if_tun.c revision 257176
1207151Smarius/* $NetBSD: if_tun.c,v 1.14 1994/06/29 06:36:25 cgd Exp $ */ 2207151Smarius 3207151Smarius/*- 4207151Smarius * Copyright (c) 1988, Julian Onions <jpo@cs.nott.ac.uk> 5207151Smarius * Nottingham University 1987. 6207151Smarius * 7207151Smarius * This source may be freely distributed, however I would be interested 8207151Smarius * in any changes that are made. 9207151Smarius * 10207151Smarius * This driver takes packets off the IP i/f and hands them up to a 11207151Smarius * user process to have its wicked way with. This driver has it's 12207151Smarius * roots in a similar driver written by Phil Cockcroft (formerly) at 13207151Smarius * UCL. This driver is based much more on read/write/poll mode of 14207151Smarius * operation though. 15207151Smarius * 16207151Smarius * $FreeBSD: head/sys/net/if_tun.c 257176 2013-10-26 17:58:36Z glebius $ 17207151Smarius */ 18207151Smarius 19207151Smarius#include "opt_atalk.h" 20207151Smarius#include "opt_inet.h" 21207151Smarius#include "opt_inet6.h" 22207151Smarius#include "opt_ipx.h" 23207151Smarius 24207151Smarius#include <sys/param.h> 25207151Smarius#include <sys/priv.h> 26207151Smarius#include <sys/proc.h> 27207151Smarius#include <sys/systm.h> 28207151Smarius#include <sys/jail.h> 29207151Smarius#include <sys/mbuf.h> 30207151Smarius#include <sys/module.h> 31207151Smarius#include <sys/socket.h> 32207151Smarius#include <sys/fcntl.h> 33207151Smarius#include <sys/filio.h> 34207151Smarius#include <sys/sockio.h> 35207151Smarius#include <sys/ttycom.h> 36207151Smarius#include <sys/poll.h> 37207151Smarius#include <sys/selinfo.h> 38207151Smarius#include <sys/signalvar.h> 39207151Smarius#include <sys/filedesc.h> 40207151Smarius#include <sys/kernel.h> 41207151Smarius#include <sys/sysctl.h> 42207151Smarius#include <sys/conf.h> 43207151Smarius#include <sys/uio.h> 44207151Smarius#include <sys/malloc.h> 45207151Smarius#include <sys/random.h> 46207151Smarius 47207151Smarius#include <net/if.h> 48207151Smarius#include <net/if_var.h> 49207151Smarius#include <net/if_clone.h> 50207151Smarius#include <net/if_types.h> 51207151Smarius#include <net/netisr.h> 52207151Smarius#include <net/route.h> 53207151Smarius#include <net/vnet.h> 54207151Smarius#ifdef INET 55207151Smarius#include <netinet/in.h> 56207151Smarius#endif 57207151Smarius#include <net/bpf.h> 58207151Smarius#include <net/if_tun.h> 59207151Smarius 60207151Smarius#include <sys/queue.h> 61207151Smarius#include <sys/condvar.h> 62207151Smarius 63207151Smarius#include <security/mac/mac_framework.h> 64 65/* 66 * tun_list is protected by global tunmtx. Other mutable fields are 67 * protected by tun->tun_mtx, or by their owning subsystem. tun_dev is 68 * static for the duration of a tunnel interface. 69 */ 70struct tun_softc { 71 TAILQ_ENTRY(tun_softc) tun_list; 72 struct cdev *tun_dev; 73 u_short tun_flags; /* misc flags */ 74#define TUN_OPEN 0x0001 75#define TUN_INITED 0x0002 76#define TUN_RCOLL 0x0004 77#define TUN_IASET 0x0008 78#define TUN_DSTADDR 0x0010 79#define TUN_LMODE 0x0020 80#define TUN_RWAIT 0x0040 81#define TUN_ASYNC 0x0080 82#define TUN_IFHEAD 0x0100 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 103 104/* 105 * All mutable global variables in if_tun are locked using tunmtx, with 106 * the exception of tundebug, which is used unlocked, and tunclones, 107 * which is static after setup. 108 */ 109static struct mtx tunmtx; 110static const char tunname[] = "tun"; 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); 119static SYSCTL_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 void tuninit(struct ifnet *); 131static int tunmodevent(module_t, int, void *); 132static int tunoutput(struct ifnet *, struct mbuf *, 133 const struct sockaddr *, struct route *ro); 134static void tunstart(struct ifnet *); 135 136static int tun_clone_create(struct if_clone *, int, caddr_t); 137static void tun_clone_destroy(struct ifnet *); 138static struct if_clone *tun_cloner; 139 140static d_open_t tunopen; 141static d_close_t tunclose; 142static d_read_t tunread; 143static d_write_t tunwrite; 144static d_ioctl_t tunioctl; 145static d_poll_t tunpoll; 146static d_kqfilter_t tunkqfilter; 147 148static int tunkqread(struct knote *, long); 149static int tunkqwrite(struct knote *, long); 150static void tunkqdetach(struct knote *); 151 152static struct filterops tun_read_filterops = { 153 .f_isfd = 1, 154 .f_attach = NULL, 155 .f_detach = tunkqdetach, 156 .f_event = tunkqread, 157}; 158 159static struct filterops tun_write_filterops = { 160 .f_isfd = 1, 161 .f_attach = NULL, 162 .f_detach = tunkqdetach, 163 .f_event = tunkqwrite, 164}; 165 166static struct cdevsw tun_cdevsw = { 167 .d_version = D_VERSION, 168 .d_flags = D_NEEDMINOR, 169 .d_open = tunopen, 170 .d_close = tunclose, 171 .d_read = tunread, 172 .d_write = tunwrite, 173 .d_ioctl = tunioctl, 174 .d_poll = tunpoll, 175 .d_kqfilter = tunkqfilter, 176 .d_name = tunname, 177}; 178 179static int 180tun_clone_create(struct if_clone *ifc, int unit, caddr_t params) 181{ 182 struct cdev *dev; 183 int i; 184 185 /* find any existing device, or allocate new unit number */ 186 i = clone_create(&tunclones, &tun_cdevsw, &unit, &dev, 0); 187 if (i) { 188 /* No preexisting struct cdev *, create one */ 189 dev = make_dev(&tun_cdevsw, unit, 190 UID_UUCP, GID_DIALER, 0600, "%s%d", tunname, unit); 191 } 192 tuncreate(tunname, dev); 193 194 return (0); 195} 196 197static void 198tunclone(void *arg, struct ucred *cred, char *name, int namelen, 199 struct cdev **dev) 200{ 201 char devname[SPECNAMELEN + 1]; 202 int u, i, append_unit; 203 204 if (*dev != NULL) 205 return; 206 207 /* 208 * If tun cloning is enabled, only the superuser can create an 209 * interface. 210 */ 211 if (!tundclone || priv_check_cred(cred, PRIV_NET_IFCREATE, 0) != 0) 212 return; 213 214 if (strcmp(name, tunname) == 0) { 215 u = -1; 216 } else if (dev_stdclone(name, NULL, tunname, &u) != 1) 217 return; /* Don't recognise the name */ 218 if (u != -1 && u > IF_MAXUNIT) 219 return; /* Unit number too high */ 220 221 if (u == -1) 222 append_unit = 1; 223 else 224 append_unit = 0; 225 226 CURVNET_SET(CRED_TO_VNET(cred)); 227 /* find any existing device, or allocate new unit number */ 228 i = clone_create(&tunclones, &tun_cdevsw, &u, dev, 0); 229 if (i) { 230 if (append_unit) { 231 namelen = snprintf(devname, sizeof(devname), "%s%d", 232 name, u); 233 name = devname; 234 } 235 /* No preexisting struct cdev *, create one */ 236 *dev = make_dev_credf(MAKEDEV_REF, &tun_cdevsw, u, cred, 237 UID_UUCP, GID_DIALER, 0600, "%s", name); 238 } 239 240 if_clone_create(name, namelen, NULL); 241 CURVNET_RESTORE(); 242} 243 244static void 245tun_destroy(struct tun_softc *tp) 246{ 247 struct cdev *dev; 248 249 mtx_lock(&tp->tun_mtx); 250 if ((tp->tun_flags & TUN_OPEN) != 0) 251 cv_wait_unlock(&tp->tun_cv, &tp->tun_mtx); 252 else 253 mtx_unlock(&tp->tun_mtx); 254 255 CURVNET_SET(TUN2IFP(tp)->if_vnet); 256 dev = tp->tun_dev; 257 bpfdetach(TUN2IFP(tp)); 258 if_detach(TUN2IFP(tp)); 259 if_free(TUN2IFP(tp)); 260 destroy_dev(dev); 261 seldrain(&tp->tun_rsel); 262 knlist_clear(&tp->tun_rsel.si_note, 0); 263 knlist_destroy(&tp->tun_rsel.si_note); 264 mtx_destroy(&tp->tun_mtx); 265 cv_destroy(&tp->tun_cv); 266 free(tp, M_TUN); 267 CURVNET_RESTORE(); 268} 269 270static void 271tun_clone_destroy(struct ifnet *ifp) 272{ 273 struct tun_softc *tp = ifp->if_softc; 274 275 mtx_lock(&tunmtx); 276 TAILQ_REMOVE(&tunhead, tp, tun_list); 277 mtx_unlock(&tunmtx); 278 tun_destroy(tp); 279} 280 281static int 282tunmodevent(module_t mod, int type, void *data) 283{ 284 static eventhandler_tag tag; 285 struct tun_softc *tp; 286 287 switch (type) { 288 case MOD_LOAD: 289 mtx_init(&tunmtx, "tunmtx", NULL, MTX_DEF); 290 clone_setup(&tunclones); 291 tag = EVENTHANDLER_REGISTER(dev_clone, tunclone, 0, 1000); 292 if (tag == NULL) 293 return (ENOMEM); 294 tun_cloner = if_clone_simple(tunname, tun_clone_create, 295 tun_clone_destroy, 0); 296 break; 297 case MOD_UNLOAD: 298 if_clone_detach(tun_cloner); 299 EVENTHANDLER_DEREGISTER(dev_clone, tag); 300 drain_dev_clone_events(); 301 302 mtx_lock(&tunmtx); 303 while ((tp = TAILQ_FIRST(&tunhead)) != NULL) { 304 TAILQ_REMOVE(&tunhead, tp, tun_list); 305 mtx_unlock(&tunmtx); 306 tun_destroy(tp); 307 mtx_lock(&tunmtx); 308 } 309 mtx_unlock(&tunmtx); 310 clone_cleanup(&tunclones); 311 mtx_destroy(&tunmtx); 312 break; 313 default: 314 return EOPNOTSUPP; 315 } 316 return 0; 317} 318 319static moduledata_t tun_mod = { 320 "if_tun", 321 tunmodevent, 322 0 323}; 324 325DECLARE_MODULE(if_tun, tun_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 326MODULE_VERSION(if_tun, 1); 327 328static void 329tunstart(struct ifnet *ifp) 330{ 331 struct tun_softc *tp = ifp->if_softc; 332 struct mbuf *m; 333 334 TUNDEBUG(ifp,"%s starting\n", ifp->if_xname); 335 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 336 IFQ_LOCK(&ifp->if_snd); 337 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 338 if (m == NULL) { 339 IFQ_UNLOCK(&ifp->if_snd); 340 return; 341 } 342 IFQ_UNLOCK(&ifp->if_snd); 343 } 344 345 mtx_lock(&tp->tun_mtx); 346 if (tp->tun_flags & TUN_RWAIT) { 347 tp->tun_flags &= ~TUN_RWAIT; 348 wakeup(tp); 349 } 350 selwakeuppri(&tp->tun_rsel, PZERO + 1); 351 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0); 352 if (tp->tun_flags & TUN_ASYNC && tp->tun_sigio) { 353 mtx_unlock(&tp->tun_mtx); 354 pgsigio(&tp->tun_sigio, SIGIO, 0); 355 } else 356 mtx_unlock(&tp->tun_mtx); 357} 358 359/* XXX: should return an error code so it can fail. */ 360static void 361tuncreate(const char *name, struct cdev *dev) 362{ 363 struct tun_softc *sc; 364 struct ifnet *ifp; 365 366 sc = malloc(sizeof(*sc), M_TUN, M_WAITOK | M_ZERO); 367 mtx_init(&sc->tun_mtx, "tun_mtx", NULL, MTX_DEF); 368 cv_init(&sc->tun_cv, "tun_condvar"); 369 sc->tun_flags = TUN_INITED; 370 sc->tun_dev = dev; 371 mtx_lock(&tunmtx); 372 TAILQ_INSERT_TAIL(&tunhead, sc, tun_list); 373 mtx_unlock(&tunmtx); 374 375 ifp = sc->tun_ifp = if_alloc(IFT_PPP); 376 if (ifp == NULL) 377 panic("%s%d: failed to if_alloc() interface.\n", 378 name, dev2unit(dev)); 379 if_initname(ifp, name, dev2unit(dev)); 380 ifp->if_mtu = TUNMTU; 381 ifp->if_ioctl = tunifioctl; 382 ifp->if_output = tunoutput; 383 ifp->if_start = tunstart; 384 ifp->if_flags = IFF_POINTOPOINT | IFF_MULTICAST; 385 ifp->if_softc = sc; 386 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 387 ifp->if_snd.ifq_drv_maxlen = 0; 388 IFQ_SET_READY(&ifp->if_snd); 389 knlist_init_mtx(&sc->tun_rsel.si_note, &sc->tun_mtx); 390 ifp->if_capabilities |= IFCAP_LINKSTATE; 391 ifp->if_capenable |= IFCAP_LINKSTATE; 392 393 if_attach(ifp); 394 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); 395 dev->si_drv1 = sc; 396 TUNDEBUG(ifp, "interface %s is created, minor = %#x\n", 397 ifp->if_xname, dev2unit(dev)); 398} 399 400static int 401tunopen(struct cdev *dev, int flag, int mode, struct thread *td) 402{ 403 struct ifnet *ifp; 404 struct tun_softc *tp; 405 406 /* 407 * XXXRW: Non-atomic test and set of dev->si_drv1 requires 408 * synchronization. 409 */ 410 tp = dev->si_drv1; 411 if (!tp) { 412 tuncreate(tunname, dev); 413 tp = dev->si_drv1; 414 } 415 416 /* 417 * XXXRW: This use of tun_pid is subject to error due to the 418 * fact that a reference to the tunnel can live beyond the 419 * death of the process that created it. Can we replace this 420 * with a simple busy flag? 421 */ 422 mtx_lock(&tp->tun_mtx); 423 if (tp->tun_pid != 0 && tp->tun_pid != td->td_proc->p_pid) { 424 mtx_unlock(&tp->tun_mtx); 425 return (EBUSY); 426 } 427 tp->tun_pid = td->td_proc->p_pid; 428 429 tp->tun_flags |= TUN_OPEN; 430 ifp = TUN2IFP(tp); 431 if_link_state_change(ifp, LINK_STATE_UP); 432 TUNDEBUG(ifp, "open\n"); 433 mtx_unlock(&tp->tun_mtx); 434 435 return (0); 436} 437 438/* 439 * tunclose - close the device - mark i/f down & delete 440 * routing info 441 */ 442static int 443tunclose(struct cdev *dev, int foo, int bar, struct thread *td) 444{ 445 struct tun_softc *tp; 446 struct ifnet *ifp; 447 448 tp = dev->si_drv1; 449 ifp = TUN2IFP(tp); 450 451 mtx_lock(&tp->tun_mtx); 452 tp->tun_flags &= ~TUN_OPEN; 453 tp->tun_pid = 0; 454 455 /* 456 * junk all pending output 457 */ 458 CURVNET_SET(ifp->if_vnet); 459 IFQ_PURGE(&ifp->if_snd); 460 461 if (ifp->if_flags & IFF_UP) { 462 mtx_unlock(&tp->tun_mtx); 463 if_down(ifp); 464 mtx_lock(&tp->tun_mtx); 465 } 466 467 /* Delete all addresses and routes which reference this interface. */ 468 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 469 struct ifaddr *ifa; 470 471 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 472 mtx_unlock(&tp->tun_mtx); 473 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 474 /* deal w/IPv4 PtP destination; unlocked read */ 475 if (ifa->ifa_addr->sa_family == AF_INET) { 476 rtinit(ifa, (int)RTM_DELETE, 477 tp->tun_flags & TUN_DSTADDR ? RTF_HOST : 0); 478 } else { 479 rtinit(ifa, (int)RTM_DELETE, 0); 480 } 481 } 482 if_purgeaddrs(ifp); 483 mtx_lock(&tp->tun_mtx); 484 } 485 if_link_state_change(ifp, LINK_STATE_DOWN); 486 CURVNET_RESTORE(); 487 488 funsetown(&tp->tun_sigio); 489 selwakeuppri(&tp->tun_rsel, PZERO + 1); 490 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0); 491 TUNDEBUG (ifp, "closed\n"); 492 493 cv_broadcast(&tp->tun_cv); 494 mtx_unlock(&tp->tun_mtx); 495 return (0); 496} 497 498static void 499tuninit(struct ifnet *ifp) 500{ 501 struct tun_softc *tp = ifp->if_softc; 502#ifdef INET 503 struct ifaddr *ifa; 504#endif 505 506 TUNDEBUG(ifp, "tuninit\n"); 507 508 mtx_lock(&tp->tun_mtx); 509 ifp->if_flags |= IFF_UP; 510 ifp->if_drv_flags |= IFF_DRV_RUNNING; 511 getmicrotime(&ifp->if_lastchange); 512 513#ifdef INET 514 if_addr_rlock(ifp); 515 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 516 if (ifa->ifa_addr->sa_family == AF_INET) { 517 struct sockaddr_in *si; 518 519 si = (struct sockaddr_in *)ifa->ifa_addr; 520 if (si->sin_addr.s_addr) 521 tp->tun_flags |= TUN_IASET; 522 523 si = (struct sockaddr_in *)ifa->ifa_dstaddr; 524 if (si && si->sin_addr.s_addr) 525 tp->tun_flags |= TUN_DSTADDR; 526 } 527 } 528 if_addr_runlock(ifp); 529#endif 530 mtx_unlock(&tp->tun_mtx); 531} 532 533/* 534 * Process an ioctl request. 535 */ 536static int 537tunifioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 538{ 539 struct ifreq *ifr = (struct ifreq *)data; 540 struct tun_softc *tp = ifp->if_softc; 541 struct ifstat *ifs; 542 int error = 0; 543 544 switch(cmd) { 545 case SIOCGIFSTATUS: 546 ifs = (struct ifstat *)data; 547 mtx_lock(&tp->tun_mtx); 548 if (tp->tun_pid) 549 sprintf(ifs->ascii + strlen(ifs->ascii), 550 "\tOpened by PID %d\n", tp->tun_pid); 551 mtx_unlock(&tp->tun_mtx); 552 break; 553 case SIOCSIFADDR: 554 tuninit(ifp); 555 TUNDEBUG(ifp, "address set\n"); 556 break; 557 case SIOCSIFMTU: 558 ifp->if_mtu = ifr->ifr_mtu; 559 TUNDEBUG(ifp, "mtu set\n"); 560 break; 561 case SIOCSIFFLAGS: 562 case SIOCADDMULTI: 563 case SIOCDELMULTI: 564 break; 565 default: 566 error = EINVAL; 567 } 568 return (error); 569} 570 571/* 572 * tunoutput - queue packets from higher level ready to put out. 573 */ 574static int 575tunoutput(struct ifnet *ifp, struct mbuf *m0, const struct sockaddr *dst, 576 struct route *ro) 577{ 578 struct tun_softc *tp = ifp->if_softc; 579 u_short cached_tun_flags; 580 int error; 581 u_int32_t af; 582 583 TUNDEBUG (ifp, "tunoutput\n"); 584 585#ifdef MAC 586 error = mac_ifnet_check_transmit(ifp, m0); 587 if (error) { 588 m_freem(m0); 589 return (error); 590 } 591#endif 592 593 /* Could be unlocked read? */ 594 mtx_lock(&tp->tun_mtx); 595 cached_tun_flags = tp->tun_flags; 596 mtx_unlock(&tp->tun_mtx); 597 if ((cached_tun_flags & TUN_READY) != TUN_READY) { 598 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 599 m_freem (m0); 600 return (EHOSTDOWN); 601 } 602 603 if ((ifp->if_flags & IFF_UP) != IFF_UP) { 604 m_freem (m0); 605 return (EHOSTDOWN); 606 } 607 608 /* BPF writes need to be handled specially. */ 609 if (dst->sa_family == AF_UNSPEC) 610 bcopy(dst->sa_data, &af, sizeof(af)); 611 else 612 af = dst->sa_family; 613 614 if (bpf_peers_present(ifp->if_bpf)) 615 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m0); 616 617 /* prepend sockaddr? this may abort if the mbuf allocation fails */ 618 if (cached_tun_flags & TUN_LMODE) { 619 /* allocate space for sockaddr */ 620 M_PREPEND(m0, dst->sa_len, M_NOWAIT); 621 622 /* if allocation failed drop packet */ 623 if (m0 == NULL) { 624 ifp->if_iqdrops++; 625 ifp->if_oerrors++; 626 return (ENOBUFS); 627 } else { 628 bcopy(dst, m0->m_data, dst->sa_len); 629 } 630 } 631 632 if (cached_tun_flags & TUN_IFHEAD) { 633 /* Prepend the address family */ 634 M_PREPEND(m0, 4, M_NOWAIT); 635 636 /* if allocation failed drop packet */ 637 if (m0 == NULL) { 638 ifp->if_iqdrops++; 639 ifp->if_oerrors++; 640 return (ENOBUFS); 641 } else 642 *(u_int32_t *)m0->m_data = htonl(af); 643 } else { 644#ifdef INET 645 if (af != AF_INET) 646#endif 647 { 648 m_freem(m0); 649 return (EAFNOSUPPORT); 650 } 651 } 652 653 error = (ifp->if_transmit)(ifp, m0); 654 if (error) 655 return (ENOBUFS); 656 ifp->if_opackets++; 657 return (0); 658} 659 660/* 661 * the cdevsw interface is now pretty minimal. 662 */ 663static int 664tunioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, 665 struct thread *td) 666{ 667 int error; 668 struct tun_softc *tp = dev->si_drv1; 669 struct tuninfo *tunp; 670 671 switch (cmd) { 672 case TUNSIFINFO: 673 tunp = (struct tuninfo *)data; 674 if (tunp->mtu < IF_MINMTU) 675 return (EINVAL); 676 if (TUN2IFP(tp)->if_mtu != tunp->mtu) { 677 error = priv_check(td, PRIV_NET_SETIFMTU); 678 if (error) 679 return (error); 680 } 681 mtx_lock(&tp->tun_mtx); 682 TUN2IFP(tp)->if_mtu = tunp->mtu; 683 TUN2IFP(tp)->if_type = tunp->type; 684 TUN2IFP(tp)->if_baudrate = tunp->baudrate; 685 mtx_unlock(&tp->tun_mtx); 686 break; 687 case TUNGIFINFO: 688 tunp = (struct tuninfo *)data; 689 mtx_lock(&tp->tun_mtx); 690 tunp->mtu = TUN2IFP(tp)->if_mtu; 691 tunp->type = TUN2IFP(tp)->if_type; 692 tunp->baudrate = TUN2IFP(tp)->if_baudrate; 693 mtx_unlock(&tp->tun_mtx); 694 break; 695 case TUNSDEBUG: 696 tundebug = *(int *)data; 697 break; 698 case TUNGDEBUG: 699 *(int *)data = tundebug; 700 break; 701 case TUNSLMODE: 702 mtx_lock(&tp->tun_mtx); 703 if (*(int *)data) { 704 tp->tun_flags |= TUN_LMODE; 705 tp->tun_flags &= ~TUN_IFHEAD; 706 } else 707 tp->tun_flags &= ~TUN_LMODE; 708 mtx_unlock(&tp->tun_mtx); 709 break; 710 case TUNSIFHEAD: 711 mtx_lock(&tp->tun_mtx); 712 if (*(int *)data) { 713 tp->tun_flags |= TUN_IFHEAD; 714 tp->tun_flags &= ~TUN_LMODE; 715 } else 716 tp->tun_flags &= ~TUN_IFHEAD; 717 mtx_unlock(&tp->tun_mtx); 718 break; 719 case TUNGIFHEAD: 720 mtx_lock(&tp->tun_mtx); 721 *(int *)data = (tp->tun_flags & TUN_IFHEAD) ? 1 : 0; 722 mtx_unlock(&tp->tun_mtx); 723 break; 724 case TUNSIFMODE: 725 /* deny this if UP */ 726 if (TUN2IFP(tp)->if_flags & IFF_UP) 727 return(EBUSY); 728 729 switch (*(int *)data & ~IFF_MULTICAST) { 730 case IFF_POINTOPOINT: 731 case IFF_BROADCAST: 732 mtx_lock(&tp->tun_mtx); 733 TUN2IFP(tp)->if_flags &= 734 ~(IFF_BROADCAST|IFF_POINTOPOINT|IFF_MULTICAST); 735 TUN2IFP(tp)->if_flags |= *(int *)data; 736 mtx_unlock(&tp->tun_mtx); 737 break; 738 default: 739 return(EINVAL); 740 } 741 break; 742 case TUNSIFPID: 743 mtx_lock(&tp->tun_mtx); 744 tp->tun_pid = curthread->td_proc->p_pid; 745 mtx_unlock(&tp->tun_mtx); 746 break; 747 case FIONBIO: 748 break; 749 case FIOASYNC: 750 mtx_lock(&tp->tun_mtx); 751 if (*(int *)data) 752 tp->tun_flags |= TUN_ASYNC; 753 else 754 tp->tun_flags &= ~TUN_ASYNC; 755 mtx_unlock(&tp->tun_mtx); 756 break; 757 case FIONREAD: 758 if (!IFQ_IS_EMPTY(&TUN2IFP(tp)->if_snd)) { 759 struct mbuf *mb; 760 IFQ_LOCK(&TUN2IFP(tp)->if_snd); 761 IFQ_POLL_NOLOCK(&TUN2IFP(tp)->if_snd, mb); 762 for (*(int *)data = 0; mb != NULL; mb = mb->m_next) 763 *(int *)data += mb->m_len; 764 IFQ_UNLOCK(&TUN2IFP(tp)->if_snd); 765 } else 766 *(int *)data = 0; 767 break; 768 case FIOSETOWN: 769 return (fsetown(*(int *)data, &tp->tun_sigio)); 770 771 case FIOGETOWN: 772 *(int *)data = fgetown(&tp->tun_sigio); 773 return (0); 774 775 /* This is deprecated, FIOSETOWN should be used instead. */ 776 case TIOCSPGRP: 777 return (fsetown(-(*(int *)data), &tp->tun_sigio)); 778 779 /* This is deprecated, FIOGETOWN should be used instead. */ 780 case TIOCGPGRP: 781 *(int *)data = -fgetown(&tp->tun_sigio); 782 return (0); 783 784 default: 785 return (ENOTTY); 786 } 787 return (0); 788} 789 790/* 791 * The cdevsw read interface - reads a packet at a time, or at 792 * least as much of a packet as can be read. 793 */ 794static int 795tunread(struct cdev *dev, struct uio *uio, int flag) 796{ 797 struct tun_softc *tp = dev->si_drv1; 798 struct ifnet *ifp = TUN2IFP(tp); 799 struct mbuf *m; 800 int error=0, len; 801 802 TUNDEBUG (ifp, "read\n"); 803 mtx_lock(&tp->tun_mtx); 804 if ((tp->tun_flags & TUN_READY) != TUN_READY) { 805 mtx_unlock(&tp->tun_mtx); 806 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 807 return (EHOSTDOWN); 808 } 809 810 tp->tun_flags &= ~TUN_RWAIT; 811 812 do { 813 IFQ_DEQUEUE(&ifp->if_snd, m); 814 if (m == NULL) { 815 if (flag & O_NONBLOCK) { 816 mtx_unlock(&tp->tun_mtx); 817 return (EWOULDBLOCK); 818 } 819 tp->tun_flags |= TUN_RWAIT; 820 error = mtx_sleep(tp, &tp->tun_mtx, PCATCH | (PZERO + 1), 821 "tunread", 0); 822 if (error != 0) { 823 mtx_unlock(&tp->tun_mtx); 824 return (error); 825 } 826 } 827 } while (m == NULL); 828 mtx_unlock(&tp->tun_mtx); 829 830 while (m && uio->uio_resid > 0 && error == 0) { 831 len = min(uio->uio_resid, m->m_len); 832 if (len != 0) 833 error = uiomove(mtod(m, void *), len, uio); 834 m = m_free(m); 835 } 836 837 if (m) { 838 TUNDEBUG(ifp, "Dropping mbuf\n"); 839 m_freem(m); 840 } 841 return (error); 842} 843 844/* 845 * the cdevsw write interface - an atomic write is a packet - or else! 846 */ 847static int 848tunwrite(struct cdev *dev, struct uio *uio, int flag) 849{ 850 struct tun_softc *tp = dev->si_drv1; 851 struct ifnet *ifp = TUN2IFP(tp); 852 struct mbuf *m; 853 uint32_t family; 854 int isr; 855 856 TUNDEBUG(ifp, "tunwrite\n"); 857 858 if ((ifp->if_flags & IFF_UP) != IFF_UP) 859 /* ignore silently */ 860 return (0); 861 862 if (uio->uio_resid == 0) 863 return (0); 864 865 if (uio->uio_resid < 0 || uio->uio_resid > TUNMRU) { 866 TUNDEBUG(ifp, "len=%zd!\n", uio->uio_resid); 867 return (EIO); 868 } 869 870 if ((m = m_uiotombuf(uio, M_NOWAIT, 0, 0, M_PKTHDR)) == NULL) { 871 ifp->if_ierrors++; 872 return (ENOBUFS); 873 } 874 875 m->m_pkthdr.rcvif = ifp; 876#ifdef MAC 877 mac_ifnet_create_mbuf(ifp, m); 878#endif 879 880 /* Could be unlocked read? */ 881 mtx_lock(&tp->tun_mtx); 882 if (tp->tun_flags & TUN_IFHEAD) { 883 mtx_unlock(&tp->tun_mtx); 884 if (m->m_len < sizeof(family) && 885 (m = m_pullup(m, sizeof(family))) == NULL) 886 return (ENOBUFS); 887 family = ntohl(*mtod(m, u_int32_t *)); 888 m_adj(m, sizeof(family)); 889 } else { 890 mtx_unlock(&tp->tun_mtx); 891 family = AF_INET; 892 } 893 894 BPF_MTAP2(ifp, &family, sizeof(family), m); 895 896 switch (family) { 897#ifdef INET 898 case AF_INET: 899 isr = NETISR_IP; 900 break; 901#endif 902#ifdef INET6 903 case AF_INET6: 904 isr = NETISR_IPV6; 905 break; 906#endif 907#ifdef IPX 908 case AF_IPX: 909 isr = NETISR_IPX; 910 break; 911#endif 912#ifdef NETATALK 913 case AF_APPLETALK: 914 isr = NETISR_ATALK2; 915 break; 916#endif 917 default: 918 m_freem(m); 919 return (EAFNOSUPPORT); 920 } 921 if (harvest.point_to_point) 922 random_harvest(&(m->m_data), 12, 2, RANDOM_NET_TUN); 923 ifp->if_ibytes += m->m_pkthdr.len; 924 ifp->if_ipackets++; 925 CURVNET_SET(ifp->if_vnet); 926 M_SETFIB(m, ifp->if_fib); 927 netisr_dispatch(isr, m); 928 CURVNET_RESTORE(); 929 return (0); 930} 931 932/* 933 * tunpoll - the poll interface, this is only useful on reads 934 * really. The write detect always returns true, write never blocks 935 * anyway, it either accepts the packet or drops it. 936 */ 937static int 938tunpoll(struct cdev *dev, int events, struct thread *td) 939{ 940 struct tun_softc *tp = dev->si_drv1; 941 struct ifnet *ifp = TUN2IFP(tp); 942 int revents = 0; 943 struct mbuf *m; 944 945 TUNDEBUG(ifp, "tunpoll\n"); 946 947 if (events & (POLLIN | POLLRDNORM)) { 948 IFQ_LOCK(&ifp->if_snd); 949 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 950 if (m != NULL) { 951 TUNDEBUG(ifp, "tunpoll q=%d\n", ifp->if_snd.ifq_len); 952 revents |= events & (POLLIN | POLLRDNORM); 953 } else { 954 TUNDEBUG(ifp, "tunpoll waiting\n"); 955 selrecord(td, &tp->tun_rsel); 956 } 957 IFQ_UNLOCK(&ifp->if_snd); 958 } 959 if (events & (POLLOUT | POLLWRNORM)) 960 revents |= events & (POLLOUT | POLLWRNORM); 961 962 return (revents); 963} 964 965/* 966 * tunkqfilter - support for the kevent() system call. 967 */ 968static int 969tunkqfilter(struct cdev *dev, struct knote *kn) 970{ 971 struct tun_softc *tp = dev->si_drv1; 972 struct ifnet *ifp = TUN2IFP(tp); 973 974 switch(kn->kn_filter) { 975 case EVFILT_READ: 976 TUNDEBUG(ifp, "%s kqfilter: EVFILT_READ, minor = %#x\n", 977 ifp->if_xname, dev2unit(dev)); 978 kn->kn_fop = &tun_read_filterops; 979 break; 980 981 case EVFILT_WRITE: 982 TUNDEBUG(ifp, "%s kqfilter: EVFILT_WRITE, minor = %#x\n", 983 ifp->if_xname, dev2unit(dev)); 984 kn->kn_fop = &tun_write_filterops; 985 break; 986 987 default: 988 TUNDEBUG(ifp, "%s kqfilter: invalid filter, minor = %#x\n", 989 ifp->if_xname, dev2unit(dev)); 990 return(EINVAL); 991 } 992 993 kn->kn_hook = tp; 994 knlist_add(&tp->tun_rsel.si_note, kn, 0); 995 996 return (0); 997} 998 999/* 1000 * Return true of there is data in the interface queue. 1001 */ 1002static int 1003tunkqread(struct knote *kn, long hint) 1004{ 1005 int ret; 1006 struct tun_softc *tp = kn->kn_hook; 1007 struct cdev *dev = tp->tun_dev; 1008 struct ifnet *ifp = TUN2IFP(tp); 1009 1010 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1011 TUNDEBUG(ifp, 1012 "%s have data in the queue. Len = %d, minor = %#x\n", 1013 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev)); 1014 ret = 1; 1015 } else { 1016 TUNDEBUG(ifp, 1017 "%s waiting for data, minor = %#x\n", ifp->if_xname, 1018 dev2unit(dev)); 1019 ret = 0; 1020 } 1021 1022 return (ret); 1023} 1024 1025/* 1026 * Always can write, always return MTU in kn->data. 1027 */ 1028static int 1029tunkqwrite(struct knote *kn, long hint) 1030{ 1031 struct tun_softc *tp = kn->kn_hook; 1032 struct ifnet *ifp = TUN2IFP(tp); 1033 1034 kn->kn_data = ifp->if_mtu; 1035 1036 return (1); 1037} 1038 1039static void 1040tunkqdetach(struct knote *kn) 1041{ 1042 struct tun_softc *tp = kn->kn_hook; 1043 1044 knlist_remove(&tp->tun_rsel.si_note, kn, 0); 1045} 1046