if_tun.c revision 255362
1/* $NetBSD: if_tun.c,v 1.14 1994/06/29 06:36:25 cgd Exp $ */ 2 3/*- 4 * Copyright (c) 1988, Julian Onions <jpo@cs.nott.ac.uk> 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: head/sys/net/if_tun.c 255362 2013-09-07 14:15:13Z markm $ 17 */ 18 19#include "opt_atalk.h" 20#include "opt_inet.h" 21#include "opt_inet6.h" 22#include "opt_ipx.h" 23 24#include <sys/param.h> 25#include <sys/priv.h> 26#include <sys/proc.h> 27#include <sys/systm.h> 28#include <sys/jail.h> 29#include <sys/mbuf.h> 30#include <sys/module.h> 31#include <sys/socket.h> 32#include <sys/fcntl.h> 33#include <sys/filio.h> 34#include <sys/sockio.h> 35#include <sys/ttycom.h> 36#include <sys/poll.h> 37#include <sys/selinfo.h> 38#include <sys/signalvar.h> 39#include <sys/filedesc.h> 40#include <sys/kernel.h> 41#include <sys/sysctl.h> 42#include <sys/conf.h> 43#include <sys/uio.h> 44#include <sys/malloc.h> 45#include <sys/random.h> 46 47#include <net/if.h> 48#include <net/if_clone.h> 49#include <net/if_types.h> 50#include <net/netisr.h> 51#include <net/route.h> 52#include <net/vnet.h> 53#ifdef INET 54#include <netinet/in.h> 55#endif 56#include <net/bpf.h> 57#include <net/if_tun.h> 58 59#include <sys/queue.h> 60#include <sys/condvar.h> 61 62#include <security/mac/mac_framework.h> 63 64/* 65 * tun_list is protected by global tunmtx. Other mutable fields are 66 * protected by tun->tun_mtx, or by their owning subsystem. tun_dev is 67 * static for the duration of a tunnel interface. 68 */ 69struct tun_softc { 70 TAILQ_ENTRY(tun_softc) tun_list; 71 struct cdev *tun_dev; 72 u_short tun_flags; /* misc flags */ 73#define TUN_OPEN 0x0001 74#define TUN_INITED 0x0002 75#define TUN_RCOLL 0x0004 76#define TUN_IASET 0x0008 77#define TUN_DSTADDR 0x0010 78#define TUN_LMODE 0x0020 79#define TUN_RWAIT 0x0040 80#define TUN_ASYNC 0x0080 81#define TUN_IFHEAD 0x0100 82 83#define TUN_READY (TUN_OPEN | TUN_INITED) 84 85 /* 86 * XXXRW: tun_pid is used to exclusively lock /dev/tun. Is this 87 * actually needed? Can we just return EBUSY if already open? 88 * Problem is that this involved inherent races when a tun device 89 * is handed off from one process to another, as opposed to just 90 * being slightly stale informationally. 91 */ 92 pid_t tun_pid; /* owning pid */ 93 struct ifnet *tun_ifp; /* the interface */ 94 struct sigio *tun_sigio; /* information for async I/O */ 95 struct selinfo tun_rsel; /* read select */ 96 struct mtx tun_mtx; /* protect mutable softc fields */ 97 struct cv tun_cv; /* protect against ref'd dev destroy */ 98}; 99#define TUN2IFP(sc) ((sc)->tun_ifp) 100 101#define TUNDEBUG if (tundebug) if_printf 102 103/* 104 * All mutable global variables in if_tun are locked using tunmtx, with 105 * the exception of tundebug, which is used unlocked, and tunclones, 106 * which is static after setup. 107 */ 108static struct mtx tunmtx; 109static const char tunname[] = "tun"; 110static MALLOC_DEFINE(M_TUN, tunname, "Tunnel Interface"); 111static int tundebug = 0; 112static int tundclone = 1; 113static struct clonedevs *tunclones; 114static TAILQ_HEAD(,tun_softc) tunhead = TAILQ_HEAD_INITIALIZER(tunhead); 115SYSCTL_INT(_debug, OID_AUTO, if_tun_debug, CTLFLAG_RW, &tundebug, 0, ""); 116 117SYSCTL_DECL(_net_link); 118static SYSCTL_NODE(_net_link, OID_AUTO, tun, CTLFLAG_RW, 0, 119 "IP tunnel software network interface."); 120SYSCTL_INT(_net_link_tun, OID_AUTO, devfs_cloning, CTLFLAG_RW, &tundclone, 0, 121 "Enable legacy devfs interface creation."); 122 123TUNABLE_INT("net.link.tun.devfs_cloning", &tundclone); 124 125static void tunclone(void *arg, struct ucred *cred, char *name, 126 int namelen, struct cdev **dev); 127static void tuncreate(const char *name, struct cdev *dev); 128static int tunifioctl(struct ifnet *, u_long, caddr_t); 129static void tuninit(struct ifnet *); 130static int tunmodevent(module_t, int, void *); 131static int tunoutput(struct ifnet *, struct mbuf *, 132 const struct sockaddr *, struct route *ro); 133static void tunstart(struct ifnet *); 134 135static int tun_clone_create(struct if_clone *, int, caddr_t); 136static void tun_clone_destroy(struct ifnet *); 137static struct if_clone *tun_cloner; 138 139static d_open_t tunopen; 140static d_close_t tunclose; 141static d_read_t tunread; 142static d_write_t tunwrite; 143static d_ioctl_t tunioctl; 144static d_poll_t tunpoll; 145static d_kqfilter_t tunkqfilter; 146 147static int tunkqread(struct knote *, long); 148static int tunkqwrite(struct knote *, long); 149static void tunkqdetach(struct knote *); 150 151static struct filterops tun_read_filterops = { 152 .f_isfd = 1, 153 .f_attach = NULL, 154 .f_detach = tunkqdetach, 155 .f_event = tunkqread, 156}; 157 158static struct filterops tun_write_filterops = { 159 .f_isfd = 1, 160 .f_attach = NULL, 161 .f_detach = tunkqdetach, 162 .f_event = tunkqwrite, 163}; 164 165static struct cdevsw tun_cdevsw = { 166 .d_version = D_VERSION, 167 .d_flags = D_NEEDMINOR, 168 .d_open = tunopen, 169 .d_close = tunclose, 170 .d_read = tunread, 171 .d_write = tunwrite, 172 .d_ioctl = tunioctl, 173 .d_poll = tunpoll, 174 .d_kqfilter = tunkqfilter, 175 .d_name = tunname, 176}; 177 178static int 179tun_clone_create(struct if_clone *ifc, int unit, caddr_t params) 180{ 181 struct cdev *dev; 182 int i; 183 184 /* find any existing device, or allocate new unit number */ 185 i = clone_create(&tunclones, &tun_cdevsw, &unit, &dev, 0); 186 if (i) { 187 /* No preexisting struct cdev *, create one */ 188 dev = make_dev(&tun_cdevsw, unit, 189 UID_UUCP, GID_DIALER, 0600, "%s%d", tunname, unit); 190 } 191 tuncreate(tunname, dev); 192 193 return (0); 194} 195 196static void 197tunclone(void *arg, struct ucred *cred, char *name, int namelen, 198 struct cdev **dev) 199{ 200 char devname[SPECNAMELEN + 1]; 201 int u, i, append_unit; 202 203 if (*dev != NULL) 204 return; 205 206 /* 207 * If tun cloning is enabled, only the superuser can create an 208 * interface. 209 */ 210 if (!tundclone || priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0) 211 return; 212 213 if (strcmp(name, tunname) == 0) { 214 u = -1; 215 } else if (dev_stdclone(name, NULL, tunname, &u) != 1) 216 return; /* Don't recognise the name */ 217 if (u != -1 && u > IF_MAXUNIT) 218 return; /* Unit number too high */ 219 220 if (u == -1) 221 append_unit = 1; 222 else 223 append_unit = 0; 224 225 CURVNET_SET(CRED_TO_VNET(cred)); 226 /* find any existing device, or allocate new unit number */ 227 i = clone_create(&tunclones, &tun_cdevsw, &u, dev, 0); 228 if (i) { 229 if (append_unit) { 230 namelen = snprintf(devname, sizeof(devname), "%s%d", 231 name, u); 232 name = devname; 233 } 234 /* No preexisting struct cdev *, create one */ 235 *dev = make_dev_credf(MAKEDEV_REF, &tun_cdevsw, u, cred, 236 UID_UUCP, GID_DIALER, 0600, "%s", name); 237 } 238 239 if_clone_create(name, namelen, NULL); 240 CURVNET_RESTORE(); 241} 242 243static void 244tun_destroy(struct tun_softc *tp) 245{ 246 struct cdev *dev; 247 248 mtx_lock(&tp->tun_mtx); 249 if ((tp->tun_flags & TUN_OPEN) != 0) 250 cv_wait_unlock(&tp->tun_cv, &tp->tun_mtx); 251 else 252 mtx_unlock(&tp->tun_mtx); 253 254 CURVNET_SET(TUN2IFP(tp)->if_vnet); 255 dev = tp->tun_dev; 256 bpfdetach(TUN2IFP(tp)); 257 if_detach(TUN2IFP(tp)); 258 if_free(TUN2IFP(tp)); 259 destroy_dev(dev); 260 seldrain(&tp->tun_rsel); 261 knlist_destroy(&tp->tun_rsel.si_note); 262 mtx_destroy(&tp->tun_mtx); 263 cv_destroy(&tp->tun_cv); 264 free(tp, M_TUN); 265 CURVNET_RESTORE(); 266} 267 268static void 269tun_clone_destroy(struct ifnet *ifp) 270{ 271 struct tun_softc *tp = ifp->if_softc; 272 273 mtx_lock(&tunmtx); 274 TAILQ_REMOVE(&tunhead, tp, tun_list); 275 mtx_unlock(&tunmtx); 276 tun_destroy(tp); 277} 278 279static int 280tunmodevent(module_t mod, int type, void *data) 281{ 282 static eventhandler_tag tag; 283 struct tun_softc *tp; 284 285 switch (type) { 286 case MOD_LOAD: 287 mtx_init(&tunmtx, "tunmtx", NULL, MTX_DEF); 288 clone_setup(&tunclones); 289 tag = EVENTHANDLER_REGISTER(dev_clone, tunclone, 0, 1000); 290 if (tag == NULL) 291 return (ENOMEM); 292 tun_cloner = if_clone_simple(tunname, tun_clone_create, 293 tun_clone_destroy, 0); 294 break; 295 case MOD_UNLOAD: 296 if_clone_detach(tun_cloner); 297 EVENTHANDLER_DEREGISTER(dev_clone, tag); 298 drain_dev_clone_events(); 299 300 mtx_lock(&tunmtx); 301 while ((tp = TAILQ_FIRST(&tunhead)) != NULL) { 302 TAILQ_REMOVE(&tunhead, tp, tun_list); 303 mtx_unlock(&tunmtx); 304 tun_destroy(tp); 305 mtx_lock(&tunmtx); 306 } 307 mtx_unlock(&tunmtx); 308 clone_cleanup(&tunclones); 309 mtx_destroy(&tunmtx); 310 break; 311 default: 312 return EOPNOTSUPP; 313 } 314 return 0; 315} 316 317static moduledata_t tun_mod = { 318 "if_tun", 319 tunmodevent, 320 0 321}; 322 323DECLARE_MODULE(if_tun, tun_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 324MODULE_VERSION(if_tun, 1); 325 326static void 327tunstart(struct ifnet *ifp) 328{ 329 struct tun_softc *tp = ifp->if_softc; 330 struct mbuf *m; 331 332 TUNDEBUG(ifp,"%s starting\n", ifp->if_xname); 333 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 334 IFQ_LOCK(&ifp->if_snd); 335 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 336 if (m == NULL) { 337 IFQ_UNLOCK(&ifp->if_snd); 338 return; 339 } 340 IFQ_UNLOCK(&ifp->if_snd); 341 } 342 343 mtx_lock(&tp->tun_mtx); 344 if (tp->tun_flags & TUN_RWAIT) { 345 tp->tun_flags &= ~TUN_RWAIT; 346 wakeup(tp); 347 } 348 selwakeuppri(&tp->tun_rsel, PZERO + 1); 349 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0); 350 if (tp->tun_flags & TUN_ASYNC && tp->tun_sigio) { 351 mtx_unlock(&tp->tun_mtx); 352 pgsigio(&tp->tun_sigio, SIGIO, 0); 353 } else 354 mtx_unlock(&tp->tun_mtx); 355} 356 357/* XXX: should return an error code so it can fail. */ 358static void 359tuncreate(const char *name, struct cdev *dev) 360{ 361 struct tun_softc *sc; 362 struct ifnet *ifp; 363 364 sc = malloc(sizeof(*sc), M_TUN, M_WAITOK | M_ZERO); 365 mtx_init(&sc->tun_mtx, "tun_mtx", NULL, MTX_DEF); 366 cv_init(&sc->tun_cv, "tun_condvar"); 367 sc->tun_flags = TUN_INITED; 368 sc->tun_dev = dev; 369 mtx_lock(&tunmtx); 370 TAILQ_INSERT_TAIL(&tunhead, sc, tun_list); 371 mtx_unlock(&tunmtx); 372 373 ifp = sc->tun_ifp = if_alloc(IFT_PPP); 374 if (ifp == NULL) 375 panic("%s%d: failed to if_alloc() interface.\n", 376 name, dev2unit(dev)); 377 if_initname(ifp, name, dev2unit(dev)); 378 ifp->if_mtu = TUNMTU; 379 ifp->if_ioctl = tunifioctl; 380 ifp->if_output = tunoutput; 381 ifp->if_start = tunstart; 382 ifp->if_flags = IFF_POINTOPOINT | IFF_MULTICAST; 383 ifp->if_softc = sc; 384 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 385 ifp->if_snd.ifq_drv_maxlen = 0; 386 IFQ_SET_READY(&ifp->if_snd); 387 knlist_init_mtx(&sc->tun_rsel.si_note, &sc->tun_mtx); 388 ifp->if_capabilities |= IFCAP_LINKSTATE; 389 ifp->if_capenable |= IFCAP_LINKSTATE; 390 391 if_attach(ifp); 392 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); 393 dev->si_drv1 = sc; 394 TUNDEBUG(ifp, "interface %s is created, minor = %#x\n", 395 ifp->if_xname, dev2unit(dev)); 396} 397 398static int 399tunopen(struct cdev *dev, int flag, int mode, struct thread *td) 400{ 401 struct ifnet *ifp; 402 struct tun_softc *tp; 403 404 /* 405 * XXXRW: Non-atomic test and set of dev->si_drv1 requires 406 * synchronization. 407 */ 408 tp = dev->si_drv1; 409 if (!tp) { 410 tuncreate(tunname, dev); 411 tp = dev->si_drv1; 412 } 413 414 /* 415 * XXXRW: This use of tun_pid is subject to error due to the 416 * fact that a reference to the tunnel can live beyond the 417 * death of the process that created it. Can we replace this 418 * with a simple busy flag? 419 */ 420 mtx_lock(&tp->tun_mtx); 421 if (tp->tun_pid != 0 && tp->tun_pid != td->td_proc->p_pid) { 422 mtx_unlock(&tp->tun_mtx); 423 return (EBUSY); 424 } 425 tp->tun_pid = td->td_proc->p_pid; 426 427 tp->tun_flags |= TUN_OPEN; 428 ifp = TUN2IFP(tp); 429 if_link_state_change(ifp, LINK_STATE_UP); 430 TUNDEBUG(ifp, "open\n"); 431 mtx_unlock(&tp->tun_mtx); 432 433 return (0); 434} 435 436/* 437 * tunclose - close the device - mark i/f down & delete 438 * routing info 439 */ 440static int 441tunclose(struct cdev *dev, int foo, int bar, struct thread *td) 442{ 443 struct tun_softc *tp; 444 struct ifnet *ifp; 445 446 tp = dev->si_drv1; 447 ifp = TUN2IFP(tp); 448 449 mtx_lock(&tp->tun_mtx); 450 tp->tun_flags &= ~TUN_OPEN; 451 tp->tun_pid = 0; 452 453 /* 454 * junk all pending output 455 */ 456 CURVNET_SET(ifp->if_vnet); 457 IFQ_PURGE(&ifp->if_snd); 458 459 if (ifp->if_flags & IFF_UP) { 460 mtx_unlock(&tp->tun_mtx); 461 if_down(ifp); 462 mtx_lock(&tp->tun_mtx); 463 } 464 465 /* Delete all addresses and routes which reference this interface. */ 466 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 467 struct ifaddr *ifa; 468 469 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 470 mtx_unlock(&tp->tun_mtx); 471 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 472 /* deal w/IPv4 PtP destination; unlocked read */ 473 if (ifa->ifa_addr->sa_family == AF_INET) { 474 rtinit(ifa, (int)RTM_DELETE, 475 tp->tun_flags & TUN_DSTADDR ? RTF_HOST : 0); 476 } else { 477 rtinit(ifa, (int)RTM_DELETE, 0); 478 } 479 } 480 if_purgeaddrs(ifp); 481 mtx_lock(&tp->tun_mtx); 482 } 483 if_link_state_change(ifp, LINK_STATE_DOWN); 484 CURVNET_RESTORE(); 485 486 funsetown(&tp->tun_sigio); 487 selwakeuppri(&tp->tun_rsel, PZERO + 1); 488 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0); 489 TUNDEBUG (ifp, "closed\n"); 490 491 cv_broadcast(&tp->tun_cv); 492 mtx_unlock(&tp->tun_mtx); 493 return (0); 494} 495 496static void 497tuninit(struct ifnet *ifp) 498{ 499 struct tun_softc *tp = ifp->if_softc; 500#ifdef INET 501 struct ifaddr *ifa; 502#endif 503 504 TUNDEBUG(ifp, "tuninit\n"); 505 506 mtx_lock(&tp->tun_mtx); 507 ifp->if_flags |= IFF_UP; 508 ifp->if_drv_flags |= IFF_DRV_RUNNING; 509 getmicrotime(&ifp->if_lastchange); 510 511#ifdef INET 512 if_addr_rlock(ifp); 513 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 514 if (ifa->ifa_addr->sa_family == AF_INET) { 515 struct sockaddr_in *si; 516 517 si = (struct sockaddr_in *)ifa->ifa_addr; 518 if (si->sin_addr.s_addr) 519 tp->tun_flags |= TUN_IASET; 520 521 si = (struct sockaddr_in *)ifa->ifa_dstaddr; 522 if (si && si->sin_addr.s_addr) 523 tp->tun_flags |= TUN_DSTADDR; 524 } 525 } 526 if_addr_runlock(ifp); 527#endif 528 mtx_unlock(&tp->tun_mtx); 529} 530 531/* 532 * Process an ioctl request. 533 */ 534static int 535tunifioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 536{ 537 struct ifreq *ifr = (struct ifreq *)data; 538 struct tun_softc *tp = ifp->if_softc; 539 struct ifstat *ifs; 540 int error = 0; 541 542 switch(cmd) { 543 case SIOCGIFSTATUS: 544 ifs = (struct ifstat *)data; 545 mtx_lock(&tp->tun_mtx); 546 if (tp->tun_pid) 547 sprintf(ifs->ascii + strlen(ifs->ascii), 548 "\tOpened by PID %d\n", tp->tun_pid); 549 mtx_unlock(&tp->tun_mtx); 550 break; 551 case SIOCSIFADDR: 552 tuninit(ifp); 553 TUNDEBUG(ifp, "address set\n"); 554 break; 555 case SIOCSIFDSTADDR: 556 tuninit(ifp); 557 TUNDEBUG(ifp, "destination address set\n"); 558 break; 559 case SIOCSIFMTU: 560 ifp->if_mtu = ifr->ifr_mtu; 561 TUNDEBUG(ifp, "mtu set\n"); 562 break; 563 case SIOCSIFFLAGS: 564 case SIOCADDMULTI: 565 case SIOCDELMULTI: 566 break; 567 default: 568 error = EINVAL; 569 } 570 return (error); 571} 572 573/* 574 * tunoutput - queue packets from higher level ready to put out. 575 */ 576static int 577tunoutput(struct ifnet *ifp, struct mbuf *m0, const struct sockaddr *dst, 578 struct route *ro) 579{ 580 struct tun_softc *tp = ifp->if_softc; 581 u_short cached_tun_flags; 582 int error; 583 u_int32_t af; 584 585 TUNDEBUG (ifp, "tunoutput\n"); 586 587#ifdef MAC 588 error = mac_ifnet_check_transmit(ifp, m0); 589 if (error) { 590 m_freem(m0); 591 return (error); 592 } 593#endif 594 595 /* Could be unlocked read? */ 596 mtx_lock(&tp->tun_mtx); 597 cached_tun_flags = tp->tun_flags; 598 mtx_unlock(&tp->tun_mtx); 599 if ((cached_tun_flags & TUN_READY) != TUN_READY) { 600 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 601 m_freem (m0); 602 return (EHOSTDOWN); 603 } 604 605 if ((ifp->if_flags & IFF_UP) != IFF_UP) { 606 m_freem (m0); 607 return (EHOSTDOWN); 608 } 609 610 /* BPF writes need to be handled specially. */ 611 if (dst->sa_family == AF_UNSPEC) 612 bcopy(dst->sa_data, &af, sizeof(af)); 613 else 614 af = dst->sa_family; 615 616 if (bpf_peers_present(ifp->if_bpf)) 617 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m0); 618 619 /* prepend sockaddr? this may abort if the mbuf allocation fails */ 620 if (cached_tun_flags & TUN_LMODE) { 621 /* allocate space for sockaddr */ 622 M_PREPEND(m0, dst->sa_len, M_NOWAIT); 623 624 /* if allocation failed drop packet */ 625 if (m0 == NULL) { 626 ifp->if_iqdrops++; 627 ifp->if_oerrors++; 628 return (ENOBUFS); 629 } else { 630 bcopy(dst, m0->m_data, dst->sa_len); 631 } 632 } 633 634 if (cached_tun_flags & TUN_IFHEAD) { 635 /* Prepend the address family */ 636 M_PREPEND(m0, 4, M_NOWAIT); 637 638 /* if allocation failed drop packet */ 639 if (m0 == NULL) { 640 ifp->if_iqdrops++; 641 ifp->if_oerrors++; 642 return (ENOBUFS); 643 } else 644 *(u_int32_t *)m0->m_data = htonl(af); 645 } else { 646#ifdef INET 647 if (af != AF_INET) 648#endif 649 { 650 m_freem(m0); 651 return (EAFNOSUPPORT); 652 } 653 } 654 655 error = (ifp->if_transmit)(ifp, m0); 656 if (error) 657 return (ENOBUFS); 658 ifp->if_opackets++; 659 return (0); 660} 661 662/* 663 * the cdevsw interface is now pretty minimal. 664 */ 665static int 666tunioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, 667 struct thread *td) 668{ 669 int error; 670 struct tun_softc *tp = dev->si_drv1; 671 struct tuninfo *tunp; 672 673 switch (cmd) { 674 case TUNSIFINFO: 675 tunp = (struct tuninfo *)data; 676 if (tunp->mtu < IF_MINMTU) 677 return (EINVAL); 678 if (TUN2IFP(tp)->if_mtu != tunp->mtu) { 679 error = priv_check(td, PRIV_NET_SETIFMTU); 680 if (error) 681 return (error); 682 } 683 mtx_lock(&tp->tun_mtx); 684 TUN2IFP(tp)->if_mtu = tunp->mtu; 685 TUN2IFP(tp)->if_type = tunp->type; 686 TUN2IFP(tp)->if_baudrate = tunp->baudrate; 687 mtx_unlock(&tp->tun_mtx); 688 break; 689 case TUNGIFINFO: 690 tunp = (struct tuninfo *)data; 691 mtx_lock(&tp->tun_mtx); 692 tunp->mtu = TUN2IFP(tp)->if_mtu; 693 tunp->type = TUN2IFP(tp)->if_type; 694 tunp->baudrate = TUN2IFP(tp)->if_baudrate; 695 mtx_unlock(&tp->tun_mtx); 696 break; 697 case TUNSDEBUG: 698 tundebug = *(int *)data; 699 break; 700 case TUNGDEBUG: 701 *(int *)data = tundebug; 702 break; 703 case TUNSLMODE: 704 mtx_lock(&tp->tun_mtx); 705 if (*(int *)data) { 706 tp->tun_flags |= TUN_LMODE; 707 tp->tun_flags &= ~TUN_IFHEAD; 708 } else 709 tp->tun_flags &= ~TUN_LMODE; 710 mtx_unlock(&tp->tun_mtx); 711 break; 712 case TUNSIFHEAD: 713 mtx_lock(&tp->tun_mtx); 714 if (*(int *)data) { 715 tp->tun_flags |= TUN_IFHEAD; 716 tp->tun_flags &= ~TUN_LMODE; 717 } else 718 tp->tun_flags &= ~TUN_IFHEAD; 719 mtx_unlock(&tp->tun_mtx); 720 break; 721 case TUNGIFHEAD: 722 mtx_lock(&tp->tun_mtx); 723 *(int *)data = (tp->tun_flags & TUN_IFHEAD) ? 1 : 0; 724 mtx_unlock(&tp->tun_mtx); 725 break; 726 case TUNSIFMODE: 727 /* deny this if UP */ 728 if (TUN2IFP(tp)->if_flags & IFF_UP) 729 return(EBUSY); 730 731 switch (*(int *)data & ~IFF_MULTICAST) { 732 case IFF_POINTOPOINT: 733 case IFF_BROADCAST: 734 mtx_lock(&tp->tun_mtx); 735 TUN2IFP(tp)->if_flags &= 736 ~(IFF_BROADCAST|IFF_POINTOPOINT|IFF_MULTICAST); 737 TUN2IFP(tp)->if_flags |= *(int *)data; 738 mtx_unlock(&tp->tun_mtx); 739 break; 740 default: 741 return(EINVAL); 742 } 743 break; 744 case TUNSIFPID: 745 mtx_lock(&tp->tun_mtx); 746 tp->tun_pid = curthread->td_proc->p_pid; 747 mtx_unlock(&tp->tun_mtx); 748 break; 749 case FIONBIO: 750 break; 751 case FIOASYNC: 752 mtx_lock(&tp->tun_mtx); 753 if (*(int *)data) 754 tp->tun_flags |= TUN_ASYNC; 755 else 756 tp->tun_flags &= ~TUN_ASYNC; 757 mtx_unlock(&tp->tun_mtx); 758 break; 759 case FIONREAD: 760 if (!IFQ_IS_EMPTY(&TUN2IFP(tp)->if_snd)) { 761 struct mbuf *mb; 762 IFQ_LOCK(&TUN2IFP(tp)->if_snd); 763 IFQ_POLL_NOLOCK(&TUN2IFP(tp)->if_snd, mb); 764 for (*(int *)data = 0; mb != NULL; mb = mb->m_next) 765 *(int *)data += mb->m_len; 766 IFQ_UNLOCK(&TUN2IFP(tp)->if_snd); 767 } else 768 *(int *)data = 0; 769 break; 770 case FIOSETOWN: 771 return (fsetown(*(int *)data, &tp->tun_sigio)); 772 773 case FIOGETOWN: 774 *(int *)data = fgetown(&tp->tun_sigio); 775 return (0); 776 777 /* This is deprecated, FIOSETOWN should be used instead. */ 778 case TIOCSPGRP: 779 return (fsetown(-(*(int *)data), &tp->tun_sigio)); 780 781 /* This is deprecated, FIOGETOWN should be used instead. */ 782 case TIOCGPGRP: 783 *(int *)data = -fgetown(&tp->tun_sigio); 784 return (0); 785 786 default: 787 return (ENOTTY); 788 } 789 return (0); 790} 791 792/* 793 * The cdevsw read interface - reads a packet at a time, or at 794 * least as much of a packet as can be read. 795 */ 796static int 797tunread(struct cdev *dev, struct uio *uio, int flag) 798{ 799 struct tun_softc *tp = dev->si_drv1; 800 struct ifnet *ifp = TUN2IFP(tp); 801 struct mbuf *m; 802 int error=0, len; 803 804 TUNDEBUG (ifp, "read\n"); 805 mtx_lock(&tp->tun_mtx); 806 if ((tp->tun_flags & TUN_READY) != TUN_READY) { 807 mtx_unlock(&tp->tun_mtx); 808 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 809 return (EHOSTDOWN); 810 } 811 812 tp->tun_flags &= ~TUN_RWAIT; 813 814 do { 815 IFQ_DEQUEUE(&ifp->if_snd, m); 816 if (m == NULL) { 817 if (flag & O_NONBLOCK) { 818 mtx_unlock(&tp->tun_mtx); 819 return (EWOULDBLOCK); 820 } 821 tp->tun_flags |= TUN_RWAIT; 822 error = mtx_sleep(tp, &tp->tun_mtx, PCATCH | (PZERO + 1), 823 "tunread", 0); 824 if (error != 0) { 825 mtx_unlock(&tp->tun_mtx); 826 return (error); 827 } 828 } 829 } while (m == NULL); 830 mtx_unlock(&tp->tun_mtx); 831 832 while (m && uio->uio_resid > 0 && error == 0) { 833 len = min(uio->uio_resid, m->m_len); 834 if (len != 0) 835 error = uiomove(mtod(m, void *), len, uio); 836 m = m_free(m); 837 } 838 839 if (m) { 840 TUNDEBUG(ifp, "Dropping mbuf\n"); 841 m_freem(m); 842 } 843 return (error); 844} 845 846/* 847 * the cdevsw write interface - an atomic write is a packet - or else! 848 */ 849static int 850tunwrite(struct cdev *dev, struct uio *uio, int flag) 851{ 852 struct tun_softc *tp = dev->si_drv1; 853 struct ifnet *ifp = TUN2IFP(tp); 854 struct mbuf *m; 855 uint32_t family; 856 int isr; 857 858 TUNDEBUG(ifp, "tunwrite\n"); 859 860 if ((ifp->if_flags & IFF_UP) != IFF_UP) 861 /* ignore silently */ 862 return (0); 863 864 if (uio->uio_resid == 0) 865 return (0); 866 867 if (uio->uio_resid < 0 || uio->uio_resid > TUNMRU) { 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 ifp->if_ierrors++; 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#ifdef IPX 910 case AF_IPX: 911 isr = NETISR_IPX; 912 break; 913#endif 914#ifdef NETATALK 915 case AF_APPLETALK: 916 isr = NETISR_ATALK2; 917 break; 918#endif 919 default: 920 m_freem(m); 921 return (EAFNOSUPPORT); 922 } 923 if (harvest.point_to_point) 924 random_harvest(&(m->m_data), 12, 3, 0, RANDOM_NET_TUN); 925 ifp->if_ibytes += m->m_pkthdr.len; 926 ifp->if_ipackets++; 927 CURVNET_SET(ifp->if_vnet); 928 M_SETFIB(m, ifp->if_fib); 929 netisr_dispatch(isr, m); 930 CURVNET_RESTORE(); 931 return (0); 932} 933 934/* 935 * tunpoll - the poll interface, this is only useful on reads 936 * really. The write detect always returns true, write never blocks 937 * anyway, it either accepts the packet or drops it. 938 */ 939static int 940tunpoll(struct cdev *dev, int events, struct thread *td) 941{ 942 struct tun_softc *tp = dev->si_drv1; 943 struct ifnet *ifp = TUN2IFP(tp); 944 int revents = 0; 945 struct mbuf *m; 946 947 TUNDEBUG(ifp, "tunpoll\n"); 948 949 if (events & (POLLIN | POLLRDNORM)) { 950 IFQ_LOCK(&ifp->if_snd); 951 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 952 if (m != NULL) { 953 TUNDEBUG(ifp, "tunpoll q=%d\n", ifp->if_snd.ifq_len); 954 revents |= events & (POLLIN | POLLRDNORM); 955 } else { 956 TUNDEBUG(ifp, "tunpoll waiting\n"); 957 selrecord(td, &tp->tun_rsel); 958 } 959 IFQ_UNLOCK(&ifp->if_snd); 960 } 961 if (events & (POLLOUT | POLLWRNORM)) 962 revents |= events & (POLLOUT | POLLWRNORM); 963 964 return (revents); 965} 966 967/* 968 * tunkqfilter - support for the kevent() system call. 969 */ 970static int 971tunkqfilter(struct cdev *dev, struct knote *kn) 972{ 973 struct tun_softc *tp = dev->si_drv1; 974 struct ifnet *ifp = TUN2IFP(tp); 975 976 switch(kn->kn_filter) { 977 case EVFILT_READ: 978 TUNDEBUG(ifp, "%s kqfilter: EVFILT_READ, minor = %#x\n", 979 ifp->if_xname, dev2unit(dev)); 980 kn->kn_fop = &tun_read_filterops; 981 break; 982 983 case EVFILT_WRITE: 984 TUNDEBUG(ifp, "%s kqfilter: EVFILT_WRITE, minor = %#x\n", 985 ifp->if_xname, dev2unit(dev)); 986 kn->kn_fop = &tun_write_filterops; 987 break; 988 989 default: 990 TUNDEBUG(ifp, "%s kqfilter: invalid filter, minor = %#x\n", 991 ifp->if_xname, dev2unit(dev)); 992 return(EINVAL); 993 } 994 995 kn->kn_hook = tp; 996 knlist_add(&tp->tun_rsel.si_note, kn, 0); 997 998 return (0); 999} 1000 1001/* 1002 * Return true of there is data in the interface queue. 1003 */ 1004static int 1005tunkqread(struct knote *kn, long hint) 1006{ 1007 int ret; 1008 struct tun_softc *tp = kn->kn_hook; 1009 struct cdev *dev = tp->tun_dev; 1010 struct ifnet *ifp = TUN2IFP(tp); 1011 1012 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1013 TUNDEBUG(ifp, 1014 "%s have data in the queue. Len = %d, minor = %#x\n", 1015 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev)); 1016 ret = 1; 1017 } else { 1018 TUNDEBUG(ifp, 1019 "%s waiting for data, minor = %#x\n", ifp->if_xname, 1020 dev2unit(dev)); 1021 ret = 0; 1022 } 1023 1024 return (ret); 1025} 1026 1027/* 1028 * Always can write, always return MTU in kn->data. 1029 */ 1030static int 1031tunkqwrite(struct knote *kn, long hint) 1032{ 1033 struct tun_softc *tp = kn->kn_hook; 1034 struct ifnet *ifp = TUN2IFP(tp); 1035 1036 kn->kn_data = ifp->if_mtu; 1037 1038 return (1); 1039} 1040 1041static void 1042tunkqdetach(struct knote *kn) 1043{ 1044 struct tun_softc *tp = kn->kn_hook; 1045 1046 knlist_remove(&tp->tun_rsel.si_note, kn, 0); 1047} 1048