1 /* $NetBSD: if_tun.c,v 1.14 1994/06/29 06:36:25 cgd Exp $ */ 2 3 /*- 4 * Copyright (c) 1988, Julian Onions <[email protected]> 5 * Nottingham University 1987. 6 * 7 * This source may be freely distributed, however I would be interested 8 * in any changes that are made. 9 * 10 * This driver takes packets off the IP i/f and hands them up to a 11 * user process to have its wicked way with. This driver has it's 12 * roots in a similar driver written by Phil Cockcroft (formerly) at 13 * UCL. This driver is based much more on read/write/poll mode of 14 * operation though. 15 * 16 * $FreeBSD$ 17 */ 18 19 #include "opt_inet.h" 20 #include "opt_inet6.h" 21 22 #include <sys/param.h> 23 #include <sys/priv.h> 24 #include <sys/proc.h> 25 #include <sys/systm.h> 26 #include <sys/jail.h> 27 #include <sys/mbuf.h> 28 #include <sys/module.h> 29 #include <sys/socket.h> 30 #include <sys/fcntl.h> 31 #include <sys/filio.h> 32 #include <sys/sockio.h> 33 #include <sys/ttycom.h> 34 #include <sys/poll.h> 35 #include <sys/selinfo.h> 36 #include <sys/signalvar.h> 37 #include <sys/filedesc.h> 38 #include <sys/kernel.h> 39 #include <sys/sysctl.h> 40 #include <sys/conf.h> 41 #include <sys/uio.h> 42 #include <sys/malloc.h> 43 #include <sys/random.h> 44 45 #include <net/if.h> 46 #include <net/if_var.h> 47 #include <net/if_clone.h> 48 #include <net/if_types.h> 49 #include <net/netisr.h> 50 #include <net/route.h> 51 #include <net/vnet.h> 52 #ifdef INET 53 #include <netinet/in.h> 54 #endif 55 #include <net/bpf.h> 56 #include <net/if_tun.h> 57 58 #include <sys/queue.h> 59 #include <sys/condvar.h> 60 61 #include <security/mac/mac_framework.h> 62 63 /* 64 * tun_list is protected by global tunmtx. Other mutable fields are 65 * protected by tun->tun_mtx, or by their owning subsystem. tun_dev is 66 * static for the duration of a tunnel interface. 67 */ 68 struct tun_softc { 69 TAILQ_ENTRY(tun_softc) tun_list; 70 struct cdev *tun_dev; 71 u_short tun_flags; /* misc flags */ 72 #define TUN_OPEN 0x0001 73 #define TUN_INITED 0x0002 74 #define TUN_RCOLL 0x0004 75 #define TUN_IASET 0x0008 76 #define TUN_DSTADDR 0x0010 77 #define TUN_LMODE 0x0020 78 #define TUN_RWAIT 0x0040 79 #define TUN_ASYNC 0x0080 80 #define TUN_IFHEAD 0x0100 81 82 #define TUN_READY (TUN_OPEN | TUN_INITED) 83 84 /* 85 * XXXRW: tun_pid is used to exclusively lock /dev/tun. Is this 86 * actually needed? Can we just return EBUSY if already open? 87 * Problem is that this involved inherent races when a tun device 88 * is handed off from one process to another, as opposed to just 89 * being slightly stale informationally. 90 */ 91 pid_t tun_pid; /* owning pid */ 92 struct ifnet *tun_ifp; /* the interface */ 93 struct sigio *tun_sigio; /* information for async I/O */ 94 struct selinfo tun_rsel; /* read select */ 95 struct mtx tun_mtx; /* protect mutable softc fields */ 96 struct cv tun_cv; /* protect against ref'd dev destroy */ 97 }; 98 #define TUN2IFP(sc) ((sc)->tun_ifp) 99 100 #define TUNDEBUG if (tundebug) if_printf 101 102 /* 103 * All mutable global variables in if_tun are locked using tunmtx, with 104 * the exception of tundebug, which is used unlocked, and tunclones, 105 * which is static after setup. 106 */ 107 static struct mtx tunmtx; 108 static const char tunname[] = "tun"; 109 static MALLOC_DEFINE(M_TUN, tunname, "Tunnel Interface"); 110 static int tundebug = 0; 111 static int tundclone = 1; 112 static struct clonedevs *tunclones; 113 static TAILQ_HEAD(,tun_softc) tunhead = TAILQ_HEAD_INITIALIZER(tunhead); 114 SYSCTL_INT(_debug, OID_AUTO, if_tun_debug, CTLFLAG_RW, &tundebug, 0, ""); 115 116 SYSCTL_DECL(_net_link); 117 static SYSCTL_NODE(_net_link, OID_AUTO, tun, CTLFLAG_RW, 0, 118 "IP tunnel software network interface."); 119 SYSCTL_INT(_net_link_tun, OID_AUTO, devfs_cloning, CTLFLAG_RWTUN, &tundclone, 0, 120 "Enable legacy devfs interface creation."); 121 122 static void tunclone(void *arg, struct ucred *cred, char *name, 123 int namelen, struct cdev **dev); 124 static void tuncreate(const char *name, struct cdev *dev); 125 static int tunifioctl(struct ifnet *, u_long, caddr_t); 126 static void tuninit(struct ifnet *); 127 static int tunmodevent(module_t, int, void *); 128 static int tunoutput(struct ifnet *, struct mbuf *, 129 const struct sockaddr *, struct route *ro); 130 static void tunstart(struct ifnet *); 131 132 static int tun_clone_create(struct if_clone *, int, caddr_t); 133 static void tun_clone_destroy(struct ifnet *); 134 static struct if_clone *tun_cloner; 135 136 static d_open_t tunopen; 137 static d_close_t tunclose; 138 static d_read_t tunread; 139 static d_write_t tunwrite; 140 static d_ioctl_t tunioctl; 141 static d_poll_t tunpoll; 142 static d_kqfilter_t tunkqfilter; 143 144 static int tunkqread(struct knote *, long); 145 static int tunkqwrite(struct knote *, long); 146 static void tunkqdetach(struct knote *); 147 148 static struct filterops tun_read_filterops = { 149 .f_isfd = 1, 150 .f_attach = NULL, 151 .f_detach = tunkqdetach, 152 .f_event = tunkqread, 153 }; 154 155 static struct filterops tun_write_filterops = { 156 .f_isfd = 1, 157 .f_attach = NULL, 158 .f_detach = tunkqdetach, 159 .f_event = tunkqwrite, 160 }; 161 162 static struct cdevsw tun_cdevsw = { 163 .d_version = D_VERSION, 164 .d_flags = D_NEEDMINOR, 165 .d_open = tunopen, 166 .d_close = tunclose, 167 .d_read = tunread, 168 .d_write = tunwrite, 169 .d_ioctl = tunioctl, 170 .d_poll = tunpoll, 171 .d_kqfilter = tunkqfilter, 172 .d_name = tunname, 173 }; 174 175 static int 176 tun_clone_create(struct if_clone *ifc, int unit, caddr_t params) 177 { 178 struct cdev *dev; 179 int i; 180 181 /* find any existing device, or allocate new unit number */ 182 i = clone_create(&tunclones, &tun_cdevsw, &unit, &dev, 0); 183 if (i) { 184 /* No preexisting struct cdev *, create one */ 185 dev = make_dev(&tun_cdevsw, unit, 186 UID_UUCP, GID_DIALER, 0600, "%s%d", tunname, unit); 187 } 188 tuncreate(tunname, dev); 189 190 return (0); 191 } 192 193 static void 194 tunclone(void *arg, struct ucred *cred, char *name, int namelen, 195 struct cdev **dev) 196 { 197 char devname[SPECNAMELEN + 1]; 198 int u, i, append_unit; 199 200 if (*dev != NULL) 201 return; 202 203 /* 204 * If tun cloning is enabled, only the superuser can create an 205 * interface. 206 */ 207 if (!tundclone || priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0) 208 return; 209 210 if (strcmp(name, tunname) == 0) { 211 u = -1; 212 } else if (dev_stdclone(name, NULL, tunname, &u) != 1) 213 return; /* Don't recognise the name */ 214 if (u != -1 && u > IF_MAXUNIT) 215 return; /* Unit number too high */ 216 217 if (u == -1) 218 append_unit = 1; 219 else 220 append_unit = 0; 221 222 CURVNET_SET(CRED_TO_VNET(cred)); 223 /* find any existing device, or allocate new unit number */ 224 i = clone_create(&tunclones, &tun_cdevsw, &u, dev, 0); 225 if (i) { 226 if (append_unit) { 227 namelen = snprintf(devname, sizeof(devname), "%s%d", 228 name, u); 229 name = devname; 230 } 231 /* No preexisting struct cdev *, create one */ 232 *dev = make_dev_credf(MAKEDEV_REF, &tun_cdevsw, u, cred, 233 UID_UUCP, GID_DIALER, 0600, "%s", name); 234 } 235 236 if_clone_create(name, namelen, NULL); 237 CURVNET_RESTORE(); 238 } 239 240 static void 241 tun_destroy(struct tun_softc *tp) 242 { 243 struct cdev *dev; 244 245 mtx_lock(&tp->tun_mtx); 246 if ((tp->tun_flags & TUN_OPEN) != 0) 247 cv_wait_unlock(&tp->tun_cv, &tp->tun_mtx); 248 else 249 mtx_unlock(&tp->tun_mtx); 250 251 CURVNET_SET(TUN2IFP(tp)->if_vnet); 252 dev = tp->tun_dev; 253 bpfdetach(TUN2IFP(tp)); 254 if_detach(TUN2IFP(tp)); 255 if_free(TUN2IFP(tp)); 256 destroy_dev(dev); 257 seldrain(&tp->tun_rsel); 258 knlist_clear(&tp->tun_rsel.si_note, 0); 259 knlist_destroy(&tp->tun_rsel.si_note); 260 mtx_destroy(&tp->tun_mtx); 261 cv_destroy(&tp->tun_cv); 262 free(tp, M_TUN); 263 CURVNET_RESTORE(); 264 } 265 266 static void 267 tun_clone_destroy(struct ifnet *ifp) 268 { 269 struct tun_softc *tp = ifp->if_softc; 270 271 mtx_lock(&tunmtx); 272 TAILQ_REMOVE(&tunhead, tp, tun_list); 273 mtx_unlock(&tunmtx); 274 tun_destroy(tp); 275 } 276 277 static int 278 tunmodevent(module_t mod, int type, void *data) 279 { 280 static eventhandler_tag tag; 281 struct tun_softc *tp; 282 283 switch (type) { 284 case MOD_LOAD: 285 mtx_init(&tunmtx, "tunmtx", NULL, MTX_DEF); 286 clone_setup(&tunclones); 287 tag = EVENTHANDLER_REGISTER(dev_clone, tunclone, 0, 1000); 288 if (tag == NULL) 289 return (ENOMEM); 290 tun_cloner = if_clone_simple(tunname, tun_clone_create, 291 tun_clone_destroy, 0); 292 break; 293 case MOD_UNLOAD: 294 if_clone_detach(tun_cloner); 295 EVENTHANDLER_DEREGISTER(dev_clone, tag); 296 drain_dev_clone_events(); 297 298 mtx_lock(&tunmtx); 299 while ((tp = TAILQ_FIRST(&tunhead)) != NULL) { 300 TAILQ_REMOVE(&tunhead, tp, tun_list); 301 mtx_unlock(&tunmtx); 302 tun_destroy(tp); 303 mtx_lock(&tunmtx); 304 } 305 mtx_unlock(&tunmtx); 306 clone_cleanup(&tunclones); 307 mtx_destroy(&tunmtx); 308 break; 309 default: 310 return EOPNOTSUPP; 311 } 312 return 0; 313 } 314 315 static moduledata_t tun_mod = { 316 "if_tun", 317 tunmodevent, 318 0 319 }; 320 321 DECLARE_MODULE(if_tun, tun_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 322 MODULE_VERSION(if_tun, 1); 323 324 static void 325 tunstart(struct ifnet *ifp) 326 { 327 struct tun_softc *tp = ifp->if_softc; 328 struct mbuf *m; 329 330 TUNDEBUG(ifp,"%s starting\n", ifp->if_xname); 331 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 332 IFQ_LOCK(&ifp->if_snd); 333 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 334 if (m == NULL) { 335 IFQ_UNLOCK(&ifp->if_snd); 336 return; 337 } 338 IFQ_UNLOCK(&ifp->if_snd); 339 } 340 341 mtx_lock(&tp->tun_mtx); 342 if (tp->tun_flags & TUN_RWAIT) { 343 tp->tun_flags &= ~TUN_RWAIT; 344 wakeup(tp); 345 } 346 selwakeuppri(&tp->tun_rsel, PZERO + 1); 347 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0); 348 if (tp->tun_flags & TUN_ASYNC && tp->tun_sigio) { 349 mtx_unlock(&tp->tun_mtx); 350 pgsigio(&tp->tun_sigio, SIGIO, 0); 351 } else 352 mtx_unlock(&tp->tun_mtx); 353 } 354 355 /* XXX: should return an error code so it can fail. */ 356 static void 357 tuncreate(const char *name, struct cdev *dev) 358 { 359 struct tun_softc *sc; 360 struct ifnet *ifp; 361 362 sc = malloc(sizeof(*sc), M_TUN, M_WAITOK | M_ZERO); 363 mtx_init(&sc->tun_mtx, "tun_mtx", NULL, MTX_DEF); 364 cv_init(&sc->tun_cv, "tun_condvar"); 365 sc->tun_flags = TUN_INITED; 366 sc->tun_dev = dev; 367 mtx_lock(&tunmtx); 368 TAILQ_INSERT_TAIL(&tunhead, sc, tun_list); 369 mtx_unlock(&tunmtx); 370 371 ifp = sc->tun_ifp = if_alloc(IFT_PPP); 372 if (ifp == NULL) 373 panic("%s%d: failed to if_alloc() interface.\n", 374 name, dev2unit(dev)); 375 if_initname(ifp, name, dev2unit(dev)); 376 ifp->if_mtu = TUNMTU; 377 ifp->if_ioctl = tunifioctl; 378 ifp->if_output = tunoutput; 379 ifp->if_start = tunstart; 380 ifp->if_flags = IFF_POINTOPOINT | IFF_MULTICAST; 381 ifp->if_softc = sc; 382 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 383 ifp->if_snd.ifq_drv_maxlen = 0; 384 IFQ_SET_READY(&ifp->if_snd); 385 knlist_init_mtx(&sc->tun_rsel.si_note, &sc->tun_mtx); 386 ifp->if_capabilities |= IFCAP_LINKSTATE; 387 ifp->if_capenable |= IFCAP_LINKSTATE; 388 389 if_attach(ifp); 390 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); 391 dev->si_drv1 = sc; 392 TUNDEBUG(ifp, "interface %s is created, minor = %#x\n", 393 ifp->if_xname, dev2unit(dev)); 394 } 395 396 static int 397 tunopen(struct cdev *dev, int flag, int mode, struct thread *td) 398 { 399 struct ifnet *ifp; 400 struct tun_softc *tp; 401 402 /* 403 * XXXRW: Non-atomic test and set of dev->si_drv1 requires 404 * synchronization. 405 */ 406 tp = dev->si_drv1; 407 if (!tp) { 408 tuncreate(tunname, dev); 409 tp = dev->si_drv1; 410 } 411 412 /* 413 * XXXRW: This use of tun_pid is subject to error due to the 414 * fact that a reference to the tunnel can live beyond the 415 * death of the process that created it. Can we replace this 416 * with a simple busy flag? 417 */ 418 mtx_lock(&tp->tun_mtx); 419 if (tp->tun_pid != 0 && tp->tun_pid != td->td_proc->p_pid) { 420 mtx_unlock(&tp->tun_mtx); 421 return (EBUSY); 422 } 423 tp->tun_pid = td->td_proc->p_pid; 424 425 tp->tun_flags |= TUN_OPEN; 426 ifp = TUN2IFP(tp); 427 if_link_state_change(ifp, LINK_STATE_UP); 428 TUNDEBUG(ifp, "open\n"); 429 mtx_unlock(&tp->tun_mtx); 430 431 return (0); 432 } 433 434 /* 435 * tunclose - close the device - mark i/f down & delete 436 * routing info 437 */ 438 static int 439 tunclose(struct cdev *dev, int foo, int bar, struct thread *td) 440 { 441 struct tun_softc *tp; 442 struct ifnet *ifp; 443 444 tp = dev->si_drv1; 445 ifp = TUN2IFP(tp); 446 447 mtx_lock(&tp->tun_mtx); 448 tp->tun_flags &= ~TUN_OPEN; 449 tp->tun_pid = 0; 450 451 /* 452 * junk all pending output 453 */ 454 CURVNET_SET(ifp->if_vnet); 455 IFQ_PURGE(&ifp->if_snd); 456 457 if (ifp->if_flags & IFF_UP) { 458 mtx_unlock(&tp->tun_mtx); 459 if_down(ifp); 460 mtx_lock(&tp->tun_mtx); 461 } 462 463 /* Delete all addresses and routes which reference this interface. */ 464 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 465 struct ifaddr *ifa; 466 467 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 468 mtx_unlock(&tp->tun_mtx); 469 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 470 /* deal w/IPv4 PtP destination; unlocked read */ 471 if (ifa->ifa_addr->sa_family == AF_INET) { 472 rtinit(ifa, (int)RTM_DELETE, 473 tp->tun_flags & TUN_DSTADDR ? RTF_HOST : 0); 474 } else { 475 rtinit(ifa, (int)RTM_DELETE, 0); 476 } 477 } 478 if_purgeaddrs(ifp); 479 mtx_lock(&tp->tun_mtx); 480 } 481 if_link_state_change(ifp, LINK_STATE_DOWN); 482 CURVNET_RESTORE(); 483 484 funsetown(&tp->tun_sigio); 485 selwakeuppri(&tp->tun_rsel, PZERO + 1); 486 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0); 487 TUNDEBUG (ifp, "closed\n"); 488 489 cv_broadcast(&tp->tun_cv); 490 mtx_unlock(&tp->tun_mtx); 491 return (0); 492 } 493 494 static void 495 tuninit(struct ifnet *ifp) 496 { 497 struct tun_softc *tp = ifp->if_softc; 498 #ifdef INET 499 struct ifaddr *ifa; 500 #endif 501 502 TUNDEBUG(ifp, "tuninit\n"); 503 504 mtx_lock(&tp->tun_mtx); 505 ifp->if_flags |= IFF_UP; 506 ifp->if_drv_flags |= IFF_DRV_RUNNING; 507 getmicrotime(&ifp->if_lastchange); 508 509 #ifdef INET 510 if_addr_rlock(ifp); 511 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 512 if (ifa->ifa_addr->sa_family == AF_INET) { 513 struct sockaddr_in *si; 514 515 si = (struct sockaddr_in *)ifa->ifa_addr; 516 if (si->sin_addr.s_addr) 517 tp->tun_flags |= TUN_IASET; 518 519 si = (struct sockaddr_in *)ifa->ifa_dstaddr; 520 if (si && si->sin_addr.s_addr) 521 tp->tun_flags |= TUN_DSTADDR; 522 } 523 } 524 if_addr_runlock(ifp); 525 #endif 526 mtx_unlock(&tp->tun_mtx); 527 } 528 529 /* 530 * Process an ioctl request. 531 */ 532 static int 533 tunifioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 534 { 535 struct ifreq *ifr = (struct ifreq *)data; 536 struct tun_softc *tp = ifp->if_softc; 537 struct ifstat *ifs; 538 int error = 0; 539 540 switch(cmd) { 541 case SIOCGIFSTATUS: 542 ifs = (struct ifstat *)data; 543 mtx_lock(&tp->tun_mtx); 544 if (tp->tun_pid) 545 snprintf(ifs->ascii, sizeof(ifs->ascii), 546 "\tOpened by PID %d\n", tp->tun_pid); 547 else 548 ifs->ascii[0] = '\0'; 549 mtx_unlock(&tp->tun_mtx); 550 break; 551 case SIOCSIFADDR: 552 tuninit(ifp); 553 TUNDEBUG(ifp, "address set\n"); 554 break; 555 case SIOCSIFMTU: 556 ifp->if_mtu = ifr->ifr_mtu; 557 TUNDEBUG(ifp, "mtu set\n"); 558 break; 559 case SIOCSIFFLAGS: 560 case SIOCADDMULTI: 561 case SIOCDELMULTI: 562 break; 563 default: 564 error = EINVAL; 565 } 566 return (error); 567 } 568 569 /* 570 * tunoutput - queue packets from higher level ready to put out. 571 */ 572 static int 573 tunoutput(struct ifnet *ifp, struct mbuf *m0, const struct sockaddr *dst, 574 struct route *ro) 575 { 576 struct tun_softc *tp = ifp->if_softc; 577 u_short cached_tun_flags; 578 int error; 579 u_int32_t af; 580 581 TUNDEBUG (ifp, "tunoutput\n"); 582 583 #ifdef MAC 584 error = mac_ifnet_check_transmit(ifp, m0); 585 if (error) { 586 m_freem(m0); 587 return (error); 588 } 589 #endif 590 591 /* Could be unlocked read? */ 592 mtx_lock(&tp->tun_mtx); 593 cached_tun_flags = tp->tun_flags; 594 mtx_unlock(&tp->tun_mtx); 595 if ((cached_tun_flags & TUN_READY) != TUN_READY) { 596 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 597 m_freem (m0); 598 return (EHOSTDOWN); 599 } 600 601 if ((ifp->if_flags & IFF_UP) != IFF_UP) { 602 m_freem (m0); 603 return (EHOSTDOWN); 604 } 605 606 /* BPF writes need to be handled specially. */ 607 if (dst->sa_family == AF_UNSPEC) 608 bcopy(dst->sa_data, &af, sizeof(af)); 609 else 610 af = dst->sa_family; 611 612 if (bpf_peers_present(ifp->if_bpf)) 613 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m0); 614 615 /* prepend sockaddr? this may abort if the mbuf allocation fails */ 616 if (cached_tun_flags & TUN_LMODE) { 617 /* allocate space for sockaddr */ 618 M_PREPEND(m0, dst->sa_len, M_NOWAIT); 619 620 /* if allocation failed drop packet */ 621 if (m0 == NULL) { 622 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1); 623 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 624 return (ENOBUFS); 625 } else { 626 bcopy(dst, m0->m_data, dst->sa_len); 627 } 628 } 629 630 if (cached_tun_flags & TUN_IFHEAD) { 631 /* Prepend the address family */ 632 M_PREPEND(m0, 4, M_NOWAIT); 633 634 /* if allocation failed drop packet */ 635 if (m0 == NULL) { 636 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1); 637 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 638 return (ENOBUFS); 639 } else 640 *(u_int32_t *)m0->m_data = htonl(af); 641 } else { 642 #ifdef INET 643 if (af != AF_INET) 644 #endif 645 { 646 m_freem(m0); 647 return (EAFNOSUPPORT); 648 } 649 } 650 651 error = (ifp->if_transmit)(ifp, m0); 652 if (error) 653 return (ENOBUFS); 654 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); 655 return (0); 656 } 657 658 /* 659 * the cdevsw interface is now pretty minimal. 660 */ 661 static int 662 tunioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, 663 struct thread *td) 664 { 665 int error; 666 struct tun_softc *tp = dev->si_drv1; 667 struct tuninfo *tunp; 668 669 switch (cmd) { 670 case TUNSIFINFO: 671 tunp = (struct tuninfo *)data; 672 if (tunp->mtu < IF_MINMTU) 673 return (EINVAL); 674 if (TUN2IFP(tp)->if_mtu != tunp->mtu) { 675 error = priv_check(td, PRIV_NET_SETIFMTU); 676 if (error) 677 return (error); 678 } 679 if (TUN2IFP(tp)->if_type != tunp->type) 680 return (EPROTOTYPE); 681 mtx_lock(&tp->tun_mtx); 682 TUN2IFP(tp)->if_mtu = tunp->mtu; 683 TUN2IFP(tp)->if_baudrate = tunp->baudrate; 684 mtx_unlock(&tp->tun_mtx); 685 break; 686 case TUNGIFINFO: 687 tunp = (struct tuninfo *)data; 688 mtx_lock(&tp->tun_mtx); 689 tunp->mtu = TUN2IFP(tp)->if_mtu; 690 tunp->type = TUN2IFP(tp)->if_type; 691 tunp->baudrate = TUN2IFP(tp)->if_baudrate; 692 mtx_unlock(&tp->tun_mtx); 693 break; 694 case TUNSDEBUG: 695 tundebug = *(int *)data; 696 break; 697 case TUNGDEBUG: 698 *(int *)data = tundebug; 699 break; 700 case TUNSLMODE: 701 mtx_lock(&tp->tun_mtx); 702 if (*(int *)data) { 703 tp->tun_flags |= TUN_LMODE; 704 tp->tun_flags &= ~TUN_IFHEAD; 705 } else 706 tp->tun_flags &= ~TUN_LMODE; 707 mtx_unlock(&tp->tun_mtx); 708 break; 709 case TUNSIFHEAD: 710 mtx_lock(&tp->tun_mtx); 711 if (*(int *)data) { 712 tp->tun_flags |= TUN_IFHEAD; 713 tp->tun_flags &= ~TUN_LMODE; 714 } else 715 tp->tun_flags &= ~TUN_IFHEAD; 716 mtx_unlock(&tp->tun_mtx); 717 break; 718 case TUNGIFHEAD: 719 mtx_lock(&tp->tun_mtx); 720 *(int *)data = (tp->tun_flags & TUN_IFHEAD) ? 1 : 0; 721 mtx_unlock(&tp->tun_mtx); 722 break; 723 case TUNSIFMODE: 724 /* deny this if UP */ 725 if (TUN2IFP(tp)->if_flags & IFF_UP) 726 return(EBUSY); 727 728 switch (*(int *)data & ~IFF_MULTICAST) { 729 case IFF_POINTOPOINT: 730 case IFF_BROADCAST: 731 mtx_lock(&tp->tun_mtx); 732 TUN2IFP(tp)->if_flags &= 733 ~(IFF_BROADCAST|IFF_POINTOPOINT|IFF_MULTICAST); 734 TUN2IFP(tp)->if_flags |= *(int *)data; 735 mtx_unlock(&tp->tun_mtx); 736 break; 737 default: 738 return(EINVAL); 739 } 740 break; 741 case TUNSIFPID: 742 mtx_lock(&tp->tun_mtx); 743 tp->tun_pid = curthread->td_proc->p_pid; 744 mtx_unlock(&tp->tun_mtx); 745 break; 746 case FIONBIO: 747 break; 748 case FIOASYNC: 749 mtx_lock(&tp->tun_mtx); 750 if (*(int *)data) 751 tp->tun_flags |= TUN_ASYNC; 752 else 753 tp->tun_flags &= ~TUN_ASYNC; 754 mtx_unlock(&tp->tun_mtx); 755 break; 756 case FIONREAD: 757 if (!IFQ_IS_EMPTY(&TUN2IFP(tp)->if_snd)) { 758 struct mbuf *mb; 759 IFQ_LOCK(&TUN2IFP(tp)->if_snd); 760 IFQ_POLL_NOLOCK(&TUN2IFP(tp)->if_snd, mb); 761 for (*(int *)data = 0; mb != NULL; mb = mb->m_next) 762 *(int *)data += mb->m_len; 763 IFQ_UNLOCK(&TUN2IFP(tp)->if_snd); 764 } else 765 *(int *)data = 0; 766 break; 767 case FIOSETOWN: 768 return (fsetown(*(int *)data, &tp->tun_sigio)); 769 770 case FIOGETOWN: 771 *(int *)data = fgetown(&tp->tun_sigio); 772 return (0); 773 774 /* This is deprecated, FIOSETOWN should be used instead. */ 775 case TIOCSPGRP: 776 return (fsetown(-(*(int *)data), &tp->tun_sigio)); 777 778 /* This is deprecated, FIOGETOWN should be used instead. */ 779 case TIOCGPGRP: 780 *(int *)data = -fgetown(&tp->tun_sigio); 781 return (0); 782 783 default: 784 return (ENOTTY); 785 } 786 return (0); 787 } 788 789 /* 790 * The cdevsw read interface - reads a packet at a time, or at 791 * least as much of a packet as can be read. 792 */ 793 static int 794 tunread(struct cdev *dev, struct uio *uio, int flag) 795 { 796 struct tun_softc *tp = dev->si_drv1; 797 struct ifnet *ifp = TUN2IFP(tp); 798 struct mbuf *m; 799 int error=0, len; 800 801 TUNDEBUG (ifp, "read\n"); 802 mtx_lock(&tp->tun_mtx); 803 if ((tp->tun_flags & TUN_READY) != TUN_READY) { 804 mtx_unlock(&tp->tun_mtx); 805 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 806 return (EHOSTDOWN); 807 } 808 809 tp->tun_flags &= ~TUN_RWAIT; 810 811 do { 812 IFQ_DEQUEUE(&ifp->if_snd, m); 813 if (m == NULL) { 814 if (flag & O_NONBLOCK) { 815 mtx_unlock(&tp->tun_mtx); 816 return (EWOULDBLOCK); 817 } 818 tp->tun_flags |= TUN_RWAIT; 819 error = mtx_sleep(tp, &tp->tun_mtx, PCATCH | (PZERO + 1), 820 "tunread", 0); 821 if (error != 0) { 822 mtx_unlock(&tp->tun_mtx); 823 return (error); 824 } 825 } 826 } while (m == NULL); 827 mtx_unlock(&tp->tun_mtx); 828 829 while (m && uio->uio_resid > 0 && error == 0) { 830 len = min(uio->uio_resid, m->m_len); 831 if (len != 0) 832 error = uiomove(mtod(m, void *), len, uio); 833 m = m_free(m); 834 } 835 836 if (m) { 837 TUNDEBUG(ifp, "Dropping mbuf\n"); 838 m_freem(m); 839 } 840 return (error); 841 } 842 843 /* 844 * the cdevsw write interface - an atomic write is a packet - or else! 845 */ 846 static int 847 tunwrite(struct cdev *dev, struct uio *uio, int flag) 848 { 849 struct tun_softc *tp = dev->si_drv1; 850 struct ifnet *ifp = TUN2IFP(tp); 851 struct mbuf *m; 852 uint32_t family, mru; 853 int isr; 854 855 TUNDEBUG(ifp, "tunwrite\n"); 856 857 if ((ifp->if_flags & IFF_UP) != IFF_UP) 858 /* ignore silently */ 859 return (0); 860 861 if (uio->uio_resid == 0) 862 return (0); 863 864 mru = TUNMRU; 865 if (tp->tun_flags & TUN_IFHEAD) 866 mru += sizeof(family); 867 if (uio->uio_resid < 0 || uio->uio_resid > mru) { 868 TUNDEBUG(ifp, "len=%zd!\n", uio->uio_resid); 869 return (EIO); 870 } 871 872 if ((m = m_uiotombuf(uio, M_NOWAIT, 0, 0, M_PKTHDR)) == NULL) { 873 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); 874 return (ENOBUFS); 875 } 876 877 m->m_pkthdr.rcvif = ifp; 878 #ifdef MAC 879 mac_ifnet_create_mbuf(ifp, m); 880 #endif 881 882 /* Could be unlocked read? */ 883 mtx_lock(&tp->tun_mtx); 884 if (tp->tun_flags & TUN_IFHEAD) { 885 mtx_unlock(&tp->tun_mtx); 886 if (m->m_len < sizeof(family) && 887 (m = m_pullup(m, sizeof(family))) == NULL) 888 return (ENOBUFS); 889 family = ntohl(*mtod(m, u_int32_t *)); 890 m_adj(m, sizeof(family)); 891 } else { 892 mtx_unlock(&tp->tun_mtx); 893 family = AF_INET; 894 } 895 896 BPF_MTAP2(ifp, &family, sizeof(family), m); 897 898 switch (family) { 899 #ifdef INET 900 case AF_INET: 901 isr = NETISR_IP; 902 break; 903 #endif 904 #ifdef INET6 905 case AF_INET6: 906 isr = NETISR_IPV6; 907 break; 908 #endif 909 default: 910 m_freem(m); 911 return (EAFNOSUPPORT); 912 } 913 random_harvest_queue(m, sizeof(*m), 2, RANDOM_NET_TUN); 914 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); 915 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1); 916 CURVNET_SET(ifp->if_vnet); 917 M_SETFIB(m, ifp->if_fib); 918 netisr_dispatch(isr, m); 919 CURVNET_RESTORE(); 920 return (0); 921 } 922 923 /* 924 * tunpoll - the poll interface, this is only useful on reads 925 * really. The write detect always returns true, write never blocks 926 * anyway, it either accepts the packet or drops it. 927 */ 928 static int 929 tunpoll(struct cdev *dev, int events, struct thread *td) 930 { 931 struct tun_softc *tp = dev->si_drv1; 932 struct ifnet *ifp = TUN2IFP(tp); 933 int revents = 0; 934 struct mbuf *m; 935 936 TUNDEBUG(ifp, "tunpoll\n"); 937 938 if (events & (POLLIN | POLLRDNORM)) { 939 IFQ_LOCK(&ifp->if_snd); 940 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 941 if (m != NULL) { 942 TUNDEBUG(ifp, "tunpoll q=%d\n", ifp->if_snd.ifq_len); 943 revents |= events & (POLLIN | POLLRDNORM); 944 } else { 945 TUNDEBUG(ifp, "tunpoll waiting\n"); 946 selrecord(td, &tp->tun_rsel); 947 } 948 IFQ_UNLOCK(&ifp->if_snd); 949 } 950 if (events & (POLLOUT | POLLWRNORM)) 951 revents |= events & (POLLOUT | POLLWRNORM); 952 953 return (revents); 954 } 955 956 /* 957 * tunkqfilter - support for the kevent() system call. 958 */ 959 static int 960 tunkqfilter(struct cdev *dev, struct knote *kn) 961 { 962 struct tun_softc *tp = dev->si_drv1; 963 struct ifnet *ifp = TUN2IFP(tp); 964 965 switch(kn->kn_filter) { 966 case EVFILT_READ: 967 TUNDEBUG(ifp, "%s kqfilter: EVFILT_READ, minor = %#x\n", 968 ifp->if_xname, dev2unit(dev)); 969 kn->kn_fop = &tun_read_filterops; 970 break; 971 972 case EVFILT_WRITE: 973 TUNDEBUG(ifp, "%s kqfilter: EVFILT_WRITE, minor = %#x\n", 974 ifp->if_xname, dev2unit(dev)); 975 kn->kn_fop = &tun_write_filterops; 976 break; 977 978 default: 979 TUNDEBUG(ifp, "%s kqfilter: invalid filter, minor = %#x\n", 980 ifp->if_xname, dev2unit(dev)); 981 return(EINVAL); 982 } 983 984 kn->kn_hook = tp; 985 knlist_add(&tp->tun_rsel.si_note, kn, 0); 986 987 return (0); 988 } 989 990 /* 991 * Return true of there is data in the interface queue. 992 */ 993 static int 994 tunkqread(struct knote *kn, long hint) 995 { 996 int ret; 997 struct tun_softc *tp = kn->kn_hook; 998 struct cdev *dev = tp->tun_dev; 999 struct ifnet *ifp = TUN2IFP(tp); 1000 1001 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1002 TUNDEBUG(ifp, 1003 "%s have data in the queue. Len = %d, minor = %#x\n", 1004 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev)); 1005 ret = 1; 1006 } else { 1007 TUNDEBUG(ifp, 1008 "%s waiting for data, minor = %#x\n", ifp->if_xname, 1009 dev2unit(dev)); 1010 ret = 0; 1011 } 1012 1013 return (ret); 1014 } 1015 1016 /* 1017 * Always can write, always return MTU in kn->data. 1018 */ 1019 static int 1020 tunkqwrite(struct knote *kn, long hint) 1021 { 1022 struct tun_softc *tp = kn->kn_hook; 1023 struct ifnet *ifp = TUN2IFP(tp); 1024 1025 kn->kn_data = ifp->if_mtu; 1026 1027 return (1); 1028 } 1029 1030 static void 1031 tunkqdetach(struct knote *kn) 1032 { 1033 struct tun_softc *tp = kn->kn_hook; 1034 1035 knlist_remove(&tp->tun_rsel.si_note, kn, 0); 1036 } 1037