 956 /* 957 * Only pass a unicast frame to ether_input(), if it would actually 958 * have been received by non-virtual hardware. 959 */ 960 if (m->m_len < sizeof(struct ether_header)) { 961 m_freem(m); 962 return (0); 963 } 964 eh = mtod(m, struct ether_header *); 965 966 if (eh && (ifp->if_flags & IFF_PROMISC) == 0 && 967 !ETHER_IS_MULTICAST(eh->ether_dhost) && 968 bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) { 969 m_freem(m); 970 return (0); 971 } 972 973 /* Pass packet up to parent. */ 974 CURVNET_SET(ifp->if_vnet); 975 (*ifp->if_input)(ifp, m); 976 CURVNET_RESTORE(); 977 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1); /* ibytes are counted in parent */ 978 979 return (0); 980} /* tapwrite */ 981 982 983/* 984 * tappoll 985 * 986 * the poll interface, this is only useful on reads 987 * really. the write detect always returns true, write never blocks 988 * anyway, it either accepts the packet or drops it 989 */ 990static int 991tappoll(struct cdev *dev, int events, struct thread *td) 992{ 993 struct tap_softc *tp = dev->si_drv1; 994 struct ifnet *ifp = tp->tap_ifp; 995 int revents = 0; 996 997 TAPDEBUG("%s polling, minor = %#x\n", 998 ifp->if_xname, dev2unit(dev)); 999 1000 if (events & (POLLIN | POLLRDNORM)) { 1001 IFQ_LOCK(&ifp->if_snd); 1002 if (!IFQ_IS_EMPTY(&ifp->if_snd)) { 1003 TAPDEBUG("%s have data in queue. len = %d, " \ 1004 "minor = %#x\n", ifp->if_xname, 1005 ifp->if_snd.ifq_len, dev2unit(dev)); 1006 1007 revents |= (events & (POLLIN | POLLRDNORM)); 1008 } else { 1009 TAPDEBUG("%s waiting for data, minor = %#x\n", 1010 ifp->if_xname, dev2unit(dev)); 1011 1012 selrecord(td, &tp->tap_rsel); 1013 } 1014 IFQ_UNLOCK(&ifp->if_snd); 1015 } 1016 1017 if (events & (POLLOUT | POLLWRNORM)) 1018 revents |= (events & (POLLOUT | POLLWRNORM)); 1019 1020 return (revents); 1021} /* tappoll */ 1022 1023 1024/* 1025 * tap_kqfilter 1026 * 1027 * support for kevent() system call 1028 */ 1029static int 1030tapkqfilter(struct cdev *dev, struct knote *kn) 1031{ 1032 struct tap_softc *tp = dev->si_drv1; 1033 struct ifnet *ifp = tp->tap_ifp; 1034 1035 switch (kn->kn_filter) { 1036 case EVFILT_READ: 1037 TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n", 1038 ifp->if_xname, dev2unit(dev)); 1039 kn->kn_fop = &tap_read_filterops; 1040 break; 1041 1042 case EVFILT_WRITE: 1043 TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n", 1044 ifp->if_xname, dev2unit(dev)); 1045 kn->kn_fop = &tap_write_filterops; 1046 break; 1047 1048 default: 1049 TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n", 1050 ifp->if_xname, dev2unit(dev)); 1051 return (EINVAL); 1052 /* NOT REACHED */ 1053 } 1054 1055 kn->kn_hook = tp; 1056 knlist_add(&tp->tap_rsel.si_note, kn, 0); 1057 1058 return (0); 1059} /* tapkqfilter */ 1060 1061 1062/* 1063 * tap_kqread 1064 * 1065 * Return true if there is data in the interface queue 1066 */ 1067static int 1068tapkqread(struct knote *kn, long hint) 1069{ 1070 int ret; 1071 struct tap_softc *tp = kn->kn_hook; 1072 struct cdev *dev = tp->tap_dev; 1073 struct ifnet *ifp = tp->tap_ifp; 1074 1075 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1076 TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n", 1077 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev)); 1078 ret = 1; 1079 } else { 1080 TAPDEBUG("%s waiting for data, minor = %#x\n", 1081 ifp->if_xname, dev2unit(dev)); 1082 ret = 0; 1083 } 1084 1085 return (ret); 1086} /* tapkqread */ 1087 1088 1089/* 1090 * tap_kqwrite 1091 * 1092 * Always can write. Return the MTU in kn->data 1093 */ 1094static int 1095tapkqwrite(struct knote *kn, long hint) 1096{ 1097 struct tap_softc *tp = kn->kn_hook; 1098 struct ifnet *ifp = tp->tap_ifp; 1099 1100 kn->kn_data = ifp->if_mtu; 1101 1102 return (1); 1103} /* tapkqwrite */ 1104 1105 1106static void 1107tapkqdetach(struct knote *kn) 1108{ 1109 struct tap_softc *tp = kn->kn_hook; 1110 1111 knlist_remove(&tp->tap_rsel.si_note, kn, 0); 1112} /* tapkqdetach */ 1113 1114