if_tun.c revision 205222
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 205222 2010-03-16 17:59:12Z qingli $ 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); 118SYSCTL_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 int 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_NEEDGIANT | 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", name, 232 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 knlist_destroy(&tp->tun_rsel.si_note); 263 mtx_destroy(&tp->tun_mtx); 264 cv_destroy(&tp->tun_cv); 265 free(tp, M_TUN); 266 CURVNET_RESTORE(); 267} 268 269static void 270tun_clone_destroy(struct ifnet *ifp) 271{ 272 struct tun_softc *tp = ifp->if_softc; 273 274 mtx_lock(&tunmtx); 275 TAILQ_REMOVE(&tunhead, tp, tun_list); 276 mtx_unlock(&tunmtx); 277 tun_destroy(tp); 278} 279 280static int 281tunmodevent(module_t mod, int type, void *data) 282{ 283 static eventhandler_tag tag; 284 struct tun_softc *tp; 285 286 switch (type) { 287 case MOD_LOAD: 288 mtx_init(&tunmtx, "tunmtx", NULL, MTX_DEF); 289 clone_setup(&tunclones); 290 tag = EVENTHANDLER_REGISTER(dev_clone, tunclone, 0, 1000); 291 if (tag == NULL) 292 return (ENOMEM); 293 if_clone_attach(&tun_cloner); 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); 324 325static void 326tunstart(struct ifnet *ifp) 327{ 328 struct tun_softc *tp = ifp->if_softc; 329 struct mbuf *m; 330 331 TUNDEBUG(ifp,"%s starting\n", ifp->if_xname); 332 if (ALTQ_IS_ENABLED(&ifp->if_snd)) { 333 IFQ_LOCK(&ifp->if_snd); 334 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 335 if (m == NULL) { 336 IFQ_UNLOCK(&ifp->if_snd); 337 return; 338 } 339 IFQ_UNLOCK(&ifp->if_snd); 340 } 341 342 mtx_lock(&tp->tun_mtx); 343 if (tp->tun_flags & TUN_RWAIT) { 344 tp->tun_flags &= ~TUN_RWAIT; 345 wakeup(tp); 346 } 347 if (tp->tun_flags & TUN_ASYNC && tp->tun_sigio) { 348 mtx_unlock(&tp->tun_mtx); 349 pgsigio(&tp->tun_sigio, SIGIO, 0); 350 } else 351 mtx_unlock(&tp->tun_mtx); 352 selwakeuppri(&tp->tun_rsel, PZERO + 1); 353 KNOTE_UNLOCKED(&tp->tun_rsel.si_note, 0); 354} 355 356/* XXX: should return an error code so it can fail. */ 357static void 358tuncreate(const char *name, struct cdev *dev) 359{ 360 struct tun_softc *sc; 361 struct ifnet *ifp; 362 363 dev->si_flags &= ~SI_CHEAPCLONE; 364 365 sc = malloc(sizeof(*sc), M_TUN, M_WAITOK | M_ZERO); 366 mtx_init(&sc->tun_mtx, "tun_mtx", NULL, MTX_DEF); 367 cv_init(&sc->tun_cv, "tun_condvar"); 368 sc->tun_flags = TUN_INITED; 369 sc->tun_dev = dev; 370 mtx_lock(&tunmtx); 371 TAILQ_INSERT_TAIL(&tunhead, sc, tun_list); 372 mtx_unlock(&tunmtx); 373 374 ifp = sc->tun_ifp = if_alloc(IFT_PPP); 375 if (ifp == NULL) 376 panic("%s%d: failed to if_alloc() interface.\n", 377 name, dev2unit(dev)); 378 if_initname(ifp, name, dev2unit(dev)); 379 ifp->if_mtu = TUNMTU; 380 ifp->if_ioctl = tunifioctl; 381 ifp->if_output = tunoutput; 382 ifp->if_start = tunstart; 383 ifp->if_flags = IFF_POINTOPOINT | IFF_MULTICAST; 384 ifp->if_softc = sc; 385 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 386 ifp->if_snd.ifq_drv_maxlen = 0; 387 IFQ_SET_READY(&ifp->if_snd); 388 knlist_init_mtx(&sc->tun_rsel.si_note, NULL); 389 ifp->if_capabilities |= IFCAP_LINKSTATE; 390 ifp->if_capenable |= IFCAP_LINKSTATE; 391 392 if_attach(ifp); 393 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); 394 dev->si_drv1 = sc; 395 TUNDEBUG(ifp, "interface %s is created, minor = %#x\n", 396 ifp->if_xname, dev2unit(dev)); 397} 398 399static int 400tunopen(struct cdev *dev, int flag, int mode, struct thread *td) 401{ 402 struct ifnet *ifp; 403 struct tun_softc *tp; 404 405 /* 406 * XXXRW: Non-atomic test and set of dev->si_drv1 requires 407 * synchronization. 408 */ 409 tp = dev->si_drv1; 410 if (!tp) { 411 tuncreate(TUNNAME, dev); 412 tp = dev->si_drv1; 413 } 414 415 /* 416 * XXXRW: This use of tun_pid is subject to error due to the 417 * fact that a reference to the tunnel can live beyond the 418 * death of the process that created it. Can we replace this 419 * with a simple busy flag? 420 */ 421 mtx_lock(&tp->tun_mtx); 422 if (tp->tun_pid != 0 && tp->tun_pid != td->td_proc->p_pid) { 423 mtx_unlock(&tp->tun_mtx); 424 return (EBUSY); 425 } 426 tp->tun_pid = td->td_proc->p_pid; 427 428 tp->tun_flags |= TUN_OPEN; 429 mtx_unlock(&tp->tun_mtx); 430 ifp = TUN2IFP(tp); 431 if_link_state_change(ifp, LINK_STATE_UP); 432 TUNDEBUG(ifp, "open\n"); 433 434 return (0); 435} 436 437/* 438 * tunclose - close the device - mark i/f down & delete 439 * routing info 440 */ 441static int 442tunclose(struct cdev *dev, int foo, int bar, struct thread *td) 443{ 444 struct tun_softc *tp; 445 struct ifnet *ifp; 446 int s; 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 mtx_unlock(&tp->tun_mtx); 455 456 /* 457 * junk all pending output 458 */ 459 CURVNET_SET(ifp->if_vnet); 460 s = splimp(); 461 IFQ_PURGE(&ifp->if_snd); 462 splx(s); 463 464 if (ifp->if_flags & IFF_UP) { 465 s = splimp(); 466 if_down(ifp); 467 splx(s); 468 } 469 470 /* Delete all addresses and routes which reference this interface. */ 471 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 472 struct ifaddr *ifa; 473 474 s = splimp(); 475 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 476 /* deal w/IPv4 PtP destination; unlocked read */ 477 if (ifa->ifa_addr->sa_family == AF_INET) { 478 rtinit(ifa, (int)RTM_DELETE, 479 tp->tun_flags & TUN_DSTADDR ? RTF_HOST : 0); 480 } else { 481 rtinit(ifa, (int)RTM_DELETE, 0); 482 } 483 } 484 if_purgeaddrs(ifp); 485 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 486 splx(s); 487 } 488 if_link_state_change(ifp, LINK_STATE_DOWN); 489 CURVNET_RESTORE(); 490 491 mtx_lock(&tp->tun_mtx); 492 funsetown(&tp->tun_sigio); 493 selwakeuppri(&tp->tun_rsel, PZERO + 1); 494 KNOTE_UNLOCKED(&tp->tun_rsel.si_note, 0); 495 TUNDEBUG (ifp, "closed\n"); 496 497 cv_broadcast(&tp->tun_cv); 498 mtx_unlock(&tp->tun_mtx); 499 return (0); 500} 501 502static int 503tuninit(struct ifnet *ifp) 504{ 505#ifdef INET 506 struct tun_softc *tp = ifp->if_softc; 507 struct ifaddr *ifa; 508#endif 509 int error = 0; 510 511 TUNDEBUG(ifp, "tuninit\n"); 512 513 ifp->if_flags |= IFF_UP; 514 ifp->if_drv_flags |= IFF_DRV_RUNNING; 515 getmicrotime(&ifp->if_lastchange); 516 517#ifdef INET 518 if_addr_rlock(ifp); 519 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 520 if (ifa->ifa_addr->sa_family == AF_INET) { 521 struct sockaddr_in *si; 522 523 si = (struct sockaddr_in *)ifa->ifa_addr; 524 mtx_lock(&tp->tun_mtx); 525 if (si->sin_addr.s_addr) 526 tp->tun_flags |= TUN_IASET; 527 528 si = (struct sockaddr_in *)ifa->ifa_dstaddr; 529 if (si && si->sin_addr.s_addr) 530 tp->tun_flags |= TUN_DSTADDR; 531 mtx_unlock(&tp->tun_mtx); 532 } 533 } 534 if_addr_runlock(ifp); 535#endif 536 return (error); 537} 538 539/* 540 * Process an ioctl request. 541 */ 542static int 543tunifioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 544{ 545 struct ifreq *ifr = (struct ifreq *)data; 546 struct tun_softc *tp = ifp->if_softc; 547 struct ifstat *ifs; 548 int error = 0, s; 549 550 s = splimp(); 551 switch(cmd) { 552 case SIOCGIFSTATUS: 553 ifs = (struct ifstat *)data; 554 mtx_lock(&tp->tun_mtx); 555 if (tp->tun_pid) 556 sprintf(ifs->ascii + strlen(ifs->ascii), 557 "\tOpened by PID %d\n", tp->tun_pid); 558 mtx_unlock(&tp->tun_mtx); 559 break; 560 case SIOCSIFADDR: 561 error = tuninit(ifp); 562 TUNDEBUG(ifp, "address set, error=%d\n", error); 563 break; 564 case SIOCSIFDSTADDR: 565 error = tuninit(ifp); 566 TUNDEBUG(ifp, "destination address set, error=%d\n", error); 567 break; 568 case SIOCSIFMTU: 569 ifp->if_mtu = ifr->ifr_mtu; 570 TUNDEBUG(ifp, "mtu set\n"); 571 break; 572 case SIOCSIFFLAGS: 573 case SIOCADDMULTI: 574 case SIOCDELMULTI: 575 break; 576 default: 577 error = EINVAL; 578 } 579 splx(s); 580 return (error); 581} 582 583/* 584 * tunoutput - queue packets from higher level ready to put out. 585 */ 586static int 587tunoutput( 588 struct ifnet *ifp, 589 struct mbuf *m0, 590 struct sockaddr *dst, 591 struct route *ro) 592{ 593 struct tun_softc *tp = ifp->if_softc; 594 u_short cached_tun_flags; 595 int error; 596 u_int32_t af; 597 598 TUNDEBUG (ifp, "tunoutput\n"); 599 600#ifdef MAC 601 error = mac_ifnet_check_transmit(ifp, m0); 602 if (error) { 603 m_freem(m0); 604 return (error); 605 } 606#endif 607 608 /* Could be unlocked read? */ 609 mtx_lock(&tp->tun_mtx); 610 cached_tun_flags = tp->tun_flags; 611 mtx_unlock(&tp->tun_mtx); 612 if ((cached_tun_flags & TUN_READY) != TUN_READY) { 613 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 614 m_freem (m0); 615 return (EHOSTDOWN); 616 } 617 618 if ((ifp->if_flags & IFF_UP) != IFF_UP) { 619 m_freem (m0); 620 return (EHOSTDOWN); 621 } 622 623 /* BPF writes need to be handled specially. */ 624 if (dst->sa_family == AF_UNSPEC) { 625 bcopy(dst->sa_data, &af, sizeof(af)); 626 dst->sa_family = af; 627 } 628 629 if (bpf_peers_present(ifp->if_bpf)) { 630 af = dst->sa_family; 631 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m0); 632 } 633 634 /* prepend sockaddr? this may abort if the mbuf allocation fails */ 635 if (cached_tun_flags & TUN_LMODE) { 636 /* allocate space for sockaddr */ 637 M_PREPEND(m0, dst->sa_len, M_DONTWAIT); 638 639 /* if allocation failed drop packet */ 640 if (m0 == NULL) { 641 ifp->if_iqdrops++; 642 ifp->if_oerrors++; 643 return (ENOBUFS); 644 } else { 645 bcopy(dst, m0->m_data, dst->sa_len); 646 } 647 } 648 649 if (cached_tun_flags & TUN_IFHEAD) { 650 /* Prepend the address family */ 651 M_PREPEND(m0, 4, M_DONTWAIT); 652 653 /* if allocation failed drop packet */ 654 if (m0 == NULL) { 655 ifp->if_iqdrops++; 656 ifp->if_oerrors++; 657 return (ENOBUFS); 658 } else 659 *(u_int32_t *)m0->m_data = htonl(dst->sa_family); 660 } else { 661#ifdef INET 662 if (dst->sa_family != AF_INET) 663#endif 664 { 665 m_freem(m0); 666 return (EAFNOSUPPORT); 667 } 668 } 669 670 error = (ifp->if_transmit)(ifp, m0); 671 if (error) { 672 ifp->if_collisions++; 673 return (ENOBUFS); 674 } 675 ifp->if_opackets++; 676 return (0); 677} 678 679/* 680 * the cdevsw interface is now pretty minimal. 681 */ 682static int 683tunioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 684{ 685 int s; 686 int error; 687 struct tun_softc *tp = dev->si_drv1; 688 struct tuninfo *tunp; 689 690 switch (cmd) { 691 case TUNSIFINFO: 692 tunp = (struct tuninfo *)data; 693 if (tunp->mtu < IF_MINMTU) 694 return (EINVAL); 695 if (TUN2IFP(tp)->if_mtu != tunp->mtu) { 696 error = priv_check(td, PRIV_NET_SETIFMTU); 697 if (error) 698 return (error); 699 } 700 TUN2IFP(tp)->if_mtu = tunp->mtu; 701 TUN2IFP(tp)->if_type = tunp->type; 702 TUN2IFP(tp)->if_baudrate = tunp->baudrate; 703 break; 704 case TUNGIFINFO: 705 tunp = (struct tuninfo *)data; 706 tunp->mtu = TUN2IFP(tp)->if_mtu; 707 tunp->type = TUN2IFP(tp)->if_type; 708 tunp->baudrate = TUN2IFP(tp)->if_baudrate; 709 break; 710 case TUNSDEBUG: 711 tundebug = *(int *)data; 712 break; 713 case TUNGDEBUG: 714 *(int *)data = tundebug; 715 break; 716 case TUNSLMODE: 717 mtx_lock(&tp->tun_mtx); 718 if (*(int *)data) { 719 tp->tun_flags |= TUN_LMODE; 720 tp->tun_flags &= ~TUN_IFHEAD; 721 } else 722 tp->tun_flags &= ~TUN_LMODE; 723 mtx_unlock(&tp->tun_mtx); 724 break; 725 case TUNSIFHEAD: 726 mtx_lock(&tp->tun_mtx); 727 if (*(int *)data) { 728 tp->tun_flags |= TUN_IFHEAD; 729 tp->tun_flags &= ~TUN_LMODE; 730 } else 731 tp->tun_flags &= ~TUN_IFHEAD; 732 mtx_unlock(&tp->tun_mtx); 733 break; 734 case TUNGIFHEAD: 735 /* Could be unlocked read? */ 736 mtx_lock(&tp->tun_mtx); 737 *(int *)data = (tp->tun_flags & TUN_IFHEAD) ? 1 : 0; 738 mtx_unlock(&tp->tun_mtx); 739 break; 740 case TUNSIFMODE: 741 /* deny this if UP */ 742 if (TUN2IFP(tp)->if_flags & IFF_UP) 743 return(EBUSY); 744 745 switch (*(int *)data & ~IFF_MULTICAST) { 746 case IFF_POINTOPOINT: 747 case IFF_BROADCAST: 748 TUN2IFP(tp)->if_flags &= 749 ~(IFF_BROADCAST|IFF_POINTOPOINT|IFF_MULTICAST); 750 TUN2IFP(tp)->if_flags |= *(int *)data; 751 break; 752 default: 753 return(EINVAL); 754 } 755 break; 756 case TUNSIFPID: 757 mtx_lock(&tp->tun_mtx); 758 tp->tun_pid = curthread->td_proc->p_pid; 759 mtx_unlock(&tp->tun_mtx); 760 break; 761 case FIONBIO: 762 break; 763 case FIOASYNC: 764 mtx_lock(&tp->tun_mtx); 765 if (*(int *)data) 766 tp->tun_flags |= TUN_ASYNC; 767 else 768 tp->tun_flags &= ~TUN_ASYNC; 769 mtx_unlock(&tp->tun_mtx); 770 break; 771 case FIONREAD: 772 s = splimp(); 773 if (!IFQ_IS_EMPTY(&TUN2IFP(tp)->if_snd)) { 774 struct mbuf *mb; 775 IFQ_LOCK(&TUN2IFP(tp)->if_snd); 776 IFQ_POLL_NOLOCK(&TUN2IFP(tp)->if_snd, mb); 777 for( *(int *)data = 0; mb != 0; mb = mb->m_next) 778 *(int *)data += mb->m_len; 779 IFQ_UNLOCK(&TUN2IFP(tp)->if_snd); 780 } else 781 *(int *)data = 0; 782 splx(s); 783 break; 784 case FIOSETOWN: 785 return (fsetown(*(int *)data, &tp->tun_sigio)); 786 787 case FIOGETOWN: 788 *(int *)data = fgetown(&tp->tun_sigio); 789 return (0); 790 791 /* This is deprecated, FIOSETOWN should be used instead. */ 792 case TIOCSPGRP: 793 return (fsetown(-(*(int *)data), &tp->tun_sigio)); 794 795 /* This is deprecated, FIOGETOWN should be used instead. */ 796 case TIOCGPGRP: 797 *(int *)data = -fgetown(&tp->tun_sigio); 798 return (0); 799 800 default: 801 return (ENOTTY); 802 } 803 return (0); 804} 805 806/* 807 * The cdevsw read interface - reads a packet at a time, or at 808 * least as much of a packet as can be read. 809 */ 810static int 811tunread(struct cdev *dev, struct uio *uio, int flag) 812{ 813 struct tun_softc *tp = dev->si_drv1; 814 struct ifnet *ifp = TUN2IFP(tp); 815 struct mbuf *m; 816 int error=0, len, s; 817 818 TUNDEBUG (ifp, "read\n"); 819 mtx_lock(&tp->tun_mtx); 820 if ((tp->tun_flags & TUN_READY) != TUN_READY) { 821 mtx_unlock(&tp->tun_mtx); 822 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 823 return (EHOSTDOWN); 824 } 825 826 tp->tun_flags &= ~TUN_RWAIT; 827 mtx_unlock(&tp->tun_mtx); 828 829 s = splimp(); 830 do { 831 IFQ_DEQUEUE(&ifp->if_snd, m); 832 if (m == NULL) { 833 if (flag & O_NONBLOCK) { 834 splx(s); 835 return (EWOULDBLOCK); 836 } 837 mtx_lock(&tp->tun_mtx); 838 tp->tun_flags |= TUN_RWAIT; 839 mtx_unlock(&tp->tun_mtx); 840 if ((error = tsleep(tp, PCATCH | (PZERO + 1), 841 "tunread", 0)) != 0) { 842 splx(s); 843 return (error); 844 } 845 } 846 } while (m == NULL); 847 splx(s); 848 849 while (m && uio->uio_resid > 0 && error == 0) { 850 len = min(uio->uio_resid, m->m_len); 851 if (len != 0) 852 error = uiomove(mtod(m, void *), len, uio); 853 m = m_free(m); 854 } 855 856 if (m) { 857 TUNDEBUG(ifp, "Dropping mbuf\n"); 858 m_freem(m); 859 } 860 return (error); 861} 862 863/* 864 * the cdevsw write interface - an atomic write is a packet - or else! 865 */ 866static int 867tunwrite(struct cdev *dev, struct uio *uio, int flag) 868{ 869 struct tun_softc *tp = dev->si_drv1; 870 struct ifnet *ifp = TUN2IFP(tp); 871 struct mbuf *m; 872 int error = 0; 873 uint32_t family; 874 int isr; 875 876 TUNDEBUG(ifp, "tunwrite\n"); 877 878 if ((ifp->if_flags & IFF_UP) != IFF_UP) 879 /* ignore silently */ 880 return (0); 881 882 if (uio->uio_resid == 0) 883 return (0); 884 885 if (uio->uio_resid < 0 || uio->uio_resid > TUNMRU) { 886 TUNDEBUG(ifp, "len=%zd!\n", uio->uio_resid); 887 return (EIO); 888 } 889 890 if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, 0, M_PKTHDR)) == NULL) { 891 ifp->if_ierrors++; 892 return (error); 893 } 894 895 m->m_pkthdr.rcvif = ifp; 896#ifdef MAC 897 mac_ifnet_create_mbuf(ifp, m); 898#endif 899 900 /* Could be unlocked read? */ 901 mtx_lock(&tp->tun_mtx); 902 if (tp->tun_flags & TUN_IFHEAD) { 903 mtx_unlock(&tp->tun_mtx); 904 if (m->m_len < sizeof(family) && 905 (m = m_pullup(m, sizeof(family))) == NULL) 906 return (ENOBUFS); 907 family = ntohl(*mtod(m, u_int32_t *)); 908 m_adj(m, sizeof(family)); 909 } else { 910 mtx_unlock(&tp->tun_mtx); 911 family = AF_INET; 912 } 913 914 BPF_MTAP2(ifp, &family, sizeof(family), m); 915 916 switch (family) { 917#ifdef INET 918 case AF_INET: 919 isr = NETISR_IP; 920 break; 921#endif 922#ifdef INET6 923 case AF_INET6: 924 isr = NETISR_IPV6; 925 break; 926#endif 927#ifdef IPX 928 case AF_IPX: 929 isr = NETISR_IPX; 930 break; 931#endif 932#ifdef NETATALK 933 case AF_APPLETALK: 934 isr = NETISR_ATALK2; 935 break; 936#endif 937 default: 938 m_freem(m); 939 return (EAFNOSUPPORT); 940 } 941 /* First chunk of an mbuf contains good junk */ 942 if (harvest.point_to_point) 943 random_harvest(m, 16, 3, 0, RANDOM_NET); 944 ifp->if_ibytes += m->m_pkthdr.len; 945 ifp->if_ipackets++; 946 CURVNET_SET(ifp->if_vnet); 947 netisr_dispatch(isr, m); 948 CURVNET_RESTORE(); 949 return (0); 950} 951 952/* 953 * tunpoll - the poll interface, this is only useful on reads 954 * really. The write detect always returns true, write never blocks 955 * anyway, it either accepts the packet or drops it. 956 */ 957static int 958tunpoll(struct cdev *dev, int events, struct thread *td) 959{ 960 int s; 961 struct tun_softc *tp = dev->si_drv1; 962 struct ifnet *ifp = TUN2IFP(tp); 963 int revents = 0; 964 struct mbuf *m; 965 966 s = splimp(); 967 TUNDEBUG(ifp, "tunpoll\n"); 968 969 if (events & (POLLIN | POLLRDNORM)) { 970 IFQ_LOCK(&ifp->if_snd); 971 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 972 if (m != NULL) { 973 TUNDEBUG(ifp, "tunpoll q=%d\n", ifp->if_snd.ifq_len); 974 revents |= events & (POLLIN | POLLRDNORM); 975 } else { 976 TUNDEBUG(ifp, "tunpoll waiting\n"); 977 selrecord(td, &tp->tun_rsel); 978 } 979 IFQ_UNLOCK(&ifp->if_snd); 980 } 981 if (events & (POLLOUT | POLLWRNORM)) 982 revents |= events & (POLLOUT | POLLWRNORM); 983 984 splx(s); 985 return (revents); 986} 987 988/* 989 * tunkqfilter - support for the kevent() system call. 990 */ 991static int 992tunkqfilter(struct cdev *dev, struct knote *kn) 993{ 994 int s; 995 struct tun_softc *tp = dev->si_drv1; 996 struct ifnet *ifp = TUN2IFP(tp); 997 998 s = splimp(); 999 switch(kn->kn_filter) { 1000 case EVFILT_READ: 1001 TUNDEBUG(ifp, "%s kqfilter: EVFILT_READ, minor = %#x\n", 1002 ifp->if_xname, dev2unit(dev)); 1003 kn->kn_fop = &tun_read_filterops; 1004 break; 1005 1006 case EVFILT_WRITE: 1007 TUNDEBUG(ifp, "%s kqfilter: EVFILT_WRITE, minor = %#x\n", 1008 ifp->if_xname, dev2unit(dev)); 1009 kn->kn_fop = &tun_write_filterops; 1010 break; 1011 1012 default: 1013 TUNDEBUG(ifp, "%s kqfilter: invalid filter, minor = %#x\n", 1014 ifp->if_xname, dev2unit(dev)); 1015 splx(s); 1016 return(EINVAL); 1017 } 1018 splx(s); 1019 1020 kn->kn_hook = (caddr_t) dev; 1021 knlist_add(&tp->tun_rsel.si_note, kn, 0); 1022 1023 return (0); 1024} 1025 1026/* 1027 * Return true of there is data in the interface queue. 1028 */ 1029static int 1030tunkqread(struct knote *kn, long hint) 1031{ 1032 int ret, s; 1033 struct cdev *dev = (struct cdev *)(kn->kn_hook); 1034 struct tun_softc *tp = dev->si_drv1; 1035 struct ifnet *ifp = TUN2IFP(tp); 1036 1037 s = splimp(); 1038 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1039 TUNDEBUG(ifp, 1040 "%s have data in the queue. Len = %d, minor = %#x\n", 1041 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev)); 1042 ret = 1; 1043 } else { 1044 TUNDEBUG(ifp, 1045 "%s waiting for data, minor = %#x\n", ifp->if_xname, 1046 dev2unit(dev)); 1047 ret = 0; 1048 } 1049 splx(s); 1050 1051 return (ret); 1052} 1053 1054/* 1055 * Always can write, always return MTU in kn->data. 1056 */ 1057static int 1058tunkqwrite(struct knote *kn, long hint) 1059{ 1060 int s; 1061 struct tun_softc *tp = ((struct cdev *)kn->kn_hook)->si_drv1; 1062 struct ifnet *ifp = TUN2IFP(tp); 1063 1064 s = splimp(); 1065 kn->kn_data = ifp->if_mtu; 1066 splx(s); 1067 1068 return (1); 1069} 1070 1071static void 1072tunkqdetach(struct knote *kn) 1073{ 1074 struct tun_softc *tp = ((struct cdev *)kn->kn_hook)->si_drv1; 1075 1076 knlist_remove(&tp->tun_rsel.si_note, kn, 0); 1077} 1078