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: src/sys/net/if_tap.c,v 1.3.2.3 2002/04/14 21:41:48 luigi Exp $ 35 * $DragonFly: src/sys/net/tap/if_tap.c,v 1.36 2007/07/03 17:40:51 dillon Exp $ 36 * $Id: if_tap.c,v 0.21 2000/07/23 21:46:02 max Exp $ 37 */ 38 39 #include "opt_inet.h" 40 41 #include <sys/param.h> 42 #include <sys/conf.h> 43 #include <sys/device.h> 44 #include <sys/filedesc.h> 45 #include <sys/filio.h> 46 #include <sys/kernel.h> 47 #include <sys/malloc.h> 48 #include <sys/mbuf.h> 49 #include <sys/poll.h> 50 #include <sys/proc.h> 51 #include <sys/signalvar.h> 52 #include <sys/socket.h> 53 #include <sys/sockio.h> 54 #include <sys/sysctl.h> 55 #include <sys/systm.h> 56 #include <sys/thread2.h> 57 #include <sys/ttycom.h> 58 #include <sys/uio.h> 59 #include <sys/vnode.h> 60 #include <sys/serialize.h> 61 62 #include <net/bpf.h> 63 #include <net/ethernet.h> 64 #include <net/if.h> 65 #include <net/ifq_var.h> 66 #include <net/if_arp.h> 67 #include <net/route.h> 68 69 #include <netinet/in.h> 70 71 #include "if_tapvar.h" 72 #include "if_tap.h" 73 74 75 #define CDEV_NAME "tap" 76 #define CDEV_MAJOR 149 77 #define TAPDEBUG if (tapdebug) if_printf 78 79 #define TAP "tap" 80 #define VMNET "vmnet" 81 #define VMNET_DEV_MASK 0x00010000 82 83 /* module */ 84 static int tapmodevent (module_t, int, void *); 85 86 /* device */ 87 static void tapcreate (cdev_t); 88 89 /* network interface */ 90 static void tapifstart (struct ifnet *); 91 static int tapifioctl (struct ifnet *, u_long, caddr_t, 92 struct ucred *); 93 static void tapifinit (void *); 94 95 /* character device */ 96 static d_open_t tapopen; 97 static d_close_t tapclose; 98 static d_read_t tapread; 99 static d_write_t tapwrite; 100 static d_ioctl_t tapioctl; 101 static d_poll_t tappoll; 102 static d_kqfilter_t tapkqfilter; 103 104 static struct dev_ops tap_ops = { 105 { CDEV_NAME, CDEV_MAJOR, 0 }, 106 .d_open = tapopen, 107 .d_close = tapclose, 108 .d_read = tapread, 109 .d_write = tapwrite, 110 .d_ioctl = tapioctl, 111 .d_poll = tappoll, 112 .d_kqfilter = tapkqfilter 113 }; 114 115 static int taprefcnt = 0; /* module ref. counter */ 116 static int taplastunit = -1; /* max. open unit number */ 117 static int tapdebug = 0; /* debug flag */ 118 119 MALLOC_DECLARE(M_TAP); 120 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface"); 121 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, ""); 122 DEV_MODULE(if_tap, tapmodevent, NULL); 123 124 /* 125 * tapmodevent 126 * 127 * module event handler 128 */ 129 static int 130 tapmodevent(module_t mod, int type, void *data) 131 { 132 static int attached = 0; 133 struct ifnet *ifp = NULL; 134 int unit; 135 136 switch (type) { 137 case MOD_LOAD: 138 if (attached) 139 return (EEXIST); 140 141 dev_ops_add(&tap_ops, 0, 0); 142 attached = 1; 143 break; 144 145 case MOD_UNLOAD: 146 if (taprefcnt > 0) 147 return (EBUSY); 148 149 dev_ops_remove(&tap_ops, 0, 0); 150 151 /* XXX: maintain tap ifs in a local list */ 152 unit = 0; 153 while (unit <= taplastunit) { 154 TAILQ_FOREACH(ifp, &ifnet, if_link) { 155 if ((strcmp(ifp->if_dname, TAP) == 0) || 156 (strcmp(ifp->if_dname, VMNET) == 0)) { 157 if (ifp->if_dunit == unit) 158 break; 159 } 160 } 161 162 if (ifp != NULL) { 163 struct tap_softc *tp = ifp->if_softc; 164 165 TAPDEBUG(ifp, "detached. minor = %#x, " \ 166 "taplastunit = %d\n", 167 minor(tp->tap_dev), taplastunit); 168 169 ether_ifdetach(ifp); 170 destroy_dev(tp->tap_dev); 171 kfree(tp, M_TAP); 172 } 173 else 174 unit ++; 175 } 176 177 attached = 0; 178 break; 179 180 default: 181 return (EOPNOTSUPP); 182 } 183 184 return (0); 185 } /* tapmodevent */ 186 187 188 /* 189 * tapcreate 190 * 191 * to create interface 192 */ 193 static void 194 tapcreate(cdev_t dev) 195 { 196 struct ifnet *ifp = NULL; 197 struct tap_softc *tp = NULL; 198 uint8_t ether_addr[ETHER_ADDR_LEN]; 199 int unit; 200 char *name = NULL; 201 202 /* allocate driver storage and create device */ 203 MALLOC(tp, struct tap_softc *, sizeof(*tp), M_TAP, M_WAITOK); 204 bzero(tp, sizeof(*tp)); 205 206 /* select device: tap or vmnet */ 207 if (minor(dev) & VMNET_DEV_MASK) { 208 name = VMNET; 209 unit = lminor(dev) & 0xff; 210 tp->tap_flags |= TAP_VMNET; 211 } 212 else { 213 name = TAP; 214 unit = lminor(dev); 215 } 216 217 tp->tap_dev = make_dev(&tap_ops, minor(dev), UID_ROOT, GID_WHEEL, 218 0600, "%s%d", name, unit); 219 tp->tap_dev->si_drv1 = dev->si_drv1 = tp; 220 reference_dev(tp->tap_dev); /* so we can destroy it later */ 221 222 /* generate fake MAC address: 00 bd xx xx xx unit_no */ 223 ether_addr[0] = 0x00; 224 ether_addr[1] = 0xbd; 225 bcopy(&ticks, ðer_addr[2], 3); 226 ether_addr[5] = (u_char)unit; 227 228 /* fill the rest and attach interface */ 229 ifp = &tp->tap_if; 230 ifp->if_softc = tp; 231 232 if_initname(ifp, name, unit); 233 if (unit > taplastunit) 234 taplastunit = unit; 235 236 ifp->if_init = tapifinit; 237 ifp->if_start = tapifstart; 238 ifp->if_ioctl = tapifioctl; 239 ifp->if_mtu = ETHERMTU; 240 ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST); 241 ifq_set_maxlen(&ifp->if_snd, ifqmaxlen); 242 ifq_set_ready(&ifp->if_snd); 243 244 ether_ifattach(ifp, ether_addr, NULL); 245 246 tp->tap_flags |= TAP_INITED; 247 248 TAPDEBUG(ifp, "created. minor = %#x\n", minor(tp->tap_dev)); 249 } /* tapcreate */ 250 251 252 /* 253 * tapopen 254 * 255 * to open tunnel. must be superuser 256 */ 257 static int 258 tapopen(struct dev_open_args *ap) 259 { 260 cdev_t dev = ap->a_head.a_dev; 261 struct tap_softc *tp = NULL; 262 struct ifnet *ifp = NULL; 263 int error; 264 265 if ((error = suser_cred(ap->a_cred, 0)) != 0) 266 return (error); 267 268 get_mplock(); 269 tp = dev->si_drv1; 270 if (tp == NULL) { 271 tapcreate(dev); 272 tp = dev->si_drv1; 273 ifp = &tp->arpcom.ac_if; 274 } else { 275 ifp = &tp->arpcom.ac_if; 276 277 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp); 278 279 /* Announce the return of the interface. */ 280 rt_ifannouncemsg(ifp, IFAN_ARRIVAL); 281 } 282 283 if (tp->tap_flags & TAP_OPEN) { 284 rel_mplock(); 285 return (EBUSY); 286 } 287 288 bcopy(tp->arpcom.ac_enaddr, tp->ether_addr, sizeof(tp->ether_addr)); 289 290 tp->tap_td = curthread; 291 tp->tap_flags |= TAP_OPEN; 292 taprefcnt ++; 293 294 TAPDEBUG(ifp, "opened. minor = %#x, refcnt = %d, taplastunit = %d\n", 295 minor(tp->tap_dev), taprefcnt, taplastunit); 296 297 rel_mplock(); 298 return (0); 299 } 300 301 302 /* 303 * tapclose 304 * 305 * close the device - mark i/f down & delete routing info 306 */ 307 static int 308 tapclose(struct dev_close_args *ap) 309 { 310 cdev_t dev = ap->a_head.a_dev; 311 struct tap_softc *tp = dev->si_drv1; 312 struct ifnet *ifp = &tp->tap_if; 313 314 /* junk all pending output */ 315 316 get_mplock(); 317 lwkt_serialize_enter(ifp->if_serializer); 318 ifq_purge(&ifp->if_snd); 319 lwkt_serialize_exit(ifp->if_serializer); 320 321 /* 322 * do not bring the interface down, and do not anything with 323 * interface, if we are in VMnet mode. just close the device. 324 */ 325 326 if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) { 327 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp); 328 329 /* Announce the departure of the interface. */ 330 rt_ifannouncemsg(ifp, IFAN_DEPARTURE); 331 332 if_down(ifp); 333 lwkt_serialize_enter(ifp->if_serializer); 334 if (ifp->if_flags & IFF_RUNNING) { 335 /* find internet addresses and delete routes */ 336 struct ifaddr *ifa = NULL; 337 338 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 339 if (ifa->ifa_addr->sa_family == AF_INET) { 340 rtinit(ifa, (int)RTM_DELETE, 0); 341 342 /* remove address from interface */ 343 bzero(ifa->ifa_addr, 344 sizeof(*(ifa->ifa_addr))); 345 bzero(ifa->ifa_dstaddr, 346 sizeof(*(ifa->ifa_dstaddr))); 347 bzero(ifa->ifa_netmask, 348 sizeof(*(ifa->ifa_netmask))); 349 } 350 } 351 352 ifp->if_flags &= ~IFF_RUNNING; 353 } 354 lwkt_serialize_exit(ifp->if_serializer); 355 } 356 357 funsetown(tp->tap_sigio); 358 selwakeup(&tp->tap_rsel); 359 360 tp->tap_flags &= ~TAP_OPEN; 361 tp->tap_td = NULL; 362 363 taprefcnt --; 364 if (taprefcnt < 0) { 365 taprefcnt = 0; 366 if_printf(ifp, "minor = %#x, refcnt = %d is out of sync. " 367 "set refcnt to 0\n", minor(tp->tap_dev), taprefcnt); 368 } 369 370 TAPDEBUG(ifp, "closed. minor = %#x, refcnt = %d, taplastunit = %d\n", 371 minor(tp->tap_dev), taprefcnt, taplastunit); 372 373 rel_mplock(); 374 return (0); 375 } 376 377 378 /* 379 * tapifinit 380 * 381 * Network interface initialization function (called with if serializer held) 382 * 383 * MPSAFE 384 */ 385 static void 386 tapifinit(void *xtp) 387 { 388 struct tap_softc *tp = (struct tap_softc *)xtp; 389 struct ifnet *ifp = &tp->tap_if; 390 391 TAPDEBUG(ifp, "initializing, minor = %#x\n", minor(tp->tap_dev)); 392 393 ifp->if_flags |= IFF_RUNNING; 394 ifp->if_flags &= ~IFF_OACTIVE; 395 396 /* attempt to start output */ 397 tapifstart(ifp); 398 } 399 400 401 /* 402 * tapifioctl 403 * 404 * Process an ioctl request on network interface (called with if serializer 405 * held). 406 * 407 * MPSAFE 408 */ 409 int 410 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr) 411 { 412 struct tap_softc *tp = (struct tap_softc *)(ifp->if_softc); 413 struct ifstat *ifs = NULL; 414 int dummy; 415 416 switch (cmd) { 417 case SIOCSIFADDR: 418 case SIOCGIFADDR: 419 case SIOCSIFMTU: 420 dummy = ether_ioctl(ifp, cmd, data); 421 return (dummy); 422 423 case SIOCSIFFLAGS: 424 if ((tp->tap_flags & TAP_VMNET) == 0) { 425 /* 426 * Only for non-vmnet tap(4) 427 */ 428 if (ifp->if_flags & IFF_UP) { 429 if ((ifp->if_flags & IFF_RUNNING) == 0) 430 tapifinit(tp); 431 } 432 } 433 break; 434 case SIOCADDMULTI: /* XXX -- just like vmnet does */ 435 case SIOCDELMULTI: 436 break; 437 438 case SIOCGIFSTATUS: 439 ifs = (struct ifstat *)data; 440 dummy = strlen(ifs->ascii); 441 if (tp->tap_td != NULL && dummy < sizeof(ifs->ascii)) { 442 if (tp->tap_td->td_proc) { 443 ksnprintf(ifs->ascii + dummy, 444 sizeof(ifs->ascii) - dummy, 445 "\tOpened by pid %d\n", 446 (int)tp->tap_td->td_proc->p_pid); 447 } else { 448 ksnprintf(ifs->ascii + dummy, 449 sizeof(ifs->ascii) - dummy, 450 "\tOpened by td %p\n", tp->tap_td); 451 } 452 } 453 break; 454 455 default: 456 return (EINVAL); 457 } 458 459 return (0); 460 } 461 462 463 /* 464 * tapifstart 465 * 466 * Queue packets from higher level ready to put out (called with if serializer 467 * held) 468 * 469 * MPSAFE 470 */ 471 static void 472 tapifstart(struct ifnet *ifp) 473 { 474 struct tap_softc *tp = ifp->if_softc; 475 476 TAPDEBUG(ifp, "starting, minor = %#x\n", minor(tp->tap_dev)); 477 478 /* 479 * do not junk pending output if we are in VMnet mode. 480 * XXX: can this do any harm because of queue overflow? 481 */ 482 483 if (((tp->tap_flags & TAP_VMNET) == 0) && 484 ((tp->tap_flags & TAP_READY) != TAP_READY)) { 485 TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n", 486 minor(tp->tap_dev), tp->tap_flags); 487 488 ifq_purge(&ifp->if_snd); 489 return; 490 } 491 492 ifp->if_flags |= IFF_OACTIVE; 493 494 if (!ifq_is_empty(&ifp->if_snd)) { 495 if (tp->tap_flags & TAP_RWAIT) { 496 tp->tap_flags &= ~TAP_RWAIT; 497 wakeup((caddr_t)tp); 498 } 499 KNOTE(&tp->tap_rsel.si_note, 0); 500 501 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) { 502 get_mplock(); 503 pgsigio(tp->tap_sigio, SIGIO, 0); 504 rel_mplock(); 505 } 506 507 /* 508 * selwakeup is not MPSAFE. tapifstart is. 509 */ 510 get_mplock(); 511 selwakeup(&tp->tap_rsel); 512 rel_mplock(); 513 ifp->if_opackets ++; /* obytes are counted in ether_output */ 514 } 515 516 ifp->if_flags &= ~IFF_OACTIVE; 517 } 518 519 520 /* 521 * tapioctl 522 * 523 * The ops interface is now pretty minimal. Called via fileops with nothing 524 * held. 525 * 526 * MPSAFE 527 */ 528 static int 529 tapioctl(struct dev_ioctl_args *ap) 530 { 531 cdev_t dev = ap->a_head.a_dev; 532 caddr_t data = ap->a_data; 533 struct tap_softc *tp = dev->si_drv1; 534 struct ifnet *ifp = &tp->tap_if; 535 struct tapinfo *tapp = NULL; 536 struct mbuf *mb; 537 short f; 538 int error; 539 540 lwkt_serialize_enter(ifp->if_serializer); 541 error = 0; 542 543 switch (ap->a_cmd) { 544 case TAPSIFINFO: 545 tapp = (struct tapinfo *)data; 546 ifp->if_mtu = tapp->mtu; 547 ifp->if_type = tapp->type; 548 ifp->if_baudrate = tapp->baudrate; 549 break; 550 551 case TAPGIFINFO: 552 tapp = (struct tapinfo *)data; 553 tapp->mtu = ifp->if_mtu; 554 tapp->type = ifp->if_type; 555 tapp->baudrate = ifp->if_baudrate; 556 break; 557 558 case TAPSDEBUG: 559 tapdebug = *(int *)data; 560 break; 561 562 case TAPGDEBUG: 563 *(int *)data = tapdebug; 564 break; 565 566 case FIOASYNC: 567 if (*(int *)data) 568 tp->tap_flags |= TAP_ASYNC; 569 else 570 tp->tap_flags &= ~TAP_ASYNC; 571 break; 572 573 case FIONREAD: 574 *(int *)data = 0; 575 if ((mb = ifq_poll(&ifp->if_snd)) != NULL) { 576 for(; mb != NULL; mb = mb->m_next) 577 *(int *)data += mb->m_len; 578 } 579 break; 580 581 case FIOSETOWN: 582 error = fsetown(*(int *)data, &tp->tap_sigio); 583 break; 584 585 case FIOGETOWN: 586 *(int *)data = fgetown(tp->tap_sigio); 587 break; 588 589 /* this is deprecated, FIOSETOWN should be used instead */ 590 case TIOCSPGRP: 591 error = fsetown(-(*(int *)data), &tp->tap_sigio); 592 break; 593 594 /* this is deprecated, FIOGETOWN should be used instead */ 595 case TIOCGPGRP: 596 *(int *)data = -fgetown(tp->tap_sigio); 597 break; 598 599 /* VMware/VMnet port ioctl's */ 600 601 case SIOCGIFFLAGS: /* get ifnet flags */ 602 bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags)); 603 break; 604 605 case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */ 606 f = *(short *)data; 607 f &= 0x0fff; 608 f &= ~IFF_CANTCHANGE; 609 f |= IFF_UP; 610 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE); 611 break; 612 613 case OSIOCGIFADDR: /* get MAC address of the remote side */ 614 case SIOCGIFADDR: 615 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr)); 616 break; 617 618 case SIOCSIFADDR: /* set MAC address of the remote side */ 619 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr)); 620 break; 621 622 default: 623 error = ENOTTY; 624 break; 625 } 626 lwkt_serialize_exit(ifp->if_serializer); 627 return (error); 628 } 629 630 631 /* 632 * tapread 633 * 634 * The ops read interface - reads a packet at a time, or at 635 * least as much of a packet as can be read. 636 * 637 * Called from the fileops interface with nothing held. 638 * 639 * MPSAFE 640 */ 641 static int 642 tapread(struct dev_read_args *ap) 643 { 644 cdev_t dev = ap->a_head.a_dev; 645 struct uio *uio = ap->a_uio; 646 struct tap_softc *tp = dev->si_drv1; 647 struct ifnet *ifp = &tp->tap_if; 648 struct mbuf *m0 = NULL; 649 int error = 0, len; 650 651 TAPDEBUG(ifp, "reading, minor = %#x\n", minor(tp->tap_dev)); 652 653 if ((tp->tap_flags & TAP_READY) != TAP_READY) { 654 TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n", 655 minor(tp->tap_dev), tp->tap_flags); 656 657 return (EHOSTDOWN); 658 } 659 660 tp->tap_flags &= ~TAP_RWAIT; 661 662 /* sleep until we get a packet */ 663 do { 664 lwkt_serialize_enter(ifp->if_serializer); 665 m0 = ifq_dequeue(&ifp->if_snd, NULL); 666 if (m0 == NULL) { 667 if (ap->a_ioflag & IO_NDELAY) { 668 lwkt_serialize_exit(ifp->if_serializer); 669 return (EWOULDBLOCK); 670 } 671 tp->tap_flags |= TAP_RWAIT; 672 crit_enter(); 673 tsleep_interlock(tp); 674 lwkt_serialize_exit(ifp->if_serializer); 675 error = tsleep(tp, PCATCH, "taprd", 0); 676 crit_exit(); 677 if (error) 678 return (error); 679 } else { 680 lwkt_serialize_exit(ifp->if_serializer); 681 } 682 } while (m0 == NULL); 683 684 BPF_MTAP(ifp, m0); 685 686 /* xfer packet to user space */ 687 while ((m0 != NULL) && (uio->uio_resid > 0) && (error == 0)) { 688 len = min(uio->uio_resid, m0->m_len); 689 if (len == 0) 690 break; 691 692 error = uiomove(mtod(m0, caddr_t), len, uio); 693 m0 = m_free(m0); 694 } 695 696 if (m0 != NULL) { 697 TAPDEBUG(ifp, "dropping mbuf, minor = %#x\n", 698 minor(tp->tap_dev)); 699 m_freem(m0); 700 } 701 702 return (error); 703 } 704 705 /* 706 * tapwrite 707 * 708 * The ops write interface - an atomic write is a packet - or else! 709 * 710 * Called from the fileops interface with nothing held. 711 * 712 * MPSAFE 713 */ 714 static int 715 tapwrite(struct dev_write_args *ap) 716 { 717 cdev_t dev = ap->a_head.a_dev; 718 struct uio *uio = ap->a_uio; 719 struct tap_softc *tp = dev->si_drv1; 720 struct ifnet *ifp = &tp->tap_if; 721 struct mbuf *top = NULL, **mp = NULL, *m = NULL; 722 int error = 0, tlen, mlen; 723 724 TAPDEBUG(ifp, "writing, minor = %#x\n", minor(tp->tap_dev)); 725 726 if (uio->uio_resid == 0) 727 return (0); 728 729 if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) { 730 TAPDEBUG(ifp, "invalid packet len = %d, minor = %#x\n", 731 uio->uio_resid, minor(tp->tap_dev)); 732 733 return (EIO); 734 } 735 tlen = uio->uio_resid; 736 737 /* get a header mbuf */ 738 MGETHDR(m, MB_DONTWAIT, MT_DATA); 739 if (m == NULL) 740 return (ENOBUFS); 741 mlen = MHLEN; 742 743 top = 0; 744 mp = ⊤ 745 while ((error == 0) && (uio->uio_resid > 0)) { 746 m->m_len = min(mlen, uio->uio_resid); 747 error = uiomove(mtod(m, caddr_t), m->m_len, uio); 748 *mp = m; 749 mp = &m->m_next; 750 if (uio->uio_resid > 0) { 751 MGET(m, MB_DONTWAIT, MT_DATA); 752 if (m == NULL) { 753 error = ENOBUFS; 754 break; 755 } 756 mlen = MLEN; 757 } 758 } 759 if (error) { 760 ifp->if_ierrors ++; 761 if (top) 762 m_freem(top); 763 return (error); 764 } 765 766 top->m_pkthdr.len = tlen; 767 top->m_pkthdr.rcvif = ifp; 768 769 /* 770 * Ethernet bridge and bpf are handled in ether_input 771 * 772 * adjust mbuf and give packet to the ether_input 773 */ 774 lwkt_serialize_enter(ifp->if_serializer); 775 ifp->if_input(ifp, top); 776 ifp->if_ipackets ++; /* ibytes are counted in ether_input */ 777 lwkt_serialize_exit(ifp->if_serializer); 778 779 return (0); 780 } 781 782 /* 783 * tappoll 784 * 785 * The poll interface, this is only useful on reads really. The write 786 * detect always returns true, write never blocks anyway, it either 787 * accepts the packet or drops it 788 * 789 * Called from the fileops interface with nothing held. 790 * 791 * MPSAFE 792 */ 793 static int 794 tappoll(struct dev_poll_args *ap) 795 { 796 cdev_t dev = ap->a_head.a_dev; 797 struct tap_softc *tp = dev->si_drv1; 798 struct ifnet *ifp = &tp->tap_if; 799 int revents = 0; 800 801 TAPDEBUG(ifp, "polling, minor = %#x\n", minor(tp->tap_dev)); 802 803 lwkt_serialize_enter(ifp->if_serializer); 804 if (ap->a_events & (POLLIN | POLLRDNORM)) { 805 if (!ifq_is_empty(&ifp->if_snd)) { 806 TAPDEBUG(ifp, 807 "has data in queue. minor = %#x\n", 808 minor(tp->tap_dev)); 809 810 revents |= (ap->a_events & (POLLIN | POLLRDNORM)); 811 } else { 812 TAPDEBUG(ifp, "waiting for data, minor = %#x\n", 813 minor(tp->tap_dev)); 814 815 get_mplock(); 816 selrecord(curthread, &tp->tap_rsel); 817 rel_mplock(); 818 } 819 } 820 lwkt_serialize_exit(ifp->if_serializer); 821 822 if (ap->a_events & (POLLOUT | POLLWRNORM)) 823 revents |= (ap->a_events & (POLLOUT | POLLWRNORM)); 824 ap->a_events = revents; 825 return(0); 826 } 827 828 /* 829 * tapkqfilter - called from the fileops interface with nothing held 830 * 831 * MPSAFE 832 */ 833 static int filt_tapread(struct knote *kn, long hint); 834 static void filt_tapdetach(struct knote *kn); 835 static struct filterops tapread_filtops = 836 { 1, NULL, filt_tapdetach, filt_tapread }; 837 838 int 839 tapkqfilter(struct dev_kqfilter_args *ap) 840 { 841 cdev_t dev = ap->a_head.a_dev; 842 struct knote *kn = ap->a_kn; 843 struct tap_softc *tp; 844 struct klist *list; 845 struct ifnet *ifp; 846 847 get_mplock(); 848 tp = dev->si_drv1; 849 ifp = &tp->tap_if; 850 ap->a_result =0; 851 852 switch(kn->kn_filter) { 853 case EVFILT_READ: 854 list = &tp->tap_rsel.si_note; 855 kn->kn_fop = &tapread_filtops; 856 kn->kn_hook = (void *)tp; 857 break; 858 case EVFILT_WRITE: 859 /* fall through */ 860 default: 861 ap->a_result = 1; 862 rel_mplock(); 863 return(0); 864 } 865 crit_enter(); 866 SLIST_INSERT_HEAD(list, kn, kn_selnext); 867 crit_exit(); 868 rel_mplock(); 869 return(0); 870 } 871 872 static int 873 filt_tapread(struct knote *kn, long hint) 874 { 875 struct tap_softc *tp = (void *)kn->kn_hook; 876 struct ifnet *ifp = &tp->tap_if; 877 878 if (ifq_is_empty(&ifp->if_snd) == 0) { 879 return(1); 880 } else { 881 return(0); 882 } 883 } 884 885 static void 886 filt_tapdetach(struct knote *kn) 887 { 888 struct tap_softc *tp = (void *)kn->kn_hook; 889 890 SLIST_REMOVE(&tp->tap_rsel.si_note, kn, knote, kn_selnext); 891 } 892