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 * $FreeBSD$ 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#define TUNNAME "tun" 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 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 *, struct sockaddr *, 132 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 *); 137 138IFC_SIMPLE_DECLARE(tun, 0); 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_PSEUDO | 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", ifc->ifc_name, unit); 191 } 192 tuncreate(ifc->ifc_name, 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 /* Unlocked read. */ 250 mtx_lock(&tp->tun_mtx); 251 if ((tp->tun_flags & TUN_OPEN) != 0) 252 cv_wait_unlock(&tp->tun_cv, &tp->tun_mtx); 253 else 254 mtx_unlock(&tp->tun_mtx); 255 256 CURVNET_SET(TUN2IFP(tp)->if_vnet); 257 dev = tp->tun_dev; 258 bpfdetach(TUN2IFP(tp)); 259 if_detach(TUN2IFP(tp)); 260 if_free(TUN2IFP(tp)); 261 destroy_dev(dev); 262 seldrain(&tp->tun_rsel); 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 if_clone_attach(&tun_cloner); 295 break; 296 case MOD_UNLOAD: 297 if_clone_detach(&tun_cloner); 298 EVENTHANDLER_DEREGISTER(dev_clone, tag); 299 drain_dev_clone_events(); 300 301 mtx_lock(&tunmtx); 302 while ((tp = TAILQ_FIRST(&tunhead)) != NULL) { 303 TAILQ_REMOVE(&tunhead, tp, tun_list); 304 mtx_unlock(&tunmtx); 305 tun_destroy(tp); 306 mtx_lock(&tunmtx); 307 } 308 mtx_unlock(&tunmtx); 309 clone_cleanup(&tunclones); 310 mtx_destroy(&tunmtx); 311 break; 312 default: 313 return EOPNOTSUPP; 314 } 315 return 0; 316} 317 318static moduledata_t tun_mod = { 319 "if_tun", 320 tunmodevent, 321 0 322}; 323 324DECLARE_MODULE(if_tun, tun_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 325MODULE_VERSION(if_tun, 1); 326 327static void 328tunstart(struct ifnet *ifp) 329{ 330 struct tun_softc *tp = ifp->if_softc; 331 struct mbuf *m; 332 333 TUNDEBUG(ifp,"%s starting\n", ifp->if_xname); 334 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 335 IFQ_LOCK(&ifp->if_snd); 336 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 337 if (m == NULL) { 338 IFQ_UNLOCK(&ifp->if_snd); 339 return; 340 } 341 IFQ_UNLOCK(&ifp->if_snd); 342 } 343 344 mtx_lock(&tp->tun_mtx); 345 if (tp->tun_flags & TUN_RWAIT) { 346 tp->tun_flags &= ~TUN_RWAIT; 347 wakeup(tp); 348 } 349 selwakeuppri(&tp->tun_rsel, PZERO + 1); 350 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0); 351 if (tp->tun_flags & TUN_ASYNC && tp->tun_sigio) { 352 mtx_unlock(&tp->tun_mtx); 353 pgsigio(&tp->tun_sigio, SIGIO, 0); 354 } else 355 mtx_unlock(&tp->tun_mtx); 356} 357 358/* XXX: should return an error code so it can fail. */ 359static void 360tuncreate(const char *name, struct cdev *dev) 361{ 362 struct tun_softc *sc; 363 struct ifnet *ifp; 364 365 dev->si_flags &= ~SI_CHEAPCLONE; 366 367 sc = malloc(sizeof(*sc), M_TUN, M_WAITOK | M_ZERO); 368 mtx_init(&sc->tun_mtx, "tun_mtx", NULL, MTX_DEF); 369 cv_init(&sc->tun_cv, "tun_condvar"); 370 sc->tun_flags = TUN_INITED; 371 sc->tun_dev = dev; 372 mtx_lock(&tunmtx); 373 TAILQ_INSERT_TAIL(&tunhead, sc, tun_list); 374 mtx_unlock(&tunmtx); 375 376 ifp = sc->tun_ifp = if_alloc(IFT_PPP); 377 if (ifp == NULL) 378 panic("%s%d: failed to if_alloc() interface.\n", 379 name, dev2unit(dev)); 380 if_initname(ifp, name, dev2unit(dev)); 381 ifp->if_mtu = TUNMTU; 382 ifp->if_ioctl = tunifioctl; 383 ifp->if_output = tunoutput; 384 ifp->if_start = tunstart; 385 ifp->if_flags = IFF_POINTOPOINT | IFF_MULTICAST; 386 ifp->if_softc = sc; 387 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 388 ifp->if_snd.ifq_drv_maxlen = 0; 389 IFQ_SET_READY(&ifp->if_snd); 390 knlist_init_mtx(&sc->tun_rsel.si_note, &sc->tun_mtx); 391 ifp->if_capabilities |= IFCAP_LINKSTATE; 392 ifp->if_capenable |= IFCAP_LINKSTATE; 393 394 if_attach(ifp); 395 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); 396 dev->si_drv1 = sc; 397 TUNDEBUG(ifp, "interface %s is created, minor = %#x\n", 398 ifp->if_xname, dev2unit(dev)); 399} 400 401static int 402tunopen(struct cdev *dev, int flag, int mode, struct thread *td) 403{ 404 struct ifnet *ifp; 405 struct tun_softc *tp; 406 407 /* 408 * XXXRW: Non-atomic test and set of dev->si_drv1 requires 409 * synchronization. 410 */ 411 tp = dev->si_drv1; 412 if (!tp) { 413 tuncreate(TUNNAME, dev); 414 tp = dev->si_drv1; 415 } 416 417 /* 418 * XXXRW: This use of tun_pid is subject to error due to the 419 * fact that a reference to the tunnel can live beyond the 420 * death of the process that created it. Can we replace this 421 * with a simple busy flag? 422 */ 423 mtx_lock(&tp->tun_mtx); 424 if (tp->tun_pid != 0 && tp->tun_pid != td->td_proc->p_pid) { 425 mtx_unlock(&tp->tun_mtx); 426 return (EBUSY); 427 } 428 tp->tun_pid = td->td_proc->p_pid; 429 430 tp->tun_flags |= TUN_OPEN; 431 ifp = TUN2IFP(tp); 432 if_link_state_change(ifp, LINK_STATE_UP); 433 TUNDEBUG(ifp, "open\n"); 434 mtx_unlock(&tp->tun_mtx); 435 436 return (0); 437} 438 439/* 440 * tunclose - close the device - mark i/f down & delete 441 * routing info 442 */ 443static int 444tunclose(struct cdev *dev, int foo, int bar, struct thread *td) 445{ 446 struct tun_softc *tp; 447 struct ifnet *ifp; 448 449 tp = dev->si_drv1; 450 ifp = TUN2IFP(tp); 451 452 mtx_lock(&tp->tun_mtx); 453 tp->tun_flags &= ~TUN_OPEN; 454 tp->tun_pid = 0; 455 456 /* 457 * junk all pending output 458 */ 459 CURVNET_SET(ifp->if_vnet); 460 IFQ_PURGE(&ifp->if_snd); 461 462 if (ifp->if_flags & IFF_UP) { 463 mtx_unlock(&tp->tun_mtx); 464 if_down(ifp); 465 mtx_lock(&tp->tun_mtx); 466 } 467 468 /* Delete all addresses and routes which reference this interface. */ 469 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 470 struct ifaddr *ifa; 471 472 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 473 mtx_unlock(&tp->tun_mtx); 474 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 475 /* deal w/IPv4 PtP destination; unlocked read */ 476 if (ifa->ifa_addr->sa_family == AF_INET) { 477 rtinit(ifa, (int)RTM_DELETE, 478 tp->tun_flags & TUN_DSTADDR ? RTF_HOST : 0); 479 } else { 480 rtinit(ifa, (int)RTM_DELETE, 0); 481 } 482 } 483 if_purgeaddrs(ifp); 484 mtx_lock(&tp->tun_mtx); 485 } 486 if_link_state_change(ifp, LINK_STATE_DOWN); 487 CURVNET_RESTORE(); 488 489 funsetown(&tp->tun_sigio); 490 selwakeuppri(&tp->tun_rsel, PZERO + 1); 491 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0); 492 TUNDEBUG (ifp, "closed\n"); 493 494 cv_broadcast(&tp->tun_cv); 495 mtx_unlock(&tp->tun_mtx); 496 return (0); 497} 498 499static void 500tuninit(struct ifnet *ifp) 501{ 502 struct tun_softc *tp = ifp->if_softc; 503#ifdef INET 504 struct ifaddr *ifa; 505#endif 506 507 TUNDEBUG(ifp, "tuninit\n"); 508 509 mtx_lock(&tp->tun_mtx); 510 ifp->if_flags |= IFF_UP; 511 ifp->if_drv_flags |= IFF_DRV_RUNNING; 512 getmicrotime(&ifp->if_lastchange); 513 514#ifdef INET 515 if_addr_rlock(ifp); 516 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 517 if (ifa->ifa_addr->sa_family == AF_INET) { 518 struct sockaddr_in *si; 519 520 si = (struct sockaddr_in *)ifa->ifa_addr; 521 if (si->sin_addr.s_addr) 522 tp->tun_flags |= TUN_IASET; 523 524 si = (struct sockaddr_in *)ifa->ifa_dstaddr; 525 if (si && si->sin_addr.s_addr) 526 tp->tun_flags |= TUN_DSTADDR; 527 } 528 } 529 if_addr_runlock(ifp); 530#endif 531 mtx_unlock(&tp->tun_mtx); 532} 533 534/* 535 * Process an ioctl request. 536 */ 537static int 538tunifioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 539{ 540 struct ifreq *ifr = (struct ifreq *)data; 541 struct tun_softc *tp = ifp->if_softc; 542 struct ifstat *ifs; 543 int error = 0; 544 545 switch(cmd) { 546 case SIOCGIFSTATUS: 547 ifs = (struct ifstat *)data; 548 mtx_lock(&tp->tun_mtx); 549 if (tp->tun_pid) 550 sprintf(ifs->ascii + strlen(ifs->ascii), 551 "\tOpened by PID %d\n", tp->tun_pid); 552 mtx_unlock(&tp->tun_mtx); 553 break; 554 case SIOCSIFADDR: 555 tuninit(ifp); 556 TUNDEBUG(ifp, "address set\n"); 557 break; 558 case SIOCSIFDSTADDR: 559 tuninit(ifp); 560 TUNDEBUG(ifp, "destination address set\n"); 561 break; 562 case SIOCSIFMTU: 563 ifp->if_mtu = ifr->ifr_mtu; 564 TUNDEBUG(ifp, "mtu set\n"); 565 break; 566 case SIOCSIFFLAGS: 567 case SIOCADDMULTI: 568 case SIOCDELMULTI: 569 break; 570 default: 571 error = EINVAL; 572 } 573 return (error); 574} 575 576/* 577 * tunoutput - queue packets from higher level ready to put out. 578 */ 579static int 580tunoutput(struct ifnet *ifp, struct mbuf *m0, struct sockaddr *dst, 581 struct route *ro) 582{ 583 struct tun_softc *tp = ifp->if_softc; 584 u_short cached_tun_flags; 585 int error; 586 u_int32_t af; 587 588 TUNDEBUG (ifp, "tunoutput\n"); 589 590#ifdef MAC 591 error = mac_ifnet_check_transmit(ifp, m0); 592 if (error) { 593 m_freem(m0); 594 return (error); 595 } 596#endif 597 598 /* Could be unlocked read? */ 599 mtx_lock(&tp->tun_mtx); 600 cached_tun_flags = tp->tun_flags; 601 mtx_unlock(&tp->tun_mtx); 602 if ((cached_tun_flags & TUN_READY) != TUN_READY) { 603 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 604 m_freem (m0); 605 return (EHOSTDOWN); 606 } 607 608 if ((ifp->if_flags & IFF_UP) != IFF_UP) { 609 m_freem (m0); 610 return (EHOSTDOWN); 611 } 612 613 /* BPF writes need to be handled specially. */ 614 if (dst->sa_family == AF_UNSPEC) { 615 bcopy(dst->sa_data, &af, sizeof(af)); 616 dst->sa_family = af; 617 } 618 619 if (bpf_peers_present(ifp->if_bpf)) { 620 af = dst->sa_family; 621 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m0); 622 } 623 624 /* prepend sockaddr? this may abort if the mbuf allocation fails */ 625 if (cached_tun_flags & TUN_LMODE) { 626 /* allocate space for sockaddr */ 627 M_PREPEND(m0, dst->sa_len, M_DONTWAIT); 628 629 /* if allocation failed drop packet */ 630 if (m0 == NULL) { 631 ifp->if_iqdrops++; 632 ifp->if_oerrors++; 633 return (ENOBUFS); 634 } else { 635 bcopy(dst, m0->m_data, dst->sa_len); 636 } 637 } 638 639 if (cached_tun_flags & TUN_IFHEAD) { 640 /* Prepend the address family */ 641 M_PREPEND(m0, 4, M_DONTWAIT); 642 643 /* if allocation failed drop packet */ 644 if (m0 == NULL) { 645 ifp->if_iqdrops++; 646 ifp->if_oerrors++; 647 return (ENOBUFS); 648 } else 649 *(u_int32_t *)m0->m_data = htonl(dst->sa_family); 650 } else { 651#ifdef INET 652 if (dst->sa_family != AF_INET) 653#endif 654 { 655 m_freem(m0); 656 return (EAFNOSUPPORT); 657 } 658 } 659 660 error = (ifp->if_transmit)(ifp, m0); 661 if (error) 662 return (ENOBUFS); 663 ifp->if_opackets++; 664 return (0); 665} 666 667/* 668 * the cdevsw interface is now pretty minimal. 669 */ 670static int 671tunioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, 672 struct thread *td) 673{ 674 int error; 675 struct tun_softc *tp = dev->si_drv1; 676 struct tuninfo *tunp; 677 678 switch (cmd) { 679 case TUNSIFINFO: 680 tunp = (struct tuninfo *)data; 681 if (tunp->mtu < IF_MINMTU) 682 return (EINVAL); 683 if (TUN2IFP(tp)->if_mtu != tunp->mtu) { 684 error = priv_check(td, PRIV_NET_SETIFMTU); 685 if (error) 686 return (error); 687 } 688 mtx_lock(&tp->tun_mtx); 689 TUN2IFP(tp)->if_mtu = tunp->mtu; 690 TUN2IFP(tp)->if_type = tunp->type; 691 TUN2IFP(tp)->if_baudrate = tunp->baudrate; 692 mtx_unlock(&tp->tun_mtx); 693 break; 694 case TUNGIFINFO: 695 tunp = (struct tuninfo *)data; 696 mtx_lock(&tp->tun_mtx); 697 tunp->mtu = TUN2IFP(tp)->if_mtu; 698 tunp->type = TUN2IFP(tp)->if_type; 699 tunp->baudrate = TUN2IFP(tp)->if_baudrate; 700 mtx_unlock(&tp->tun_mtx); 701 break; 702 case TUNSDEBUG: 703 tundebug = *(int *)data; 704 break; 705 case TUNGDEBUG: 706 *(int *)data = tundebug; 707 break; 708 case TUNSLMODE: 709 mtx_lock(&tp->tun_mtx); 710 if (*(int *)data) { 711 tp->tun_flags |= TUN_LMODE; 712 tp->tun_flags &= ~TUN_IFHEAD; 713 } else 714 tp->tun_flags &= ~TUN_LMODE; 715 mtx_unlock(&tp->tun_mtx); 716 break; 717 case TUNSIFHEAD: 718 mtx_lock(&tp->tun_mtx); 719 if (*(int *)data) { 720 tp->tun_flags |= TUN_IFHEAD; 721 tp->tun_flags &= ~TUN_LMODE; 722 } else 723 tp->tun_flags &= ~TUN_IFHEAD; 724 mtx_unlock(&tp->tun_mtx); 725 break; 726 case TUNGIFHEAD: 727 mtx_lock(&tp->tun_mtx); 728 *(int *)data = (tp->tun_flags & TUN_IFHEAD) ? 1 : 0; 729 mtx_unlock(&tp->tun_mtx); 730 break; 731 case TUNSIFMODE: 732 /* deny this if UP */ 733 if (TUN2IFP(tp)->if_flags & IFF_UP) 734 return(EBUSY); 735 736 switch (*(int *)data & ~IFF_MULTICAST) { 737 case IFF_POINTOPOINT: 738 case IFF_BROADCAST: 739 mtx_lock(&tp->tun_mtx); 740 TUN2IFP(tp)->if_flags &= 741 ~(IFF_BROADCAST|IFF_POINTOPOINT|IFF_MULTICAST); 742 TUN2IFP(tp)->if_flags |= *(int *)data; 743 mtx_unlock(&tp->tun_mtx); 744 break; 745 default: 746 return(EINVAL); 747 } 748 break; 749 case TUNSIFPID: 750 mtx_lock(&tp->tun_mtx); 751 tp->tun_pid = curthread->td_proc->p_pid; 752 mtx_unlock(&tp->tun_mtx); 753 break; 754 case FIONBIO: 755 break; 756 case FIOASYNC: 757 mtx_lock(&tp->tun_mtx); 758 if (*(int *)data) 759 tp->tun_flags |= TUN_ASYNC; 760 else 761 tp->tun_flags &= ~TUN_ASYNC; 762 mtx_unlock(&tp->tun_mtx); 763 break; 764 case FIONREAD: 765 if (!IFQ_IS_EMPTY(&TUN2IFP(tp)->if_snd)) { 766 struct mbuf *mb; 767 IFQ_LOCK(&TUN2IFP(tp)->if_snd); 768 IFQ_POLL_NOLOCK(&TUN2IFP(tp)->if_snd, mb); 769 for (*(int *)data = 0; mb != NULL; mb = mb->m_next) 770 *(int *)data += mb->m_len; 771 IFQ_UNLOCK(&TUN2IFP(tp)->if_snd); 772 } else 773 *(int *)data = 0; 774 break; 775 case FIOSETOWN: 776 return (fsetown(*(int *)data, &tp->tun_sigio)); 777 778 case FIOGETOWN: 779 *(int *)data = fgetown(&tp->tun_sigio); 780 return (0); 781 782 /* This is deprecated, FIOSETOWN should be used instead. */ 783 case TIOCSPGRP: 784 return (fsetown(-(*(int *)data), &tp->tun_sigio)); 785 786 /* This is deprecated, FIOGETOWN should be used instead. */ 787 case TIOCGPGRP: 788 *(int *)data = -fgetown(&tp->tun_sigio); 789 return (0); 790 791 default: 792 return (ENOTTY); 793 } 794 return (0); 795} 796 797/* 798 * The cdevsw read interface - reads a packet at a time, or at 799 * least as much of a packet as can be read. 800 */ 801static int 802tunread(struct cdev *dev, struct uio *uio, int flag) 803{ 804 struct tun_softc *tp = dev->si_drv1; 805 struct ifnet *ifp = TUN2IFP(tp); 806 struct mbuf *m; 807 int error=0, len; 808 809 TUNDEBUG (ifp, "read\n"); 810 mtx_lock(&tp->tun_mtx); 811 if ((tp->tun_flags & TUN_READY) != TUN_READY) { 812 mtx_unlock(&tp->tun_mtx); 813 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 814 return (EHOSTDOWN); 815 } 816 817 tp->tun_flags &= ~TUN_RWAIT; 818 819 do { 820 IFQ_DEQUEUE(&ifp->if_snd, m); 821 if (m == NULL) { 822 if (flag & O_NONBLOCK) { 823 mtx_unlock(&tp->tun_mtx); 824 return (EWOULDBLOCK); 825 } 826 tp->tun_flags |= TUN_RWAIT; 827 error = mtx_sleep(tp, &tp->tun_mtx, PCATCH | (PZERO + 1), 828 "tunread", 0); 829 if (error != 0) { 830 mtx_unlock(&tp->tun_mtx); 831 return (error); 832 } 833 } 834 } while (m == NULL); 835 mtx_unlock(&tp->tun_mtx); 836 837 while (m && uio->uio_resid > 0 && error == 0) { 838 len = min(uio->uio_resid, m->m_len); 839 if (len != 0) 840 error = uiomove(mtod(m, void *), len, uio); 841 m = m_free(m); 842 } 843 844 if (m) { 845 TUNDEBUG(ifp, "Dropping mbuf\n"); 846 m_freem(m); 847 } 848 return (error); 849} 850 851/* 852 * the cdevsw write interface - an atomic write is a packet - or else! 853 */ 854static int 855tunwrite(struct cdev *dev, struct uio *uio, int flag) 856{ 857 struct tun_softc *tp = dev->si_drv1; 858 struct ifnet *ifp = TUN2IFP(tp); 859 struct mbuf *m; 860 uint32_t family; 861 int isr; 862 863 TUNDEBUG(ifp, "tunwrite\n"); 864 865 if ((ifp->if_flags & IFF_UP) != IFF_UP) 866 /* ignore silently */ 867 return (0); 868 869 if (uio->uio_resid == 0) 870 return (0); 871 872 if (uio->uio_resid < 0 || uio->uio_resid > TUNMRU) { 873 TUNDEBUG(ifp, "len=%zd!\n", uio->uio_resid); 874 return (EIO); 875 } 876 877 if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, 0, M_PKTHDR)) == NULL) { 878 ifp->if_ierrors++; 879 return (ENOBUFS); 880 } 881 882 m->m_pkthdr.rcvif = ifp; 883#ifdef MAC 884 mac_ifnet_create_mbuf(ifp, m); 885#endif 886 887 /* Could be unlocked read? */ 888 mtx_lock(&tp->tun_mtx); 889 if (tp->tun_flags & TUN_IFHEAD) { 890 mtx_unlock(&tp->tun_mtx); 891 if (m->m_len < sizeof(family) && 892 (m = m_pullup(m, sizeof(family))) == NULL) 893 return (ENOBUFS); 894 family = ntohl(*mtod(m, u_int32_t *)); 895 m_adj(m, sizeof(family)); 896 } else { 897 mtx_unlock(&tp->tun_mtx); 898 family = AF_INET; 899 } 900 901 BPF_MTAP2(ifp, &family, sizeof(family), m); 902 903 switch (family) { 904#ifdef INET 905 case AF_INET: 906 isr = NETISR_IP; 907 break; 908#endif 909#ifdef INET6 910 case AF_INET6: 911 isr = NETISR_IPV6; 912 break; 913#endif 914#ifdef IPX 915 case AF_IPX: 916 isr = NETISR_IPX; 917 break; 918#endif 919#ifdef NETATALK 920 case AF_APPLETALK: 921 isr = NETISR_ATALK2; 922 break; 923#endif 924 default: 925 m_freem(m); 926 return (EAFNOSUPPORT); 927 } 928 /* First chunk of an mbuf contains good junk */ 929 if (harvest.point_to_point) 930 random_harvest(m, 16, 3, 0, RANDOM_NET); 931 ifp->if_ibytes += m->m_pkthdr.len; 932 ifp->if_ipackets++; 933 CURVNET_SET(ifp->if_vnet); 934 M_SETFIB(m, ifp->if_fib); 935 netisr_dispatch(isr, m); 936 CURVNET_RESTORE(); 937 return (0); 938} 939 940/* 941 * tunpoll - the poll interface, this is only useful on reads 942 * really. The write detect always returns true, write never blocks 943 * anyway, it either accepts the packet or drops it. 944 */ 945static int 946tunpoll(struct cdev *dev, int events, struct thread *td) 947{ 948 struct tun_softc *tp = dev->si_drv1; 949 struct ifnet *ifp = TUN2IFP(tp); 950 int revents = 0; 951 struct mbuf *m; 952 953 TUNDEBUG(ifp, "tunpoll\n"); 954 955 if (events & (POLLIN | POLLRDNORM)) { 956 IFQ_LOCK(&ifp->if_snd); 957 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 958 if (m != NULL) { 959 TUNDEBUG(ifp, "tunpoll q=%d\n", ifp->if_snd.ifq_len); 960 revents |= events & (POLLIN | POLLRDNORM); 961 } else { 962 TUNDEBUG(ifp, "tunpoll waiting\n"); 963 selrecord(td, &tp->tun_rsel); 964 } 965 IFQ_UNLOCK(&ifp->if_snd); 966 } 967 if (events & (POLLOUT | POLLWRNORM)) 968 revents |= events & (POLLOUT | POLLWRNORM); 969 970 return (revents); 971} 972 973/* 974 * tunkqfilter - support for the kevent() system call. 975 */ 976static int 977tunkqfilter(struct cdev *dev, struct knote *kn) 978{ 979 struct tun_softc *tp = dev->si_drv1; 980 struct ifnet *ifp = TUN2IFP(tp); 981 982 switch(kn->kn_filter) { 983 case EVFILT_READ: 984 TUNDEBUG(ifp, "%s kqfilter: EVFILT_READ, minor = %#x\n", 985 ifp->if_xname, dev2unit(dev)); 986 kn->kn_fop = &tun_read_filterops; 987 break; 988 989 case EVFILT_WRITE: 990 TUNDEBUG(ifp, "%s kqfilter: EVFILT_WRITE, minor = %#x\n", 991 ifp->if_xname, dev2unit(dev)); 992 kn->kn_fop = &tun_write_filterops; 993 break; 994 995 default: 996 TUNDEBUG(ifp, "%s kqfilter: invalid filter, minor = %#x\n", 997 ifp->if_xname, dev2unit(dev)); 998 return(EINVAL); 999 } 1000 1001 kn->kn_hook = tp; 1002 knlist_add(&tp->tun_rsel.si_note, kn, 0); 1003 1004 return (0); 1005} 1006 1007/* 1008 * Return true of there is data in the interface queue. 1009 */ 1010static int 1011tunkqread(struct knote *kn, long hint) 1012{ 1013 int ret; 1014 struct tun_softc *tp = kn->kn_hook; 1015 struct cdev *dev = tp->tun_dev; 1016 struct ifnet *ifp = TUN2IFP(tp); 1017 1018 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1019 TUNDEBUG(ifp, 1020 "%s have data in the queue. Len = %d, minor = %#x\n", 1021 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev)); 1022 ret = 1; 1023 } else { 1024 TUNDEBUG(ifp, 1025 "%s waiting for data, minor = %#x\n", ifp->if_xname, 1026 dev2unit(dev)); 1027 ret = 0; 1028 } 1029 1030 return (ret); 1031} 1032 1033/* 1034 * Always can write, always return MTU in kn->data. 1035 */ 1036static int 1037tunkqwrite(struct knote *kn, long hint) 1038{ 1039 struct tun_softc *tp = kn->kn_hook; 1040 struct ifnet *ifp = TUN2IFP(tp); 1041 1042 kn->kn_data = ifp->if_mtu; 1043 1044 return (1); 1045} 1046 1047static void 1048tunkqdetach(struct knote *kn) 1049{ 1050 struct tun_softc *tp = kn->kn_hook; 1051 1052 knlist_remove(&tp->tun_rsel.si_note, kn, 0); 1053} 1054