bpf.c revision 104090
1298707Sjhb/* 2298707Sjhb * Copyright (c) 1990, 1991, 1993 3298707Sjhb * The Regents of the University of California. All rights reserved. 4298707Sjhb * 5298707Sjhb * This code is derived from the Stanford/CMU enet packet filter, 6298707Sjhb * (net/enet.c) distributed as part of 4.3BSD, and code contributed 7298707Sjhb * to Berkeley by Steven McCanne and Van Jacobson both of Lawrence 8298707Sjhb * Berkeley Laboratory. 9298707Sjhb * 10298707Sjhb * Redistribution and use in source and binary forms, with or without 11298707Sjhb * modification, are permitted provided that the following conditions 12298707Sjhb * are met: 13298707Sjhb * 1. Redistributions of source code must retain the above copyright 14298707Sjhb * notice, this list of conditions and the following disclaimer. 15298707Sjhb * 2. Redistributions in binary form must reproduce the above copyright 16298707Sjhb * notice, this list of conditions and the following disclaimer in the 17298707Sjhb * documentation and/or other materials provided with the distribution. 18298707Sjhb * 3. All advertising materials mentioning features or use of this software 19298707Sjhb * must display the following acknowledgement: 20298707Sjhb * This product includes software developed by the University of 21298707Sjhb * California, Berkeley and its contributors. 22298707Sjhb * 4. Neither the name of the University nor the names of its contributors 23298707Sjhb * may be used to endorse or promote products derived from this software 24298707Sjhb * without specific prior written permission. 25298707Sjhb * 26298707Sjhb * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27298707Sjhb * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28298707Sjhb * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29298707Sjhb * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30298707Sjhb * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31298707Sjhb * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32298707Sjhb * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33298707Sjhb * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34298707Sjhb * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35298707Sjhb * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36298707Sjhb * SUCH DAMAGE. 37298707Sjhb * 38298707Sjhb * @(#)bpf.c 8.4 (Berkeley) 1/9/95 39298707Sjhb * 40298707Sjhb * $FreeBSD: head/sys/net/bpf.c 104090 2002-09-28 14:03:27Z phk $ 41298707Sjhb */ 42298707Sjhb 43298707Sjhb#include "opt_bpf.h" 44298707Sjhb#include "opt_mac.h" 45298707Sjhb#include "opt_netgraph.h" 46298707Sjhb 47298707Sjhb#include <sys/param.h> 48298707Sjhb#include <sys/systm.h> 49298707Sjhb#include <sys/conf.h> 50298707Sjhb#include <sys/mac.h> 51#include <sys/malloc.h> 52#include <sys/mbuf.h> 53#include <sys/time.h> 54#include <sys/proc.h> 55#include <sys/signalvar.h> 56#include <sys/filio.h> 57#include <sys/sockio.h> 58#include <sys/ttycom.h> 59#include <sys/filedesc.h> 60 61#include <sys/poll.h> 62 63#include <sys/socket.h> 64#include <sys/vnode.h> 65 66#include <net/if.h> 67#include <net/bpf.h> 68#include <net/bpfdesc.h> 69 70#include <netinet/in.h> 71#include <netinet/if_ether.h> 72#include <sys/kernel.h> 73#include <sys/sysctl.h> 74 75static MALLOC_DEFINE(M_BPF, "BPF", "BPF data"); 76 77#if defined(DEV_BPF) || defined(NETGRAPH_BPF) 78 79#define PRINET 26 /* interruptible */ 80 81/* 82 * The default read buffer size is patchable. 83 */ 84static int bpf_bufsize = 4096; 85SYSCTL_INT(_debug, OID_AUTO, bpf_bufsize, CTLFLAG_RW, 86 &bpf_bufsize, 0, ""); 87static int bpf_maxbufsize = BPF_MAXBUFSIZE; 88SYSCTL_INT(_debug, OID_AUTO, bpf_maxbufsize, CTLFLAG_RW, 89 &bpf_maxbufsize, 0, ""); 90 91/* 92 * bpf_iflist is the list of interfaces; each corresponds to an ifnet 93 */ 94static struct bpf_if *bpf_iflist; 95static struct mtx bpf_mtx; /* bpf global lock */ 96 97static int bpf_allocbufs(struct bpf_d *); 98static void bpf_attachd(struct bpf_d *d, struct bpf_if *bp); 99static void bpf_detachd(struct bpf_d *d); 100static void bpf_freed(struct bpf_d *); 101static void bpf_mcopy(const void *, void *, size_t); 102static int bpf_movein(struct uio *, int, 103 struct mbuf **, struct sockaddr *, int *); 104static int bpf_setif(struct bpf_d *, struct ifreq *); 105static void bpf_timed_out(void *); 106static __inline void 107 bpf_wakeup(struct bpf_d *); 108static void catchpacket(struct bpf_d *, u_char *, u_int, 109 u_int, void (*)(const void *, void *, size_t)); 110static void reset_d(struct bpf_d *); 111static int bpf_setf(struct bpf_d *, struct bpf_program *); 112 113static d_open_t bpfopen; 114static d_close_t bpfclose; 115static d_read_t bpfread; 116static d_write_t bpfwrite; 117static d_ioctl_t bpfioctl; 118static d_poll_t bpfpoll; 119 120#define CDEV_MAJOR 23 121static struct cdevsw bpf_cdevsw = { 122 /* open */ bpfopen, 123 /* close */ bpfclose, 124 /* read */ bpfread, 125 /* write */ bpfwrite, 126 /* ioctl */ bpfioctl, 127 /* poll */ bpfpoll, 128 /* mmap */ nommap, 129 /* strategy */ nostrategy, 130 /* name */ "bpf", 131 /* maj */ CDEV_MAJOR, 132 /* dump */ nodump, 133 /* psize */ nopsize, 134 /* flags */ 0, 135}; 136 137 138static int 139bpf_movein(uio, linktype, mp, sockp, datlen) 140 register struct uio *uio; 141 int linktype, *datlen; 142 register struct mbuf **mp; 143 register struct sockaddr *sockp; 144{ 145 struct mbuf *m; 146 int error; 147 int len; 148 int hlen; 149 150 /* 151 * Build a sockaddr based on the data link layer type. 152 * We do this at this level because the ethernet header 153 * is copied directly into the data field of the sockaddr. 154 * In the case of SLIP, there is no header and the packet 155 * is forwarded as is. 156 * Also, we are careful to leave room at the front of the mbuf 157 * for the link level header. 158 */ 159 switch (linktype) { 160 161 case DLT_SLIP: 162 sockp->sa_family = AF_INET; 163 hlen = 0; 164 break; 165 166 case DLT_EN10MB: 167 sockp->sa_family = AF_UNSPEC; 168 /* XXX Would MAXLINKHDR be better? */ 169 hlen = sizeof(struct ether_header); 170 break; 171 172 case DLT_FDDI: 173 sockp->sa_family = AF_IMPLINK; 174 hlen = 0; 175 break; 176 177 case DLT_RAW: 178 case DLT_NULL: 179 sockp->sa_family = AF_UNSPEC; 180 hlen = 0; 181 break; 182 183 case DLT_ATM_RFC1483: 184 /* 185 * en atm driver requires 4-byte atm pseudo header. 186 * though it isn't standard, vpi:vci needs to be 187 * specified anyway. 188 */ 189 sockp->sa_family = AF_UNSPEC; 190 hlen = 12; /* XXX 4(ATM_PH) + 3(LLC) + 5(SNAP) */ 191 break; 192 193 case DLT_PPP: 194 sockp->sa_family = AF_UNSPEC; 195 hlen = 4; /* This should match PPP_HDRLEN */ 196 break; 197 198 default: 199 return (EIO); 200 } 201 202 len = uio->uio_resid; 203 *datlen = len - hlen; 204 if ((unsigned)len > MCLBYTES) 205 return (EIO); 206 207 MGETHDR(m, M_TRYWAIT, MT_DATA); 208 if (m == 0) 209 return (ENOBUFS); 210 if (len > MHLEN) { 211 MCLGET(m, M_TRYWAIT); 212 if ((m->m_flags & M_EXT) == 0) { 213 error = ENOBUFS; 214 goto bad; 215 } 216 } 217 m->m_pkthdr.len = m->m_len = len; 218 m->m_pkthdr.rcvif = NULL; 219 *mp = m; 220 /* 221 * Make room for link header. 222 */ 223 if (hlen != 0) { 224 m->m_pkthdr.len -= hlen; 225 m->m_len -= hlen; 226#if BSD >= 199103 227 m->m_data += hlen; /* XXX */ 228#else 229 m->m_off += hlen; 230#endif 231 error = uiomove((caddr_t)sockp->sa_data, hlen, uio); 232 if (error) 233 goto bad; 234 } 235 error = uiomove(mtod(m, caddr_t), len - hlen, uio); 236 if (!error) 237 return (0); 238 bad: 239 m_freem(m); 240 return (error); 241} 242 243/* 244 * Attach file to the bpf interface, i.e. make d listen on bp. 245 */ 246static void 247bpf_attachd(d, bp) 248 struct bpf_d *d; 249 struct bpf_if *bp; 250{ 251 /* 252 * Point d at bp, and add d to the interface's list of listeners. 253 * Finally, point the driver's bpf cookie at the interface so 254 * it will divert packets to bpf. 255 */ 256 BPFIF_LOCK(bp); 257 d->bd_bif = bp; 258 d->bd_next = bp->bif_dlist; 259 bp->bif_dlist = d; 260 261 bp->bif_ifp->if_bpf = bp; 262 BPFIF_UNLOCK(bp); 263} 264 265/* 266 * Detach a file from its interface. 267 */ 268static void 269bpf_detachd(d) 270 struct bpf_d *d; 271{ 272 int error; 273 struct bpf_d **p; 274 struct bpf_if *bp; 275 276 bp = d->bd_bif; 277 /* 278 * Check if this descriptor had requested promiscuous mode. 279 * If so, turn it off. 280 */ 281 if (d->bd_promisc) { 282 d->bd_promisc = 0; 283 error = ifpromisc(bp->bif_ifp, 0); 284 if (error != 0 && error != ENXIO) { 285 /* 286 * ENXIO can happen if a pccard is unplugged 287 * Something is really wrong if we were able to put 288 * the driver into promiscuous mode, but can't 289 * take it out. 290 */ 291 printf("%s%d: ifpromisc failed %d\n", 292 bp->bif_ifp->if_name, bp->bif_ifp->if_unit, error); 293 } 294 } 295 /* Remove d from the interface's descriptor list. */ 296 BPFIF_LOCK(bp); 297 p = &bp->bif_dlist; 298 while (*p != d) { 299 p = &(*p)->bd_next; 300 if (*p == 0) 301 panic("bpf_detachd: descriptor not in list"); 302 } 303 *p = (*p)->bd_next; 304 if (bp->bif_dlist == 0) 305 /* 306 * Let the driver know that there are no more listeners. 307 */ 308 d->bd_bif->bif_ifp->if_bpf = 0; 309 BPFIF_UNLOCK(bp); 310 d->bd_bif = 0; 311} 312 313/* 314 * Open ethernet device. Returns ENXIO for illegal minor device number, 315 * EBUSY if file is open by another process. 316 */ 317/* ARGSUSED */ 318static int 319bpfopen(dev, flags, fmt, td) 320 dev_t dev; 321 int flags; 322 int fmt; 323 struct thread *td; 324{ 325 struct bpf_d *d; 326 327 mtx_lock(&bpf_mtx); 328 d = dev->si_drv1; 329 /* 330 * Each minor can be opened by only one process. If the requested 331 * minor is in use, return EBUSY. 332 */ 333 if (d) { 334 mtx_unlock(&bpf_mtx); 335 return (EBUSY); 336 } 337 dev->si_drv1 = (struct bpf_d *)~0; /* mark device in use */ 338 mtx_unlock(&bpf_mtx); 339 340 if ((dev->si_flags & SI_NAMED) == 0) 341 make_dev(&bpf_cdevsw, minor(dev), UID_ROOT, GID_WHEEL, 0600, 342 "bpf%d", dev2unit(dev)); 343 MALLOC(d, struct bpf_d *, sizeof(*d), M_BPF, M_WAITOK | M_ZERO); 344 dev->si_drv1 = d; 345 d->bd_bufsize = bpf_bufsize; 346 d->bd_sig = SIGIO; 347 d->bd_seesent = 1; 348#ifdef MAC 349 mac_init_bpfdesc(d); 350 mac_create_bpfdesc(td->td_ucred, d); 351#endif 352 mtx_init(&d->bd_mtx, devtoname(dev), "bpf cdev lock", MTX_DEF); 353 callout_init(&d->bd_callout, 1); 354 355 return (0); 356} 357 358/* 359 * Close the descriptor by detaching it from its interface, 360 * deallocating its buffers, and marking it free. 361 */ 362/* ARGSUSED */ 363static int 364bpfclose(dev, flags, fmt, td) 365 dev_t dev; 366 int flags; 367 int fmt; 368 struct thread *td; 369{ 370 struct bpf_d *d = dev->si_drv1; 371 372 BPFD_LOCK(d); 373 if (d->bd_state == BPF_WAITING) 374 callout_stop(&d->bd_callout); 375 d->bd_state = BPF_IDLE; 376 BPFD_UNLOCK(d); 377 funsetown(&d->bd_sigio); 378 mtx_lock(&bpf_mtx); 379 if (d->bd_bif) 380 bpf_detachd(d); 381 mtx_unlock(&bpf_mtx); 382#ifdef MAC 383 mac_destroy_bpfdesc(d); 384#endif /* MAC */ 385 bpf_freed(d); 386 dev->si_drv1 = 0; 387 free(d, M_BPF); 388 389 return (0); 390} 391 392 393/* 394 * Rotate the packet buffers in descriptor d. Move the store buffer 395 * into the hold slot, and the free buffer into the store slot. 396 * Zero the length of the new store buffer. 397 */ 398#define ROTATE_BUFFERS(d) \ 399 (d)->bd_hbuf = (d)->bd_sbuf; \ 400 (d)->bd_hlen = (d)->bd_slen; \ 401 (d)->bd_sbuf = (d)->bd_fbuf; \ 402 (d)->bd_slen = 0; \ 403 (d)->bd_fbuf = 0; 404/* 405 * bpfread - read next chunk of packets from buffers 406 */ 407static int 408bpfread(dev, uio, ioflag) 409 dev_t dev; 410 register struct uio *uio; 411 int ioflag; 412{ 413 struct bpf_d *d = dev->si_drv1; 414 int timed_out; 415 int error; 416 417 /* 418 * Restrict application to use a buffer the same size as 419 * as kernel buffers. 420 */ 421 if (uio->uio_resid != d->bd_bufsize) 422 return (EINVAL); 423 424 BPFD_LOCK(d); 425 if (d->bd_state == BPF_WAITING) 426 callout_stop(&d->bd_callout); 427 timed_out = (d->bd_state == BPF_TIMED_OUT); 428 d->bd_state = BPF_IDLE; 429 /* 430 * If the hold buffer is empty, then do a timed sleep, which 431 * ends when the timeout expires or when enough packets 432 * have arrived to fill the store buffer. 433 */ 434 while (d->bd_hbuf == 0) { 435 if ((d->bd_immediate || timed_out) && d->bd_slen != 0) { 436 /* 437 * A packet(s) either arrived since the previous 438 * read or arrived while we were asleep. 439 * Rotate the buffers and return what's here. 440 */ 441 ROTATE_BUFFERS(d); 442 break; 443 } 444 445 /* 446 * No data is available, check to see if the bpf device 447 * is still pointed at a real interface. If not, return 448 * ENXIO so that the userland process knows to rebind 449 * it before using it again. 450 */ 451 if (d->bd_bif == NULL) { 452 BPFD_UNLOCK(d); 453 return (ENXIO); 454 } 455 456 if (ioflag & IO_NDELAY) { 457 BPFD_UNLOCK(d); 458 return (EWOULDBLOCK); 459 } 460 error = msleep((caddr_t)d, &d->bd_mtx, PRINET|PCATCH, 461 "bpf", d->bd_rtout); 462 if (error == EINTR || error == ERESTART) { 463 BPFD_UNLOCK(d); 464 return (error); 465 } 466 if (error == EWOULDBLOCK) { 467 /* 468 * On a timeout, return what's in the buffer, 469 * which may be nothing. If there is something 470 * in the store buffer, we can rotate the buffers. 471 */ 472 if (d->bd_hbuf) 473 /* 474 * We filled up the buffer in between 475 * getting the timeout and arriving 476 * here, so we don't need to rotate. 477 */ 478 break; 479 480 if (d->bd_slen == 0) { 481 BPFD_UNLOCK(d); 482 return (0); 483 } 484 ROTATE_BUFFERS(d); 485 break; 486 } 487 } 488 /* 489 * At this point, we know we have something in the hold slot. 490 */ 491 BPFD_UNLOCK(d); 492 493 /* 494 * Move data from hold buffer into user space. 495 * We know the entire buffer is transferred since 496 * we checked above that the read buffer is bpf_bufsize bytes. 497 */ 498 error = uiomove(d->bd_hbuf, d->bd_hlen, uio); 499 500 BPFD_LOCK(d); 501 d->bd_fbuf = d->bd_hbuf; 502 d->bd_hbuf = 0; 503 d->bd_hlen = 0; 504 BPFD_UNLOCK(d); 505 506 return (error); 507} 508 509 510/* 511 * If there are processes sleeping on this descriptor, wake them up. 512 */ 513static __inline void 514bpf_wakeup(d) 515 register struct bpf_d *d; 516{ 517 if (d->bd_state == BPF_WAITING) { 518 callout_stop(&d->bd_callout); 519 d->bd_state = BPF_IDLE; 520 } 521 wakeup((caddr_t)d); 522 if (d->bd_async && d->bd_sig && d->bd_sigio) 523 pgsigio(&d->bd_sigio, d->bd_sig, 0); 524 525 selwakeup(&d->bd_sel); 526} 527 528static void 529bpf_timed_out(arg) 530 void *arg; 531{ 532 struct bpf_d *d = (struct bpf_d *)arg; 533 534 BPFD_LOCK(d); 535 if (d->bd_state == BPF_WAITING) { 536 d->bd_state = BPF_TIMED_OUT; 537 if (d->bd_slen != 0) 538 bpf_wakeup(d); 539 } 540 BPFD_UNLOCK(d); 541} 542 543static int 544bpfwrite(dev, uio, ioflag) 545 dev_t dev; 546 struct uio *uio; 547 int ioflag; 548{ 549 struct bpf_d *d = dev->si_drv1; 550 struct ifnet *ifp; 551 struct mbuf *m; 552 int error; 553 static struct sockaddr dst; 554 int datlen; 555 556 if (d->bd_bif == 0) 557 return (ENXIO); 558 559 ifp = d->bd_bif->bif_ifp; 560 561 if (uio->uio_resid == 0) 562 return (0); 563 564 error = bpf_movein(uio, (int)d->bd_bif->bif_dlt, &m, &dst, &datlen); 565 if (error) 566 return (error); 567 568 if (datlen > ifp->if_mtu) 569 return (EMSGSIZE); 570 571 if (d->bd_hdrcmplt) 572 dst.sa_family = pseudo_AF_HDRCMPLT; 573 574 mtx_lock(&Giant); 575#ifdef MAC 576 mac_create_mbuf_from_bpfdesc(d, m); 577#endif 578 error = (*ifp->if_output)(ifp, m, &dst, (struct rtentry *)0); 579 mtx_unlock(&Giant); 580 /* 581 * The driver frees the mbuf. 582 */ 583 return (error); 584} 585 586/* 587 * Reset a descriptor by flushing its packet buffer and clearing the 588 * receive and drop counts. 589 */ 590static void 591reset_d(d) 592 struct bpf_d *d; 593{ 594 595 mtx_assert(&d->bd_mtx, MA_OWNED); 596 if (d->bd_hbuf) { 597 /* Free the hold buffer. */ 598 d->bd_fbuf = d->bd_hbuf; 599 d->bd_hbuf = 0; 600 } 601 d->bd_slen = 0; 602 d->bd_hlen = 0; 603 d->bd_rcount = 0; 604 d->bd_dcount = 0; 605} 606 607/* 608 * FIONREAD Check for read packet available. 609 * SIOCGIFADDR Get interface address - convenient hook to driver. 610 * BIOCGBLEN Get buffer len [for read()]. 611 * BIOCSETF Set ethernet read filter. 612 * BIOCFLUSH Flush read packet buffer. 613 * BIOCPROMISC Put interface into promiscuous mode. 614 * BIOCGDLT Get link layer type. 615 * BIOCGETIF Get interface name. 616 * BIOCSETIF Set interface. 617 * BIOCSRTIMEOUT Set read timeout. 618 * BIOCGRTIMEOUT Get read timeout. 619 * BIOCGSTATS Get packet stats. 620 * BIOCIMMEDIATE Set immediate mode. 621 * BIOCVERSION Get filter language version. 622 * BIOCGHDRCMPLT Get "header already complete" flag 623 * BIOCSHDRCMPLT Set "header already complete" flag 624 * BIOCGSEESENT Get "see packets sent" flag 625 * BIOCSSEESENT Set "see packets sent" flag 626 */ 627/* ARGSUSED */ 628static int 629bpfioctl(dev, cmd, addr, flags, td) 630 dev_t dev; 631 u_long cmd; 632 caddr_t addr; 633 int flags; 634 struct thread *td; 635{ 636 struct bpf_d *d = dev->si_drv1; 637 int error = 0; 638 639 BPFD_LOCK(d); 640 if (d->bd_state == BPF_WAITING) 641 callout_stop(&d->bd_callout); 642 d->bd_state = BPF_IDLE; 643 BPFD_UNLOCK(d); 644 645 switch (cmd) { 646 647 default: 648 error = EINVAL; 649 break; 650 651 /* 652 * Check for read packet available. 653 */ 654 case FIONREAD: 655 { 656 int n; 657 658 BPFD_LOCK(d); 659 n = d->bd_slen; 660 if (d->bd_hbuf) 661 n += d->bd_hlen; 662 BPFD_UNLOCK(d); 663 664 *(int *)addr = n; 665 break; 666 } 667 668 case SIOCGIFADDR: 669 { 670 struct ifnet *ifp; 671 672 if (d->bd_bif == 0) 673 error = EINVAL; 674 else { 675 ifp = d->bd_bif->bif_ifp; 676 error = (*ifp->if_ioctl)(ifp, cmd, addr); 677 } 678 break; 679 } 680 681 /* 682 * Get buffer len [for read()]. 683 */ 684 case BIOCGBLEN: 685 *(u_int *)addr = d->bd_bufsize; 686 break; 687 688 /* 689 * Set buffer length. 690 */ 691 case BIOCSBLEN: 692 if (d->bd_bif != 0) 693 error = EINVAL; 694 else { 695 register u_int size = *(u_int *)addr; 696 697 if (size > bpf_maxbufsize) 698 *(u_int *)addr = size = bpf_maxbufsize; 699 else if (size < BPF_MINBUFSIZE) 700 *(u_int *)addr = size = BPF_MINBUFSIZE; 701 d->bd_bufsize = size; 702 } 703 break; 704 705 /* 706 * Set link layer read filter. 707 */ 708 case BIOCSETF: 709 error = bpf_setf(d, (struct bpf_program *)addr); 710 break; 711 712 /* 713 * Flush read packet buffer. 714 */ 715 case BIOCFLUSH: 716 BPFD_LOCK(d); 717 reset_d(d); 718 BPFD_UNLOCK(d); 719 break; 720 721 /* 722 * Put interface into promiscuous mode. 723 */ 724 case BIOCPROMISC: 725 if (d->bd_bif == 0) { 726 /* 727 * No interface attached yet. 728 */ 729 error = EINVAL; 730 break; 731 } 732 if (d->bd_promisc == 0) { 733 mtx_lock(&Giant); 734 error = ifpromisc(d->bd_bif->bif_ifp, 1); 735 mtx_unlock(&Giant); 736 if (error == 0) 737 d->bd_promisc = 1; 738 } 739 break; 740 741 /* 742 * Get device parameters. 743 */ 744 case BIOCGDLT: 745 if (d->bd_bif == 0) 746 error = EINVAL; 747 else 748 *(u_int *)addr = d->bd_bif->bif_dlt; 749 break; 750 751 /* 752 * Get interface name. 753 */ 754 case BIOCGETIF: 755 if (d->bd_bif == 0) 756 error = EINVAL; 757 else { 758 struct ifnet *const ifp = d->bd_bif->bif_ifp; 759 struct ifreq *const ifr = (struct ifreq *)addr; 760 761 snprintf(ifr->ifr_name, sizeof(ifr->ifr_name), 762 "%s%d", ifp->if_name, ifp->if_unit); 763 } 764 break; 765 766 /* 767 * Set interface. 768 */ 769 case BIOCSETIF: 770 error = bpf_setif(d, (struct ifreq *)addr); 771 break; 772 773 /* 774 * Set read timeout. 775 */ 776 case BIOCSRTIMEOUT: 777 { 778 struct timeval *tv = (struct timeval *)addr; 779 780 /* 781 * Subtract 1 tick from tvtohz() since this isn't 782 * a one-shot timer. 783 */ 784 if ((error = itimerfix(tv)) == 0) 785 d->bd_rtout = tvtohz(tv) - 1; 786 break; 787 } 788 789 /* 790 * Get read timeout. 791 */ 792 case BIOCGRTIMEOUT: 793 { 794 struct timeval *tv = (struct timeval *)addr; 795 796 tv->tv_sec = d->bd_rtout / hz; 797 tv->tv_usec = (d->bd_rtout % hz) * tick; 798 break; 799 } 800 801 /* 802 * Get packet stats. 803 */ 804 case BIOCGSTATS: 805 { 806 struct bpf_stat *bs = (struct bpf_stat *)addr; 807 808 bs->bs_recv = d->bd_rcount; 809 bs->bs_drop = d->bd_dcount; 810 break; 811 } 812 813 /* 814 * Set immediate mode. 815 */ 816 case BIOCIMMEDIATE: 817 d->bd_immediate = *(u_int *)addr; 818 break; 819 820 case BIOCVERSION: 821 { 822 struct bpf_version *bv = (struct bpf_version *)addr; 823 824 bv->bv_major = BPF_MAJOR_VERSION; 825 bv->bv_minor = BPF_MINOR_VERSION; 826 break; 827 } 828 829 /* 830 * Get "header already complete" flag 831 */ 832 case BIOCGHDRCMPLT: 833 *(u_int *)addr = d->bd_hdrcmplt; 834 break; 835 836 /* 837 * Set "header already complete" flag 838 */ 839 case BIOCSHDRCMPLT: 840 d->bd_hdrcmplt = *(u_int *)addr ? 1 : 0; 841 break; 842 843 /* 844 * Get "see sent packets" flag 845 */ 846 case BIOCGSEESENT: 847 *(u_int *)addr = d->bd_seesent; 848 break; 849 850 /* 851 * Set "see sent packets" flag 852 */ 853 case BIOCSSEESENT: 854 d->bd_seesent = *(u_int *)addr; 855 break; 856 857 case FIONBIO: /* Non-blocking I/O */ 858 break; 859 860 case FIOASYNC: /* Send signal on receive packets */ 861 d->bd_async = *(int *)addr; 862 break; 863 864 case FIOSETOWN: 865 error = fsetown(*(int *)addr, &d->bd_sigio); 866 break; 867 868 case FIOGETOWN: 869 *(int *)addr = fgetown(d->bd_sigio); 870 break; 871 872 /* This is deprecated, FIOSETOWN should be used instead. */ 873 case TIOCSPGRP: 874 error = fsetown(-(*(int *)addr), &d->bd_sigio); 875 break; 876 877 /* This is deprecated, FIOGETOWN should be used instead. */ 878 case TIOCGPGRP: 879 *(int *)addr = -fgetown(d->bd_sigio); 880 break; 881 882 case BIOCSRSIG: /* Set receive signal */ 883 { 884 u_int sig; 885 886 sig = *(u_int *)addr; 887 888 if (sig >= NSIG) 889 error = EINVAL; 890 else 891 d->bd_sig = sig; 892 break; 893 } 894 case BIOCGRSIG: 895 *(u_int *)addr = d->bd_sig; 896 break; 897 } 898 return (error); 899} 900 901/* 902 * Set d's packet filter program to fp. If this file already has a filter, 903 * free it and replace it. Returns EINVAL for bogus requests. 904 */ 905static int 906bpf_setf(d, fp) 907 struct bpf_d *d; 908 struct bpf_program *fp; 909{ 910 struct bpf_insn *fcode, *old; 911 u_int flen, size; 912 913 old = d->bd_filter; 914 if (fp->bf_insns == 0) { 915 if (fp->bf_len != 0) 916 return (EINVAL); 917 BPFD_LOCK(d); 918 d->bd_filter = 0; 919 reset_d(d); 920 BPFD_UNLOCK(d); 921 if (old != 0) 922 free((caddr_t)old, M_BPF); 923 return (0); 924 } 925 flen = fp->bf_len; 926 if (flen > BPF_MAXINSNS) 927 return (EINVAL); 928 929 size = flen * sizeof(*fp->bf_insns); 930 fcode = (struct bpf_insn *)malloc(size, M_BPF, M_WAITOK); 931 if (copyin((caddr_t)fp->bf_insns, (caddr_t)fcode, size) == 0 && 932 bpf_validate(fcode, (int)flen)) { 933 BPFD_LOCK(d); 934 d->bd_filter = fcode; 935 reset_d(d); 936 BPFD_UNLOCK(d); 937 if (old != 0) 938 free((caddr_t)old, M_BPF); 939 940 return (0); 941 } 942 free((caddr_t)fcode, M_BPF); 943 return (EINVAL); 944} 945 946/* 947 * Detach a file from its current interface (if attached at all) and attach 948 * to the interface indicated by the name stored in ifr. 949 * Return an errno or 0. 950 */ 951static int 952bpf_setif(d, ifr) 953 struct bpf_d *d; 954 struct ifreq *ifr; 955{ 956 struct bpf_if *bp; 957 int error; 958 struct ifnet *theywant; 959 960 theywant = ifunit(ifr->ifr_name); 961 if (theywant == 0) 962 return ENXIO; 963 964 /* 965 * Look through attached interfaces for the named one. 966 */ 967 mtx_lock(&bpf_mtx); 968 for (bp = bpf_iflist; bp != 0; bp = bp->bif_next) { 969 struct ifnet *ifp = bp->bif_ifp; 970 971 if (ifp == 0 || ifp != theywant) 972 continue; 973 974 mtx_unlock(&bpf_mtx); 975 /* 976 * We found the requested interface. 977 * If it's not up, return an error. 978 * Allocate the packet buffers if we need to. 979 * If we're already attached to requested interface, 980 * just flush the buffer. 981 */ 982 if ((ifp->if_flags & IFF_UP) == 0) 983 return (ENETDOWN); 984 985 if (d->bd_sbuf == 0) { 986 error = bpf_allocbufs(d); 987 if (error != 0) 988 return (error); 989 } 990 if (bp != d->bd_bif) { 991 if (d->bd_bif) 992 /* 993 * Detach if attached to something else. 994 */ 995 bpf_detachd(d); 996 997 bpf_attachd(d, bp); 998 } 999 BPFD_LOCK(d); 1000 reset_d(d); 1001 BPFD_UNLOCK(d); 1002 return (0); 1003 } 1004 mtx_unlock(&bpf_mtx); 1005 /* Not found. */ 1006 return (ENXIO); 1007} 1008 1009/* 1010 * Support for select() and poll() system calls 1011 * 1012 * Return true iff the specific operation will not block indefinitely. 1013 * Otherwise, return false but make a note that a selwakeup() must be done. 1014 */ 1015int 1016bpfpoll(dev, events, td) 1017 register dev_t dev; 1018 int events; 1019 struct thread *td; 1020{ 1021 struct bpf_d *d; 1022 int revents; 1023 1024 d = dev->si_drv1; 1025 if (d->bd_bif == NULL) 1026 return (ENXIO); 1027 1028 revents = events & (POLLOUT | POLLWRNORM); 1029 BPFD_LOCK(d); 1030 if (events & (POLLIN | POLLRDNORM)) { 1031 /* 1032 * An imitation of the FIONREAD ioctl code. 1033 * XXX not quite. An exact imitation: 1034 * if (d->b_slen != 0 || 1035 * (d->bd_hbuf != NULL && d->bd_hlen != 0) 1036 */ 1037 if (d->bd_hlen != 0 || 1038 ((d->bd_immediate || d->bd_state == BPF_TIMED_OUT) && 1039 d->bd_slen != 0)) 1040 revents |= events & (POLLIN | POLLRDNORM); 1041 else { 1042 selrecord(td, &d->bd_sel); 1043 /* Start the read timeout if necessary. */ 1044 if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) { 1045 callout_reset(&d->bd_callout, d->bd_rtout, 1046 bpf_timed_out, d); 1047 d->bd_state = BPF_WAITING; 1048 } 1049 } 1050 } 1051 BPFD_UNLOCK(d); 1052 return (revents); 1053} 1054 1055/* 1056 * Incoming linkage from device drivers. Process the packet pkt, of length 1057 * pktlen, which is stored in a contiguous buffer. The packet is parsed 1058 * by each process' filter, and if accepted, stashed into the corresponding 1059 * buffer. 1060 */ 1061void 1062bpf_tap(ifp, pkt, pktlen) 1063 struct ifnet *ifp; 1064 register u_char *pkt; 1065 register u_int pktlen; 1066{ 1067 struct bpf_if *bp; 1068 register struct bpf_d *d; 1069 register u_int slen; 1070 1071 bp = ifp->if_bpf; 1072 BPFIF_LOCK(bp); 1073 for (d = bp->bif_dlist; d != 0; d = d->bd_next) { 1074 BPFD_LOCK(d); 1075 ++d->bd_rcount; 1076 slen = bpf_filter(d->bd_filter, pkt, pktlen, pktlen); 1077 if (slen != 0) { 1078#ifdef MAC 1079 if (mac_check_bpfdesc_receive(d, ifp) == 0) 1080#endif 1081 catchpacket(d, pkt, pktlen, slen, bcopy); 1082 } 1083 BPFD_UNLOCK(d); 1084 } 1085 BPFIF_UNLOCK(bp); 1086} 1087 1088/* 1089 * Copy data from an mbuf chain into a buffer. This code is derived 1090 * from m_copydata in sys/uipc_mbuf.c. 1091 */ 1092static void 1093bpf_mcopy(src_arg, dst_arg, len) 1094 const void *src_arg; 1095 void *dst_arg; 1096 register size_t len; 1097{ 1098 register const struct mbuf *m; 1099 register u_int count; 1100 u_char *dst; 1101 1102 m = src_arg; 1103 dst = dst_arg; 1104 while (len > 0) { 1105 if (m == 0) 1106 panic("bpf_mcopy"); 1107 count = min(m->m_len, len); 1108 bcopy(mtod(m, void *), dst, count); 1109 m = m->m_next; 1110 dst += count; 1111 len -= count; 1112 } 1113} 1114 1115/* 1116 * Incoming linkage from device drivers, when packet is in an mbuf chain. 1117 */ 1118void 1119bpf_mtap(ifp, m) 1120 struct ifnet *ifp; 1121 struct mbuf *m; 1122{ 1123 struct bpf_if *bp = ifp->if_bpf; 1124 struct bpf_d *d; 1125 u_int pktlen, slen; 1126 1127 pktlen = m_length(m, NULL); 1128 if (pktlen == m->m_len) { 1129 bpf_tap(ifp, mtod(m, u_char *), pktlen); 1130 return; 1131 } 1132 1133 BPFIF_LOCK(bp); 1134 for (d = bp->bif_dlist; d != 0; d = d->bd_next) { 1135 if (!d->bd_seesent && (m->m_pkthdr.rcvif == NULL)) 1136 continue; 1137 BPFD_LOCK(d); 1138 ++d->bd_rcount; 1139 slen = bpf_filter(d->bd_filter, (u_char *)m, pktlen, 0); 1140 if (slen != 0) 1141#ifdef MAC 1142 if (mac_check_bpfdesc_receive(d, ifp) == 0) 1143#endif 1144 catchpacket(d, (u_char *)m, pktlen, slen, 1145 bpf_mcopy); 1146 BPFD_UNLOCK(d); 1147 } 1148 BPFIF_UNLOCK(bp); 1149} 1150 1151/* 1152 * Move the packet data from interface memory (pkt) into the 1153 * store buffer. Return 1 if it's time to wakeup a listener (buffer full), 1154 * otherwise 0. "copy" is the routine called to do the actual data 1155 * transfer. bcopy is passed in to copy contiguous chunks, while 1156 * bpf_mcopy is passed in to copy mbuf chains. In the latter case, 1157 * pkt is really an mbuf. 1158 */ 1159static void 1160catchpacket(d, pkt, pktlen, snaplen, cpfn) 1161 register struct bpf_d *d; 1162 register u_char *pkt; 1163 register u_int pktlen, snaplen; 1164 register void (*cpfn)(const void *, void *, size_t); 1165{ 1166 register struct bpf_hdr *hp; 1167 register int totlen, curlen; 1168 register int hdrlen = d->bd_bif->bif_hdrlen; 1169 /* 1170 * Figure out how many bytes to move. If the packet is 1171 * greater or equal to the snapshot length, transfer that 1172 * much. Otherwise, transfer the whole packet (unless 1173 * we hit the buffer size limit). 1174 */ 1175 totlen = hdrlen + min(snaplen, pktlen); 1176 if (totlen > d->bd_bufsize) 1177 totlen = d->bd_bufsize; 1178 1179 /* 1180 * Round up the end of the previous packet to the next longword. 1181 */ 1182 curlen = BPF_WORDALIGN(d->bd_slen); 1183 if (curlen + totlen > d->bd_bufsize) { 1184 /* 1185 * This packet will overflow the storage buffer. 1186 * Rotate the buffers if we can, then wakeup any 1187 * pending reads. 1188 */ 1189 if (d->bd_fbuf == 0) { 1190 /* 1191 * We haven't completed the previous read yet, 1192 * so drop the packet. 1193 */ 1194 ++d->bd_dcount; 1195 return; 1196 } 1197 ROTATE_BUFFERS(d); 1198 bpf_wakeup(d); 1199 curlen = 0; 1200 } 1201 else if (d->bd_immediate || d->bd_state == BPF_TIMED_OUT) 1202 /* 1203 * Immediate mode is set, or the read timeout has 1204 * already expired during a select call. A packet 1205 * arrived, so the reader should be woken up. 1206 */ 1207 bpf_wakeup(d); 1208 1209 /* 1210 * Append the bpf header. 1211 */ 1212 hp = (struct bpf_hdr *)(d->bd_sbuf + curlen); 1213 microtime(&hp->bh_tstamp); 1214 hp->bh_datalen = pktlen; 1215 hp->bh_hdrlen = hdrlen; 1216 /* 1217 * Copy the packet data into the store buffer and update its length. 1218 */ 1219 (*cpfn)(pkt, (u_char *)hp + hdrlen, (hp->bh_caplen = totlen - hdrlen)); 1220 d->bd_slen = curlen + totlen; 1221} 1222 1223/* 1224 * Initialize all nonzero fields of a descriptor. 1225 */ 1226static int 1227bpf_allocbufs(d) 1228 register struct bpf_d *d; 1229{ 1230 d->bd_fbuf = (caddr_t)malloc(d->bd_bufsize, M_BPF, M_WAITOK); 1231 if (d->bd_fbuf == 0) 1232 return (ENOBUFS); 1233 1234 d->bd_sbuf = (caddr_t)malloc(d->bd_bufsize, M_BPF, M_WAITOK); 1235 if (d->bd_sbuf == 0) { 1236 free(d->bd_fbuf, M_BPF); 1237 return (ENOBUFS); 1238 } 1239 d->bd_slen = 0; 1240 d->bd_hlen = 0; 1241 return (0); 1242} 1243 1244/* 1245 * Free buffers currently in use by a descriptor. 1246 * Called on close. 1247 */ 1248static void 1249bpf_freed(d) 1250 register struct bpf_d *d; 1251{ 1252 /* 1253 * We don't need to lock out interrupts since this descriptor has 1254 * been detached from its interface and it yet hasn't been marked 1255 * free. 1256 */ 1257 if (d->bd_sbuf != 0) { 1258 free(d->bd_sbuf, M_BPF); 1259 if (d->bd_hbuf != 0) 1260 free(d->bd_hbuf, M_BPF); 1261 if (d->bd_fbuf != 0) 1262 free(d->bd_fbuf, M_BPF); 1263 } 1264 if (d->bd_filter) 1265 free((caddr_t)d->bd_filter, M_BPF); 1266 mtx_destroy(&d->bd_mtx); 1267} 1268 1269/* 1270 * Attach an interface to bpf. ifp is a pointer to the structure 1271 * defining the interface to be attached, dlt is the link layer type, 1272 * and hdrlen is the fixed size of the link header (variable length 1273 * headers are not yet supporrted). 1274 */ 1275void 1276bpfattach(ifp, dlt, hdrlen) 1277 struct ifnet *ifp; 1278 u_int dlt, hdrlen; 1279{ 1280 struct bpf_if *bp; 1281 bp = (struct bpf_if *)malloc(sizeof(*bp), M_BPF, M_NOWAIT | M_ZERO); 1282 if (bp == 0) 1283 panic("bpfattach"); 1284 1285 bp->bif_ifp = ifp; 1286 bp->bif_dlt = dlt; 1287 mtx_init(&bp->bif_mtx, "bpf interface lock", NULL, MTX_DEF); 1288 1289 mtx_lock(&bpf_mtx); 1290 bp->bif_next = bpf_iflist; 1291 bpf_iflist = bp; 1292 mtx_unlock(&bpf_mtx); 1293 1294 bp->bif_ifp->if_bpf = 0; 1295 1296 /* 1297 * Compute the length of the bpf header. This is not necessarily 1298 * equal to SIZEOF_BPF_HDR because we want to insert spacing such 1299 * that the network layer header begins on a longword boundary (for 1300 * performance reasons and to alleviate alignment restrictions). 1301 */ 1302 bp->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen; 1303 1304 if (bootverbose) 1305 printf("bpf: %s%d attached\n", ifp->if_name, ifp->if_unit); 1306} 1307 1308/* 1309 * Detach bpf from an interface. This involves detaching each descriptor 1310 * associated with the interface, and leaving bd_bif NULL. Notify each 1311 * descriptor as it's detached so that any sleepers wake up and get 1312 * ENXIO. 1313 */ 1314void 1315bpfdetach(ifp) 1316 struct ifnet *ifp; 1317{ 1318 struct bpf_if *bp, *bp_prev; 1319 struct bpf_d *d; 1320 1321 mtx_lock(&bpf_mtx); 1322 1323 /* Locate BPF interface information */ 1324 bp_prev = NULL; 1325 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) { 1326 if (ifp == bp->bif_ifp) 1327 break; 1328 bp_prev = bp; 1329 } 1330 1331 /* Interface wasn't attached */ 1332 if (bp->bif_ifp == NULL) { 1333 mtx_unlock(&bpf_mtx); 1334 printf("bpfdetach: %s%d was not attached\n", ifp->if_name, 1335 ifp->if_unit); 1336 return; 1337 } 1338 1339 if (bp_prev) { 1340 bp_prev->bif_next = bp->bif_next; 1341 } else { 1342 bpf_iflist = bp->bif_next; 1343 } 1344 1345 while ((d = bp->bif_dlist) != NULL) { 1346 bpf_detachd(d); 1347 BPFD_LOCK(d); 1348 bpf_wakeup(d); 1349 BPFD_UNLOCK(d); 1350 } 1351 1352 mtx_destroy(&bp->bif_mtx); 1353 free(bp, M_BPF); 1354 1355 mtx_unlock(&bpf_mtx); 1356} 1357 1358static void bpf_drvinit(void *unused); 1359 1360static void bpf_clone(void *arg, char *name, int namelen, dev_t *dev); 1361 1362static void 1363bpf_clone(arg, name, namelen, dev) 1364 void *arg; 1365 char *name; 1366 int namelen; 1367 dev_t *dev; 1368{ 1369 int u; 1370 1371 if (*dev != NODEV) 1372 return; 1373 if (dev_stdclone(name, NULL, "bpf", &u) != 1) 1374 return; 1375 *dev = make_dev(&bpf_cdevsw, unit2minor(u), UID_ROOT, GID_WHEEL, 0600, 1376 "bpf%d", u); 1377 (*dev)->si_flags |= SI_CHEAPCLONE; 1378 return; 1379} 1380 1381static void 1382bpf_drvinit(unused) 1383 void *unused; 1384{ 1385 1386 mtx_init(&bpf_mtx, "bpf global lock", NULL, MTX_DEF); 1387 EVENTHANDLER_REGISTER(dev_clone, bpf_clone, 0, 1000); 1388 cdevsw_add(&bpf_cdevsw); 1389} 1390 1391SYSINIT(bpfdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,bpf_drvinit,NULL) 1392 1393#else /* !DEV_BPF && !NETGRAPH_BPF */ 1394/* 1395 * NOP stubs to allow bpf-using drivers to load and function. 1396 * 1397 * A 'better' implementation would allow the core bpf functionality 1398 * to be loaded at runtime. 1399 */ 1400 1401void 1402bpf_tap(ifp, pkt, pktlen) 1403 struct ifnet *ifp; 1404 register u_char *pkt; 1405 register u_int pktlen; 1406{ 1407} 1408 1409void 1410bpf_mtap(ifp, m) 1411 struct ifnet *ifp; 1412 struct mbuf *m; 1413{ 1414} 1415 1416void 1417bpfattach(ifp, dlt, hdrlen) 1418 struct ifnet *ifp; 1419 u_int dlt, hdrlen; 1420{ 1421} 1422 1423void 1424bpfdetach(ifp) 1425 struct ifnet *ifp; 1426{ 1427} 1428 1429u_int 1430bpf_filter(pc, p, wirelen, buflen) 1431 register const struct bpf_insn *pc; 1432 register u_char *p; 1433 u_int wirelen; 1434 register u_int buflen; 1435{ 1436 return -1; /* "no filter" behaviour */ 1437} 1438 1439int 1440bpf_validate(f, len) 1441 const struct bpf_insn *f; 1442 int len; 1443{ 1444 return 0; /* false */ 1445} 1446 1447#endif /* !DEV_BPF && !NETGRAPH_BPF */ 1448