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 1.37 2007/09/16 17:02:49 pavalos 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 | M_ZERO);
 204
 205         /* select device: tap or vmnet */
 206         if (minor(dev) & VMNET_DEV_MASK) {
 207                 name = VMNET;
 208                 unit = lminor(dev) & 0xff;
 209                 tp->tap_flags |= TAP_VMNET;
 210         }
 211         else {
 212                 name = TAP;
 213                 unit = lminor(dev);
 214         }
 215
 216         tp->tap_dev = make_dev(&tap_ops, minor(dev), UID_ROOT, GID_WHEEL,
 217                                                 0600, "%s%d", name, unit);
 218         tp->tap_dev->si_drv1 = dev->si_drv1 = tp;
 219         reference_dev(tp->tap_dev);     /* so we can destroy it later */
 220
 221         /* generate fake MAC address: 00 bd xx xx xx unit_no */
 222         ether_addr[0] = 0x00;
 223         ether_addr[1] = 0xbd;
 224         bcopy(&ticks, &ether_addr[2], 3);
 225         ether_addr[5] = (u_char)unit;
 226
 227         /* fill the rest and attach interface */
 228         ifp = &tp->tap_if;
 229         ifp->if_softc = tp;
 230
 231         if_initname(ifp, name, unit);
 232         if (unit > taplastunit)
 233                 taplastunit = unit;
 234
 235         ifp->if_init = tapifinit;
 236         ifp->if_start = tapifstart;
 237         ifp->if_ioctl = tapifioctl;
 238         ifp->if_mtu = ETHERMTU;
 239         ifp->if_flags = (IFF_BROADCAST|IFF_SIMPLEX|IFF_MULTICAST);
 240         ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
 241         ifq_set_ready(&ifp->if_snd);
 242
 243         ether_ifattach(ifp, ether_addr, NULL);
 244
 245         tp->tap_flags |= TAP_INITED;
 246
 247         TAPDEBUG(ifp, "created. minor = %#x\n", minor(tp->tap_dev));
 248 } /* tapcreate */
 249
 250
 251 /*
 252  * tapopen
 253  *
 254  * to open tunnel. must be superuser
 255  */
 256 static int
 257 tapopen(struct dev_open_args *ap)
 258 {
 259         cdev_t dev = ap->a_head.a_dev;
 260         struct tap_softc *tp = NULL;
 261         struct ifnet *ifp = NULL;
 262         int error;
 263
 264         if ((error = suser_cred(ap->a_cred, 0)) != 0)
 265                 return (error);
 266
 267         get_mplock();
 268         tp = dev->si_drv1;
 269         if (tp == NULL) {
 270                 tapcreate(dev);
 271                 tp = dev->si_drv1;
 272                 ifp = &tp->arpcom.ac_if;
 273         } else {
 274                 ifp = &tp->arpcom.ac_if;
 275
 276                 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
 277
 278                 /* Announce the return of the interface. */
 279                 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
 280         }
 281
 282         if (tp->tap_flags & TAP_OPEN) {
 283                 rel_mplock();
 284                 return (EBUSY);
 285         }
 286
 287         bcopy(tp->arpcom.ac_enaddr, tp->ether_addr, sizeof(tp->ether_addr));
 288
 289         tp->tap_td = curthread;
 290         tp->tap_flags |= TAP_OPEN;
 291         taprefcnt ++;
 292
 293         TAPDEBUG(ifp, "opened. minor = %#x, refcnt = %d, taplastunit = %d\n",
 294                  minor(tp->tap_dev), taprefcnt, taplastunit);
 295
 296         rel_mplock();
 297         return (0);
 298 }
 299
 300
 301 /*
 302  * tapclose
 303  *
 304  * close the device - mark i/f down & delete routing info
 305  */
 306 static int
 307 tapclose(struct dev_close_args *ap)
 308 {
 309         cdev_t dev = ap->a_head.a_dev;
 310         struct tap_softc        *tp = dev->si_drv1;
 311         struct ifnet            *ifp = &tp->tap_if;
 312
 313         /* junk all pending output */
 314
 315         get_mplock();
 316         lwkt_serialize_enter(ifp->if_serializer);
 317         ifq_purge(&ifp->if_snd);
 318         lwkt_serialize_exit(ifp->if_serializer);
 319
 320         /*
 321          * do not bring the interface down, and do not anything with
 322          * interface, if we are in VMnet mode. just close the device.
 323          */
 324
 325         if (((tp->tap_flags & TAP_VMNET) == 0) && (ifp->if_flags & IFF_UP)) {
 326                 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
 327
 328                 /* Announce the departure of the interface. */
 329                 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
 330
 331                 if_down(ifp);
 332                 lwkt_serialize_enter(ifp->if_serializer);
 333                 if (ifp->if_flags & IFF_RUNNING) {
 334                         /* find internet addresses and delete routes */
 335                         struct ifaddr   *ifa = NULL;
 336
 337                         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 338                                 if (ifa->ifa_addr->sa_family == AF_INET) {
 339                                         rtinit(ifa, (int)RTM_DELETE, 0);
 340
 341                                         /* remove address from interface */
 342                                         bzero(ifa->ifa_addr,
 343                                                    sizeof(*(ifa->ifa_addr)));
 344                                         bzero(ifa->ifa_dstaddr,
 345                                                    sizeof(*(ifa->ifa_dstaddr)));
 346                                         bzero(ifa->ifa_netmask,
 347                                                    sizeof(*(ifa->ifa_netmask)));
 348                                 }
 349                         }
 350
 351                         ifp->if_flags &= ~IFF_RUNNING;
 352                 }
 353                 lwkt_serialize_exit(ifp->if_serializer);
 354         }
 355
 356         funsetown(tp->tap_sigio);
 357         selwakeup(&tp->tap_rsel);
 358
 359         tp->tap_flags &= ~TAP_OPEN;
 360         tp->tap_td = NULL;
 361
 362         taprefcnt --;
 363         if (taprefcnt < 0) {
 364                 taprefcnt = 0;
 365                 if_printf(ifp, "minor = %#x, refcnt = %d is out of sync. "
 366                         "set refcnt to 0\n", minor(tp->tap_dev), taprefcnt);
 367         }
 368
 369         TAPDEBUG(ifp, "closed. minor = %#x, refcnt = %d, taplastunit = %d\n",
 370                  minor(tp->tap_dev), taprefcnt, taplastunit);
 371
 372         rel_mplock();
 373         return (0);
 374 }
 375
 376
 377 /*
 378  * tapifinit
 379  *
 380  * Network interface initialization function (called with if serializer held)
 381  *
 382  * MPSAFE
 383  */
 384 static void
 385 tapifinit(void *xtp)
 386 {
 387         struct tap_softc        *tp = (struct tap_softc *)xtp;
 388         struct ifnet            *ifp = &tp->tap_if;
 389
 390         TAPDEBUG(ifp, "initializing, minor = %#x\n", minor(tp->tap_dev));
 391
 392         ifp->if_flags |= IFF_RUNNING;
 393         ifp->if_flags &= ~IFF_OACTIVE;
 394
 395         /* attempt to start output */
 396         tapifstart(ifp);
 397 }
 398
 399
 400 /*
 401  * tapifioctl
 402  *
 403  * Process an ioctl request on network interface (called with if serializer
 404  * held).
 405  *
 406  * MPSAFE
 407  */
 408 static int
 409 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
 410 {
 411         struct tap_softc        *tp = (struct tap_softc *)(ifp->if_softc);
 412         struct ifstat           *ifs = NULL;
 413         int                      dummy;
 414
 415         switch (cmd) {
 416                 case SIOCSIFADDR:
 417                 case SIOCGIFADDR:
 418                 case SIOCSIFMTU:
 419                         dummy = ether_ioctl(ifp, cmd, data);
 420                         return (dummy);
 421
 422                 case SIOCSIFFLAGS:
 423                         if ((tp->tap_flags & TAP_VMNET) == 0) {
 424                                 /*
 425                                  * Only for non-vmnet tap(4)
 426                                  */
 427                                 if (ifp->if_flags & IFF_UP) {
 428                                         if ((ifp->if_flags & IFF_RUNNING) == 0)
 429                                                 tapifinit(tp);
 430                                 }
 431                         }
 432                         break;
 433                 case SIOCADDMULTI: /* XXX -- just like vmnet does */
 434                 case SIOCDELMULTI:
 435                         break;
 436
 437                 case SIOCGIFSTATUS:
 438                         ifs = (struct ifstat *)data;
 439                         dummy = strlen(ifs->ascii);
 440                         if (tp->tap_td != NULL && dummy < sizeof(ifs->ascii)) {
 441                                 if (tp->tap_td->td_proc) {
 442                                     ksnprintf(ifs->ascii + dummy,
 443                                         sizeof(ifs->ascii) - dummy,
 444                                         "\tOpened by pid %d\n",
 445                                         (int)tp->tap_td->td_proc->p_pid);
 446                                 } else {
 447                                     ksnprintf(ifs->ascii + dummy,
 448                                         sizeof(ifs->ascii) - dummy,
 449                                         "\tOpened by td %p\n", tp->tap_td);
 450                                 }
 451                         }
 452                         break;
 453
 454                 default:
 455                         return (EINVAL);
 456         }
 457
 458         return (0);
 459 }
 460
 461
 462 /*
 463  * tapifstart
 464  *
 465  * Queue packets from higher level ready to put out (called with if serializer
 466  * held)
 467  *
 468  * MPSAFE
 469  */
 470 static void
 471 tapifstart(struct ifnet *ifp)
 472 {
 473         struct tap_softc        *tp = ifp->if_softc;
 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
 487                 ifq_purge(&ifp->if_snd);
 488                 return;
 489         }
 490
 491         ifp->if_flags |= IFF_OACTIVE;
 492
 493         if (!ifq_is_empty(&ifp->if_snd)) {
 494                 if (tp->tap_flags & TAP_RWAIT) {
 495                         tp->tap_flags &= ~TAP_RWAIT;
 496                         wakeup((caddr_t)tp);
 497                 }
 498                 KNOTE(&tp->tap_rsel.si_note, 0);
 499
 500                 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
 501                         get_mplock();
 502                         pgsigio(tp->tap_sigio, SIGIO, 0);
 503                         rel_mplock();
 504                 }
 505
 506                 /*
 507                  * selwakeup is not MPSAFE.  tapifstart is.
 508                  */
 509                 get_mplock();
 510                 selwakeup(&tp->tap_rsel);
 511                 rel_mplock();
 512                 ifp->if_opackets ++; /* obytes are counted in ether_output */
 513         }
 514
 515         ifp->if_flags &= ~IFF_OACTIVE;
 516 }
 517
 518
 519 /*
 520  * tapioctl
 521  *
 522  * The ops interface is now pretty minimal.  Called via fileops with nothing
 523  * held.
 524  *
 525  * MPSAFE
 526  */
 527 static int
 528 tapioctl(struct dev_ioctl_args *ap)
 529 {
 530         cdev_t dev = ap->a_head.a_dev;
 531         caddr_t data = ap->a_data;
 532         struct tap_softc        *tp = dev->si_drv1;
 533         struct ifnet            *ifp = &tp->tap_if;
 534         struct tapinfo          *tapp = NULL;
 535         struct mbuf *mb;
 536         short f;
 537         int error;
 538
 539         lwkt_serialize_enter(ifp->if_serializer);
 540         error = 0;
 541
 542         switch (ap->a_cmd) {
 543         case TAPSIFINFO:
 544                 tapp = (struct tapinfo *)data;
 545                 ifp->if_mtu = tapp->mtu;
 546                 ifp->if_type = tapp->type;
 547                 ifp->if_baudrate = tapp->baudrate;
 548                 break;
 549
 550         case TAPGIFINFO:
 551                 tapp = (struct tapinfo *)data;
 552                 tapp->mtu = ifp->if_mtu;
 553                 tapp->type = ifp->if_type;
 554                 tapp->baudrate = ifp->if_baudrate;
 555                 break;
 556
 557         case TAPSDEBUG:
 558                 tapdebug = *(int *)data;
 559                 break;
 560
 561         case TAPGDEBUG:
 562                 *(int *)data = tapdebug;
 563                 break;
 564
 565         case FIOASYNC:
 566                 if (*(int *)data)
 567                         tp->tap_flags |= TAP_ASYNC;
 568                 else
 569                         tp->tap_flags &= ~TAP_ASYNC;
 570                 break;
 571
 572         case FIONREAD:
 573                 *(int *)data = 0;
 574                 if ((mb = ifq_poll(&ifp->if_snd)) != NULL) {
 575                         for(; mb != NULL; mb = mb->m_next)
 576                                 *(int *)data += mb->m_len;
 577                 }
 578                 break;
 579
 580         case FIOSETOWN:
 581                 error = fsetown(*(int *)data, &tp->tap_sigio);
 582                 break;
 583
 584         case FIOGETOWN:
 585                 *(int *)data = fgetown(tp->tap_sigio);
 586                 break;
 587
 588         /* this is deprecated, FIOSETOWN should be used instead */
 589         case TIOCSPGRP:
 590                 error = fsetown(-(*(int *)data), &tp->tap_sigio);
 591                 break;
 592
 593         /* this is deprecated, FIOGETOWN should be used instead */
 594         case TIOCGPGRP:
 595                 *(int *)data = -fgetown(tp->tap_sigio);
 596                 break;
 597
 598         /* VMware/VMnet port ioctl's */
 599
 600         case SIOCGIFFLAGS:      /* get ifnet flags */
 601                 bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags));
 602                 break;
 603
 604         case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
 605                 f = *(short *)data;
 606                 f &= 0x0fff;
 607                 f &= ~IFF_CANTCHANGE;
 608                 f |= IFF_UP;
 609                 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
 610                 break;
 611
 612         case OSIOCGIFADDR:      /* get MAC address of the remote side */
 613         case SIOCGIFADDR:
 614                 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
 615                 break;
 616
 617         case SIOCSIFADDR:       /* set MAC address of the remote side */
 618                 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
 619                 break;
 620
 621         default:
 622                 error = ENOTTY;
 623                 break;
 624         }
 625         lwkt_serialize_exit(ifp->if_serializer);
 626         return (error);
 627 }
 628
 629
 630 /*
 631  * tapread
 632  *
 633  * The ops read interface - reads a packet at a time, or at
 634  * least as much of a packet as can be read.
 635  *
 636  * Called from the fileops interface with nothing held.
 637  *
 638  * MPSAFE
 639  */
 640 static int
 641 tapread(struct dev_read_args *ap)
 642 {
 643         cdev_t dev = ap->a_head.a_dev;
 644         struct uio *uio = ap->a_uio;
 645         struct tap_softc        *tp = dev->si_drv1;
 646         struct ifnet            *ifp = &tp->tap_if;
 647         struct mbuf             *m0 = NULL;
 648         int                      error = 0, len;
 649
 650         TAPDEBUG(ifp, "reading, minor = %#x\n", minor(tp->tap_dev));
 651
 652         if ((tp->tap_flags & TAP_READY) != TAP_READY) {
 653                 TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
 654                          minor(tp->tap_dev), tp->tap_flags);
 655
 656                 return (EHOSTDOWN);
 657         }
 658
 659         tp->tap_flags &= ~TAP_RWAIT;
 660
 661         /* sleep until we get a packet */
 662         do {
 663                 lwkt_serialize_enter(ifp->if_serializer);
 664                 m0 = ifq_dequeue(&ifp->if_snd, NULL);
 665                 if (m0 == NULL) {
 666                         if (ap->a_ioflag & IO_NDELAY) {
 667                                 lwkt_serialize_exit(ifp->if_serializer);
 668                                 return (EWOULDBLOCK);
 669                         }
 670                         tp->tap_flags |= TAP_RWAIT;
 671                         crit_enter();
 672                         tsleep_interlock(tp);
 673                         lwkt_serialize_exit(ifp->if_serializer);
 674                         error = tsleep(tp, PCATCH, "taprd", 0);
 675                         crit_exit();
 676                         if (error)
 677                                 return (error);
 678                 } else {
 679                         lwkt_serialize_exit(ifp->if_serializer);
 680                 }
 681         } while (m0 == NULL);
 682
 683         BPF_MTAP(ifp, m0);
 684
 685         /* xfer packet to user space */
 686         while ((m0 != NULL) && (uio->uio_resid > 0) && (error == 0)) {
 687                 len = min(uio->uio_resid, m0->m_len);
 688                 if (len == 0)
 689                         break;
 690
 691                 error = uiomove(mtod(m0, caddr_t), len, uio);
 692                 m0 = m_free(m0);
 693         }
 694
 695         if (m0 != NULL) {
 696                 TAPDEBUG(ifp, "dropping mbuf, minor = %#x\n",
 697                          minor(tp->tap_dev));
 698                 m_freem(m0);
 699         }
 700
 701         return (error);
 702 }
 703
 704 /*
 705  * tapwrite
 706  *
 707  * The ops write interface - an atomic write is a packet - or else!
 708  *
 709  * Called from the fileops interface with nothing held.
 710  *
 711  * MPSAFE
 712  */
 713 static int
 714 tapwrite(struct dev_write_args *ap)
 715 {
 716         cdev_t dev = ap->a_head.a_dev;
 717         struct uio *uio = ap->a_uio;
 718         struct tap_softc        *tp = dev->si_drv1;
 719         struct ifnet            *ifp = &tp->tap_if;
 720         struct mbuf             *top = NULL, **mp = NULL, *m = NULL;
 721         int                      error = 0, tlen, mlen;
 722
 723         TAPDEBUG(ifp, "writing, minor = %#x\n", minor(tp->tap_dev));
 724
 725         if (uio->uio_resid == 0)
 726                 return (0);
 727
 728         if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
 729                 TAPDEBUG(ifp, "invalid packet len = %d, minor = %#x\n",
 730                          uio->uio_resid, minor(tp->tap_dev));
 731
 732                 return (EIO);
 733         }
 734         tlen = uio->uio_resid;
 735
 736         /* get a header mbuf */
 737         MGETHDR(m, MB_DONTWAIT, MT_DATA);
 738         if (m == NULL)
 739                 return (ENOBUFS);
 740         mlen = MHLEN;
 741
 742         top = 0;
 743         mp = &top;
 744         while ((error == 0) && (uio->uio_resid > 0)) {
 745                 m->m_len = min(mlen, uio->uio_resid);
 746                 error = uiomove(mtod(m, caddr_t), m->m_len, uio);
 747                 *mp = m;
 748                 mp = &m->m_next;
 749                 if (uio->uio_resid > 0) {
 750                         MGET(m, MB_DONTWAIT, MT_DATA);
 751                         if (m == NULL) {
 752                                 error = ENOBUFS;
 753                                 break;
 754                         }
 755                         mlen = MLEN;
 756                 }
 757         }
 758         if (error) {
 759                 ifp->if_ierrors ++;
 760                 if (top)
 761                         m_freem(top);
 762                 return (error);
 763         }
 764
 765         top->m_pkthdr.len = tlen;
 766         top->m_pkthdr.rcvif = ifp;
 767
 768         /*
 769          * Ethernet bridge and bpf are handled in ether_input
 770          *
 771          * adjust mbuf and give packet to the ether_input
 772          */
 773         lwkt_serialize_enter(ifp->if_serializer);
 774         ifp->if_input(ifp, top);
 775         ifp->if_ipackets ++; /* ibytes are counted in ether_input */
 776         lwkt_serialize_exit(ifp->if_serializer);
 777
 778         return (0);
 779 }
 780
 781 /*
 782  * tappoll
 783  *
 784  * The poll interface, this is only useful on reads really. The write
 785  * detect always returns true, write never blocks anyway, it either
 786  * accepts the packet or drops it
 787  *
 788  * Called from the fileops interface with nothing held.
 789  *
 790  * MPSAFE
 791  */
 792 static int
 793 tappoll(struct dev_poll_args *ap)
 794 {
 795         cdev_t dev = ap->a_head.a_dev;
 796         struct tap_softc        *tp = dev->si_drv1;
 797         struct ifnet            *ifp = &tp->tap_if;
 798         int                      revents = 0;
 799
 800         TAPDEBUG(ifp, "polling, minor = %#x\n", minor(tp->tap_dev));
 801
 802         lwkt_serialize_enter(ifp->if_serializer);
 803         if (ap->a_events & (POLLIN | POLLRDNORM)) {
 804                 if (!ifq_is_empty(&ifp->if_snd)) {
 805                         TAPDEBUG(ifp,
 806                                  "has data in queue. minor = %#x\n",
 807                                  minor(tp->tap_dev));
 808
 809                         revents |= (ap->a_events & (POLLIN | POLLRDNORM));
 810                 } else {
 811                         TAPDEBUG(ifp, "waiting for data, minor = %#x\n",
 812                                  minor(tp->tap_dev));
 813
 814                         get_mplock();
 815                         selrecord(curthread, &tp->tap_rsel);
 816                         rel_mplock();
 817                 }
 818         }
 819         lwkt_serialize_exit(ifp->if_serializer);
 820
 821         if (ap->a_events & (POLLOUT | POLLWRNORM))
 822                 revents |= (ap->a_events & (POLLOUT | POLLWRNORM));
 823         ap->a_events = revents;
 824         return(0);
 825 }
 826
 827 /*
 828  * tapkqfilter - called from the fileops interface with nothing held
 829  *
 830  * MPSAFE
 831  */
 832 static int filt_tapread(struct knote *kn, long hint);
 833 static void filt_tapdetach(struct knote *kn);
 834 static struct filterops tapread_filtops =
 835         { 1, NULL, filt_tapdetach, filt_tapread };
 836
 837 static int
 838 tapkqfilter(struct dev_kqfilter_args *ap)
 839 {
 840         cdev_t dev = ap->a_head.a_dev;
 841         struct knote *kn = ap->a_kn;
 842         struct tap_softc *tp;
 843         struct klist *list;
 844         struct ifnet *ifp;
 845
 846         get_mplock();
 847         tp = dev->si_drv1;
 848         ifp = &tp->tap_if;
 849         ap->a_result =0;
 850
 851         switch(kn->kn_filter) {
 852         case EVFILT_READ:
 853                 list = &tp->tap_rsel.si_note;
 854                 kn->kn_fop = &tapread_filtops;
 855                 kn->kn_hook = (void *)tp;
 856                 break;
 857         case EVFILT_WRITE:
 858                 /* fall through */
 859         default:
 860                 ap->a_result = 1;
 861                 rel_mplock();
 862                 return(0);
 863         }
 864         crit_enter();
 865         SLIST_INSERT_HEAD(list, kn, kn_selnext);
 866         crit_exit();
 867         rel_mplock();
 868         return(0);
 869 }
 870
 871 static int
 872 filt_tapread(struct knote *kn, long hint)
 873 {
 874         struct tap_softc *tp = (void *)kn->kn_hook;
 875         struct ifnet *ifp = &tp->tap_if;
 876
 877         if (ifq_is_empty(&ifp->if_snd) == 0) {
 878                 return(1);
 879         } else {
 880                 return(0);
 881         }
 882 }
 883
 884 static void
 885 filt_tapdetach(struct knote *kn)
 886 {
 887         struct tap_softc *tp = (void *)kn->kn_hook;
 888
 889         SLIST_REMOVE(&tp->tap_rsel.si_note, kn, knote, kn_selnext);
 890 }