if_tun.c revision 186483
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 186483 2008-12-25 02:14:25Z kmacy $ 17 */ 18 19#include "opt_atalk.h" 20#include "opt_inet.h" 21#include "opt_inet6.h" 22#include "opt_ipx.h" 23#include "opt_mac.h" 24 25#include <sys/param.h> 26#include <sys/priv.h> 27#include <sys/proc.h> 28#include <sys/systm.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#include <sys/vimage.h> 47 48#include <net/if.h> 49#include <net/if_clone.h> 50#include <net/if_types.h> 51#include <net/netisr.h> 52#include <net/route.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#define TUN_CLOSED 0x0200 83 84#define TUN_READY (TUN_OPEN | TUN_INITED) 85 86 /* 87 * XXXRW: tun_pid is used to exclusively lock /dev/tun. Is this 88 * actually needed? Can we just return EBUSY if already open? 89 * Problem is that this involved inherent races when a tun device 90 * is handed off from one process to another, as opposed to just 91 * being slightly stale informationally. 92 */ 93 pid_t tun_pid; /* owning pid */ 94 struct ifnet *tun_ifp; /* the interface */ 95 struct sigio *tun_sigio; /* information for async I/O */ 96 struct selinfo tun_rsel; /* read select */ 97 struct mtx tun_mtx; /* protect mutable softc fields */ 98 struct cv tun_cv; /* protect against ref'd dev destroy */ 99}; 100#define TUN2IFP(sc) ((sc)->tun_ifp) 101 102#define TUNDEBUG if (tundebug) if_printf 103#define TUNNAME "tun" 104 105/* 106 * All mutable global variables in if_tun are locked using tunmtx, with 107 * the exception of tundebug, which is used unlocked, and tunclones, 108 * which is static after setup. 109 */ 110static struct mtx tunmtx; 111static MALLOC_DEFINE(M_TUN, TUNNAME, "Tunnel Interface"); 112static int tundebug = 0; 113static int tundclone = 1; 114static struct clonedevs *tunclones; 115static TAILQ_HEAD(,tun_softc) tunhead = TAILQ_HEAD_INITIALIZER(tunhead); 116SYSCTL_INT(_debug, OID_AUTO, if_tun_debug, CTLFLAG_RW, &tundebug, 0, ""); 117 118SYSCTL_DECL(_net_link); 119SYSCTL_NODE(_net_link, OID_AUTO, tun, CTLFLAG_RW, 0, 120 "IP tunnel software network interface."); 121SYSCTL_INT(_net_link_tun, OID_AUTO, devfs_cloning, CTLFLAG_RW, &tundclone, 0, 122 "Enable legacy devfs interface creation."); 123 124TUNABLE_INT("net.link.tun.devfs_cloning", &tundclone); 125 126static void tunclone(void *arg, struct ucred *cred, char *name, 127 int namelen, struct cdev **dev); 128static void tuncreate(const char *name, struct cdev *dev); 129static int tunifioctl(struct ifnet *, u_long, caddr_t); 130static int tuninit(struct ifnet *); 131static int tunmodevent(module_t, int, void *); 132static int tunoutput(struct ifnet *, struct mbuf *, struct sockaddr *, 133 struct rtentry *rt); 134static void tunstart(struct ifnet *); 135 136static int tun_clone_create(struct if_clone *, int, caddr_t); 137static void tun_clone_destroy(struct ifnet *); 138 139IFC_SIMPLE_DECLARE(tun, 0); 140 141static d_open_t tunopen; 142static d_close_t tunclose; 143static d_read_t tunread; 144static d_write_t tunwrite; 145static d_ioctl_t tunioctl; 146static d_poll_t tunpoll; 147static d_kqfilter_t tunkqfilter; 148 149static int tunkqread(struct knote *, long); 150static int tunkqwrite(struct knote *, long); 151static void tunkqdetach(struct knote *); 152 153static struct filterops tun_read_filterops = { 154 .f_isfd = 1, 155 .f_attach = NULL, 156 .f_detach = tunkqdetach, 157 .f_event = tunkqread, 158}; 159 160static struct filterops tun_write_filterops = { 161 .f_isfd = 1, 162 .f_attach = NULL, 163 .f_detach = tunkqdetach, 164 .f_event = tunkqwrite, 165}; 166 167static struct cdevsw tun_cdevsw = { 168 .d_version = D_VERSION, 169 .d_flags = D_PSEUDO | D_NEEDGIANT | D_NEEDMINOR, 170 .d_open = tunopen, 171 .d_close = tunclose, 172 .d_read = tunread, 173 .d_write = tunwrite, 174 .d_ioctl = tunioctl, 175 .d_poll = tunpoll, 176 .d_kqfilter = tunkqfilter, 177 .d_name = TUNNAME, 178}; 179 180static int 181tun_clone_create(struct if_clone *ifc, int unit, caddr_t params) 182{ 183 struct cdev *dev; 184 int i; 185 186 /* find any existing device, or allocate new unit number */ 187 i = clone_create(&tunclones, &tun_cdevsw, &unit, &dev, 0); 188 if (i) { 189 /* No preexisting struct cdev *, create one */ 190 dev = make_dev(&tun_cdevsw, unit, 191 UID_UUCP, GID_DIALER, 0600, "%s%d", ifc->ifc_name, unit); 192 if (dev != NULL) { 193 dev_ref(dev); 194 dev->si_flags |= SI_CHEAPCLONE; 195 } 196 } 197 tuncreate(ifc->ifc_name, dev); 198 199 return (0); 200} 201 202static void 203tunclone(void *arg, struct ucred *cred, char *name, int namelen, 204 struct cdev **dev) 205{ 206 char devname[SPECNAMELEN + 1]; 207 int u, i, append_unit; 208 209 if (*dev != NULL) 210 return; 211 212 /* 213 * If tun cloning is enabled, only the superuser can create an 214 * interface. 215 */ 216 if (!tundclone || priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0) 217 return; 218 219 if (strcmp(name, TUNNAME) == 0) { 220 u = -1; 221 } else if (dev_stdclone(name, NULL, TUNNAME, &u) != 1) 222 return; /* Don't recognise the name */ 223 if (u != -1 && u > IF_MAXUNIT) 224 return; /* Unit number too high */ 225 226 if (u == -1) 227 append_unit = 1; 228 else 229 append_unit = 0; 230 231 CURVNET_SET(TD_TO_VNET(curthread)); 232 /* find any existing device, or allocate new unit number */ 233 i = clone_create(&tunclones, &tun_cdevsw, &u, dev, 0); 234 if (i) { 235 if (append_unit) { 236 namelen = snprintf(devname, sizeof(devname), "%s%d", name, 237 u); 238 name = devname; 239 } 240 /* No preexisting struct cdev *, create one */ 241 *dev = make_dev(&tun_cdevsw, u, 242 UID_UUCP, GID_DIALER, 0600, "%s", name); 243 if (*dev != NULL) { 244 dev_ref(*dev); 245 (*dev)->si_flags |= SI_CHEAPCLONE; 246 } 247 } 248 249 if_clone_create(name, namelen, NULL); 250 CURVNET_RESTORE(); 251} 252 253static void 254tun_destroy(struct tun_softc *tp) 255{ 256 struct cdev *dev; 257 258 /* Unlocked read. */ 259 mtx_lock(&tp->tun_mtx); 260 if ((tp->tun_flags & (TUN_OPEN|TUN_CLOSED)) != TUN_CLOSED) 261 cv_wait_unlock(&tp->tun_cv, &tp->tun_mtx); 262 else 263 mtx_unlock(&tp->tun_mtx); 264 265 CURVNET_SET(TUN2IFP(tp)->if_vnet); 266 dev = tp->tun_dev; 267 bpfdetach(TUN2IFP(tp)); 268 if_detach(TUN2IFP(tp)); 269 if_free(TUN2IFP(tp)); 270 destroy_dev(dev); 271 knlist_destroy(&tp->tun_rsel.si_note); 272 mtx_destroy(&tp->tun_mtx); 273 cv_destroy(&tp->tun_cv); 274 free(tp, M_TUN); 275 CURVNET_RESTORE(); 276} 277 278static void 279tun_clone_destroy(struct ifnet *ifp) 280{ 281 struct tun_softc *tp = ifp->if_softc; 282 283 mtx_lock(&tunmtx); 284 TAILQ_REMOVE(&tunhead, tp, tun_list); 285 mtx_unlock(&tunmtx); 286 tun_destroy(tp); 287} 288 289static int 290tunmodevent(module_t mod, int type, void *data) 291{ 292 static eventhandler_tag tag; 293 struct tun_softc *tp; 294 295 switch (type) { 296 case MOD_LOAD: 297 mtx_init(&tunmtx, "tunmtx", NULL, MTX_DEF); 298 clone_setup(&tunclones); 299 tag = EVENTHANDLER_REGISTER(dev_clone, tunclone, 0, 1000); 300 if (tag == NULL) 301 return (ENOMEM); 302 if_clone_attach(&tun_cloner); 303 break; 304 case MOD_UNLOAD: 305 if_clone_detach(&tun_cloner); 306 EVENTHANDLER_DEREGISTER(dev_clone, tag); 307 308 mtx_lock(&tunmtx); 309 while ((tp = TAILQ_FIRST(&tunhead)) != NULL) { 310 TAILQ_REMOVE(&tunhead, tp, tun_list); 311 mtx_unlock(&tunmtx); 312 tun_destroy(tp); 313 mtx_lock(&tunmtx); 314 } 315 mtx_unlock(&tunmtx); 316 clone_cleanup(&tunclones); 317 mtx_destroy(&tunmtx); 318 break; 319 default: 320 return EOPNOTSUPP; 321 } 322 return 0; 323} 324 325static moduledata_t tun_mod = { 326 "if_tun", 327 tunmodevent, 328 0 329}; 330 331DECLARE_MODULE(if_tun, tun_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 332 333static void 334tunstart(struct ifnet *ifp) 335{ 336 struct tun_softc *tp = ifp->if_softc; 337 struct mbuf *m; 338 339 TUNDEBUG(ifp,"%s starting\n", ifp->if_xname); 340 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 341 IFQ_LOCK(&ifp->if_snd); 342 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 343 if (m == NULL) { 344 IFQ_UNLOCK(&ifp->if_snd); 345 return; 346 } 347 IFQ_UNLOCK(&ifp->if_snd); 348 } 349 350 mtx_lock(&tp->tun_mtx); 351 if (tp->tun_flags & TUN_RWAIT) { 352 tp->tun_flags &= ~TUN_RWAIT; 353 wakeup(tp); 354 } 355 if (tp->tun_flags & TUN_ASYNC && tp->tun_sigio) { 356 mtx_unlock(&tp->tun_mtx); 357 pgsigio(&tp->tun_sigio, SIGIO, 0); 358 } else 359 mtx_unlock(&tp->tun_mtx); 360 selwakeuppri(&tp->tun_rsel, PZERO + 1); 361 KNOTE_UNLOCKED(&tp->tun_rsel.si_note, 0); 362} 363 364/* XXX: should return an error code so it can fail. */ 365static void 366tuncreate(const char *name, struct cdev *dev) 367{ 368 struct tun_softc *sc; 369 struct ifnet *ifp; 370 371 dev->si_flags &= ~SI_CHEAPCLONE; 372 373 sc = malloc(sizeof(*sc), M_TUN, M_WAITOK | M_ZERO); 374 mtx_init(&sc->tun_mtx, "tun_mtx", NULL, MTX_DEF); 375 cv_init(&sc->tun_cv, "tun_condvar"); 376 sc->tun_flags = TUN_INITED; 377 sc->tun_dev = dev; 378 mtx_lock(&tunmtx); 379 TAILQ_INSERT_TAIL(&tunhead, sc, tun_list); 380 mtx_unlock(&tunmtx); 381 382 ifp = sc->tun_ifp = if_alloc(IFT_PPP); 383 if (ifp == NULL) 384 panic("%s%d: failed to if_alloc() interface.\n", 385 name, dev2unit(dev)); 386 if_initname(ifp, name, dev2unit(dev)); 387 ifp->if_mtu = TUNMTU; 388 ifp->if_ioctl = tunifioctl; 389 ifp->if_output = tunoutput; 390 ifp->if_start = tunstart; 391 ifp->if_flags = IFF_POINTOPOINT | IFF_MULTICAST; 392 ifp->if_softc = sc; 393 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 394 ifp->if_snd.ifq_drv_maxlen = 0; 395 IFQ_SET_READY(&ifp->if_snd); 396 knlist_init(&sc->tun_rsel.si_note, NULL, NULL, NULL, NULL); 397 398 if_attach(ifp); 399 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); 400 dev->si_drv1 = sc; 401 TUNDEBUG(ifp, "interface %s is created, minor = %#x\n", 402 ifp->if_xname, dev2unit(dev)); 403} 404 405static int 406tunopen(struct cdev *dev, int flag, int mode, struct thread *td) 407{ 408 struct ifnet *ifp; 409 struct tun_softc *tp; 410 411 /* 412 * XXXRW: Non-atomic test and set of dev->si_drv1 requires 413 * synchronization. 414 */ 415 tp = dev->si_drv1; 416 if (!tp) { 417 tuncreate(TUNNAME, dev); 418 tp = dev->si_drv1; 419 } 420 421 /* 422 * XXXRW: This use of tun_pid is subject to error due to the 423 * fact that a reference to the tunnel can live beyond the 424 * death of the process that created it. Can we replace this 425 * with a simple busy flag? 426 */ 427 mtx_lock(&tp->tun_mtx); 428 if (tp->tun_pid != 0 && tp->tun_pid != td->td_proc->p_pid) { 429 mtx_unlock(&tp->tun_mtx); 430 return (EBUSY); 431 } 432 tp->tun_pid = td->td_proc->p_pid; 433 434 tp->tun_flags |= TUN_OPEN; 435 mtx_unlock(&tp->tun_mtx); 436 ifp = TUN2IFP(tp); 437 if_link_state_change(ifp, LINK_STATE_UP); 438 TUNDEBUG(ifp, "open\n"); 439 440 return (0); 441} 442 443/* 444 * tunclose - close the device - mark i/f down & delete 445 * routing info 446 */ 447static int 448tunclose(struct cdev *dev, int foo, int bar, struct thread *td) 449{ 450 struct tun_softc *tp; 451 struct ifnet *ifp; 452 int s; 453 454 tp = dev->si_drv1; 455 ifp = TUN2IFP(tp); 456 457 mtx_lock(&tp->tun_mtx); 458 tp->tun_flags &= ~TUN_OPEN; 459 tp->tun_pid = 0; 460 mtx_unlock(&tp->tun_mtx); 461 462 /* 463 * junk all pending output 464 */ 465 CURVNET_SET(ifp->if_vnet); 466 s = splimp(); 467 IFQ_PURGE(&ifp->if_snd); 468 splx(s); 469 470 if (ifp->if_flags & IFF_UP) { 471 s = splimp(); 472 if_down(ifp); 473 splx(s); 474 } 475 476 /* Delete all addresses and routes which reference this interface. */ 477 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 478 struct ifaddr *ifa; 479 480 s = splimp(); 481 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 482 /* deal w/IPv4 PtP destination; unlocked read */ 483 if (ifa->ifa_addr->sa_family == AF_INET) { 484 rtinit(ifa, (int)RTM_DELETE, 485 tp->tun_flags & TUN_DSTADDR ? RTF_HOST : 0); 486 } else { 487 rtinit(ifa, (int)RTM_DELETE, 0); 488 } 489 } 490 if_purgeaddrs(ifp); 491 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 492 splx(s); 493 } 494 if_link_state_change(ifp, LINK_STATE_DOWN); 495 CURVNET_RESTORE(); 496 497 mtx_lock(&tp->tun_mtx); 498 funsetown(&tp->tun_sigio); 499 selwakeuppri(&tp->tun_rsel, PZERO + 1); 500 KNOTE_UNLOCKED(&tp->tun_rsel.si_note, 0); 501 TUNDEBUG (ifp, "closed\n"); 502 503 tp->tun_flags |= TUN_CLOSED; 504 cv_broadcast(&tp->tun_cv); 505 mtx_unlock(&tp->tun_mtx); 506 return (0); 507} 508 509static int 510tuninit(struct ifnet *ifp) 511{ 512#ifdef INET 513 struct tun_softc *tp = ifp->if_softc; 514 struct ifaddr *ifa; 515#endif 516 int error = 0; 517 518 TUNDEBUG(ifp, "tuninit\n"); 519 520 ifp->if_flags |= IFF_UP; 521 ifp->if_drv_flags |= IFF_DRV_RUNNING; 522 getmicrotime(&ifp->if_lastchange); 523 524#ifdef INET 525 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 526 if (ifa->ifa_addr->sa_family == AF_INET) { 527 struct sockaddr_in *si; 528 529 si = (struct sockaddr_in *)ifa->ifa_addr; 530 mtx_lock(&tp->tun_mtx); 531 if (si->sin_addr.s_addr) 532 tp->tun_flags |= TUN_IASET; 533 534 si = (struct sockaddr_in *)ifa->ifa_dstaddr; 535 if (si && si->sin_addr.s_addr) 536 tp->tun_flags |= TUN_DSTADDR; 537 mtx_unlock(&tp->tun_mtx); 538 } 539 } 540#endif 541 return (error); 542} 543 544/* 545 * Process an ioctl request. 546 */ 547static int 548tunifioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 549{ 550 struct ifreq *ifr = (struct ifreq *)data; 551 struct tun_softc *tp = ifp->if_softc; 552 struct ifstat *ifs; 553 int error = 0, s; 554 555 s = splimp(); 556 switch(cmd) { 557 case SIOCGIFSTATUS: 558 ifs = (struct ifstat *)data; 559 mtx_lock(&tp->tun_mtx); 560 if (tp->tun_pid) 561 sprintf(ifs->ascii + strlen(ifs->ascii), 562 "\tOpened by PID %d\n", tp->tun_pid); 563 mtx_unlock(&tp->tun_mtx); 564 break; 565 case SIOCSIFADDR: 566 error = tuninit(ifp); 567 TUNDEBUG(ifp, "address set, error=%d\n", error); 568 break; 569 case SIOCSIFDSTADDR: 570 error = tuninit(ifp); 571 TUNDEBUG(ifp, "destination address set, error=%d\n", error); 572 break; 573 case SIOCSIFMTU: 574 ifp->if_mtu = ifr->ifr_mtu; 575 TUNDEBUG(ifp, "mtu set\n"); 576 break; 577 case SIOCSIFFLAGS: 578 case SIOCADDMULTI: 579 case SIOCDELMULTI: 580 break; 581 default: 582 error = EINVAL; 583 } 584 splx(s); 585 return (error); 586} 587 588/* 589 * tunoutput - queue packets from higher level ready to put out. 590 */ 591static int 592tunoutput( 593 struct ifnet *ifp, 594 struct mbuf *m0, 595 struct sockaddr *dst, 596 struct rtentry *rt) 597{ 598 struct tun_softc *tp = ifp->if_softc; 599 u_short cached_tun_flags; 600 int error; 601 u_int32_t af; 602 603 TUNDEBUG (ifp, "tunoutput\n"); 604 605#ifdef MAC 606 error = mac_ifnet_check_transmit(ifp, m0); 607 if (error) { 608 m_freem(m0); 609 return (error); 610 } 611#endif 612 613 /* Could be unlocked read? */ 614 mtx_lock(&tp->tun_mtx); 615 cached_tun_flags = tp->tun_flags; 616 mtx_unlock(&tp->tun_mtx); 617 if ((cached_tun_flags & TUN_READY) != TUN_READY) { 618 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 619 m_freem (m0); 620 return (EHOSTDOWN); 621 } 622 623 if ((ifp->if_flags & IFF_UP) != IFF_UP) { 624 m_freem (m0); 625 return (EHOSTDOWN); 626 } 627 628 /* BPF writes need to be handled specially. */ 629 if (dst->sa_family == AF_UNSPEC) { 630 bcopy(dst->sa_data, &af, sizeof(af)); 631 dst->sa_family = af; 632 } 633 634 if (bpf_peers_present(ifp->if_bpf)) { 635 af = dst->sa_family; 636 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m0); 637 } 638 639 /* prepend sockaddr? this may abort if the mbuf allocation fails */ 640 if (cached_tun_flags & TUN_LMODE) { 641 /* allocate space for sockaddr */ 642 M_PREPEND(m0, dst->sa_len, M_DONTWAIT); 643 644 /* if allocation failed drop packet */ 645 if (m0 == NULL) { 646 ifp->if_iqdrops++; 647 ifp->if_oerrors++; 648 return (ENOBUFS); 649 } else { 650 bcopy(dst, m0->m_data, dst->sa_len); 651 } 652 } 653 654 if (cached_tun_flags & TUN_IFHEAD) { 655 /* Prepend the address family */ 656 M_PREPEND(m0, 4, M_DONTWAIT); 657 658 /* if allocation failed drop packet */ 659 if (m0 == NULL) { 660 ifp->if_iqdrops++; 661 ifp->if_oerrors++; 662 return (ENOBUFS); 663 } else 664 *(u_int32_t *)m0->m_data = htonl(dst->sa_family); 665 } else { 666#ifdef INET 667 if (dst->sa_family != AF_INET) 668#endif 669 { 670 m_freem(m0); 671 return (EAFNOSUPPORT); 672 } 673 } 674 675 error = (ifp->if_transmit)(ifp, m0); 676 if (error) { 677 ifp->if_collisions++; 678 return (ENOBUFS); 679 } 680 ifp->if_opackets++; 681 return (0); 682} 683 684/* 685 * the cdevsw interface is now pretty minimal. 686 */ 687static int 688tunioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 689{ 690 int s; 691 int error; 692 struct tun_softc *tp = dev->si_drv1; 693 struct tuninfo *tunp; 694 695 switch (cmd) { 696 case TUNSIFINFO: 697 tunp = (struct tuninfo *)data; 698 if (tunp->mtu < IF_MINMTU) 699 return (EINVAL); 700 if (TUN2IFP(tp)->if_mtu != tunp->mtu) { 701 error = priv_check(td, PRIV_NET_SETIFMTU); 702 if (error) 703 return (error); 704 } 705 TUN2IFP(tp)->if_mtu = tunp->mtu; 706 TUN2IFP(tp)->if_type = tunp->type; 707 TUN2IFP(tp)->if_baudrate = tunp->baudrate; 708 break; 709 case TUNGIFINFO: 710 tunp = (struct tuninfo *)data; 711 tunp->mtu = TUN2IFP(tp)->if_mtu; 712 tunp->type = TUN2IFP(tp)->if_type; 713 tunp->baudrate = TUN2IFP(tp)->if_baudrate; 714 break; 715 case TUNSDEBUG: 716 tundebug = *(int *)data; 717 break; 718 case TUNGDEBUG: 719 *(int *)data = tundebug; 720 break; 721 case TUNSLMODE: 722 mtx_lock(&tp->tun_mtx); 723 if (*(int *)data) { 724 tp->tun_flags |= TUN_LMODE; 725 tp->tun_flags &= ~TUN_IFHEAD; 726 } else 727 tp->tun_flags &= ~TUN_LMODE; 728 mtx_unlock(&tp->tun_mtx); 729 break; 730 case TUNSIFHEAD: 731 mtx_lock(&tp->tun_mtx); 732 if (*(int *)data) { 733 tp->tun_flags |= TUN_IFHEAD; 734 tp->tun_flags &= ~TUN_LMODE; 735 } else 736 tp->tun_flags &= ~TUN_IFHEAD; 737 mtx_unlock(&tp->tun_mtx); 738 break; 739 case TUNGIFHEAD: 740 /* Could be unlocked read? */ 741 mtx_lock(&tp->tun_mtx); 742 *(int *)data = (tp->tun_flags & TUN_IFHEAD) ? 1 : 0; 743 mtx_unlock(&tp->tun_mtx); 744 break; 745 case TUNSIFMODE: 746 /* deny this if UP */ 747 if (TUN2IFP(tp)->if_flags & IFF_UP) 748 return(EBUSY); 749 750 switch (*(int *)data & ~IFF_MULTICAST) { 751 case IFF_POINTOPOINT: 752 case IFF_BROADCAST: 753 TUN2IFP(tp)->if_flags &= 754 ~(IFF_BROADCAST|IFF_POINTOPOINT|IFF_MULTICAST); 755 TUN2IFP(tp)->if_flags |= *(int *)data; 756 break; 757 default: 758 return(EINVAL); 759 } 760 break; 761 case TUNSIFPID: 762 mtx_lock(&tp->tun_mtx); 763 tp->tun_pid = curthread->td_proc->p_pid; 764 mtx_unlock(&tp->tun_mtx); 765 break; 766 case FIONBIO: 767 break; 768 case FIOASYNC: 769 mtx_lock(&tp->tun_mtx); 770 if (*(int *)data) 771 tp->tun_flags |= TUN_ASYNC; 772 else 773 tp->tun_flags &= ~TUN_ASYNC; 774 mtx_unlock(&tp->tun_mtx); 775 break; 776 case FIONREAD: 777 s = splimp(); 778 if (!IFQ_IS_EMPTY(&TUN2IFP(tp)->if_snd)) { 779 struct mbuf *mb; 780 IFQ_LOCK(&TUN2IFP(tp)->if_snd); 781 IFQ_POLL_NOLOCK(&TUN2IFP(tp)->if_snd, mb); 782 for( *(int *)data = 0; mb != 0; mb = mb->m_next) 783 *(int *)data += mb->m_len; 784 IFQ_UNLOCK(&TUN2IFP(tp)->if_snd); 785 } else 786 *(int *)data = 0; 787 splx(s); 788 break; 789 case FIOSETOWN: 790 return (fsetown(*(int *)data, &tp->tun_sigio)); 791 792 case FIOGETOWN: 793 *(int *)data = fgetown(&tp->tun_sigio); 794 return (0); 795 796 /* This is deprecated, FIOSETOWN should be used instead. */ 797 case TIOCSPGRP: 798 return (fsetown(-(*(int *)data), &tp->tun_sigio)); 799 800 /* This is deprecated, FIOGETOWN should be used instead. */ 801 case TIOCGPGRP: 802 *(int *)data = -fgetown(&tp->tun_sigio); 803 return (0); 804 805 default: 806 return (ENOTTY); 807 } 808 return (0); 809} 810 811/* 812 * The cdevsw read interface - reads a packet at a time, or at 813 * least as much of a packet as can be read. 814 */ 815static int 816tunread(struct cdev *dev, struct uio *uio, int flag) 817{ 818 struct tun_softc *tp = dev->si_drv1; 819 struct ifnet *ifp = TUN2IFP(tp); 820 struct mbuf *m; 821 int error=0, len, s; 822 823 TUNDEBUG (ifp, "read\n"); 824 mtx_lock(&tp->tun_mtx); 825 if ((tp->tun_flags & TUN_READY) != TUN_READY) { 826 mtx_unlock(&tp->tun_mtx); 827 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 828 return (EHOSTDOWN); 829 } 830 831 tp->tun_flags &= ~TUN_RWAIT; 832 mtx_unlock(&tp->tun_mtx); 833 834 s = splimp(); 835 do { 836 IFQ_DEQUEUE(&ifp->if_snd, m); 837 if (m == NULL) { 838 if (flag & O_NONBLOCK) { 839 splx(s); 840 return (EWOULDBLOCK); 841 } 842 mtx_lock(&tp->tun_mtx); 843 tp->tun_flags |= TUN_RWAIT; 844 mtx_unlock(&tp->tun_mtx); 845 if ((error = tsleep(tp, PCATCH | (PZERO + 1), 846 "tunread", 0)) != 0) { 847 splx(s); 848 return (error); 849 } 850 } 851 } while (m == NULL); 852 splx(s); 853 854 while (m && uio->uio_resid > 0 && error == 0) { 855 len = min(uio->uio_resid, m->m_len); 856 if (len != 0) 857 error = uiomove(mtod(m, void *), len, uio); 858 m = m_free(m); 859 } 860 861 if (m) { 862 TUNDEBUG(ifp, "Dropping mbuf\n"); 863 m_freem(m); 864 } 865 return (error); 866} 867 868/* 869 * the cdevsw write interface - an atomic write is a packet - or else! 870 */ 871static int 872tunwrite(struct cdev *dev, struct uio *uio, int flag) 873{ 874 struct tun_softc *tp = dev->si_drv1; 875 struct ifnet *ifp = TUN2IFP(tp); 876 struct mbuf *m; 877 int error = 0; 878 uint32_t family; 879 int isr; 880 881 TUNDEBUG(ifp, "tunwrite\n"); 882 883 if ((ifp->if_flags & IFF_UP) != IFF_UP) 884 /* ignore silently */ 885 return (0); 886 887 if (uio->uio_resid == 0) 888 return (0); 889 890 if (uio->uio_resid < 0 || uio->uio_resid > TUNMRU) { 891 TUNDEBUG(ifp, "len=%d!\n", uio->uio_resid); 892 return (EIO); 893 } 894 895 if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, 0, M_PKTHDR)) == NULL) { 896 ifp->if_ierrors++; 897 return (error); 898 } 899 900 m->m_pkthdr.rcvif = ifp; 901#ifdef MAC 902 mac_ifnet_create_mbuf(ifp, m); 903#endif 904 905 /* Could be unlocked read? */ 906 mtx_lock(&tp->tun_mtx); 907 if (tp->tun_flags & TUN_IFHEAD) { 908 mtx_unlock(&tp->tun_mtx); 909 if (m->m_len < sizeof(family) && 910 (m = m_pullup(m, sizeof(family))) == NULL) 911 return (ENOBUFS); 912 family = ntohl(*mtod(m, u_int32_t *)); 913 m_adj(m, sizeof(family)); 914 } else { 915 mtx_unlock(&tp->tun_mtx); 916 family = AF_INET; 917 } 918 919 BPF_MTAP2(ifp, &family, sizeof(family), m); 920 921 switch (family) { 922#ifdef INET 923 case AF_INET: 924 isr = NETISR_IP; 925 break; 926#endif 927#ifdef INET6 928 case AF_INET6: 929 isr = NETISR_IPV6; 930 break; 931#endif 932#ifdef IPX 933 case AF_IPX: 934 isr = NETISR_IPX; 935 break; 936#endif 937#ifdef NETATALK 938 case AF_APPLETALK: 939 isr = NETISR_ATALK2; 940 break; 941#endif 942 default: 943 m_freem(m); 944 return (EAFNOSUPPORT); 945 } 946 /* First chunk of an mbuf contains good junk */ 947 if (harvest.point_to_point) 948 random_harvest(m, 16, 3, 0, RANDOM_NET); 949 ifp->if_ibytes += m->m_pkthdr.len; 950 ifp->if_ipackets++; 951 CURVNET_SET(ifp->if_vnet); 952 netisr_dispatch(isr, m); 953 CURVNET_RESTORE(); 954 return (0); 955} 956 957/* 958 * tunpoll - the poll interface, this is only useful on reads 959 * really. The write detect always returns true, write never blocks 960 * anyway, it either accepts the packet or drops it. 961 */ 962static int 963tunpoll(struct cdev *dev, int events, struct thread *td) 964{ 965 int s; 966 struct tun_softc *tp = dev->si_drv1; 967 struct ifnet *ifp = TUN2IFP(tp); 968 int revents = 0; 969 struct mbuf *m; 970 971 s = splimp(); 972 TUNDEBUG(ifp, "tunpoll\n"); 973 974 if (events & (POLLIN | POLLRDNORM)) { 975 IFQ_LOCK(&ifp->if_snd); 976 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 977 if (m != NULL) { 978 TUNDEBUG(ifp, "tunpoll q=%d\n", ifp->if_snd.ifq_len); 979 revents |= events & (POLLIN | POLLRDNORM); 980 } else { 981 TUNDEBUG(ifp, "tunpoll waiting\n"); 982 selrecord(td, &tp->tun_rsel); 983 } 984 IFQ_UNLOCK(&ifp->if_snd); 985 } 986 if (events & (POLLOUT | POLLWRNORM)) 987 revents |= events & (POLLOUT | POLLWRNORM); 988 989 splx(s); 990 return (revents); 991} 992 993/* 994 * tunkqfilter - support for the kevent() system call. 995 */ 996static int 997tunkqfilter(struct cdev *dev, struct knote *kn) 998{ 999 int s; 1000 struct tun_softc *tp = dev->si_drv1; 1001 struct ifnet *ifp = TUN2IFP(tp); 1002 1003 s = splimp(); 1004 switch(kn->kn_filter) { 1005 case EVFILT_READ: 1006 TUNDEBUG(ifp, "%s kqfilter: EVFILT_READ, minor = %#x\n", 1007 ifp->if_xname, dev2unit(dev)); 1008 kn->kn_fop = &tun_read_filterops; 1009 break; 1010 1011 case EVFILT_WRITE: 1012 TUNDEBUG(ifp, "%s kqfilter: EVFILT_WRITE, minor = %#x\n", 1013 ifp->if_xname, dev2unit(dev)); 1014 kn->kn_fop = &tun_write_filterops; 1015 break; 1016 1017 default: 1018 TUNDEBUG(ifp, "%s kqfilter: invalid filter, minor = %#x\n", 1019 ifp->if_xname, dev2unit(dev)); 1020 splx(s); 1021 return(EINVAL); 1022 } 1023 splx(s); 1024 1025 kn->kn_hook = (caddr_t) dev; 1026 knlist_add(&tp->tun_rsel.si_note, kn, 0); 1027 1028 return (0); 1029} 1030 1031/* 1032 * Return true of there is data in the interface queue. 1033 */ 1034static int 1035tunkqread(struct knote *kn, long hint) 1036{ 1037 int ret, s; 1038 struct cdev *dev = (struct cdev *)(kn->kn_hook); 1039 struct tun_softc *tp = dev->si_drv1; 1040 struct ifnet *ifp = TUN2IFP(tp); 1041 1042 s = splimp(); 1043 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1044 TUNDEBUG(ifp, 1045 "%s have data in the queue. Len = %d, minor = %#x\n", 1046 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev)); 1047 ret = 1; 1048 } else { 1049 TUNDEBUG(ifp, 1050 "%s waiting for data, minor = %#x\n", ifp->if_xname, 1051 dev2unit(dev)); 1052 ret = 0; 1053 } 1054 splx(s); 1055 1056 return (ret); 1057} 1058 1059/* 1060 * Always can write, always return MTU in kn->data. 1061 */ 1062static int 1063tunkqwrite(struct knote *kn, long hint) 1064{ 1065 int s; 1066 struct tun_softc *tp = ((struct cdev *)kn->kn_hook)->si_drv1; 1067 struct ifnet *ifp = TUN2IFP(tp); 1068 1069 s = splimp(); 1070 kn->kn_data = ifp->if_mtu; 1071 splx(s); 1072 1073 return (1); 1074} 1075 1076static void 1077tunkqdetach(struct knote *kn) 1078{ 1079 struct tun_softc *tp = ((struct cdev *)kn->kn_hook)->si_drv1; 1080 1081 knlist_remove(&tp->tun_rsel.si_note, kn, 0); 1082} 1083