sys/net/tap/if_tap.c

   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 1.40 2008/05/18 05:12:08 sephe 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 static void             tapifstop(struct tap_softc *, int);
  95
  96 /* character device */
  97 static d_open_t         tapopen;
  98 static d_close_t        tapclose;
  99 static d_read_t         tapread;
 100 static d_write_t        tapwrite;
 101 static d_ioctl_t        tapioctl;
 102 static d_poll_t         tappoll;
 103 static d_kqfilter_t     tapkqfilter;
 104
 105 static struct dev_ops   tap_ops = {
 106         { CDEV_NAME, CDEV_MAJOR, 0 },
 107         .d_open =       tapopen,
 108         .d_close =      tapclose,
 109         .d_read =       tapread,
 110         .d_write =      tapwrite,
 111         .d_ioctl =      tapioctl,
 112         .d_poll =       tappoll,
 113         .d_kqfilter =   tapkqfilter
 114 };
 115
 116 static int              taprefcnt = 0;          /* module ref. counter   */
 117 static int              taplastunit = -1;       /* max. open unit number */
 118 static int              tapdebug = 0;           /* debug flag            */
 119
 120 MALLOC_DECLARE(M_TAP);
 121 MALLOC_DEFINE(M_TAP, CDEV_NAME, "Ethernet tunnel interface");
 122 SYSCTL_INT(_debug, OID_AUTO, if_tap_debug, CTLFLAG_RW, &tapdebug, 0, "");
 123 DEV_MODULE(if_tap, tapmodevent, NULL);
 124
 125 /*
 126  * tapmodevent
 127  *
 128  * module event handler
 129  */
 130 static int
 131 tapmodevent(module_t mod, int type, void *data)
 132 {
 133         static int               attached = 0;
 134         struct ifnet            *ifp = NULL;
 135         int                      unit;
 136
 137         switch (type) {
 138         case MOD_LOAD:
 139                 if (attached)
 140                         return (EEXIST);
 141
 142                 dev_ops_add(&tap_ops, 0, 0);
 143                 attached = 1;
 144                 break;
 145
 146         case MOD_UNLOAD:
 147                 if (taprefcnt > 0)
 148                         return (EBUSY);
 149
 150                 dev_ops_remove(&tap_ops, 0, 0);
 151
 152                 /* XXX: maintain tap ifs in a local list */
 153                 unit = 0;
 154                 while (unit <= taplastunit) {
 155                         TAILQ_FOREACH(ifp, &ifnet, if_link) {
 156                                 if ((strcmp(ifp->if_dname, TAP) == 0) ||
 157                                     (strcmp(ifp->if_dname, VMNET) == 0)) {
 158                                         if (ifp->if_dunit == unit)
 159                                                 break;
 160                                 }
 161                         }
 162
 163                         if (ifp != NULL) {
 164                                 struct tap_softc        *tp = ifp->if_softc;
 165
 166                                 TAPDEBUG(ifp, "detached. minor = %#x, " \
 167                                         "taplastunit = %d\n",
 168                                         minor(tp->tap_dev), taplastunit);
 169
 170                                 lwkt_serialize_enter(ifp->if_serializer);
 171                                 tapifstop(tp, 1);
 172                                 lwkt_serialize_exit(ifp->if_serializer);
 173
 174                                 ether_ifdetach(ifp);
 175                                 destroy_dev(tp->tap_dev);
 176                                 kfree(tp, M_TAP);
 177                         } else {
 178                                 unit++;
 179                         }
 180                 }
 181                 attached = 0;
 182                 break;
 183
 184         default:
 185                 return (EOPNOTSUPP);
 186         }
 187
 188         return (0);
 189 } /* tapmodevent */
 190
 191
 192 /*
 193  * tapcreate
 194  *
 195  * to create interface
 196  */
 197 static void
 198 tapcreate(cdev_t dev)
 199 {
 200         struct ifnet            *ifp = NULL;
 201         struct tap_softc        *tp = NULL;
 202         uint8_t                 ether_addr[ETHER_ADDR_LEN];
 203         int                      unit;
 204         char                    *name = NULL;
 205
 206         /* allocate driver storage and create device */
 207         MALLOC(tp, struct tap_softc *, sizeof(*tp), M_TAP, M_WAITOK | M_ZERO);
 208
 209         /* select device: tap or vmnet */
 210         if (minor(dev) & VMNET_DEV_MASK) {
 211                 name = VMNET;
 212                 unit = lminor(dev) & 0xff;
 213                 tp->tap_flags |= TAP_VMNET;
 214         }
 215         else {
 216                 name = TAP;
 217                 unit = lminor(dev);
 218         }
 219
 220         tp->tap_dev = make_dev(&tap_ops, minor(dev), UID_ROOT, GID_WHEEL,
 221                                                 0600, "%s%d", name, unit);
 222         tp->tap_dev->si_drv1 = dev->si_drv1 = tp;
 223         reference_dev(tp->tap_dev);     /* so we can destroy it later */
 224
 225         /* generate fake MAC address: 00 bd xx xx xx unit_no */
 226         ether_addr[0] = 0x00;
 227         ether_addr[1] = 0xbd;
 228         bcopy(&ticks, &ether_addr[2], 3);
 229         ether_addr[5] = (u_char)unit;
 230
 231         /* fill the rest and attach interface */
 232         ifp = &tp->tap_if;
 233         ifp->if_softc = tp;
 234
 235         if_initname(ifp, name, unit);
 236         if (unit > taplastunit)
 237                 taplastunit = unit;
 238
 239         ifp->if_init = tapifinit;
 240         ifp->if_start = tapifstart;
 241         ifp->if_ioctl = tapifioctl;
 242         ifp->if_mtu = ETHERMTU;
 243         ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
 244         ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
 245         ifq_set_ready(&ifp->if_snd);
 246
 247         ether_ifattach(ifp, ether_addr, NULL);
 248
 249         tp->tap_flags |= TAP_INITED;
 250         tp->tap_devq.ifq_maxlen = ifqmaxlen;
 251
 252         TAPDEBUG(ifp, "created. minor = %#x\n", minor(tp->tap_dev));
 253 } /* tapcreate */
 254
 255
 256 /*
 257  * tapopen
 258  *
 259  * to open tunnel. must be superuser
 260  */
 261 static int
 262 tapopen(struct dev_open_args *ap)
 263 {
 264         cdev_t dev = ap->a_head.a_dev;
 265         struct tap_softc *tp = NULL;
 266         struct ifnet *ifp = NULL;
 267         int error;
 268
 269         if ((error = suser_cred(ap->a_cred, 0)) != 0)
 270                 return (error);
 271
 272         get_mplock();
 273         tp = dev->si_drv1;
 274         if (tp == NULL) {
 275                 tapcreate(dev);
 276                 tp = dev->si_drv1;
 277                 ifp = &tp->arpcom.ac_if;
 278         } else {
 279                 if (tp->tap_flags & TAP_OPEN) {
 280                         rel_mplock();
 281                         return (EBUSY);
 282                 }
 283
 284                 ifp = &tp->arpcom.ac_if;
 285
 286                 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
 287
 288                 /* Announce the return of the interface. */
 289                 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
 290         }
 291
 292         bcopy(tp->arpcom.ac_enaddr, tp->ether_addr, sizeof(tp->ether_addr));
 293
 294         tp->tap_td = curthread;
 295         tp->tap_flags |= TAP_OPEN;
 296         taprefcnt ++;
 297
 298         TAPDEBUG(ifp, "opened. minor = %#x, refcnt = %d, taplastunit = %d\n",
 299                  minor(tp->tap_dev), taprefcnt, taplastunit);
 300
 301         rel_mplock();
 302         return (0);
 303 }
 304
 305
 306 /*
 307  * tapclose
 308  *
 309  * close the device - mark i/f down & delete routing info
 310  */
 311 static int
 312 tapclose(struct dev_close_args *ap)
 313 {
 314         cdev_t dev = ap->a_head.a_dev;
 315         struct tap_softc        *tp = dev->si_drv1;
 316         struct ifnet            *ifp = &tp->tap_if;
 317         int clear_flags = 1;
 318
 319         /* junk all pending output */
 320
 321         get_mplock();
 322         ifq_purge(&ifp->if_snd);
 323
 324         /*
 325          * do not bring the interface down, and do not anything with
 326          * interface, if we are in VMnet mode. just close the device.
 327          */
 328
 329         if ((tp->tap_flags & TAP_VMNET) == 0) {
 330                 if (ifp->if_flags & IFF_UP)
 331                         if_down(ifp);
 332                 clear_flags = 0;
 333         }
 334         lwkt_serialize_enter(ifp->if_serializer);
 335         tapifstop(tp, clear_flags);
 336         lwkt_serialize_exit(ifp->if_serializer);
 337
 338         if_purgeaddrs_nolink(ifp);
 339
 340         EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
 341
 342         /* Announce the departure of the interface. */
 343         rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
 344
 345         funsetown(tp->tap_sigio);
 346         selwakeup(&tp->tap_rsel);
 347
 348         tp->tap_flags &= ~TAP_OPEN;
 349         tp->tap_td = NULL;
 350
 351         taprefcnt --;
 352         if (taprefcnt < 0) {
 353                 taprefcnt = 0;
 354                 if_printf(ifp, "minor = %#x, refcnt = %d is out of sync. "
 355                         "set refcnt to 0\n", minor(tp->tap_dev), taprefcnt);
 356         }
 357
 358         TAPDEBUG(ifp, "closed. minor = %#x, refcnt = %d, taplastunit = %d\n",
 359                  minor(tp->tap_dev), taprefcnt, taplastunit);
 360
 361         rel_mplock();
 362         return (0);
 363 }
 364
 365
 366 /*
 367  * tapifinit
 368  *
 369  * Network interface initialization function (called with if serializer held)
 370  *
 371  * MPSAFE
 372  */
 373 static void
 374 tapifinit(void *xtp)
 375 {
 376         struct tap_softc *tp = xtp;
 377         struct ifnet *ifp = &tp->tap_if;
 378
 379         TAPDEBUG(ifp, "initializing, minor = %#x\n", minor(tp->tap_dev));
 380
 381         ASSERT_SERIALIZED(ifp->if_serializer);
 382
 383         tapifstop(tp, 1);
 384
 385         ifp->if_flags |= IFF_RUNNING;
 386         ifp->if_flags &= ~IFF_OACTIVE;
 387
 388         /* attempt to start output */
 389         tapifstart(ifp);
 390 }
 391
 392
 393 /*
 394  * tapifioctl
 395  *
 396  * Process an ioctl request on network interface (called with if serializer
 397  * held).
 398  *
 399  * MPSAFE
 400  */
 401 static int
 402 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
 403 {
 404         struct tap_softc        *tp = (struct tap_softc *)(ifp->if_softc);
 405         struct ifstat           *ifs = NULL;
 406         int                      dummy;
 407
 408         switch (cmd) {
 409                 case SIOCSIFADDR:
 410                 case SIOCGIFADDR:
 411                 case SIOCSIFMTU:
 412                         dummy = ether_ioctl(ifp, cmd, data);
 413                         return (dummy);
 414
 415                 case SIOCSIFFLAGS:
 416                         if ((tp->tap_flags & TAP_VMNET) == 0) {
 417                                 /*
 418                                  * Only for non-vmnet tap(4)
 419                                  */
 420                                 if (ifp->if_flags & IFF_UP) {
 421                                         if ((ifp->if_flags & IFF_RUNNING) == 0)
 422                                                 tapifinit(tp);
 423                                 } else {
 424                                         tapifstop(tp, 1);
 425                                 }
 426                         } else {
 427                                 /* XXX */
 428                         }
 429                         break;
 430                 case SIOCADDMULTI: /* XXX -- just like vmnet does */
 431                 case SIOCDELMULTI:
 432                         break;
 433
 434                 case SIOCGIFSTATUS:
 435                         ifs = (struct ifstat *)data;
 436                         dummy = strlen(ifs->ascii);
 437                         if (tp->tap_td != NULL && dummy < sizeof(ifs->ascii)) {
 438                                 if (tp->tap_td->td_proc) {
 439                                     ksnprintf(ifs->ascii + dummy,
 440                                         sizeof(ifs->ascii) - dummy,
 441                                         "\tOpened by pid %d\n",
 442                                         (int)tp->tap_td->td_proc->p_pid);
 443                                 } else {
 444                                     ksnprintf(ifs->ascii + dummy,
 445                                         sizeof(ifs->ascii) - dummy,
 446                                         "\tOpened by td %p\n", tp->tap_td);
 447                                 }
 448                         }
 449                         break;
 450
 451                 default:
 452                         return (EINVAL);
 453         }
 454
 455         return (0);
 456 }
 457
 458
 459 /*
 460  * tapifstart
 461  *
 462  * Queue packets from higher level ready to put out (called with if serializer
 463  * held)
 464  *
 465  * MPSAFE
 466  */
 467 static void
 468 tapifstart(struct ifnet *ifp)
 469 {
 470         struct tap_softc *tp = ifp->if_softc;
 471         struct ifqueue *ifq;
 472         struct mbuf *m;
 473         int has_data = 0;
 474
 475         TAPDEBUG(ifp, "starting, minor = %#x\n", minor(tp->tap_dev));
 476
 477         /*
 478          * do not junk pending output if we are in VMnet mode.
 479          * XXX: can this do any harm because of queue overflow?
 480          */
 481
 482         if (((tp->tap_flags & TAP_VMNET) == 0) &&
 483             ((tp->tap_flags & TAP_READY) != TAP_READY)) {
 484                 TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
 485                          minor(tp->tap_dev), tp->tap_flags);
 486                 ifq_purge(&ifp->if_snd);
 487                 return;
 488         }
 489
 490         ifp->if_flags |= IFF_OACTIVE;
 491
 492         ifq = &tp->tap_devq;
 493         while ((m = ifq_dequeue(&ifp->if_snd, NULL)) != NULL) {
 494                 if (IF_QFULL(ifq)) {
 495                         IF_DROP(ifq);
 496                         ifp->if_oerrors++;
 497                         m_freem(m);
 498                 } else {
 499                         IF_ENQUEUE(ifq, m);
 500                         ifp->if_opackets++;
 501                         has_data = 1;
 502                 }
 503         }
 504
 505         if (has_data) {
 506                 if (tp->tap_flags & TAP_RWAIT) {
 507                         tp->tap_flags &= ~TAP_RWAIT;
 508                         wakeup((caddr_t)tp);
 509                 }
 510
 511                 get_mplock();
 512                 KNOTE(&tp->tap_rsel.si_note, 0);
 513                 rel_mplock();
 514
 515                 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
 516                         get_mplock();
 517                         pgsigio(tp->tap_sigio, SIGIO, 0);
 518                         rel_mplock();
 519                 }
 520
 521                 /*
 522                  * selwakeup is not MPSAFE.  tapifstart is.
 523                  */
 524                 get_mplock();
 525                 selwakeup(&tp->tap_rsel);
 526                 rel_mplock();
 527         }
 528
 529         ifp->if_flags &= ~IFF_OACTIVE;
 530 }
 531
 532
 533 /*
 534  * tapioctl
 535  *
 536  * The ops interface is now pretty minimal.  Called via fileops with nothing
 537  * held.
 538  *
 539  * MPSAFE
 540  */
 541 static int
 542 tapioctl(struct dev_ioctl_args *ap)
 543 {
 544         cdev_t dev = ap->a_head.a_dev;
 545         caddr_t data = ap->a_data;
 546         struct tap_softc        *tp = dev->si_drv1;
 547         struct ifnet            *ifp = &tp->tap_if;
 548         struct tapinfo          *tapp = NULL;
 549         struct mbuf *mb;
 550         short f;
 551         int error;
 552
 553         lwkt_serialize_enter(ifp->if_serializer);
 554         error = 0;
 555
 556         switch (ap->a_cmd) {
 557         case TAPSIFINFO:
 558                 tapp = (struct tapinfo *)data;
 559                 ifp->if_mtu = tapp->mtu;
 560                 ifp->if_type = tapp->type;
 561                 ifp->if_baudrate = tapp->baudrate;
 562                 break;
 563
 564         case TAPGIFINFO:
 565                 tapp = (struct tapinfo *)data;
 566                 tapp->mtu = ifp->if_mtu;
 567                 tapp->type = ifp->if_type;
 568                 tapp->baudrate = ifp->if_baudrate;
 569                 break;
 570
 571         case TAPSDEBUG:
 572                 tapdebug = *(int *)data;
 573                 break;
 574
 575         case TAPGDEBUG:
 576                 *(int *)data = tapdebug;
 577                 break;
 578
 579         case FIOASYNC:
 580                 if (*(int *)data)
 581                         tp->tap_flags |= TAP_ASYNC;
 582                 else
 583                         tp->tap_flags &= ~TAP_ASYNC;
 584                 break;
 585
 586         case FIONREAD:
 587                 *(int *)data = 0;
 588
 589                 /* Take a look at devq first */
 590                 IF_POLL(&tp->tap_devq, mb);
 591                 if (mb == NULL)
 592                         mb = ifq_poll(&ifp->if_snd);
 593
 594                 if (mb != NULL) {
 595                         for(; mb != NULL; mb = mb->m_next)
 596                                 *(int *)data += mb->m_len;
 597                 }
 598                 break;
 599
 600         case FIOSETOWN:
 601                 error = fsetown(*(int *)data, &tp->tap_sigio);
 602                 break;
 603
 604         case FIOGETOWN:
 605                 *(int *)data = fgetown(tp->tap_sigio);
 606                 break;
 607
 608         /* this is deprecated, FIOSETOWN should be used instead */
 609         case TIOCSPGRP:
 610                 error = fsetown(-(*(int *)data), &tp->tap_sigio);
 611                 break;
 612
 613         /* this is deprecated, FIOGETOWN should be used instead */
 614         case TIOCGPGRP:
 615                 *(int *)data = -fgetown(tp->tap_sigio);
 616                 break;
 617
 618         /* VMware/VMnet port ioctl's */
 619
 620         case SIOCGIFFLAGS:      /* get ifnet flags */
 621                 bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags));
 622                 break;
 623
 624         case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
 625                 f = *(short *)data;
 626                 f &= 0x0fff;
 627                 f &= ~IFF_CANTCHANGE;
 628                 f |= IFF_UP;
 629                 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
 630                 break;
 631
 632         case OSIOCGIFADDR:      /* get MAC address of the remote side */
 633         case SIOCGIFADDR:
 634                 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
 635                 break;
 636
 637         case SIOCSIFADDR:       /* set MAC address of the remote side */
 638                 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
 639                 break;
 640
 641         default:
 642                 error = ENOTTY;
 643                 break;
 644         }
 645         lwkt_serialize_exit(ifp->if_serializer);
 646         return (error);
 647 }
 648
 649
 650 /*
 651  * tapread
 652  *
 653  * The ops read interface - reads a packet at a time, or at
 654  * least as much of a packet as can be read.
 655  *
 656  * Called from the fileops interface with nothing held.
 657  *
 658  * MPSAFE
 659  */
 660 static int
 661 tapread(struct dev_read_args *ap)
 662 {
 663         cdev_t dev = ap->a_head.a_dev;
 664         struct uio *uio = ap->a_uio;
 665         struct tap_softc        *tp = dev->si_drv1;
 666         struct ifnet            *ifp = &tp->tap_if;
 667         struct mbuf             *m0 = NULL;
 668         int                      error = 0, len;
 669
 670         TAPDEBUG(ifp, "reading, minor = %#x\n", minor(tp->tap_dev));
 671
 672         if ((tp->tap_flags & TAP_READY) != TAP_READY) {
 673                 TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
 674                          minor(tp->tap_dev), tp->tap_flags);
 675
 676                 return (EHOSTDOWN);
 677         }
 678
 679         tp->tap_flags &= ~TAP_RWAIT;
 680
 681         /* sleep until we get a packet */
 682         do {
 683                 lwkt_serialize_enter(ifp->if_serializer);
 684                 IF_DEQUEUE(&tp->tap_devq, m0);
 685                 if (m0 == NULL) {
 686                         if (ap->a_ioflag & IO_NDELAY) {
 687                                 lwkt_serialize_exit(ifp->if_serializer);
 688                                 return (EWOULDBLOCK);
 689                         }
 690                         tp->tap_flags |= TAP_RWAIT;
 691                         crit_enter();
 692                         tsleep_interlock(tp);
 693                         lwkt_serialize_exit(ifp->if_serializer);
 694                         error = tsleep(tp, PCATCH, "taprd", 0);
 695                         crit_exit();
 696                         if (error)
 697                                 return (error);
 698                 } else {
 699                         lwkt_serialize_exit(ifp->if_serializer);
 700                 }
 701         } while (m0 == NULL);
 702
 703         BPF_MTAP(ifp, m0);
 704
 705         /* xfer packet to user space */
 706         while ((m0 != NULL) && (uio->uio_resid > 0) && (error == 0)) {
 707                 len = min(uio->uio_resid, m0->m_len);
 708                 if (len == 0)
 709                         break;
 710
 711                 error = uiomove(mtod(m0, caddr_t), len, uio);
 712                 m0 = m_free(m0);
 713         }
 714
 715         if (m0 != NULL) {
 716                 TAPDEBUG(ifp, "dropping mbuf, minor = %#x\n",
 717                          minor(tp->tap_dev));
 718                 m_freem(m0);
 719         }
 720
 721         return (error);
 722 }
 723
 724 /*
 725  * tapwrite
 726  *
 727  * The ops write interface - an atomic write is a packet - or else!
 728  *
 729  * Called from the fileops interface with nothing held.
 730  *
 731  * MPSAFE
 732  */
 733 static int
 734 tapwrite(struct dev_write_args *ap)
 735 {
 736         cdev_t dev = ap->a_head.a_dev;
 737         struct uio *uio = ap->a_uio;
 738         struct tap_softc        *tp = dev->si_drv1;
 739         struct ifnet            *ifp = &tp->tap_if;
 740         struct mbuf             *top = NULL, **mp = NULL, *m = NULL;
 741         int                      error = 0, tlen, mlen;
 742
 743         TAPDEBUG(ifp, "writing, minor = %#x\n", minor(tp->tap_dev));
 744
 745         if (uio->uio_resid == 0)
 746                 return (0);
 747
 748         if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
 749                 TAPDEBUG(ifp, "invalid packet len = %d, minor = %#x\n",
 750                          uio->uio_resid, minor(tp->tap_dev));
 751
 752                 return (EIO);
 753         }
 754         tlen = uio->uio_resid;
 755
 756         /* get a header mbuf */
 757         MGETHDR(m, MB_DONTWAIT, MT_DATA);
 758         if (m == NULL)
 759                 return (ENOBUFS);
 760         mlen = MHLEN;
 761
 762         top = 0;
 763         mp = &top;
 764         while ((error == 0) && (uio->uio_resid > 0)) {
 765                 m->m_len = min(mlen, uio->uio_resid);
 766                 error = uiomove(mtod(m, caddr_t), m->m_len, uio);
 767                 *mp = m;
 768                 mp = &m->m_next;
 769                 if (uio->uio_resid > 0) {
 770                         MGET(m, MB_DONTWAIT, MT_DATA);
 771                         if (m == NULL) {
 772                                 error = ENOBUFS;
 773                                 break;
 774                         }
 775                         mlen = MLEN;
 776                 }
 777         }
 778         if (error) {
 779                 ifp->if_ierrors ++;
 780                 if (top)
 781                         m_freem(top);
 782                 return (error);
 783         }
 784
 785         top->m_pkthdr.len = tlen;
 786         top->m_pkthdr.rcvif = ifp;
 787
 788         /*
 789          * Ethernet bridge and bpf are handled in ether_input
 790          *
 791          * adjust mbuf and give packet to the ether_input
 792          */
 793         lwkt_serialize_enter(ifp->if_serializer);
 794         ifp->if_input(ifp, top);
 795         ifp->if_ipackets ++; /* ibytes are counted in ether_input */
 796         lwkt_serialize_exit(ifp->if_serializer);
 797
 798         return (0);
 799 }
 800
 801 /*
 802  * tappoll
 803  *
 804  * The poll interface, this is only useful on reads really. The write
 805  * detect always returns true, write never blocks anyway, it either
 806  * accepts the packet or drops it
 807  *
 808  * Called from the fileops interface with nothing held.
 809  *
 810  * MPSAFE
 811  */
 812 static int
 813 tappoll(struct dev_poll_args *ap)
 814 {
 815         cdev_t dev = ap->a_head.a_dev;
 816         struct tap_softc        *tp = dev->si_drv1;
 817         struct ifnet            *ifp = &tp->tap_if;
 818         int                      revents = 0;
 819
 820         TAPDEBUG(ifp, "polling, minor = %#x\n", minor(tp->tap_dev));
 821
 822         if (ap->a_events & (POLLIN | POLLRDNORM)) {
 823                 if (!IF_QEMPTY(&tp->tap_devq)) {
 824                         TAPDEBUG(ifp,
 825                                  "has data in queue. minor = %#x\n",
 826                                  minor(tp->tap_dev));
 827
 828                         revents |= (ap->a_events & (POLLIN | POLLRDNORM));
 829                 } else {
 830                         TAPDEBUG(ifp, "waiting for data, minor = %#x\n",
 831                                  minor(tp->tap_dev));
 832
 833                         get_mplock();
 834                         selrecord(curthread, &tp->tap_rsel);
 835                         rel_mplock();
 836                 }
 837         }
 838
 839         if (ap->a_events & (POLLOUT | POLLWRNORM))
 840                 revents |= (ap->a_events & (POLLOUT | POLLWRNORM));
 841         ap->a_events = revents;
 842         return(0);
 843 }
 844
 845 /*
 846  * tapkqfilter - called from the fileops interface with nothing held
 847  *
 848  * MPSAFE
 849  */
 850 static int filt_tapread(struct knote *kn, long hint);
 851 static void filt_tapdetach(struct knote *kn);
 852 static struct filterops tapread_filtops =
 853         { 1, NULL, filt_tapdetach, filt_tapread };
 854
 855 static int
 856 tapkqfilter(struct dev_kqfilter_args *ap)
 857 {
 858         cdev_t dev = ap->a_head.a_dev;
 859         struct knote *kn = ap->a_kn;
 860         struct tap_softc *tp;
 861         struct klist *list;
 862         struct ifnet *ifp;
 863
 864         get_mplock();
 865         tp = dev->si_drv1;
 866         ifp = &tp->tap_if;
 867         ap->a_result =0;
 868
 869         switch(kn->kn_filter) {
 870         case EVFILT_READ:
 871                 list = &tp->tap_rsel.si_note;
 872                 kn->kn_fop = &tapread_filtops;
 873                 kn->kn_hook = (void *)tp;
 874                 break;
 875         case EVFILT_WRITE:
 876                 /* fall through */
 877         default:
 878                 ap->a_result = 1;
 879                 rel_mplock();
 880                 return(0);
 881         }
 882         crit_enter();
 883         SLIST_INSERT_HEAD(list, kn, kn_selnext);
 884         crit_exit();
 885         rel_mplock();
 886         return(0);
 887 }
 888
 889 static int
 890 filt_tapread(struct knote *kn, long hint)
 891 {
 892         struct tap_softc *tp = (void *)kn->kn_hook;
 893
 894         if (IF_QEMPTY(&tp->tap_devq) == 0)      /* XXX serializer */
 895                 return(1);
 896         else
 897                 return(0);
 898 }
 899
 900 static void
 901 filt_tapdetach(struct knote *kn)
 902 {
 903         struct tap_softc *tp = (void *)kn->kn_hook;
 904
 905         SLIST_REMOVE(&tp->tap_rsel.si_note, kn, knote, kn_selnext);
 906 }
 907
 908 static void
 909 tapifstop(struct tap_softc *tp, int clear_flags)
 910 {
 911         struct ifnet *ifp = &tp->tap_if;
 912
 913         ASSERT_SERIALIZED(ifp->if_serializer);
 914         IF_DRAIN(&tp->tap_devq);
 915         if (clear_flags)
 916                 ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
 917 }