if_tun.c revision 20407
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 it's 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/select mode of 14 * operation though. 15 */ 16 17#include "tun.h" 18#if NTUN > 0 19 20#include <sys/param.h> 21#include <sys/proc.h> 22#include <sys/systm.h> 23#include <sys/mbuf.h> 24#include <sys/buf.h> 25#include <sys/protosw.h> 26#include <sys/socket.h> 27#include <sys/ioctl.h> 28#include <sys/errno.h> 29#include <sys/syslog.h> 30#include <sys/select.h> 31#include <sys/file.h> 32#include <sys/signalvar.h> 33#include <sys/kernel.h> 34#include <sys/sysctl.h> 35#ifdef DEVFS 36#include <sys/devfsext.h> 37#endif /*DEVFS*/ 38#include <sys/conf.h> 39 40#include <net/if.h> 41#include <net/netisr.h> 42#include <net/route.h> 43 44#ifdef INET 45#include <netinet/in.h> 46#include <netinet/in_systm.h> 47#include <netinet/in_var.h> 48#include <netinet/ip.h> 49#include <netinet/if_ether.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 <sys/time.h> 60#include <net/bpf.h> 61#endif 62 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 *, int, 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_select_t tunselect; 85 86#define CDEV_MAJOR 52 87static struct cdevsw tun_cdevsw = { 88 tunopen, tunclose, tunread, tunwrite, 89 tunioctl, nullstop, noreset, nodevtotty, 90 tunselect, nommap, nostrategy, "tun", NULL, -1 91}; 92 93 94static tun_devsw_installed = 0; 95#ifdef DEVFS 96static void *tun_devfs_token[NTUN]; 97#endif 98 99static void 100tunattach(dummy) 101 void *dummy; 102{ 103 register int i; 104 struct ifnet *ifp; 105 dev_t dev; 106 107 if ( tun_devsw_installed ) 108 return; 109 dev = makedev(CDEV_MAJOR, 0); 110 cdevsw_add(&dev, &tun_cdevsw, NULL); 111 tun_devsw_installed = 1; 112 for ( i = 0; i < NTUN; i++ ) { 113#ifdef DEVFS 114 tun_devfs_token[i] = devfs_add_devswf(&tun_cdevsw, i, DV_CHR, 115 UID_UUCP, GID_DIALER, 116 0600, "tun%d", i); 117#endif 118 tunctl[i].tun_flags = TUN_INITED; 119 120 ifp = &tunctl[i].tun_if; 121 ifp->if_unit = i; 122 ifp->if_name = "tun"; 123 ifp->if_mtu = TUNMTU; 124 ifp->if_ioctl = tunifioctl; 125 ifp->if_output = tunoutput; 126 ifp->if_flags = IFF_POINTOPOINT | IFF_MULTICAST; 127 ifp->if_snd.ifq_maxlen = ifqmaxlen; 128 ifp->if_collisions = 0; 129 ifp->if_ierrors = 0; 130 ifp->if_oerrors = 0; 131 ifp->if_ipackets = 0; 132 ifp->if_opackets = 0; 133 if_attach(ifp); 134#if NBPFILTER > 0 135 bpfattach(ifp, DLT_NULL, sizeof(u_int)); 136#endif 137 } 138} 139 140/* 141 * tunnel open - must be superuser & the device must be 142 * configured in 143 */ 144static int 145tunopen(dev, flag, mode, p) 146 dev_t dev; 147 int flag, mode; 148 struct proc *p; 149{ 150 struct ifnet *ifp; 151 struct tun_softc *tp; 152 register int unit, error; 153 154 error = suser(p->p_ucred, &p->p_acflag); 155 if (error) 156 return (error); 157 158 if ((unit = minor(dev)) >= NTUN) 159 return (ENXIO); 160 tp = &tunctl[unit]; 161 if (tp->tun_flags & TUN_OPEN) 162 return EBUSY; 163 ifp = &tp->tun_if; 164 tp->tun_flags |= TUN_OPEN; 165 TUNDEBUG("%s%d: open\n", ifp->if_name, ifp->if_unit); 166 return (0); 167} 168 169/* 170 * tunclose - close the device - mark i/f down & delete 171 * routing info 172 */ 173static int 174tunclose(dev_t dev, int foo, int bar, struct proc *p) 175{ 176 register int unit = minor(dev), s; 177 struct tun_softc *tp = &tunctl[unit]; 178 struct ifnet *ifp = &tp->tun_if; 179 struct mbuf *m; 180 181 tp->tun_flags &= ~TUN_OPEN; 182 183 /* 184 * junk all pending output 185 */ 186 do { 187 s = splimp(); 188 IF_DEQUEUE(&ifp->if_snd, m); 189 splx(s); 190 if (m) 191 m_freem(m); 192 } while (m); 193 194 if (ifp->if_flags & IFF_UP) { 195 s = splimp(); 196 if_down(ifp); 197 if (ifp->if_flags & IFF_RUNNING) { 198 /* find internet addresses and delete routes */ 199 register struct ifaddr *ifa; 200 for (ifa = ifp->if_addrhead.tqh_first; ifa; 201 ifa = ifa->ifa_link.tqe_next) { 202 if (ifa->ifa_addr->sa_family == AF_INET) { 203 rtinit(ifa, (int)RTM_DELETE, 204 tp->tun_flags & TUN_DSTADDR ? RTF_HOST : 0); 205 } 206 } 207 } 208 splx(s); 209 } 210 tp->tun_pgrp = 0; 211 selwakeup(&tp->tun_rsel); 212 213 TUNDEBUG ("%s%d: closed\n", ifp->if_name, ifp->if_unit); 214 return (0); 215} 216 217static int 218tuninit(unit) 219 int unit; 220{ 221 struct tun_softc *tp = &tunctl[unit]; 222 struct ifnet *ifp = &tp->tun_if; 223 register struct ifaddr *ifa; 224 225 TUNDEBUG("%s%d: tuninit\n", ifp->if_name, ifp->if_unit); 226 227 ifp->if_flags |= IFF_UP | IFF_RUNNING; 228 microtime(&ifp->if_lastchange); 229 230 for (ifa = ifp->if_addrhead.tqh_first; ifa; 231 ifa = ifa->ifa_link.tqe_next) 232 if (ifa->ifa_addr->sa_family == AF_INET) { 233 struct sockaddr_in *si; 234 235 si = (struct sockaddr_in *)ifa->ifa_addr; 236 if (si && si->sin_addr.s_addr) 237 tp->tun_flags |= TUN_IASET; 238 239 si = (struct sockaddr_in *)ifa->ifa_dstaddr; 240 if (si && si->sin_addr.s_addr) 241 tp->tun_flags |= TUN_DSTADDR; 242 } 243 244 return 0; 245} 246 247/* 248 * Process an ioctl request. 249 */ 250int 251tunifioctl(ifp, cmd, data) 252 struct ifnet *ifp; 253 int cmd; 254 caddr_t data; 255{ 256 register struct ifreq *ifr = (struct ifreq *)data; 257 int error = 0, s; 258 259 s = splimp(); 260 switch(cmd) { 261 case SIOCSIFADDR: 262 tuninit(ifp->if_unit); 263 TUNDEBUG("%s%d: address set\n", 264 ifp->if_name, ifp->if_unit); 265 break; 266 case SIOCSIFDSTADDR: 267 tuninit(ifp->if_unit); 268 TUNDEBUG("%s%d: destination address set\n", 269 ifp->if_name, ifp->if_unit); 270 break; 271 case SIOCADDMULTI: 272 case SIOCDELMULTI: 273 if (ifr == 0) { 274 error = EAFNOSUPPORT; /* XXX */ 275 break; 276 } 277 switch (ifr->ifr_addr.sa_family) { 278 279#ifdef INET 280 case AF_INET: 281 break; 282#endif 283 284 default: 285 error = EAFNOSUPPORT; 286 break; 287 } 288 break; 289 290 291 default: 292 error = EINVAL; 293 } 294 splx(s); 295 return (error); 296} 297 298/* 299 * tunoutput - queue packets from higher level ready to put out. 300 */ 301int 302tunoutput(ifp, m0, dst, rt) 303 struct ifnet *ifp; 304 struct mbuf *m0; 305 struct sockaddr *dst; 306 struct rtentry *rt; 307{ 308 struct tun_softc *tp = &tunctl[ifp->if_unit]; 309 struct proc *p; 310 int s; 311 312 TUNDEBUG ("%s%d: tunoutput\n", ifp->if_name, ifp->if_unit); 313 314 if ((tp->tun_flags & TUN_READY) != TUN_READY) { 315 TUNDEBUG ("%s%d: not ready 0%o\n", ifp->if_name, 316 ifp->if_unit, tp->tun_flags); 317 m_freem (m0); 318 return EHOSTDOWN; 319 } 320 321#if NBPFILTER > 0 322 /* BPF write needs to be handled specially */ 323 if (dst->sa_family == AF_UNSPEC) { 324 dst->sa_family = *(mtod(m0, int *)); 325 m0->m_len -= sizeof(int); 326 m0->m_pkthdr.len -= sizeof(int); 327 m0->m_data += sizeof(int); 328 } 329 330 if (ifp->if_bpf) { 331 /* 332 * We need to prepend the address family as 333 * a four byte field. Cons up a dummy header 334 * to pacify bpf. This is safe because bpf 335 * will only read from the mbuf (i.e., it won't 336 * try to free it or keep a pointer to it). 337 */ 338 struct mbuf m; 339 u_int af = dst->sa_family; 340 341 m.m_next = m0; 342 m.m_len = 4; 343 m.m_data = (char *)⁡ 344 345 bpf_mtap(ifp, &m); 346 } 347#endif 348 349 switch(dst->sa_family) { 350#ifdef INET 351 case AF_INET: 352 s = splimp(); 353 if (IF_QFULL(&ifp->if_snd)) { 354 IF_DROP(&ifp->if_snd); 355 m_freem(m0); 356 splx(s); 357 ifp->if_collisions++; 358 return (ENOBUFS); 359 } 360 ifp->if_obytes += m0->m_pkthdr.len; 361 IF_ENQUEUE(&ifp->if_snd, m0); 362 splx(s); 363 ifp->if_opackets++; 364 break; 365#endif 366 default: 367 m_freem(m0); 368 return EAFNOSUPPORT; 369 } 370 371 if (tp->tun_flags & TUN_RWAIT) { 372 tp->tun_flags &= ~TUN_RWAIT; 373 wakeup((caddr_t)tp); 374 } 375 if (tp->tun_flags & TUN_ASYNC && tp->tun_pgrp) { 376 if (tp->tun_pgrp > 0) 377 gsignal(tp->tun_pgrp, SIGIO); 378 else if ((p = pfind(-tp->tun_pgrp)) != 0) 379 psignal(p, SIGIO); 380 } 381 selwakeup(&tp->tun_rsel); 382 return 0; 383} 384 385/* 386 * the cdevsw interface is now pretty minimal. 387 */ 388static int 389tunioctl(dev, cmd, data, flag, p) 390 dev_t dev; 391 int cmd; 392 caddr_t data; 393 int flag; 394 struct proc *p; 395{ 396 int unit = minor(dev), s; 397 struct tun_softc *tp = &tunctl[unit]; 398 struct tuninfo *tunp; 399 400 switch (cmd) { 401 case TUNSIFINFO: 402 tunp = (struct tuninfo *)data; 403 tp->tun_if.if_mtu = tunp->mtu; 404 tp->tun_if.if_type = tunp->type; 405 tp->tun_if.if_baudrate = tunp->baudrate; 406 break; 407 case TUNGIFINFO: 408 tunp = (struct tuninfo *)data; 409 tunp->mtu = tp->tun_if.if_mtu; 410 tunp->type = tp->tun_if.if_type; 411 tunp->baudrate = tp->tun_if.if_baudrate; 412 break; 413 case TUNSDEBUG: 414 tundebug = *(int *)data; 415 break; 416 case TUNGDEBUG: 417 *(int *)data = tundebug; 418 break; 419 case FIONBIO: 420 if (*(int *)data) 421 tp->tun_flags |= TUN_NBIO; 422 else 423 tp->tun_flags &= ~TUN_NBIO; 424 break; 425 case FIOASYNC: 426 if (*(int *)data) 427 tp->tun_flags |= TUN_ASYNC; 428 else 429 tp->tun_flags &= ~TUN_ASYNC; 430 break; 431 case FIONREAD: 432 s = splimp(); 433 if (tp->tun_if.if_snd.ifq_head) { 434 struct mbuf *mb = tp->tun_if.if_snd.ifq_head; 435 for( *(int *)data = 0; mb != 0; mb = mb->m_next) 436 *(int *)data += mb->m_len; 437 } else 438 *(int *)data = 0; 439 splx(s); 440 break; 441 case TIOCSPGRP: 442 tp->tun_pgrp = *(int *)data; 443 break; 444 case TIOCGPGRP: 445 *(int *)data = tp->tun_pgrp; 446 break; 447 default: 448 return (ENOTTY); 449 } 450 return (0); 451} 452 453/* 454 * The cdevsw read interface - reads a packet at a time, or at 455 * least as much of a packet as can be read. 456 */ 457static int 458tunread(dev_t dev, struct uio *uio, int flag) 459{ 460 int unit = minor(dev); 461 struct tun_softc *tp = &tunctl[unit]; 462 struct ifnet *ifp = &tp->tun_if; 463 struct mbuf *m, *m0; 464 int error=0, len, s; 465 466 TUNDEBUG ("%s%d: read\n", ifp->if_name, ifp->if_unit); 467 if ((tp->tun_flags & TUN_READY) != TUN_READY) { 468 TUNDEBUG ("%s%d: not ready 0%o\n", ifp->if_name, 469 ifp->if_unit, tp->tun_flags); 470 return EHOSTDOWN; 471 } 472 473 tp->tun_flags &= ~TUN_RWAIT; 474 475 s = splimp(); 476 do { 477 IF_DEQUEUE(&ifp->if_snd, m0); 478 if (m0 == 0) { 479 if (tp->tun_flags & TUN_NBIO) { 480 splx(s); 481 return EWOULDBLOCK; 482 } 483 tp->tun_flags |= TUN_RWAIT; 484 if( error = tsleep((caddr_t)tp, PCATCH | (PZERO + 1), 485 "tunread", 0)) { 486 splx(s); 487 return error; 488 } 489 } 490 } while (m0 == 0); 491 splx(s); 492 493 while (m0 && uio->uio_resid > 0 && error == 0) { 494 len = min(uio->uio_resid, m0->m_len); 495 if (len == 0) 496 break; 497 error = uiomove(mtod(m0, caddr_t), len, uio); 498 MFREE(m0, m); 499 m0 = m; 500 } 501 502 if (m0) { 503 TUNDEBUG("Dropping mbuf\n"); 504 m_freem(m0); 505 } 506 return error; 507} 508 509/* 510 * the cdevsw write interface - an atomic write is a packet - or else! 511 */ 512static int 513tunwrite(dev_t dev, struct uio *uio, int flag) 514{ 515 int unit = minor (dev); 516 struct ifnet *ifp = &tunctl[unit].tun_if; 517 struct mbuf *top, **mp, *m; 518 int error=0, s, tlen, mlen; 519 520 TUNDEBUG("%s%d: tunwrite\n", ifp->if_name, ifp->if_unit); 521 522 if (uio->uio_resid < 0 || uio->uio_resid > TUNMTU) { 523 TUNDEBUG("%s%d: len=%d!\n", ifp->if_name, ifp->if_unit, 524 uio->uio_resid); 525 return EIO; 526 } 527 tlen = uio->uio_resid; 528 529 /* get a header mbuf */ 530 MGETHDR(m, M_DONTWAIT, MT_DATA); 531 if (m == NULL) 532 return ENOBUFS; 533 mlen = MHLEN; 534 535 top = 0; 536 mp = ⊤ 537 while (error == 0 && uio->uio_resid > 0) { 538 m->m_len = min(mlen, uio->uio_resid); 539 error = uiomove(mtod (m, caddr_t), m->m_len, uio); 540 *mp = m; 541 mp = &m->m_next; 542 if (uio->uio_resid > 0) { 543 MGET (m, M_DONTWAIT, MT_DATA); 544 if (m == 0) { 545 error = ENOBUFS; 546 break; 547 } 548 mlen = MLEN; 549 } 550 } 551 if (error) { 552 if (top) 553 m_freem (top); 554 return error; 555 } 556 557 top->m_pkthdr.len = tlen; 558 top->m_pkthdr.rcvif = ifp; 559 560#if NBPFILTER > 0 561 if (ifp->if_bpf) { 562 /* 563 * We need to prepend the address family as 564 * a four byte field. Cons up a dummy header 565 * to pacify bpf. This is safe because bpf 566 * will only read from the mbuf (i.e., it won't 567 * try to free it or keep a pointer to it). 568 */ 569 struct mbuf m; 570 u_int af = AF_INET; 571 572 m.m_next = top; 573 m.m_len = 4; 574 m.m_data = (char *)⁡ 575 576 bpf_mtap(ifp, &m); 577 } 578#endif 579 580 s = splimp(); 581 if (IF_QFULL (&ipintrq)) { 582 IF_DROP(&ipintrq); 583 splx(s); 584 ifp->if_collisions++; 585 m_freem(top); 586 return ENOBUFS; 587 } 588 IF_ENQUEUE(&ipintrq, top); 589 splx(s); 590 ifp->if_ibytes += tlen; 591 ifp->if_ipackets++; 592 schednetisr(NETISR_IP); 593 return error; 594} 595 596/* 597 * tunselect - the select interface, this is only useful on reads 598 * really. The write detect always returns true, write never blocks 599 * anyway, it either accepts the packet or drops it. 600 */ 601static int 602tunselect(dev_t dev, int rw, struct proc *p) 603{ 604 int unit = minor(dev), s; 605 struct tun_softc *tp = &tunctl[unit]; 606 struct ifnet *ifp = &tp->tun_if; 607 608 s = splimp(); 609 TUNDEBUG("%s%d: tunselect\n", ifp->if_name, ifp->if_unit); 610 611 switch (rw) { 612 case FREAD: 613 if (ifp->if_snd.ifq_len > 0) { 614 splx(s); 615 TUNDEBUG("%s%d: tunselect q=%d\n", ifp->if_name, 616 ifp->if_unit, ifp->if_snd.ifq_len); 617 return 1; 618 } 619 selrecord(p, &tp->tun_rsel); 620 break; 621 case FWRITE: 622 splx(s); 623 return 1; 624 } 625 splx(s); 626 TUNDEBUG("%s%d: tunselect waiting\n", ifp->if_name, ifp->if_unit); 627 return 0; 628} 629 630 631#endif /* NTUN */ 632