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 1.39 2008/05/18 02:40:41 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
  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) {
 326                 if (ifp->if_flags & IFF_UP) {
 327                         lwkt_serialize_enter(ifp->if_serializer);
 328                         if_down(ifp);
 329                         lwkt_serialize_exit(ifp->if_serializer);
 330                 }
 331                 ifp->if_flags &= ~IFF_RUNNING;
 332         }
 333         if_purgeaddrs_nolink(ifp);
 334
 335         EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
 336
 337         /* Announce the departure of the interface. */
 338         rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
 339
 340         funsetown(tp->tap_sigio);
 341         selwakeup(&tp->tap_rsel);
 342
 343         tp->tap_flags &= ~TAP_OPEN;
 344         tp->tap_td = NULL;
 345
 346         taprefcnt --;
 347         if (taprefcnt < 0) {
 348                 taprefcnt = 0;
 349                 if_printf(ifp, "minor = %#x, refcnt = %d is out of sync. "
 350                         "set refcnt to 0\n", minor(tp->tap_dev), taprefcnt);
 351         }
 352
 353         TAPDEBUG(ifp, "closed. minor = %#x, refcnt = %d, taplastunit = %d\n",
 354                  minor(tp->tap_dev), taprefcnt, taplastunit);
 355
 356         rel_mplock();
 357         return (0);
 358 }
 359
 360
 361 /*
 362  * tapifinit
 363  *
 364  * Network interface initialization function (called with if serializer held)
 365  *
 366  * MPSAFE
 367  */
 368 static void
 369 tapifinit(void *xtp)
 370 {
 371         struct tap_softc        *tp = (struct tap_softc *)xtp;
 372         struct ifnet            *ifp = &tp->tap_if;
 373
 374         TAPDEBUG(ifp, "initializing, minor = %#x\n", minor(tp->tap_dev));
 375
 376         ifp->if_flags |= IFF_RUNNING;
 377         ifp->if_flags &= ~IFF_OACTIVE;
 378
 379         /* attempt to start output */
 380         tapifstart(ifp);
 381 }
 382
 383
 384 /*
 385  * tapifioctl
 386  *
 387  * Process an ioctl request on network interface (called with if serializer
 388  * held).
 389  *
 390  * MPSAFE
 391  */
 392 static int
 393 tapifioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
 394 {
 395         struct tap_softc        *tp = (struct tap_softc *)(ifp->if_softc);
 396         struct ifstat           *ifs = NULL;
 397         int                      dummy;
 398
 399         switch (cmd) {
 400                 case SIOCSIFADDR:
 401                 case SIOCGIFADDR:
 402                 case SIOCSIFMTU:
 403                         dummy = ether_ioctl(ifp, cmd, data);
 404                         return (dummy);
 405
 406                 case SIOCSIFFLAGS:
 407                         if ((tp->tap_flags & TAP_VMNET) == 0) {
 408                                 /*
 409                                  * Only for non-vmnet tap(4)
 410                                  */
 411                                 if (ifp->if_flags & IFF_UP) {
 412                                         if ((ifp->if_flags & IFF_RUNNING) == 0)
 413                                                 tapifinit(tp);
 414                                 } else {
 415                                         /* We don't have a tapifstop() */
 416                                         ifp->if_flags &= ~IFF_RUNNING;
 417                                 }
 418                         }
 419                         break;
 420                 case SIOCADDMULTI: /* XXX -- just like vmnet does */
 421                 case SIOCDELMULTI:
 422                         break;
 423
 424                 case SIOCGIFSTATUS:
 425                         ifs = (struct ifstat *)data;
 426                         dummy = strlen(ifs->ascii);
 427                         if (tp->tap_td != NULL && dummy < sizeof(ifs->ascii)) {
 428                                 if (tp->tap_td->td_proc) {
 429                                     ksnprintf(ifs->ascii + dummy,
 430                                         sizeof(ifs->ascii) - dummy,
 431                                         "\tOpened by pid %d\n",
 432                                         (int)tp->tap_td->td_proc->p_pid);
 433                                 } else {
 434                                     ksnprintf(ifs->ascii + dummy,
 435                                         sizeof(ifs->ascii) - dummy,
 436                                         "\tOpened by td %p\n", tp->tap_td);
 437                                 }
 438                         }
 439                         break;
 440
 441                 default:
 442                         return (EINVAL);
 443         }
 444
 445         return (0);
 446 }
 447
 448
 449 /*
 450  * tapifstart
 451  *
 452  * Queue packets from higher level ready to put out (called with if serializer
 453  * held)
 454  *
 455  * MPSAFE
 456  */
 457 static void
 458 tapifstart(struct ifnet *ifp)
 459 {
 460         struct tap_softc        *tp = ifp->if_softc;
 461
 462         TAPDEBUG(ifp, "starting, minor = %#x\n", minor(tp->tap_dev));
 463
 464         /*
 465          * do not junk pending output if we are in VMnet mode.
 466          * XXX: can this do any harm because of queue overflow?
 467          */
 468
 469         if (((tp->tap_flags & TAP_VMNET) == 0) &&
 470             ((tp->tap_flags & TAP_READY) != TAP_READY)) {
 471                 TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
 472                          minor(tp->tap_dev), tp->tap_flags);
 473
 474                 ifq_purge(&ifp->if_snd);
 475                 return;
 476         }
 477
 478         ifp->if_flags |= IFF_OACTIVE;
 479
 480         if (!ifq_is_empty(&ifp->if_snd)) {
 481                 if (tp->tap_flags & TAP_RWAIT) {
 482                         tp->tap_flags &= ~TAP_RWAIT;
 483                         wakeup((caddr_t)tp);
 484                 }
 485
 486                 get_mplock();
 487                 KNOTE(&tp->tap_rsel.si_note, 0);
 488                 rel_mplock();
 489
 490                 if ((tp->tap_flags & TAP_ASYNC) && (tp->tap_sigio != NULL)) {
 491                         get_mplock();
 492                         pgsigio(tp->tap_sigio, SIGIO, 0);
 493                         rel_mplock();
 494                 }
 495
 496                 /*
 497                  * selwakeup is not MPSAFE.  tapifstart is.
 498                  */
 499                 get_mplock();
 500                 selwakeup(&tp->tap_rsel);
 501                 rel_mplock();
 502                 ifp->if_opackets ++; /* obytes are counted in ether_output */
 503         }
 504
 505         ifp->if_flags &= ~IFF_OACTIVE;
 506 }
 507
 508
 509 /*
 510  * tapioctl
 511  *
 512  * The ops interface is now pretty minimal.  Called via fileops with nothing
 513  * held.
 514  *
 515  * MPSAFE
 516  */
 517 static int
 518 tapioctl(struct dev_ioctl_args *ap)
 519 {
 520         cdev_t dev = ap->a_head.a_dev;
 521         caddr_t data = ap->a_data;
 522         struct tap_softc        *tp = dev->si_drv1;
 523         struct ifnet            *ifp = &tp->tap_if;
 524         struct tapinfo          *tapp = NULL;
 525         struct mbuf *mb;
 526         short f;
 527         int error;
 528
 529         lwkt_serialize_enter(ifp->if_serializer);
 530         error = 0;
 531
 532         switch (ap->a_cmd) {
 533         case TAPSIFINFO:
 534                 tapp = (struct tapinfo *)data;
 535                 ifp->if_mtu = tapp->mtu;
 536                 ifp->if_type = tapp->type;
 537                 ifp->if_baudrate = tapp->baudrate;
 538                 break;
 539
 540         case TAPGIFINFO:
 541                 tapp = (struct tapinfo *)data;
 542                 tapp->mtu = ifp->if_mtu;
 543                 tapp->type = ifp->if_type;
 544                 tapp->baudrate = ifp->if_baudrate;
 545                 break;
 546
 547         case TAPSDEBUG:
 548                 tapdebug = *(int *)data;
 549                 break;
 550
 551         case TAPGDEBUG:
 552                 *(int *)data = tapdebug;
 553                 break;
 554
 555         case FIOASYNC:
 556                 if (*(int *)data)
 557                         tp->tap_flags |= TAP_ASYNC;
 558                 else
 559                         tp->tap_flags &= ~TAP_ASYNC;
 560                 break;
 561
 562         case FIONREAD:
 563                 *(int *)data = 0;
 564                 if ((mb = ifq_poll(&ifp->if_snd)) != NULL) {
 565                         for(; mb != NULL; mb = mb->m_next)
 566                                 *(int *)data += mb->m_len;
 567                 }
 568                 break;
 569
 570         case FIOSETOWN:
 571                 error = fsetown(*(int *)data, &tp->tap_sigio);
 572                 break;
 573
 574         case FIOGETOWN:
 575                 *(int *)data = fgetown(tp->tap_sigio);
 576                 break;
 577
 578         /* this is deprecated, FIOSETOWN should be used instead */
 579         case TIOCSPGRP:
 580                 error = fsetown(-(*(int *)data), &tp->tap_sigio);
 581                 break;
 582
 583         /* this is deprecated, FIOGETOWN should be used instead */
 584         case TIOCGPGRP:
 585                 *(int *)data = -fgetown(tp->tap_sigio);
 586                 break;
 587
 588         /* VMware/VMnet port ioctl's */
 589
 590         case SIOCGIFFLAGS:      /* get ifnet flags */
 591                 bcopy(&ifp->if_flags, data, sizeof(ifp->if_flags));
 592                 break;
 593
 594         case VMIO_SIOCSIFFLAGS: /* VMware/VMnet SIOCSIFFLAGS */
 595                 f = *(short *)data;
 596                 f &= 0x0fff;
 597                 f &= ~IFF_CANTCHANGE;
 598                 f |= IFF_UP;
 599                 ifp->if_flags = f | (ifp->if_flags & IFF_CANTCHANGE);
 600                 break;
 601
 602         case OSIOCGIFADDR:      /* get MAC address of the remote side */
 603         case SIOCGIFADDR:
 604                 bcopy(tp->ether_addr, data, sizeof(tp->ether_addr));
 605                 break;
 606
 607         case SIOCSIFADDR:       /* set MAC address of the remote side */
 608                 bcopy(data, tp->ether_addr, sizeof(tp->ether_addr));
 609                 break;
 610
 611         default:
 612                 error = ENOTTY;
 613                 break;
 614         }
 615         lwkt_serialize_exit(ifp->if_serializer);
 616         return (error);
 617 }
 618
 619
 620 /*
 621  * tapread
 622  *
 623  * The ops read interface - reads a packet at a time, or at
 624  * least as much of a packet as can be read.
 625  *
 626  * Called from the fileops interface with nothing held.
 627  *
 628  * MPSAFE
 629  */
 630 static int
 631 tapread(struct dev_read_args *ap)
 632 {
 633         cdev_t dev = ap->a_head.a_dev;
 634         struct uio *uio = ap->a_uio;
 635         struct tap_softc        *tp = dev->si_drv1;
 636         struct ifnet            *ifp = &tp->tap_if;
 637         struct mbuf             *m0 = NULL;
 638         int                      error = 0, len;
 639
 640         TAPDEBUG(ifp, "reading, minor = %#x\n", minor(tp->tap_dev));
 641
 642         if ((tp->tap_flags & TAP_READY) != TAP_READY) {
 643                 TAPDEBUG(ifp, "not ready. minor = %#x, tap_flags = 0x%x\n",
 644                          minor(tp->tap_dev), tp->tap_flags);
 645
 646                 return (EHOSTDOWN);
 647         }
 648
 649         tp->tap_flags &= ~TAP_RWAIT;
 650
 651         /* sleep until we get a packet */
 652         do {
 653                 lwkt_serialize_enter(ifp->if_serializer);
 654                 m0 = ifq_dequeue(&ifp->if_snd, NULL);
 655                 if (m0 == NULL) {
 656                         if (ap->a_ioflag & IO_NDELAY) {
 657                                 lwkt_serialize_exit(ifp->if_serializer);
 658                                 return (EWOULDBLOCK);
 659                         }
 660                         tp->tap_flags |= TAP_RWAIT;
 661                         crit_enter();
 662                         tsleep_interlock(tp);
 663                         lwkt_serialize_exit(ifp->if_serializer);
 664                         error = tsleep(tp, PCATCH, "taprd", 0);
 665                         crit_exit();
 666                         if (error)
 667                                 return (error);
 668                 } else {
 669                         lwkt_serialize_exit(ifp->if_serializer);
 670                 }
 671         } while (m0 == NULL);
 672
 673         BPF_MTAP(ifp, m0);
 674
 675         /* xfer packet to user space */
 676         while ((m0 != NULL) && (uio->uio_resid > 0) && (error == 0)) {
 677                 len = min(uio->uio_resid, m0->m_len);
 678                 if (len == 0)
 679                         break;
 680
 681                 error = uiomove(mtod(m0, caddr_t), len, uio);
 682                 m0 = m_free(m0);
 683         }
 684
 685         if (m0 != NULL) {
 686                 TAPDEBUG(ifp, "dropping mbuf, minor = %#x\n",
 687                          minor(tp->tap_dev));
 688                 m_freem(m0);
 689         }
 690
 691         return (error);
 692 }
 693
 694 /*
 695  * tapwrite
 696  *
 697  * The ops write interface - an atomic write is a packet - or else!
 698  *
 699  * Called from the fileops interface with nothing held.
 700  *
 701  * MPSAFE
 702  */
 703 static int
 704 tapwrite(struct dev_write_args *ap)
 705 {
 706         cdev_t dev = ap->a_head.a_dev;
 707         struct uio *uio = ap->a_uio;
 708         struct tap_softc        *tp = dev->si_drv1;
 709         struct ifnet            *ifp = &tp->tap_if;
 710         struct mbuf             *top = NULL, **mp = NULL, *m = NULL;
 711         int                      error = 0, tlen, mlen;
 712
 713         TAPDEBUG(ifp, "writing, minor = %#x\n", minor(tp->tap_dev));
 714
 715         if (uio->uio_resid == 0)
 716                 return (0);
 717
 718         if ((uio->uio_resid < 0) || (uio->uio_resid > TAPMRU)) {
 719                 TAPDEBUG(ifp, "invalid packet len = %d, minor = %#x\n",
 720                          uio->uio_resid, minor(tp->tap_dev));
 721
 722                 return (EIO);
 723         }
 724         tlen = uio->uio_resid;
 725
 726         /* get a header mbuf */
 727         MGETHDR(m, MB_DONTWAIT, MT_DATA);
 728         if (m == NULL)
 729                 return (ENOBUFS);
 730         mlen = MHLEN;
 731
 732         top = 0;
 733         mp = &top;
 734         while ((error == 0) && (uio->uio_resid > 0)) {
 735                 m->m_len = min(mlen, uio->uio_resid);
 736                 error = uiomove(mtod(m, caddr_t), m->m_len, uio);
 737                 *mp = m;
 738                 mp = &m->m_next;
 739                 if (uio->uio_resid > 0) {
 740                         MGET(m, MB_DONTWAIT, MT_DATA);
 741                         if (m == NULL) {
 742                                 error = ENOBUFS;
 743                                 break;
 744                         }
 745                         mlen = MLEN;
 746                 }
 747         }
 748         if (error) {
 749                 ifp->if_ierrors ++;
 750                 if (top)
 751                         m_freem(top);
 752                 return (error);
 753         }
 754
 755         top->m_pkthdr.len = tlen;
 756         top->m_pkthdr.rcvif = ifp;
 757
 758         /*
 759          * Ethernet bridge and bpf are handled in ether_input
 760          *
 761          * adjust mbuf and give packet to the ether_input
 762          */
 763         lwkt_serialize_enter(ifp->if_serializer);
 764         ifp->if_input(ifp, top);
 765         ifp->if_ipackets ++; /* ibytes are counted in ether_input */
 766         lwkt_serialize_exit(ifp->if_serializer);
 767
 768         return (0);
 769 }
 770
 771 /*
 772  * tappoll
 773  *
 774  * The poll interface, this is only useful on reads really. The write
 775  * detect always returns true, write never blocks anyway, it either
 776  * accepts the packet or drops it
 777  *
 778  * Called from the fileops interface with nothing held.
 779  *
 780  * MPSAFE
 781  */
 782 static int
 783 tappoll(struct dev_poll_args *ap)
 784 {
 785         cdev_t dev = ap->a_head.a_dev;
 786         struct tap_softc        *tp = dev->si_drv1;
 787         struct ifnet            *ifp = &tp->tap_if;
 788         int                      revents = 0;
 789
 790         TAPDEBUG(ifp, "polling, minor = %#x\n", minor(tp->tap_dev));
 791
 792         lwkt_serialize_enter(ifp->if_serializer);
 793         if (ap->a_events & (POLLIN | POLLRDNORM)) {
 794                 if (!ifq_is_empty(&ifp->if_snd)) {
 795                         TAPDEBUG(ifp,
 796                                  "has data in queue. minor = %#x\n",
 797                                  minor(tp->tap_dev));
 798
 799                         revents |= (ap->a_events & (POLLIN | POLLRDNORM));
 800                 } else {
 801                         TAPDEBUG(ifp, "waiting for data, minor = %#x\n",
 802                                  minor(tp->tap_dev));
 803
 804                         get_mplock();
 805                         selrecord(curthread, &tp->tap_rsel);
 806                         rel_mplock();
 807                 }
 808         }
 809         lwkt_serialize_exit(ifp->if_serializer);
 810
 811         if (ap->a_events & (POLLOUT | POLLWRNORM))
 812                 revents |= (ap->a_events & (POLLOUT | POLLWRNORM));
 813         ap->a_events = revents;
 814         return(0);
 815 }
 816
 817 /*
 818  * tapkqfilter - called from the fileops interface with nothing held
 819  *
 820  * MPSAFE
 821  */
 822 static int filt_tapread(struct knote *kn, long hint);
 823 static void filt_tapdetach(struct knote *kn);
 824 static struct filterops tapread_filtops =
 825         { 1, NULL, filt_tapdetach, filt_tapread };
 826
 827 static int
 828 tapkqfilter(struct dev_kqfilter_args *ap)
 829 {
 830         cdev_t dev = ap->a_head.a_dev;
 831         struct knote *kn = ap->a_kn;
 832         struct tap_softc *tp;
 833         struct klist *list;
 834         struct ifnet *ifp;
 835
 836         get_mplock();
 837         tp = dev->si_drv1;
 838         ifp = &tp->tap_if;
 839         ap->a_result =0;
 840
 841         switch(kn->kn_filter) {
 842         case EVFILT_READ:
 843                 list = &tp->tap_rsel.si_note;
 844                 kn->kn_fop = &tapread_filtops;
 845                 kn->kn_hook = (void *)tp;
 846                 break;
 847         case EVFILT_WRITE:
 848                 /* fall through */
 849         default:
 850                 ap->a_result = 1;
 851                 rel_mplock();
 852                 return(0);
 853         }
 854         crit_enter();
 855         SLIST_INSERT_HEAD(list, kn, kn_selnext);
 856         crit_exit();
 857         rel_mplock();
 858         return(0);
 859 }
 860
 861 static int
 862 filt_tapread(struct knote *kn, long hint)
 863 {
 864         struct tap_softc *tp = (void *)kn->kn_hook;
 865         struct ifnet *ifp = &tp->tap_if;
 866
 867         if (ifq_is_empty(&ifp->if_snd) == 0) {
 868                 return(1);
 869         } else {
 870                 return(0);
 871         }
 872 }
 873
 874 static void
 875 filt_tapdetach(struct knote *kn)
 876 {
 877         struct tap_softc *tp = (void *)kn->kn_hook;
 878
 879         SLIST_REMOVE(&tp->tap_rsel.si_note, kn, knote, kn_selnext);
 880 }