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 17#include "tun.h" 18#if NTUN > 0 19 20#include "opt_devfs.h" 21#include "opt_inet.h" 22 23#include <sys/param.h> 24#include <sys/proc.h> 25#include <sys/systm.h> 26#include <sys/mbuf.h> 27#include <sys/socket.h> 28#include <sys/filio.h> 29#include <sys/sockio.h> 30#include <sys/ttycom.h> 31#include <sys/poll.h> 32#include <sys/signalvar.h> 33#include <sys/filedesc.h> 34#include <sys/kernel.h> 35#include <sys/sysctl.h> 36#ifdef DEVFS 37#include <sys/devfsext.h> 38#endif /*DEVFS*/ 39#include <sys/conf.h> 40#include <sys/uio.h> 41#include <sys/vnode.h> 42 43#include <net/if.h> 44#include <net/netisr.h> 45#include <net/route.h> 46 47#ifdef INET 48#include <netinet/in.h> 49#include <netinet/in_var.h> 50#endif 51 52#ifdef NS 53#include <netns/ns.h> 54#include <netns/ns_if.h> 55#endif 56 57#include "bpfilter.h" 58#if NBPFILTER > 0 59#include <net/bpf.h> 60#endif 61 62#include <net/if_tunvar.h> 63#include <net/if_tun.h> 64 65static void tunattach __P((void *)); 66PSEUDO_SET(tunattach, if_tun); 67 68#define TUNDEBUG if (tundebug) printf 69static int tundebug = 0; 70SYSCTL_INT(_debug, OID_AUTO, if_tun_debug, CTLFLAG_RW, &tundebug, 0, ""); 71 72static struct tun_softc tunctl[NTUN]; 73 74static int tunoutput __P((struct ifnet *, struct mbuf *, struct sockaddr *, 75 struct rtentry *rt)); 76static int tunifioctl __P((struct ifnet *, u_long, caddr_t)); 77static int tuninit __P((int)); 78 79static d_open_t tunopen; 80static d_close_t tunclose; 81static d_read_t tunread; 82static d_write_t tunwrite; 83static d_ioctl_t tunioctl; 84static d_poll_t tunpoll; 85 86#define CDEV_MAJOR 52 87static struct cdevsw tun_cdevsw = { 88 tunopen, tunclose, tunread, tunwrite, 89 tunioctl, nullstop, noreset, nodevtotty, 90 tunpoll, nommap, nostrategy, "tun", NULL, -1 91}; 92 93 94static int tun_devsw_installed; 95#ifdef DEVFS 96static void *tun_devfs_token[NTUN]; 97#endif 98 99#define minor_val(n) ((((n) & ~0xff) << 8) | ((n) & 0xff)) 100#define dev_val(n) (((n) >> 8) | ((n) & 0xff)) 101 102static void 103tunattach(dummy) 104 void *dummy; 105{ 106 register int i; 107 struct ifnet *ifp; 108 dev_t dev; 109 110 if ( tun_devsw_installed ) 111 return; 112 dev = makedev(CDEV_MAJOR, 0); 113 cdevsw_add(&dev, &tun_cdevsw, NULL); 114 tun_devsw_installed = 1; 115 for ( i = 0; i < NTUN; i++ ) { 116#ifdef DEVFS 117 tun_devfs_token[i] = devfs_add_devswf(&tun_cdevsw, minor_val(i), 118 DV_CHR, UID_UUCP, 119 GID_DIALER, 0600, 120 "tun%d", i); 121#endif 122 tunctl[i].tun_flags = TUN_INITED; 123 124 ifp = &tunctl[i].tun_if; 125 ifp->if_unit = i; 126 ifp->if_name = "tun"; 127 ifp->if_mtu = TUNMTU; 128 ifp->if_ioctl = tunifioctl; 129 ifp->if_output = tunoutput; 130 ifp->if_flags = IFF_POINTOPOINT | IFF_MULTICAST; 131 ifp->if_snd.ifq_maxlen = ifqmaxlen; 132 if_attach(ifp); 133#if NBPFILTER > 0 134 bpfattach(ifp, DLT_NULL, sizeof(u_int)); 135#endif 136 } 137} 138 139/* 140 * tunnel open - must be superuser & the device must be 141 * configured in 142 */ 143static int 144tunopen(dev, flag, mode, p) 145 dev_t dev; 146 int flag, mode; 147 struct proc *p; 148{ 149 struct ifnet *ifp; 150 struct tun_softc *tp; 151 register int unit, error; 152 153 error = suser(p); 154 if (error) 155 return (error); 156 157 if ((unit = dev_val(minor(dev))) >= NTUN) 158 return (ENXIO); 159 tp = &tunctl[unit]; 160 if (tp->tun_flags & TUN_OPEN) 161 return EBUSY; 162 ifp = &tp->tun_if; 163 tp->tun_flags |= TUN_OPEN; 164 TUNDEBUG("%s%d: open\n", ifp->if_name, ifp->if_unit); 165 return (0); 166} 167 168/* 169 * tunclose - close the device - mark i/f down & delete 170 * routing info 171 */ 172static int 173tunclose(dev, foo, bar, p) 174 dev_t dev; 175 int foo; 176 int bar; 177 struct proc *p; 178{ 179 register int unit = dev_val(minor(dev)), s; 180 struct tun_softc *tp = &tunctl[unit]; 181 struct ifnet *ifp = &tp->tun_if; 182 struct mbuf *m; 183 184 tp->tun_flags &= ~TUN_OPEN; 185 186 /* 187 * junk all pending output 188 */ 189 do { 190 s = splimp(); 191 IF_DEQUEUE(&ifp->if_snd, m); 192 splx(s); 193 if (m) 194 m_freem(m); 195 } while (m); 196 197 if (ifp->if_flags & IFF_UP) { 198 s = splimp(); 199 if_down(ifp); 200 if (ifp->if_flags & IFF_RUNNING) { 201 /* find internet addresses and delete routes */ 202 register struct ifaddr *ifa; 203 for (ifa = ifp->if_addrhead.tqh_first; ifa; 204 ifa = ifa->ifa_link.tqe_next) { 205 if (ifa->ifa_addr->sa_family == AF_INET) { 206 rtinit(ifa, (int)RTM_DELETE, 207 tp->tun_flags & TUN_DSTADDR ? RTF_HOST : 0); 208 } 209 } 210 } 211 splx(s); 212 } 213 funsetown(tp->tun_sigio); 214 selwakeup(&tp->tun_rsel); 215 216 TUNDEBUG ("%s%d: closed\n", ifp->if_name, ifp->if_unit); 217 return (0); 218} 219 220static int 221tuninit(unit) 222 int unit; 223{ 224 struct tun_softc *tp = &tunctl[unit]; 225 struct ifnet *ifp = &tp->tun_if; 226 register struct ifaddr *ifa; 227 228 TUNDEBUG("%s%d: tuninit\n", ifp->if_name, ifp->if_unit); 229 230 ifp->if_flags |= IFF_UP | IFF_RUNNING; 231 getmicrotime(&ifp->if_lastchange); 232 233 for (ifa = ifp->if_addrhead.tqh_first; ifa; 234 ifa = ifa->ifa_link.tqe_next) { 235#ifdef INET 236 if (ifa->ifa_addr->sa_family == AF_INET) { 237 struct sockaddr_in *si; 238 239 si = (struct sockaddr_in *)ifa->ifa_addr; 240 if (si && si->sin_addr.s_addr) 241 tp->tun_flags |= TUN_IASET; 242 243 si = (struct sockaddr_in *)ifa->ifa_dstaddr; 244 if (si && si->sin_addr.s_addr) 245 tp->tun_flags |= TUN_DSTADDR; 246 } 247#endif 248 } 249 return 0; 250} 251 252/* 253 * Process an ioctl request. 254 */ 255int 256tunifioctl(ifp, cmd, data) 257 struct ifnet *ifp; 258 u_long cmd; 259 caddr_t data; 260{ 261 register struct ifreq *ifr = (struct ifreq *)data; 262 int error = 0, s; 263 264 s = splimp(); 265 switch(cmd) { 266 case SIOCSIFADDR: 267 tuninit(ifp->if_unit); 268 TUNDEBUG("%s%d: address set\n", 269 ifp->if_name, ifp->if_unit); 270 break; 271 case SIOCSIFDSTADDR: 272 tuninit(ifp->if_unit); 273 TUNDEBUG("%s%d: destination address set\n", 274 ifp->if_name, ifp->if_unit); 275 break; 276 case SIOCSIFMTU: 277 ifp->if_mtu = ifr->ifr_mtu; 278 TUNDEBUG("%s%d: mtu set\n", 279 ifp->if_name, ifp->if_unit); 280 break; 281 case SIOCADDMULTI: 282 case SIOCDELMULTI: 283 break; 284 285 286 default: 287 error = EINVAL; 288 } 289 splx(s); 290 return (error); 291} 292 293/* 294 * tunoutput - queue packets from higher level ready to put out. 295 */ 296int 297tunoutput(ifp, m0, dst, rt) 298 struct ifnet *ifp; 299 struct mbuf *m0; 300 struct sockaddr *dst; 301 struct rtentry *rt; 302{ 303 struct tun_softc *tp = &tunctl[ifp->if_unit]; 304 int s; 305 306 TUNDEBUG ("%s%d: tunoutput\n", ifp->if_name, ifp->if_unit); 307 308 if ((tp->tun_flags & TUN_READY) != TUN_READY) { 309 TUNDEBUG ("%s%d: not ready 0%o\n", ifp->if_name, 310 ifp->if_unit, tp->tun_flags); 311 m_freem (m0); 312 return EHOSTDOWN; 313 } 314 315#if NBPFILTER > 0 316 /* BPF write needs to be handled specially */ 317 if (dst->sa_family == AF_UNSPEC) { 318 dst->sa_family = *(mtod(m0, int *)); 319 m0->m_len -= sizeof(int); 320 m0->m_pkthdr.len -= sizeof(int); 321 m0->m_data += sizeof(int); 322 } 323 324 if (ifp->if_bpf) { 325 /* 326 * We need to prepend the address family as 327 * a four byte field. Cons up a dummy header 328 * to pacify bpf. This is safe because bpf 329 * will only read from the mbuf (i.e., it won't 330 * try to free it or keep a pointer to it). 331 */ 332 struct mbuf m; 333 u_int af = dst->sa_family; 334 335 m.m_next = m0; 336 m.m_len = 4; 337 m.m_data = (char *)⁡ 338 339 bpf_mtap(ifp, &m); 340 } 341#endif /* NBPFILTER > 0 */ 342 343 /* prepend sockaddr? this may abort if the mbuf allocation fails */ 344 if (tp->tun_flags & TUN_LMODE) { 345 /* allocate space for sockaddr */ 346 M_PREPEND(m0, dst->sa_len, M_DONTWAIT); 347 348 /* if allocation failed drop packet */ 349 if (m0 == NULL){ 350 s = splimp(); /* spl on queue manipulation */ 351 IF_DROP(&ifp->if_snd); 352 splx(s); 353 ifp->if_oerrors++; 354 return (ENOBUFS); 355 } else { 356 bcopy(dst, m0->m_data, dst->sa_len); 357 } 358 } 359 360 switch(dst->sa_family) { 361#ifdef INET 362 case AF_INET: 363 s = splimp(); 364 if (IF_QFULL(&ifp->if_snd)) { 365 IF_DROP(&ifp->if_snd); 366 m_freem(m0); 367 splx(s); 368 ifp->if_collisions++; 369 return (ENOBUFS); 370 } 371 ifp->if_obytes += m0->m_pkthdr.len; 372 IF_ENQUEUE(&ifp->if_snd, m0); 373 splx(s); 374 ifp->if_opackets++; 375 break; 376#endif 377 default: 378 m_freem(m0); 379 return EAFNOSUPPORT; 380 } 381 382 if (tp->tun_flags & TUN_RWAIT) { 383 tp->tun_flags &= ~TUN_RWAIT; 384 wakeup((caddr_t)tp); 385 } 386 if (tp->tun_flags & TUN_ASYNC && tp->tun_sigio) 387 pgsigio(tp->tun_sigio, SIGIO, 0); 388 selwakeup(&tp->tun_rsel); 389 return 0; 390} 391 392/* 393 * the cdevsw interface is now pretty minimal. 394 */ 395static int 396tunioctl(dev, cmd, data, flag, p) 397 dev_t dev; 398 u_long cmd; 399 caddr_t data; 400 int flag; 401 struct proc *p; 402{ 403 int unit = dev_val(minor(dev)), s; 404 struct tun_softc *tp = &tunctl[unit]; 405 struct tuninfo *tunp; 406 407 switch (cmd) { 408 case TUNSIFINFO: 409 tunp = (struct tuninfo *)data; 410 tp->tun_if.if_mtu = tunp->mtu; 411 tp->tun_if.if_type = tunp->type; 412 tp->tun_if.if_baudrate = tunp->baudrate; 413 break; 414 case TUNGIFINFO: 415 tunp = (struct tuninfo *)data; 416 tunp->mtu = tp->tun_if.if_mtu; 417 tunp->type = tp->tun_if.if_type; 418 tunp->baudrate = tp->tun_if.if_baudrate; 419 break; 420 case TUNSDEBUG: 421 tundebug = *(int *)data; 422 break; 423 case TUNGDEBUG: 424 *(int *)data = tundebug; 425 break; 426 case TUNSLMODE: 427 if (*(int *)data) 428 tp->tun_flags |= TUN_LMODE; 429 else 430 tp->tun_flags &= ~TUN_LMODE; 431 break; 432 case TUNSIFMODE: 433 /* deny this if UP */ 434 if (tp->tun_if.if_flags & IFF_UP) 435 return(EBUSY); 436 437 switch (*(int *)data) { 438 case IFF_POINTOPOINT: 439 tp->tun_if.if_flags |= IFF_POINTOPOINT; 440 tp->tun_if.if_flags &= ~IFF_BROADCAST; 441 break; 442 case IFF_BROADCAST: 443 tp->tun_if.if_flags &= ~IFF_POINTOPOINT; 444 tp->tun_if.if_flags |= IFF_BROADCAST; 445 break; 446 default: 447 return(EINVAL); 448 } 449 break; 450 case FIONBIO: 451 break; 452 case FIOASYNC: 453 if (*(int *)data) 454 tp->tun_flags |= TUN_ASYNC; 455 else 456 tp->tun_flags &= ~TUN_ASYNC; 457 break; 458 case FIONREAD: 459 s = splimp(); 460 if (tp->tun_if.if_snd.ifq_head) { 461 struct mbuf *mb = tp->tun_if.if_snd.ifq_head; 462 for( *(int *)data = 0; mb != 0; mb = mb->m_next) 463 *(int *)data += mb->m_len; 464 } else 465 *(int *)data = 0; 466 splx(s); 467 break; 468 case FIOSETOWN: 469 return (fsetown(*(int *)data, &tp->tun_sigio)); 470 471 case FIOGETOWN: 472 *(int *)data = fgetown(tp->tun_sigio); 473 return (0); 474 475 /* This is deprecated, FIOSETOWN should be used instead. */ 476 case TIOCSPGRP: 477 return (fsetown(-(*(int *)data), &tp->tun_sigio)); 478 479 /* This is deprecated, FIOGETOWN should be used instead. */ 480 case TIOCGPGRP: 481 *(int *)data = -fgetown(tp->tun_sigio); 482 return (0); 483 484 default: 485 return (ENOTTY); 486 } 487 return (0); 488} 489 490/* 491 * The cdevsw read interface - reads a packet at a time, or at 492 * least as much of a packet as can be read. 493 */ 494static int 495tunread(dev, uio, flag) 496 dev_t dev; 497 struct uio *uio; 498 int flag; 499{ 500 int unit = dev_val(minor(dev)); 501 struct tun_softc *tp = &tunctl[unit]; 502 struct ifnet *ifp = &tp->tun_if; 503 struct mbuf *m, *m0; 504 int error=0, len, s; 505 506 TUNDEBUG ("%s%d: read\n", ifp->if_name, ifp->if_unit); 507 if ((tp->tun_flags & TUN_READY) != TUN_READY) { 508 TUNDEBUG ("%s%d: not ready 0%o\n", ifp->if_name, 509 ifp->if_unit, tp->tun_flags); 510 return EHOSTDOWN; 511 } 512 513 tp->tun_flags &= ~TUN_RWAIT; 514 515 s = splimp(); 516 do { 517 IF_DEQUEUE(&ifp->if_snd, m0); 518 if (m0 == 0) { 519 if (flag & IO_NDELAY) { 520 splx(s); 521 return EWOULDBLOCK; 522 } 523 tp->tun_flags |= TUN_RWAIT; 524 if((error = tsleep((caddr_t)tp, PCATCH | (PZERO + 1), 525 "tunread", 0)) != 0) { 526 splx(s); 527 return error; 528 } 529 } 530 } while (m0 == 0); 531 splx(s); 532 533 while (m0 && uio->uio_resid > 0 && error == 0) { 534 len = min(uio->uio_resid, m0->m_len); 535 if (len == 0) 536 break; 537 error = uiomove(mtod(m0, caddr_t), len, uio); 538 MFREE(m0, m); 539 m0 = m; 540 } 541 542 if (m0) { 543 TUNDEBUG("Dropping mbuf\n"); 544 m_freem(m0); 545 } 546 return error; 547} 548 549/* 550 * the cdevsw write interface - an atomic write is a packet - or else! 551 */ 552static int 553tunwrite(dev, uio, flag) 554 dev_t dev; 555 struct uio *uio; 556 int flag; 557{ 558 int unit = dev_val(minor(dev)); 559 struct ifnet *ifp = &tunctl[unit].tun_if; 560 struct mbuf *top, **mp, *m; 561 int error=0, s, tlen, mlen; 562 563 TUNDEBUG("%s%d: tunwrite\n", ifp->if_name, ifp->if_unit); 564 565 if (uio->uio_resid < 0 || uio->uio_resid > TUNMRU) { 566 TUNDEBUG("%s%d: len=%d!\n", ifp->if_name, ifp->if_unit, 567 uio->uio_resid); 568 return EIO; 569 } 570 tlen = uio->uio_resid; 571 572 /* get a header mbuf */ 573 MGETHDR(m, M_DONTWAIT, MT_DATA); 574 if (m == NULL) 575 return ENOBUFS; 576 mlen = MHLEN; 577 578 top = 0; 579 mp = ⊤ 580 while (error == 0 && uio->uio_resid > 0) { 581 m->m_len = min(mlen, uio->uio_resid); 582 error = uiomove(mtod (m, caddr_t), m->m_len, uio); 583 *mp = m; 584 mp = &m->m_next; 585 if (uio->uio_resid > 0) { 586 MGET (m, M_DONTWAIT, MT_DATA); 587 if (m == 0) { 588 error = ENOBUFS; 589 break; 590 } 591 mlen = MLEN; 592 } 593 } 594 if (error) { 595 if (top) 596 m_freem (top); 597 return error; 598 } 599 600 top->m_pkthdr.len = tlen; 601 top->m_pkthdr.rcvif = ifp; 602 603#if NBPFILTER > 0 604 if (ifp->if_bpf) { 605 /* 606 * We need to prepend the address family as 607 * a four byte field. Cons up a dummy header 608 * to pacify bpf. This is safe because bpf 609 * will only read from the mbuf (i.e., it won't 610 * try to free it or keep a pointer to it). 611 */ 612 struct mbuf m; 613 u_int af = AF_INET; 614 615 m.m_next = top; 616 m.m_len = 4; 617 m.m_data = (char *)⁡ 618 619 bpf_mtap(ifp, &m); 620 } 621#endif 622 623#ifdef INET 624 s = splimp(); 625 if (IF_QFULL (&ipintrq)) { 626 IF_DROP(&ipintrq); 627 splx(s); 628 ifp->if_collisions++; 629 m_freem(top); 630 return ENOBUFS; 631 } 632 IF_ENQUEUE(&ipintrq, top); 633 splx(s); 634 ifp->if_ibytes += tlen; 635 ifp->if_ipackets++; 636 schednetisr(NETISR_IP); 637#endif 638 return error; 639} 640 641/* 642 * tunpoll - the poll interface, this is only useful on reads 643 * really. The write detect always returns true, write never blocks 644 * anyway, it either accepts the packet or drops it. 645 */ 646static int 647tunpoll(dev, events, p) 648 dev_t dev; 649 int events; 650 struct proc *p; 651{ 652 int unit = dev_val(minor(dev)), s; 653 struct tun_softc *tp = &tunctl[unit]; 654 struct ifnet *ifp = &tp->tun_if; 655 int revents = 0; 656 657 s = splimp(); 658 TUNDEBUG("%s%d: tunpoll\n", ifp->if_name, ifp->if_unit); 659
|