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 | D_NEEDMINOR, 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, 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, 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 uint32_t macaddr_mid; 408 int unit, s; 409 char *name = NULL; 410 u_char eaddr[6]; 411 412 dev->si_flags &= ~SI_CHEAPCLONE; 413 414 /* allocate driver storage and create device */ 415 tp = malloc(sizeof(*tp), M_TAP, M_WAITOK | M_ZERO); 416 mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF); 417 mtx_lock(&tapmtx); 418 SLIST_INSERT_HEAD(&taphead, tp, tap_next); 419 mtx_unlock(&tapmtx); 420 421 unit = dev2unit(dev); 422 423 /* select device: tap or vmnet */ 424 if (unit & VMNET_DEV_MASK) { 425 name = VMNET; 426 tp->tap_flags |= TAP_VMNET; 427 } else 428 name = TAP; 429 430 unit &= TAPMAXUNIT; 431 432 TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, dev2unit(dev)); 433 434 /* generate fake MAC address: 00 bd xx xx xx unit_no */ 435 macaddr_hi = htons(0x00bd); 436 macaddr_mid = (uint32_t) ticks; 437 bcopy(&macaddr_hi, eaddr, sizeof(short)); 438 bcopy(&macaddr_mid, &eaddr[2], sizeof(uint32_t)); 439 eaddr[5] = (u_char)unit; 440 441 /* fill the rest and attach interface */ 442 ifp = tp->tap_ifp = if_alloc(IFT_ETHER); 443 if (ifp == NULL) 444 panic("%s%d: can not if_alloc()", name, unit); 445 ifp->if_softc = tp; 446 if_initname(ifp, name, unit); 447 ifp->if_init = tapifinit; 448 ifp->if_start = tapifstart; 449 ifp->if_ioctl = tapifioctl; 450 ifp->if_mtu = ETHERMTU; 451 ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST); 452 ifp->if_snd.ifq_maxlen = ifqmaxlen; 453 454 dev->si_drv1 = tp; 455 tp->tap_dev = dev; 456 457 s = splimp(); 458 ether_ifattach(ifp, eaddr); 459 splx(s); 460 461 mtx_lock(&tp->tap_mtx); 462 tp->tap_flags |= TAP_INITED; 463 mtx_unlock(&tp->tap_mtx); 464 465 knlist_init_mtx(&tp->tap_rsel.si_note, NULL); 466 467 TAPDEBUG("interface %s is created. minor = %#x\n", 468 ifp->if_xname, dev2unit(dev)); 469} /* tapcreate */ 470 471 472/* 473 * tapopen 474 * 475 * to open tunnel. must be superuser 476 */ 477static int 478tapopen(struct cdev *dev, int flag, int mode, struct thread *td) 479{ 480 struct tap_softc *tp = NULL; 481 struct ifnet *ifp = NULL; 482 int error, s; 483 484 if (tapuopen == 0) { 485 error = priv_check(td, PRIV_NET_TAP); 486 if (error) 487 return (error); 488 } 489 490 if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT) 491 return (ENXIO); 492 493 tp = dev->si_drv1; 494 495 mtx_lock(&tp->tap_mtx); 496 if (tp->tap_flags & TAP_OPEN) { 497 mtx_unlock(&tp->tap_mtx); 498 return (EBUSY); 499 } 500 501 bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr)); 502 tp->tap_pid = td->td_proc->p_pid; 503 tp->tap_flags |= TAP_OPEN; 504 ifp = tp->tap_ifp; 505 mtx_unlock(&tp->tap_mtx); 506 507 s = splimp(); 508 ifp->if_drv_flags |= IFF_DRV_RUNNING; 509 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 510 if (tapuponopen) 511 ifp->if_flags |= IFF_UP; 512 splx(s); 513 514 TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, dev2unit(dev)); 515 516 return (0); 517} /* tapopen */ 518 519 520/* 521 * tapclose 522 * 523 * close the device - mark i/f down & delete routing info 524 */ 525static int 526tapclose(struct cdev *dev, int foo, int bar, struct thread *td) 527{ 528 struct ifaddr *ifa; 529 struct tap_softc *tp = dev->si_drv1; 530 struct ifnet *ifp = tp->tap_ifp; 531 int s; 532 533 /* junk all pending output */ 534 IF_DRAIN(&ifp->if_snd); 535 536 /* 537 * do not bring the interface down, and do not anything with 538 * interface, if we are in VMnet mode. just close the device. 539 */ 540 541 mtx_lock(&tp->tap_mtx); 542 if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) { 543 mtx_unlock(&tp->tap_mtx); 544 s = splimp(); 545 if_down(ifp); 546 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 547 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 548 rtinit(ifa, (int)RTM_DELETE, 0); 549 } 550 if_purgeaddrs(ifp); 551 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 552 } 553 splx(s); 554 } else 555 mtx_unlock(&tp->tap_mtx); 556 557 funsetown(&tp->tap_sigio); 558 selwakeuppri(&tp->tap_rsel, PZERO+1); 559 KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0); 560 561 mtx_lock(&tp->tap_mtx); 562 tp->tap_flags &= ~TAP_OPEN; 563 tp->tap_pid = 0; 564 mtx_unlock(&tp->tap_mtx); 565 566 TAPDEBUG("%s is closed. minor = %#x\n", 567 ifp->if_xname, dev2unit(dev)); 568 569 return (0); 570} /* tapclose */ 571 572 573/* 574 * tapifinit 575 * 576 * network interface initialization function 577 */ 578static void 579tapifinit(void *xtp) 580{ 581 struct tap_softc *tp = (struct tap_softc *)xtp; 582 struct ifnet *ifp = tp->tap_ifp; 583 584 TAPDEBUG("initializing %s\n", ifp->if_xname); 585 586 ifp->if_drv_flags |= IFF_DRV_RUNNING; 587 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 588 589 /* attempt to start output */ 590 tapifstart(ifp); 591} /* tapifinit */ 592 593 594/* 595 * tapifioctl 596 * 597 * Process an ioctl request on network interface 598 */ 599static int 600tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 601{ 602 struct tap_softc *tp = ifp->if_softc; 603 struct ifreq *ifr = (struct ifreq *)data; 604 struct ifstat *ifs = NULL; 605 int s, dummy; 606 607 switch (cmd) { 608 case SIOCSIFFLAGS: /* XXX -- just like vmnet does */ 609 case SIOCADDMULTI: 610 case SIOCDELMULTI: 611 break; 612 613 case SIOCSIFMTU: 614 ifp->if_mtu = ifr->ifr_mtu; 615 break; 616 617 case SIOCGIFSTATUS: 618 s = splimp(); 619 ifs = (struct ifstat *)data; 620 dummy = strlen(ifs->ascii); 621 mtx_lock(&tp->tap_mtx); 622 if (tp->tap_pid != 0 && dummy < sizeof(ifs->ascii)) 623 snprintf(ifs->ascii + dummy, 624 sizeof(ifs->ascii) - dummy, 625 "\tOpened by PID %d\n", tp->tap_pid); 626 mtx_unlock(&tp->tap_mtx); 627 splx(s); 628 break; 629 630 default: 631 s = splimp(); 632 dummy = ether_ioctl(ifp, cmd, data); 633 splx(s); 634 return (dummy); 635 /* NOT REACHED */ 636 } 637 638 return (0); 639} /* tapifioctl */ 640 641 642/* 643 * tapifstart 644 * 645 * queue packets from higher level ready to put out 646 */ 647static void 648tapifstart(struct ifnet *ifp) 649{ 650 struct tap_softc *tp = ifp->if_softc; 651 int s; 652 653 TAPDEBUG("%s starting\n", ifp->if_xname); 654 655 /* 656 * do not junk pending output if we are in VMnet mode. 657 * XXX: can this do any harm because of queue overflow? 658 */ 659 660 mtx_lock(&tp->tap_mtx); 661 if (((tp->tap_flags & TAP_VMNET) == 0) && 662 ((tp->tap_flags & TAP_READY) != TAP_READY)) { 663 struct mbuf *m = NULL; 664 665 mtx_unlock(&tp->tap_mtx); 666 667 /* Unlocked read. */ 668 TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname, 669 tp->tap_flags); 670 671 s = splimp(); 672 do { 673 IF_DEQUEUE(&ifp->if_snd, m); 674 if (m != NULL) 675 m_freem(m); 676 ifp->if_oerrors ++; 677 } while (m != NULL); 678 splx(s); 679 680 return; 681 } 682 mtx_unlock(&tp->tap_mtx); 683 684 s = splimp(); 685 ifp->if_drv_flags |= IFF_DRV_OACTIVE; 686 687 if (ifp->if_snd.ifq_len != 0) { 688 mtx_lock(&tp->tap_mtx); 689 if (tp->tap_flags & TAP_RWAIT) { 690 tp->tap_flags &= ~TAP_RWAIT; 691 wakeup(tp); 692 } 693 694 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) { 695 mtx_unlock(&tp->tap_mtx); 696 pgsigio(&tp->tap_sigio, SIGIO, 0); 697 } else 698 mtx_unlock(&tp->tap_mtx); 699 700 selwakeuppri(&tp->tap_rsel, PZERO+1); 701 KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0); 702 ifp->if_opackets ++; /* obytes are counted in ether_output */ 703 } 704 705 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 706 splx(s); 707} /* tapifstart */ 708 709 710/* 711 * tapioctl 712 * 713 * the cdevsw interface is now pretty minimal 714 */ 715static int 716tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 717{ 718 struct tap_softc *tp = dev->si_drv1; 719 struct ifnet *ifp = tp->tap_ifp; 720 struct tapinfo *tapp = NULL; 721 int s; 722 int f; 723#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \ 724 defined(COMPAT_FREEBSD4) 725 int ival; 726#endif 727 728 switch (cmd) { 729 case TAPSIFINFO: 730 s = splimp(); 731 tapp = (struct tapinfo *)data; 732 ifp->if_mtu = tapp->mtu; 733 ifp->if_type = tapp->type; 734 ifp->if_baudrate = tapp->baudrate; 735 splx(s); 736 break; 737 738 case TAPGIFINFO: 739 tapp = (struct tapinfo *)data; 740 tapp->mtu = ifp->if_mtu; 741 tapp->type = ifp->if_type; 742 tapp->baudrate = ifp->if_baudrate; 743 break; 744 745 case TAPSDEBUG: 746 tapdebug = *(int *)data; 747 break; 748 749 case TAPGDEBUG: 750 *(int *)data = tapdebug; 751 break; 752 753 case TAPGIFNAME: { 754 struct ifreq *ifr = (struct ifreq *) data; 755 756 strlcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ); 757 } break; 758 759 case FIONBIO: 760 break; 761 762 case FIOASYNC: 763 s = splimp(); 764 mtx_lock(&tp->tap_mtx); 765 if (*(int *)data) 766 tp->tap_flags |= TAP_ASYNC; 767 else 768 tp->tap_flags &= ~TAP_ASYNC; 769 mtx_unlock(&tp->tap_mtx); 770 splx(s); 771 break; 772 773 case FIONREAD: 774 s = splimp(); 775 if (ifp->if_snd.ifq_head) { 776 struct mbuf *mb = ifp->if_snd.ifq_head; 777 778 for(*(int *)data = 0;mb != NULL;mb = mb->m_next) 779 *(int *)data += mb->m_len; 780 } else 781 *(int *)data = 0; 782 splx(s); 783 break; 784 785 case FIOSETOWN: 786 return (fsetown(*(int *)data, &tp->tap_sigio)); 787 788 case FIOGETOWN: 789 *(int *)data = fgetown(&tp->tap_sigio); 790 return (0); 791 792 /* this is deprecated, FIOSETOWN should be used instead */ 793 case TIOCSPGRP: 794 return (fsetown(-(*(int *)data), &tp->tap_sigio)); 795 796 /* this is deprecated, FIOGETOWN should be used instead */ 797 case TIOCGPGRP: 798 *(int *)data = -fgetown(&tp->tap_sigio); 799 return (0); 800 801 /* VMware/VMnet port ioctl's */ 802 803 case SIOCGIFFLAGS: /* get ifnet flags */ 804 bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags)); 805 break; 806 807#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \ 808 defined(COMPAT_FREEBSD4) 809 case _IO('V', 0): 810 ival = IOCPARM_IVAL(data); 811 data = (caddr_t)&ival; 812 /* FALLTHROUGH */ 813#endif 814 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */ 815 f = *(int *)data; 816 f &= 0x0fff; 817 f &= ~IFF_CANTCHANGE; 818 f |= IFF_UP; 819 820 s = splimp(); 821 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE); 822 splx(s); 823 break; 824 825 case OSIOCGIFADDR: /* get MAC address of the remote side */ 826 case SIOCGIFADDR: 827 mtx_lock(&tp->tap_mtx); 828 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr)); 829 mtx_unlock(&tp->tap_mtx); 830 break; 831 832 case SIOCSIFADDR: /* set MAC address of the remote side */ 833 mtx_lock(&tp->tap_mtx); 834 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr)); 835 mtx_unlock(&tp->tap_mtx); 836 break; 837 838 default: 839 return (ENOTTY); 840 } 841 return (0); 842} /* tapioctl */ 843 844 845/* 846 * tapread 847 * 848 * the cdevsw read interface - reads a packet at a time, or at 849 * least as much of a packet as can be read 850 */ 851static int 852tapread(struct cdev *dev, struct uio *uio, int flag) 853{ 854 struct tap_softc *tp = dev->si_drv1; 855 struct ifnet *ifp = tp->tap_ifp; 856 struct mbuf *m = NULL; 857 int error = 0, len, s; 858 859 TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, dev2unit(dev)); 860 861 mtx_lock(&tp->tap_mtx); 862 if ((tp->tap_flags & TAP_READY) != TAP_READY) { 863 mtx_unlock(&tp->tap_mtx); 864 865 /* Unlocked read. */ 866 TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n", 867 ifp->if_xname, dev2unit(dev), tp->tap_flags); 868 869 return (EHOSTDOWN); 870 } 871 872 tp->tap_flags &= ~TAP_RWAIT; 873 mtx_unlock(&tp->tap_mtx); 874 875 /* sleep until we get a packet */ 876 do { 877 s = splimp(); 878 IF_DEQUEUE(&ifp->if_snd, m); 879 splx(s); 880 881 if (m == NULL) { 882 if (flag & O_NONBLOCK) 883 return (EWOULDBLOCK); 884 885 mtx_lock(&tp->tap_mtx); 886 tp->tap_flags |= TAP_RWAIT; 887 mtx_unlock(&tp->tap_mtx); 888 error = tsleep(tp,PCATCH|(PZERO+1),"taprd",0); 889 if (error) 890 return (error); 891 } 892 } while (m == NULL); 893 894 /* feed packet to bpf */ 895 BPF_MTAP(ifp, m); 896 897 /* xfer packet to user space */ 898 while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) { 899 len = min(uio->uio_resid, m->m_len); 900 if (len == 0) 901 break; 902 903 error = uiomove(mtod(m, void *), len, uio); 904 m = m_free(m); 905 } 906 907 if (m != NULL) { 908 TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname, 909 dev2unit(dev)); 910 m_freem(m); 911 } 912 913 return (error); 914} /* tapread */ 915 916 917/* 918 * tapwrite 919 * 920 * the cdevsw write interface - an atomic write is a packet - or else! 921 */ 922static int 923tapwrite(struct cdev *dev, struct uio *uio, int flag) 924{ 925 struct ether_header *eh; 926 struct tap_softc *tp = dev->si_drv1; 927 struct ifnet *ifp = tp->tap_ifp; 928 struct mbuf *m; 929 930 TAPDEBUG("%s writting, minor = %#x\n", 931 ifp->if_xname, dev2unit(dev)); 932 933 if (uio->uio_resid == 0) 934 return (0); 935 936 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
| 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 | D_NEEDMINOR, 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, 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, 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 uint32_t macaddr_mid; 408 int unit, s; 409 char *name = NULL; 410 u_char eaddr[6]; 411 412 dev->si_flags &= ~SI_CHEAPCLONE; 413 414 /* allocate driver storage and create device */ 415 tp = malloc(sizeof(*tp), M_TAP, M_WAITOK | M_ZERO); 416 mtx_init(&tp->tap_mtx, "tap_mtx", NULL, MTX_DEF); 417 mtx_lock(&tapmtx); 418 SLIST_INSERT_HEAD(&taphead, tp, tap_next); 419 mtx_unlock(&tapmtx); 420 421 unit = dev2unit(dev); 422 423 /* select device: tap or vmnet */ 424 if (unit & VMNET_DEV_MASK) { 425 name = VMNET; 426 tp->tap_flags |= TAP_VMNET; 427 } else 428 name = TAP; 429 430 unit &= TAPMAXUNIT; 431 432 TAPDEBUG("tapcreate(%s%d). minor = %#x\n", name, unit, dev2unit(dev)); 433 434 /* generate fake MAC address: 00 bd xx xx xx unit_no */ 435 macaddr_hi = htons(0x00bd); 436 macaddr_mid = (uint32_t) ticks; 437 bcopy(&macaddr_hi, eaddr, sizeof(short)); 438 bcopy(&macaddr_mid, &eaddr[2], sizeof(uint32_t)); 439 eaddr[5] = (u_char)unit; 440 441 /* fill the rest and attach interface */ 442 ifp = tp->tap_ifp = if_alloc(IFT_ETHER); 443 if (ifp == NULL) 444 panic("%s%d: can not if_alloc()", name, unit); 445 ifp->if_softc = tp; 446 if_initname(ifp, name, unit); 447 ifp->if_init = tapifinit; 448 ifp->if_start = tapifstart; 449 ifp->if_ioctl = tapifioctl; 450 ifp->if_mtu = ETHERMTU; 451 ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST); 452 ifp->if_snd.ifq_maxlen = ifqmaxlen; 453 454 dev->si_drv1 = tp; 455 tp->tap_dev = dev; 456 457 s = splimp(); 458 ether_ifattach(ifp, eaddr); 459 splx(s); 460 461 mtx_lock(&tp->tap_mtx); 462 tp->tap_flags |= TAP_INITED; 463 mtx_unlock(&tp->tap_mtx); 464 465 knlist_init_mtx(&tp->tap_rsel.si_note, NULL); 466 467 TAPDEBUG("interface %s is created. minor = %#x\n", 468 ifp->if_xname, dev2unit(dev)); 469} /* tapcreate */ 470 471 472/* 473 * tapopen 474 * 475 * to open tunnel. must be superuser 476 */ 477static int 478tapopen(struct cdev *dev, int flag, int mode, struct thread *td) 479{ 480 struct tap_softc *tp = NULL; 481 struct ifnet *ifp = NULL; 482 int error, s; 483 484 if (tapuopen == 0) { 485 error = priv_check(td, PRIV_NET_TAP); 486 if (error) 487 return (error); 488 } 489 490 if ((dev2unit(dev) & CLONE_UNITMASK) > TAPMAXUNIT) 491 return (ENXIO); 492 493 tp = dev->si_drv1; 494 495 mtx_lock(&tp->tap_mtx); 496 if (tp->tap_flags & TAP_OPEN) { 497 mtx_unlock(&tp->tap_mtx); 498 return (EBUSY); 499 } 500 501 bcopy(IF_LLADDR(tp->tap_ifp), tp->ether_addr, sizeof(tp->ether_addr)); 502 tp->tap_pid = td->td_proc->p_pid; 503 tp->tap_flags |= TAP_OPEN; 504 ifp = tp->tap_ifp; 505 mtx_unlock(&tp->tap_mtx); 506 507 s = splimp(); 508 ifp->if_drv_flags |= IFF_DRV_RUNNING; 509 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 510 if (tapuponopen) 511 ifp->if_flags |= IFF_UP; 512 splx(s); 513 514 TAPDEBUG("%s is open. minor = %#x\n", ifp->if_xname, dev2unit(dev)); 515 516 return (0); 517} /* tapopen */ 518 519 520/* 521 * tapclose 522 * 523 * close the device - mark i/f down & delete routing info 524 */ 525static int 526tapclose(struct cdev *dev, int foo, int bar, struct thread *td) 527{ 528 struct ifaddr *ifa; 529 struct tap_softc *tp = dev->si_drv1; 530 struct ifnet *ifp = tp->tap_ifp; 531 int s; 532 533 /* junk all pending output */ 534 IF_DRAIN(&ifp->if_snd); 535 536 /* 537 * do not bring the interface down, and do not anything with 538 * interface, if we are in VMnet mode. just close the device. 539 */ 540 541 mtx_lock(&tp->tap_mtx); 542 if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) { 543 mtx_unlock(&tp->tap_mtx); 544 s = splimp(); 545 if_down(ifp); 546 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 547 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 548 rtinit(ifa, (int)RTM_DELETE, 0); 549 } 550 if_purgeaddrs(ifp); 551 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 552 } 553 splx(s); 554 } else 555 mtx_unlock(&tp->tap_mtx); 556 557 funsetown(&tp->tap_sigio); 558 selwakeuppri(&tp->tap_rsel, PZERO+1); 559 KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0); 560 561 mtx_lock(&tp->tap_mtx); 562 tp->tap_flags &= ~TAP_OPEN; 563 tp->tap_pid = 0; 564 mtx_unlock(&tp->tap_mtx); 565 566 TAPDEBUG("%s is closed. minor = %#x\n", 567 ifp->if_xname, dev2unit(dev)); 568 569 return (0); 570} /* tapclose */ 571 572 573/* 574 * tapifinit 575 * 576 * network interface initialization function 577 */ 578static void 579tapifinit(void *xtp) 580{ 581 struct tap_softc *tp = (struct tap_softc *)xtp; 582 struct ifnet *ifp = tp->tap_ifp; 583 584 TAPDEBUG("initializing %s\n", ifp->if_xname); 585 586 ifp->if_drv_flags |= IFF_DRV_RUNNING; 587 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 588 589 /* attempt to start output */ 590 tapifstart(ifp); 591} /* tapifinit */ 592 593 594/* 595 * tapifioctl 596 * 597 * Process an ioctl request on network interface 598 */ 599static int 600tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 601{ 602 struct tap_softc *tp = ifp->if_softc; 603 struct ifreq *ifr = (struct ifreq *)data; 604 struct ifstat *ifs = NULL; 605 int s, dummy; 606 607 switch (cmd) { 608 case SIOCSIFFLAGS: /* XXX -- just like vmnet does */ 609 case SIOCADDMULTI: 610 case SIOCDELMULTI: 611 break; 612 613 case SIOCSIFMTU: 614 ifp->if_mtu = ifr->ifr_mtu; 615 break; 616 617 case SIOCGIFSTATUS: 618 s = splimp(); 619 ifs = (struct ifstat *)data; 620 dummy = strlen(ifs->ascii); 621 mtx_lock(&tp->tap_mtx); 622 if (tp->tap_pid != 0 && dummy < sizeof(ifs->ascii)) 623 snprintf(ifs->ascii + dummy, 624 sizeof(ifs->ascii) - dummy, 625 "\tOpened by PID %d\n", tp->tap_pid); 626 mtx_unlock(&tp->tap_mtx); 627 splx(s); 628 break; 629 630 default: 631 s = splimp(); 632 dummy = ether_ioctl(ifp, cmd, data); 633 splx(s); 634 return (dummy); 635 /* NOT REACHED */ 636 } 637 638 return (0); 639} /* tapifioctl */ 640 641 642/* 643 * tapifstart 644 * 645 * queue packets from higher level ready to put out 646 */ 647static void 648tapifstart(struct ifnet *ifp) 649{ 650 struct tap_softc *tp = ifp->if_softc; 651 int s; 652 653 TAPDEBUG("%s starting\n", ifp->if_xname); 654 655 /* 656 * do not junk pending output if we are in VMnet mode. 657 * XXX: can this do any harm because of queue overflow? 658 */ 659 660 mtx_lock(&tp->tap_mtx); 661 if (((tp->tap_flags & TAP_VMNET) == 0) && 662 ((tp->tap_flags & TAP_READY) != TAP_READY)) { 663 struct mbuf *m = NULL; 664 665 mtx_unlock(&tp->tap_mtx); 666 667 /* Unlocked read. */ 668 TAPDEBUG("%s not ready, tap_flags = 0x%x\n", ifp->if_xname, 669 tp->tap_flags); 670 671 s = splimp(); 672 do { 673 IF_DEQUEUE(&ifp->if_snd, m); 674 if (m != NULL) 675 m_freem(m); 676 ifp->if_oerrors ++; 677 } while (m != NULL); 678 splx(s); 679 680 return; 681 } 682 mtx_unlock(&tp->tap_mtx); 683 684 s = splimp(); 685 ifp->if_drv_flags |= IFF_DRV_OACTIVE; 686 687 if (ifp->if_snd.ifq_len != 0) { 688 mtx_lock(&tp->tap_mtx); 689 if (tp->tap_flags & TAP_RWAIT) { 690 tp->tap_flags &= ~TAP_RWAIT; 691 wakeup(tp); 692 } 693 694 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) { 695 mtx_unlock(&tp->tap_mtx); 696 pgsigio(&tp->tap_sigio, SIGIO, 0); 697 } else 698 mtx_unlock(&tp->tap_mtx); 699 700 selwakeuppri(&tp->tap_rsel, PZERO+1); 701 KNOTE_UNLOCKED(&tp->tap_rsel.si_note, 0); 702 ifp->if_opackets ++; /* obytes are counted in ether_output */ 703 } 704 705 ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; 706 splx(s); 707} /* tapifstart */ 708 709 710/* 711 * tapioctl 712 * 713 * the cdevsw interface is now pretty minimal 714 */ 715static int 716tapioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 717{ 718 struct tap_softc *tp = dev->si_drv1; 719 struct ifnet *ifp = tp->tap_ifp; 720 struct tapinfo *tapp = NULL; 721 int s; 722 int f; 723#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \ 724 defined(COMPAT_FREEBSD4) 725 int ival; 726#endif 727 728 switch (cmd) { 729 case TAPSIFINFO: 730 s = splimp(); 731 tapp = (struct tapinfo *)data; 732 ifp->if_mtu = tapp->mtu; 733 ifp->if_type = tapp->type; 734 ifp->if_baudrate = tapp->baudrate; 735 splx(s); 736 break; 737 738 case TAPGIFINFO: 739 tapp = (struct tapinfo *)data; 740 tapp->mtu = ifp->if_mtu; 741 tapp->type = ifp->if_type; 742 tapp->baudrate = ifp->if_baudrate; 743 break; 744 745 case TAPSDEBUG: 746 tapdebug = *(int *)data; 747 break; 748 749 case TAPGDEBUG: 750 *(int *)data = tapdebug; 751 break; 752 753 case TAPGIFNAME: { 754 struct ifreq *ifr = (struct ifreq *) data; 755 756 strlcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ); 757 } break; 758 759 case FIONBIO: 760 break; 761 762 case FIOASYNC: 763 s = splimp(); 764 mtx_lock(&tp->tap_mtx); 765 if (*(int *)data) 766 tp->tap_flags |= TAP_ASYNC; 767 else 768 tp->tap_flags &= ~TAP_ASYNC; 769 mtx_unlock(&tp->tap_mtx); 770 splx(s); 771 break; 772 773 case FIONREAD: 774 s = splimp(); 775 if (ifp->if_snd.ifq_head) { 776 struct mbuf *mb = ifp->if_snd.ifq_head; 777 778 for(*(int *)data = 0;mb != NULL;mb = mb->m_next) 779 *(int *)data += mb->m_len; 780 } else 781 *(int *)data = 0; 782 splx(s); 783 break; 784 785 case FIOSETOWN: 786 return (fsetown(*(int *)data, &tp->tap_sigio)); 787 788 case FIOGETOWN: 789 *(int *)data = fgetown(&tp->tap_sigio); 790 return (0); 791 792 /* this is deprecated, FIOSETOWN should be used instead */ 793 case TIOCSPGRP: 794 return (fsetown(-(*(int *)data), &tp->tap_sigio)); 795 796 /* this is deprecated, FIOGETOWN should be used instead */ 797 case TIOCGPGRP: 798 *(int *)data = -fgetown(&tp->tap_sigio); 799 return (0); 800 801 /* VMware/VMnet port ioctl's */ 802 803 case SIOCGIFFLAGS: /* get ifnet flags */ 804 bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags)); 805 break; 806 807#if defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD5) || \ 808 defined(COMPAT_FREEBSD4) 809 case _IO('V', 0): 810 ival = IOCPARM_IVAL(data); 811 data = (caddr_t)&ival; 812 /* FALLTHROUGH */ 813#endif 814 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */ 815 f = *(int *)data; 816 f &= 0x0fff; 817 f &= ~IFF_CANTCHANGE; 818 f |= IFF_UP; 819 820 s = splimp(); 821 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE); 822 splx(s); 823 break; 824 825 case OSIOCGIFADDR: /* get MAC address of the remote side */ 826 case SIOCGIFADDR: 827 mtx_lock(&tp->tap_mtx); 828 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr)); 829 mtx_unlock(&tp->tap_mtx); 830 break; 831 832 case SIOCSIFADDR: /* set MAC address of the remote side */ 833 mtx_lock(&tp->tap_mtx); 834 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr)); 835 mtx_unlock(&tp->tap_mtx); 836 break; 837 838 default: 839 return (ENOTTY); 840 } 841 return (0); 842} /* tapioctl */ 843 844 845/* 846 * tapread 847 * 848 * the cdevsw read interface - reads a packet at a time, or at 849 * least as much of a packet as can be read 850 */ 851static int 852tapread(struct cdev *dev, struct uio *uio, int flag) 853{ 854 struct tap_softc *tp = dev->si_drv1; 855 struct ifnet *ifp = tp->tap_ifp; 856 struct mbuf *m = NULL; 857 int error = 0, len, s; 858 859 TAPDEBUG("%s reading, minor = %#x\n", ifp->if_xname, dev2unit(dev)); 860 861 mtx_lock(&tp->tap_mtx); 862 if ((tp->tap_flags & TAP_READY) != TAP_READY) { 863 mtx_unlock(&tp->tap_mtx); 864 865 /* Unlocked read. */ 866 TAPDEBUG("%s not ready. minor = %#x, tap_flags = 0x%x\n", 867 ifp->if_xname, dev2unit(dev), tp->tap_flags); 868 869 return (EHOSTDOWN); 870 } 871 872 tp->tap_flags &= ~TAP_RWAIT; 873 mtx_unlock(&tp->tap_mtx); 874 875 /* sleep until we get a packet */ 876 do { 877 s = splimp(); 878 IF_DEQUEUE(&ifp->if_snd, m); 879 splx(s); 880 881 if (m == NULL) { 882 if (flag & O_NONBLOCK) 883 return (EWOULDBLOCK); 884 885 mtx_lock(&tp->tap_mtx); 886 tp->tap_flags |= TAP_RWAIT; 887 mtx_unlock(&tp->tap_mtx); 888 error = tsleep(tp,PCATCH|(PZERO+1),"taprd",0); 889 if (error) 890 return (error); 891 } 892 } while (m == NULL); 893 894 /* feed packet to bpf */ 895 BPF_MTAP(ifp, m); 896 897 /* xfer packet to user space */ 898 while ((m != NULL) && (uio->uio_resid > 0) && (error == 0)) { 899 len = min(uio->uio_resid, m->m_len); 900 if (len == 0) 901 break; 902 903 error = uiomove(mtod(m, void *), len, uio); 904 m = m_free(m); 905 } 906 907 if (m != NULL) { 908 TAPDEBUG("%s dropping mbuf, minor = %#x\n", ifp->if_xname, 909 dev2unit(dev)); 910 m_freem(m); 911 } 912 913 return (error); 914} /* tapread */ 915 916 917/* 918 * tapwrite 919 * 920 * the cdevsw write interface - an atomic write is a packet - or else! 921 */ 922static int 923tapwrite(struct cdev *dev, struct uio *uio, int flag) 924{ 925 struct ether_header *eh; 926 struct tap_softc *tp = dev->si_drv1; 927 struct ifnet *ifp = tp->tap_ifp; 928 struct mbuf *m; 929 930 TAPDEBUG("%s writting, minor = %#x\n", 931 ifp->if_xname, dev2unit(dev)); 932 933 if (uio->uio_resid == 0) 934 return (0); 935 936 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
|
938 ifp->if_xname, uio->uio_resid, dev2unit(dev)); 939 940 return (EIO); 941 } 942 943 if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, ETHER_ALIGN, 944 M_PKTHDR)) == NULL) { 945 ifp->if_ierrors ++; 946 return (ENOBUFS); 947 } 948 949 m->m_pkthdr.rcvif = ifp; 950 951 /* 952 * Only pass a unicast frame to ether_input(), if it would actually 953 * have been received by non-virtual hardware. 954 */ 955 if (m->m_len < sizeof(struct ether_header)) { 956 m_freem(m); 957 return (0); 958 } 959 eh = mtod(m, struct ether_header *); 960 961 if (eh && (ifp->if_flags & IFF_PROMISC) == 0 && 962 !ETHER_IS_MULTICAST(eh->ether_dhost) && 963 bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) { 964 m_freem(m); 965 return (0); 966 } 967 968 /* Pass packet up to parent. */ 969 (*ifp->if_input)(ifp, m); 970 ifp->if_ipackets ++; /* ibytes are counted in parent */ 971 972 return (0); 973} /* tapwrite */ 974 975 976/* 977 * tappoll 978 * 979 * the poll interface, this is only useful on reads 980 * really. the write detect always returns true, write never blocks 981 * anyway, it either accepts the packet or drops it 982 */ 983static int 984tappoll(struct cdev *dev, int events, struct thread *td) 985{ 986 struct tap_softc *tp = dev->si_drv1; 987 struct ifnet *ifp = tp->tap_ifp; 988 int s, revents = 0; 989 990 TAPDEBUG("%s polling, minor = %#x\n", 991 ifp->if_xname, dev2unit(dev)); 992 993 s = splimp(); 994 if (events & (POLLIN | POLLRDNORM)) { 995 if (ifp->if_snd.ifq_len > 0) { 996 TAPDEBUG("%s have data in queue. len = %d, " \ 997 "minor = %#x\n", ifp->if_xname, 998 ifp->if_snd.ifq_len, dev2unit(dev)); 999 1000 revents |= (events & (POLLIN | POLLRDNORM)); 1001 } else { 1002 TAPDEBUG("%s waiting for data, minor = %#x\n", 1003 ifp->if_xname, dev2unit(dev)); 1004 1005 selrecord(td, &tp->tap_rsel); 1006 } 1007 } 1008 1009 if (events & (POLLOUT | POLLWRNORM)) 1010 revents |= (events & (POLLOUT | POLLWRNORM)); 1011 1012 splx(s); 1013 return (revents); 1014} /* tappoll */ 1015 1016 1017/* 1018 * tap_kqfilter 1019 * 1020 * support for kevent() system call 1021 */ 1022static int 1023tapkqfilter(struct cdev *dev, struct knote *kn) 1024{ 1025 int s; 1026 struct tap_softc *tp = dev->si_drv1; 1027 struct ifnet *ifp = tp->tap_ifp; 1028 1029 s = splimp(); 1030 switch (kn->kn_filter) { 1031 case EVFILT_READ: 1032 TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n", 1033 ifp->if_xname, dev2unit(dev)); 1034 kn->kn_fop = &tap_read_filterops; 1035 break; 1036 1037 case EVFILT_WRITE: 1038 TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n", 1039 ifp->if_xname, dev2unit(dev)); 1040 kn->kn_fop = &tap_write_filterops; 1041 break; 1042 1043 default: 1044 TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n", 1045 ifp->if_xname, dev2unit(dev)); 1046 splx(s); 1047 return (EINVAL); 1048 /* NOT REACHED */ 1049 } 1050 splx(s); 1051 1052 kn->kn_hook = (caddr_t) dev; 1053 knlist_add(&tp->tap_rsel.si_note, kn, 0); 1054 1055 return (0); 1056} /* tapkqfilter */ 1057 1058 1059/* 1060 * tap_kqread 1061 * 1062 * Return true if there is data in the interface queue 1063 */ 1064static int 1065tapkqread(struct knote *kn, long hint) 1066{ 1067 int ret, s; 1068 struct cdev *dev = (struct cdev *)(kn->kn_hook); 1069 struct tap_softc *tp = dev->si_drv1; 1070 struct ifnet *ifp = tp->tap_ifp; 1071 1072 s = splimp(); 1073 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1074 TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n", 1075 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev)); 1076 ret = 1; 1077 } else { 1078 TAPDEBUG("%s waiting for data, minor = %#x\n", 1079 ifp->if_xname, dev2unit(dev)); 1080 ret = 0; 1081 } 1082 splx(s); 1083 1084 return (ret); 1085} /* tapkqread */ 1086 1087 1088/* 1089 * tap_kqwrite 1090 * 1091 * Always can write. Return the MTU in kn->data 1092 */ 1093static int 1094tapkqwrite(struct knote *kn, long hint) 1095{ 1096 int s; 1097 struct tap_softc *tp = ((struct cdev *) kn->kn_hook)->si_drv1; 1098 struct ifnet *ifp = tp->tap_ifp; 1099 1100 s = splimp(); 1101 kn->kn_data = ifp->if_mtu; 1102 splx(s); 1103 1104 return (1); 1105} /* tapkqwrite */ 1106 1107 1108static void 1109tapkqdetach(struct knote *kn) 1110{ 1111 struct tap_softc *tp = ((struct cdev *) kn->kn_hook)->si_drv1; 1112 1113 knlist_remove(&tp->tap_rsel.si_note, kn, 0); 1114} /* tapkqdetach */ 1115
| 938 ifp->if_xname, uio->uio_resid, dev2unit(dev)); 939 940 return (EIO); 941 } 942 943 if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, ETHER_ALIGN, 944 M_PKTHDR)) == NULL) { 945 ifp->if_ierrors ++; 946 return (ENOBUFS); 947 } 948 949 m->m_pkthdr.rcvif = ifp; 950 951 /* 952 * Only pass a unicast frame to ether_input(), if it would actually 953 * have been received by non-virtual hardware. 954 */ 955 if (m->m_len < sizeof(struct ether_header)) { 956 m_freem(m); 957 return (0); 958 } 959 eh = mtod(m, struct ether_header *); 960 961 if (eh && (ifp->if_flags & IFF_PROMISC) == 0 && 962 !ETHER_IS_MULTICAST(eh->ether_dhost) && 963 bcmp(eh->ether_dhost, IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0) { 964 m_freem(m); 965 return (0); 966 } 967 968 /* Pass packet up to parent. */ 969 (*ifp->if_input)(ifp, m); 970 ifp->if_ipackets ++; /* ibytes are counted in parent */ 971 972 return (0); 973} /* tapwrite */ 974 975 976/* 977 * tappoll 978 * 979 * the poll interface, this is only useful on reads 980 * really. the write detect always returns true, write never blocks 981 * anyway, it either accepts the packet or drops it 982 */ 983static int 984tappoll(struct cdev *dev, int events, struct thread *td) 985{ 986 struct tap_softc *tp = dev->si_drv1; 987 struct ifnet *ifp = tp->tap_ifp; 988 int s, revents = 0; 989 990 TAPDEBUG("%s polling, minor = %#x\n", 991 ifp->if_xname, dev2unit(dev)); 992 993 s = splimp(); 994 if (events & (POLLIN | POLLRDNORM)) { 995 if (ifp->if_snd.ifq_len > 0) { 996 TAPDEBUG("%s have data in queue. len = %d, " \ 997 "minor = %#x\n", ifp->if_xname, 998 ifp->if_snd.ifq_len, dev2unit(dev)); 999 1000 revents |= (events & (POLLIN | POLLRDNORM)); 1001 } else { 1002 TAPDEBUG("%s waiting for data, minor = %#x\n", 1003 ifp->if_xname, dev2unit(dev)); 1004 1005 selrecord(td, &tp->tap_rsel); 1006 } 1007 } 1008 1009 if (events & (POLLOUT | POLLWRNORM)) 1010 revents |= (events & (POLLOUT | POLLWRNORM)); 1011 1012 splx(s); 1013 return (revents); 1014} /* tappoll */ 1015 1016 1017/* 1018 * tap_kqfilter 1019 * 1020 * support for kevent() system call 1021 */ 1022static int 1023tapkqfilter(struct cdev *dev, struct knote *kn) 1024{ 1025 int s; 1026 struct tap_softc *tp = dev->si_drv1; 1027 struct ifnet *ifp = tp->tap_ifp; 1028 1029 s = splimp(); 1030 switch (kn->kn_filter) { 1031 case EVFILT_READ: 1032 TAPDEBUG("%s kqfilter: EVFILT_READ, minor = %#x\n", 1033 ifp->if_xname, dev2unit(dev)); 1034 kn->kn_fop = &tap_read_filterops; 1035 break; 1036 1037 case EVFILT_WRITE: 1038 TAPDEBUG("%s kqfilter: EVFILT_WRITE, minor = %#x\n", 1039 ifp->if_xname, dev2unit(dev)); 1040 kn->kn_fop = &tap_write_filterops; 1041 break; 1042 1043 default: 1044 TAPDEBUG("%s kqfilter: invalid filter, minor = %#x\n", 1045 ifp->if_xname, dev2unit(dev)); 1046 splx(s); 1047 return (EINVAL); 1048 /* NOT REACHED */ 1049 } 1050 splx(s); 1051 1052 kn->kn_hook = (caddr_t) dev; 1053 knlist_add(&tp->tap_rsel.si_note, kn, 0); 1054 1055 return (0); 1056} /* tapkqfilter */ 1057 1058 1059/* 1060 * tap_kqread 1061 * 1062 * Return true if there is data in the interface queue 1063 */ 1064static int 1065tapkqread(struct knote *kn, long hint) 1066{ 1067 int ret, s; 1068 struct cdev *dev = (struct cdev *)(kn->kn_hook); 1069 struct tap_softc *tp = dev->si_drv1; 1070 struct ifnet *ifp = tp->tap_ifp; 1071 1072 s = splimp(); 1073 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1074 TAPDEBUG("%s have data in queue. len = %d, minor = %#x\n", 1075 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev)); 1076 ret = 1; 1077 } else { 1078 TAPDEBUG("%s waiting for data, minor = %#x\n", 1079 ifp->if_xname, dev2unit(dev)); 1080 ret = 0; 1081 } 1082 splx(s); 1083 1084 return (ret); 1085} /* tapkqread */ 1086 1087 1088/* 1089 * tap_kqwrite 1090 * 1091 * Always can write. Return the MTU in kn->data 1092 */ 1093static int 1094tapkqwrite(struct knote *kn, long hint) 1095{ 1096 int s; 1097 struct tap_softc *tp = ((struct cdev *) kn->kn_hook)->si_drv1; 1098 struct ifnet *ifp = tp->tap_ifp; 1099 1100 s = splimp(); 1101 kn->kn_data = ifp->if_mtu; 1102 splx(s); 1103 1104 return (1); 1105} /* tapkqwrite */ 1106 1107 1108static void 1109tapkqdetach(struct knote *kn) 1110{ 1111 struct tap_softc *tp = ((struct cdev *) kn->kn_hook)->si_drv1; 1112 1113 knlist_remove(&tp->tap_rsel.si_note, kn, 0); 1114} /* tapkqdetach */ 1115
|