if_tun.c revision 213328
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 213328 2010-10-01 15:14:14Z bz $ 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_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 selwakeuppri(&tp->tun_rsel, PZERO + 1); 348 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0); 349 if (tp->tun_flags & TUN_ASYNC && tp->tun_sigio) { 350 mtx_unlock(&tp->tun_mtx); 351 pgsigio(&tp->tun_sigio, SIGIO, 0); 352 } else 353 mtx_unlock(&tp->tun_mtx); 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, &sc->tun_mtx); 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 ifp = TUN2IFP(tp); 430 if_link_state_change(ifp, LINK_STATE_UP); 431 TUNDEBUG(ifp, "open\n"); 432 mtx_unlock(&tp->tun_mtx); 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; 446 447 tp = dev->si_drv1; 448 ifp = TUN2IFP(tp); 449 450 mtx_lock(&tp->tun_mtx); 451 tp->tun_flags &= ~TUN_OPEN; 452 tp->tun_pid = 0; 453 454 /* 455 * junk all pending output 456 */ 457 CURVNET_SET(ifp->if_vnet); 458 IFQ_PURGE(&ifp->if_snd); 459 460 if (ifp->if_flags & IFF_UP) { 461 mtx_unlock(&tp->tun_mtx); 462 if_down(ifp); 463 mtx_lock(&tp->tun_mtx); 464 } 465 466 /* Delete all addresses and routes which reference this interface. */ 467 if (ifp->if_drv_flags & IFF_DRV_RUNNING) { 468 struct ifaddr *ifa; 469 470 ifp->if_drv_flags &= ~IFF_DRV_RUNNING; 471 mtx_unlock(&tp->tun_mtx); 472 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 473 /* deal w/IPv4 PtP destination; unlocked read */ 474 if (ifa->ifa_addr->sa_family == AF_INET) { 475 rtinit(ifa, (int)RTM_DELETE, 476 tp->tun_flags & TUN_DSTADDR ? RTF_HOST : 0); 477 } else { 478 rtinit(ifa, (int)RTM_DELETE, 0); 479 } 480 } 481 if_purgeaddrs(ifp); 482 mtx_lock(&tp->tun_mtx); 483 } 484 if_link_state_change(ifp, LINK_STATE_DOWN); 485 CURVNET_RESTORE(); 486 487 funsetown(&tp->tun_sigio); 488 selwakeuppri(&tp->tun_rsel, PZERO + 1); 489 KNOTE_LOCKED(&tp->tun_rsel.si_note, 0); 490 TUNDEBUG (ifp, "closed\n"); 491 492 cv_broadcast(&tp->tun_cv); 493 mtx_unlock(&tp->tun_mtx); 494 return (0); 495} 496 497static int 498tuninit(struct ifnet *ifp) 499{ 500 struct tun_softc *tp = ifp->if_softc; 501#ifdef INET 502 struct ifaddr *ifa; 503#endif 504 int error = 0; 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 return (error); 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 error = tuninit(ifp); 556 TUNDEBUG(ifp, "address set, error=%d\n", error); 557 break; 558 case SIOCSIFDSTADDR: 559 error = tuninit(ifp); 560 TUNDEBUG(ifp, "destination address set, error=%d\n", error); 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( 581 struct ifnet *ifp, 582 struct mbuf *m0, 583 struct sockaddr *dst, 584 struct route *ro) 585{ 586 struct tun_softc *tp = ifp->if_softc; 587 u_short cached_tun_flags; 588 int error; 589 u_int32_t af; 590 591 TUNDEBUG (ifp, "tunoutput\n"); 592 593#ifdef MAC 594 error = mac_ifnet_check_transmit(ifp, m0); 595 if (error) { 596 m_freem(m0); 597 return (error); 598 } 599#endif 600 601 /* Could be unlocked read? */ 602 mtx_lock(&tp->tun_mtx); 603 cached_tun_flags = tp->tun_flags; 604 mtx_unlock(&tp->tun_mtx); 605 if ((cached_tun_flags & TUN_READY) != TUN_READY) { 606 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 607 m_freem (m0); 608 return (EHOSTDOWN); 609 } 610 611 if ((ifp->if_flags & IFF_UP) != IFF_UP) { 612 m_freem (m0); 613 return (EHOSTDOWN); 614 } 615 616 /* BPF writes need to be handled specially. */ 617 if (dst->sa_family == AF_UNSPEC) { 618 bcopy(dst->sa_data, &af, sizeof(af)); 619 dst->sa_family = af; 620 } 621 622 if (bpf_peers_present(ifp->if_bpf)) { 623 af = dst->sa_family; 624 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m0); 625 } 626 627 /* prepend sockaddr? this may abort if the mbuf allocation fails */ 628 if (cached_tun_flags & TUN_LMODE) { 629 /* allocate space for sockaddr */ 630 M_PREPEND(m0, dst->sa_len, M_DONTWAIT); 631 632 /* if allocation failed drop packet */ 633 if (m0 == NULL) { 634 ifp->if_iqdrops++; 635 ifp->if_oerrors++; 636 return (ENOBUFS); 637 } else { 638 bcopy(dst, m0->m_data, dst->sa_len); 639 } 640 } 641 642 if (cached_tun_flags & TUN_IFHEAD) { 643 /* Prepend the address family */ 644 M_PREPEND(m0, 4, M_DONTWAIT); 645 646 /* if allocation failed drop packet */ 647 if (m0 == NULL) { 648 ifp->if_iqdrops++; 649 ifp->if_oerrors++; 650 return (ENOBUFS); 651 } else 652 *(u_int32_t *)m0->m_data = htonl(dst->sa_family); 653 } else { 654#ifdef INET 655 if (dst->sa_family != AF_INET) 656#endif 657 { 658 m_freem(m0); 659 return (EAFNOSUPPORT); 660 } 661 } 662 663 error = (ifp->if_transmit)(ifp, m0); 664 if (error) { 665 ifp->if_collisions++; 666 return (ENOBUFS); 667 } 668 ifp->if_opackets++; 669 return (0); 670} 671 672/* 673 * the cdevsw interface is now pretty minimal. 674 */ 675static int 676tunioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td) 677{ 678 int error; 679 struct tun_softc *tp = dev->si_drv1; 680 struct tuninfo *tunp; 681 682 switch (cmd) { 683 case TUNSIFINFO: 684 tunp = (struct tuninfo *)data; 685 if (tunp->mtu < IF_MINMTU) 686 return (EINVAL); 687 if (TUN2IFP(tp)->if_mtu != tunp->mtu) { 688 error = priv_check(td, PRIV_NET_SETIFMTU); 689 if (error) 690 return (error); 691 } 692 mtx_lock(&tp->tun_mtx); 693 TUN2IFP(tp)->if_mtu = tunp->mtu; 694 TUN2IFP(tp)->if_type = tunp->type; 695 TUN2IFP(tp)->if_baudrate = tunp->baudrate; 696 mtx_unlock(&tp->tun_mtx); 697 break; 698 case TUNGIFINFO: 699 tunp = (struct tuninfo *)data; 700 mtx_lock(&tp->tun_mtx); 701 tunp->mtu = TUN2IFP(tp)->if_mtu; 702 tunp->type = TUN2IFP(tp)->if_type; 703 tunp->baudrate = TUN2IFP(tp)->if_baudrate; 704 mtx_unlock(&tp->tun_mtx); 705 break; 706 case TUNSDEBUG: 707 tundebug = *(int *)data; 708 break; 709 case TUNGDEBUG: 710 *(int *)data = tundebug; 711 break; 712 case TUNSLMODE: 713 mtx_lock(&tp->tun_mtx); 714 if (*(int *)data) { 715 tp->tun_flags |= TUN_LMODE; 716 tp->tun_flags &= ~TUN_IFHEAD; 717 } else 718 tp->tun_flags &= ~TUN_LMODE; 719 mtx_unlock(&tp->tun_mtx); 720 break; 721 case TUNSIFHEAD: 722 mtx_lock(&tp->tun_mtx); 723 if (*(int *)data) { 724 tp->tun_flags |= TUN_IFHEAD; 725 tp->tun_flags &= ~TUN_LMODE; 726 } else 727 tp->tun_flags &= ~TUN_IFHEAD; 728 mtx_unlock(&tp->tun_mtx); 729 break; 730 case TUNGIFHEAD: 731 mtx_lock(&tp->tun_mtx); 732 *(int *)data = (tp->tun_flags & TUN_IFHEAD) ? 1 : 0; 733 mtx_unlock(&tp->tun_mtx); 734 break; 735 case TUNSIFMODE: 736 /* deny this if UP */ 737 if (TUN2IFP(tp)->if_flags & IFF_UP) 738 return(EBUSY); 739 740 switch (*(int *)data & ~IFF_MULTICAST) { 741 case IFF_POINTOPOINT: 742 case IFF_BROADCAST: 743 mtx_lock(&tp->tun_mtx); 744 TUN2IFP(tp)->if_flags &= 745 ~(IFF_BROADCAST|IFF_POINTOPOINT|IFF_MULTICAST); 746 TUN2IFP(tp)->if_flags |= *(int *)data; 747 mtx_unlock(&tp->tun_mtx); 748 break; 749 default: 750 return(EINVAL); 751 } 752 break; 753 case TUNSIFPID: 754 mtx_lock(&tp->tun_mtx); 755 tp->tun_pid = curthread->td_proc->p_pid; 756 mtx_unlock(&tp->tun_mtx); 757 break; 758 case FIONBIO: 759 break; 760 case FIOASYNC: 761 mtx_lock(&tp->tun_mtx); 762 if (*(int *)data) 763 tp->tun_flags |= TUN_ASYNC; 764 else 765 tp->tun_flags &= ~TUN_ASYNC; 766 mtx_unlock(&tp->tun_mtx); 767 break; 768 case FIONREAD: 769 if (!IFQ_IS_EMPTY(&TUN2IFP(tp)->if_snd)) { 770 struct mbuf *mb; 771 IFQ_LOCK(&TUN2IFP(tp)->if_snd); 772 IFQ_POLL_NOLOCK(&TUN2IFP(tp)->if_snd, mb); 773 for (*(int *)data = 0; mb != NULL; mb = mb->m_next) 774 *(int *)data += mb->m_len; 775 IFQ_UNLOCK(&TUN2IFP(tp)->if_snd); 776 } else 777 *(int *)data = 0; 778 break; 779 case FIOSETOWN: 780 return (fsetown(*(int *)data, &tp->tun_sigio)); 781 782 case FIOGETOWN: 783 *(int *)data = fgetown(&tp->tun_sigio); 784 return (0); 785 786 /* This is deprecated, FIOSETOWN should be used instead. */ 787 case TIOCSPGRP: 788 return (fsetown(-(*(int *)data), &tp->tun_sigio)); 789 790 /* This is deprecated, FIOGETOWN should be used instead. */ 791 case TIOCGPGRP: 792 *(int *)data = -fgetown(&tp->tun_sigio); 793 return (0); 794 795 default: 796 return (ENOTTY); 797 } 798 return (0); 799} 800 801/* 802 * The cdevsw read interface - reads a packet at a time, or at 803 * least as much of a packet as can be read. 804 */ 805static int 806tunread(struct cdev *dev, struct uio *uio, int flag) 807{ 808 struct tun_softc *tp = dev->si_drv1; 809 struct ifnet *ifp = TUN2IFP(tp); 810 struct mbuf *m; 811 int error=0, len; 812 813 TUNDEBUG (ifp, "read\n"); 814 mtx_lock(&tp->tun_mtx); 815 if ((tp->tun_flags & TUN_READY) != TUN_READY) { 816 mtx_unlock(&tp->tun_mtx); 817 TUNDEBUG (ifp, "not ready 0%o\n", tp->tun_flags); 818 return (EHOSTDOWN); 819 } 820 821 tp->tun_flags &= ~TUN_RWAIT; 822 823 do { 824 IFQ_DEQUEUE(&ifp->if_snd, m); 825 if (m == NULL) { 826 if (flag & O_NONBLOCK) { 827 mtx_unlock(&tp->tun_mtx); 828 return (EWOULDBLOCK); 829 } 830 tp->tun_flags |= TUN_RWAIT; 831 error = mtx_sleep(tp, &tp->tun_mtx, PCATCH | (PZERO + 1), 832 "tunread", 0); 833 if (error != 0) { 834 mtx_unlock(&tp->tun_mtx); 835 return (error); 836 } 837 } 838 } while (m == NULL); 839 mtx_unlock(&tp->tun_mtx); 840 841 while (m && uio->uio_resid > 0 && error == 0) { 842 len = min(uio->uio_resid, m->m_len); 843 if (len != 0) 844 error = uiomove(mtod(m, void *), len, uio); 845 m = m_free(m); 846 } 847 848 if (m) { 849 TUNDEBUG(ifp, "Dropping mbuf\n"); 850 m_freem(m); 851 } 852 return (error); 853} 854 855/* 856 * the cdevsw write interface - an atomic write is a packet - or else! 857 */ 858static int 859tunwrite(struct cdev *dev, struct uio *uio, int flag) 860{ 861 struct tun_softc *tp = dev->si_drv1; 862 struct ifnet *ifp = TUN2IFP(tp); 863 struct mbuf *m; 864 int error = 0; 865 uint32_t family; 866 int isr; 867 868 TUNDEBUG(ifp, "tunwrite\n"); 869 870 if ((ifp->if_flags & IFF_UP) != IFF_UP) 871 /* ignore silently */ 872 return (0); 873 874 if (uio->uio_resid == 0) 875 return (0); 876 877 if (uio->uio_resid < 0 || uio->uio_resid > TUNMRU) { 878 TUNDEBUG(ifp, "len=%zd!\n", uio->uio_resid); 879 return (EIO); 880 } 881 882 if ((m = m_uiotombuf(uio, M_DONTWAIT, 0, 0, M_PKTHDR)) == NULL) { 883 ifp->if_ierrors++; 884 return (error); 885 } 886 887 m->m_pkthdr.rcvif = ifp; 888#ifdef MAC 889 mac_ifnet_create_mbuf(ifp, m); 890#endif 891 892 /* Could be unlocked read? */ 893 mtx_lock(&tp->tun_mtx); 894 if (tp->tun_flags & TUN_IFHEAD) { 895 mtx_unlock(&tp->tun_mtx); 896 if (m->m_len < sizeof(family) && 897 (m = m_pullup(m, sizeof(family))) == NULL) 898 return (ENOBUFS); 899 family = ntohl(*mtod(m, u_int32_t *)); 900 m_adj(m, sizeof(family)); 901 } else { 902 mtx_unlock(&tp->tun_mtx); 903 family = AF_INET; 904 } 905 906 BPF_MTAP2(ifp, &family, sizeof(family), m); 907 908 switch (family) { 909#ifdef INET 910 case AF_INET: 911 isr = NETISR_IP; 912 break; 913#endif 914#ifdef INET6 915 case AF_INET6: 916 isr = NETISR_IPV6; 917 break; 918#endif 919#ifdef IPX 920 case AF_IPX: 921 isr = NETISR_IPX; 922 break; 923#endif 924#ifdef NETATALK 925 case AF_APPLETALK: 926 isr = NETISR_ATALK2; 927 break; 928#endif 929 default: 930 m_freem(m); 931 return (EAFNOSUPPORT); 932 } 933 /* First chunk of an mbuf contains good junk */ 934 if (harvest.point_to_point) 935 random_harvest(m, 16, 3, 0, RANDOM_NET); 936 ifp->if_ibytes += m->m_pkthdr.len; 937 ifp->if_ipackets++; 938 CURVNET_SET(ifp->if_vnet); 939 netisr_dispatch(isr, m); 940 CURVNET_RESTORE(); 941 return (0); 942} 943 944/* 945 * tunpoll - the poll interface, this is only useful on reads 946 * really. The write detect always returns true, write never blocks 947 * anyway, it either accepts the packet or drops it. 948 */ 949static int 950tunpoll(struct cdev *dev, int events, struct thread *td) 951{ 952 struct tun_softc *tp = dev->si_drv1; 953 struct ifnet *ifp = TUN2IFP(tp); 954 int revents = 0; 955 struct mbuf *m; 956 957 TUNDEBUG(ifp, "tunpoll\n"); 958 959 if (events & (POLLIN | POLLRDNORM)) { 960 IFQ_LOCK(&ifp->if_snd); 961 IFQ_POLL_NOLOCK(&ifp->if_snd, m); 962 if (m != NULL) { 963 TUNDEBUG(ifp, "tunpoll q=%d\n", ifp->if_snd.ifq_len); 964 revents |= events & (POLLIN | POLLRDNORM); 965 } else { 966 TUNDEBUG(ifp, "tunpoll waiting\n"); 967 selrecord(td, &tp->tun_rsel); 968 } 969 IFQ_UNLOCK(&ifp->if_snd); 970 } 971 if (events & (POLLOUT | POLLWRNORM)) 972 revents |= events & (POLLOUT | POLLWRNORM); 973 974 return (revents); 975} 976 977/* 978 * tunkqfilter - support for the kevent() system call. 979 */ 980static int 981tunkqfilter(struct cdev *dev, struct knote *kn) 982{ 983 struct tun_softc *tp = dev->si_drv1; 984 struct ifnet *ifp = TUN2IFP(tp); 985 986 switch(kn->kn_filter) { 987 case EVFILT_READ: 988 TUNDEBUG(ifp, "%s kqfilter: EVFILT_READ, minor = %#x\n", 989 ifp->if_xname, dev2unit(dev)); 990 kn->kn_fop = &tun_read_filterops; 991 break; 992 993 case EVFILT_WRITE: 994 TUNDEBUG(ifp, "%s kqfilter: EVFILT_WRITE, minor = %#x\n", 995 ifp->if_xname, dev2unit(dev)); 996 kn->kn_fop = &tun_write_filterops; 997 break; 998 999 default: 1000 TUNDEBUG(ifp, "%s kqfilter: invalid filter, minor = %#x\n", 1001 ifp->if_xname, dev2unit(dev)); 1002 return(EINVAL); 1003 } 1004 1005 kn->kn_hook = tp; 1006 knlist_add(&tp->tun_rsel.si_note, kn, 0); 1007 1008 return (0); 1009} 1010 1011/* 1012 * Return true of there is data in the interface queue. 1013 */ 1014static int 1015tunkqread(struct knote *kn, long hint) 1016{ 1017 int ret; 1018 struct tun_softc *tp = kn->kn_hook; 1019 struct cdev *dev = tp->tun_dev; 1020 struct ifnet *ifp = TUN2IFP(tp); 1021 1022 if ((kn->kn_data = ifp->if_snd.ifq_len) > 0) { 1023 TUNDEBUG(ifp, 1024 "%s have data in the queue. Len = %d, minor = %#x\n", 1025 ifp->if_xname, ifp->if_snd.ifq_len, dev2unit(dev)); 1026 ret = 1; 1027 } else { 1028 TUNDEBUG(ifp, 1029 "%s waiting for data, minor = %#x\n", ifp->if_xname, 1030 dev2unit(dev)); 1031 ret = 0; 1032 } 1033 1034 return (ret); 1035} 1036 1037/* 1038 * Always can write, always return MTU in kn->data. 1039 */ 1040static int 1041tunkqwrite(struct knote *kn, long hint) 1042{ 1043 struct tun_softc *tp = kn->kn_hook; 1044 struct ifnet *ifp = TUN2IFP(tp); 1045 1046 kn->kn_data = ifp->if_mtu; 1047 1048 return (1); 1049} 1050 1051static void 1052tunkqdetach(struct knote *kn) 1053{ 1054 struct tun_softc *tp = kn->kn_hook; 1055 1056 knlist_remove(&tp->tun_rsel.si_note, kn, 0); 1057} 1058