1 /* 2 * Copyright (c) 1990, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from the Stanford/CMU enet packet filter, 6 * (net/enet.c) distributed as part of 4.3BSD, and code contributed 7 * to Berkeley by Steven McCanne and Van Jacobson both of Lawrence 8 * Berkeley Laboratory. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)bpf.c 8.2 (Berkeley) 3/28/94 39 * 40 * $FreeBSD: src/sys/net/bpf.c,v 1.59.2.12 2002/04/14 21:41:48 luigi Exp $ 41 * $DragonFly: src/sys/net/bpf.c,v 1.45 2008/05/14 11:59:23 sephe Exp $ 42 */ 43 44 #include "use_bpf.h" 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/conf.h> 49 #include <sys/device.h> 50 #include <sys/malloc.h> 51 #include <sys/mbuf.h> 52 #include <sys/time.h> 53 #include <sys/proc.h> 54 #include <sys/signalvar.h> 55 #include <sys/filio.h> 56 #include <sys/sockio.h> 57 #include <sys/ttycom.h> 58 #include <sys/filedesc.h> 59 60 #include <sys/poll.h> 61 62 #include <sys/socket.h> 63 #include <sys/vnode.h> 64 65 #include <sys/thread2.h> 66 67 #include <net/if.h> 68 #include <net/bpf.h> 69 #include <net/bpfdesc.h> 70 71 #include <netinet/in.h> 72 #include <netinet/if_ether.h> 73 #include <sys/kernel.h> 74 #include <sys/sysctl.h> 75 76 MALLOC_DEFINE(M_BPF, "BPF", "BPF data"); 77 78 #if NBPF > 0 79 80 /* 81 * The default read buffer size is patchable. 82 */ 83 static int bpf_bufsize = BPF_DEFAULTBUFSIZE; 84 SYSCTL_INT(_debug, OID_AUTO, bpf_bufsize, CTLFLAG_RW, 85 &bpf_bufsize, 0, ""); 86 int bpf_maxbufsize = BPF_MAXBUFSIZE; 87 SYSCTL_INT(_debug, OID_AUTO, bpf_maxbufsize, CTLFLAG_RW, 88 &bpf_maxbufsize, 0, ""); 89 90 /* 91 * bpf_iflist is the list of interfaces; each corresponds to an ifnet 92 */ 93 static struct bpf_if *bpf_iflist; 94 95 static int bpf_allocbufs(struct bpf_d *); 96 static void bpf_attachd(struct bpf_d *d, struct bpf_if *bp); 97 static void bpf_detachd(struct bpf_d *d); 98 static void bpf_resetd(struct bpf_d *); 99 static void bpf_freed(struct bpf_d *); 100 static void bpf_mcopy(const void *, void *, size_t); 101 static int bpf_movein(struct uio *, int, struct mbuf **, 102 struct sockaddr *, int *, struct bpf_insn *); 103 static int bpf_setif(struct bpf_d *, struct ifreq *); 104 static void bpf_timed_out(void *); 105 static void bpf_wakeup(struct bpf_d *); 106 static void catchpacket(struct bpf_d *, u_char *, u_int, u_int, 107 void (*)(const void *, void *, size_t), 108 const struct timeval *); 109 static int bpf_setf(struct bpf_d *, struct bpf_program *, u_long cmd); 110 static int bpf_getdltlist(struct bpf_d *, struct bpf_dltlist *); 111 static int bpf_setdlt(struct bpf_d *, u_int); 112 static void bpf_drvinit(void *unused); 113 114 static d_open_t bpfopen; 115 static d_close_t bpfclose; 116 static d_read_t bpfread; 117 static d_write_t bpfwrite; 118 static d_ioctl_t bpfioctl; 119 static d_poll_t bpfpoll; 120 121 #define CDEV_MAJOR 23 122 static struct dev_ops bpf_ops = { 123 { "bpf", CDEV_MAJOR, 0 }, 124 .d_open = bpfopen, 125 .d_close = bpfclose, 126 .d_read = bpfread, 127 .d_write = bpfwrite, 128 .d_ioctl = bpfioctl, 129 .d_poll = bpfpoll, 130 }; 131 132 133 static int 134 bpf_movein(struct uio *uio, int linktype, struct mbuf **mp, 135 struct sockaddr *sockp, int *datlen, struct bpf_insn *wfilter) 136 { 137 struct mbuf *m; 138 int error; 139 int len; 140 int hlen; 141 int slen; 142 143 *datlen = 0; 144 *mp = NULL; 145 146 /* 147 * Build a sockaddr based on the data link layer type. 148 * We do this at this level because the ethernet header 149 * is copied directly into the data field of the sockaddr. 150 * In the case of SLIP, there is no header and the packet 151 * is forwarded as is. 152 * Also, we are careful to leave room at the front of the mbuf 153 * for the link level header. 154 */ 155 switch (linktype) { 156 case DLT_SLIP: 157 sockp->sa_family = AF_INET; 158 hlen = 0; 159 break; 160 161 case DLT_EN10MB: 162 sockp->sa_family = AF_UNSPEC; 163 /* XXX Would MAXLINKHDR be better? */ 164 hlen = sizeof(struct ether_header); 165 break; 166 167 case DLT_RAW: 168 case DLT_NULL: 169 sockp->sa_family = AF_UNSPEC; 170 hlen = 0; 171 break; 172 173 case DLT_ATM_RFC1483: 174 /* 175 * en atm driver requires 4-byte atm pseudo header. 176 * though it isn't standard, vpi:vci needs to be 177 * specified anyway. 178 */ 179 sockp->sa_family = AF_UNSPEC; 180 hlen = 12; /* XXX 4(ATM_PH) + 3(LLC) + 5(SNAP) */ 181 break; 182 183 case DLT_PPP: 184 sockp->sa_family = AF_UNSPEC; 185 hlen = 4; /* This should match PPP_HDRLEN */ 186 break; 187 188 default: 189 return(EIO); 190 } 191 192 len = uio->uio_resid; 193 *datlen = len - hlen; 194 if ((unsigned)len > MCLBYTES) 195 return(EIO); 196 197 m = m_getl(len, MB_WAIT, MT_DATA, M_PKTHDR, NULL); 198 if (m == NULL) 199 return(ENOBUFS); 200 m->m_pkthdr.len = m->m_len = len; 201 m->m_pkthdr.rcvif = NULL; 202 *mp = m; 203 204 if (m->m_len < hlen) { 205 error = EPERM; 206 goto bad; 207 } 208 209 error = uiomove(mtod(m, u_char *), len, uio); 210 if (error) 211 goto bad; 212 213 slen = bpf_filter(wfilter, mtod(m, u_char *), len, len); 214 if (slen == 0) { 215 error = EPERM; 216 goto bad; 217 } 218 219 /* 220 * Make room for link header, and copy it to sockaddr. 221 */ 222 if (hlen != 0) { 223 bcopy(m->m_data, sockp->sa_data, hlen); 224 m->m_pkthdr.len -= hlen; 225 m->m_len -= hlen; 226 m->m_data += hlen; /* XXX */ 227 } 228 return (0); 229 bad: 230 m_freem(m); 231 return(error); 232 } 233 234 /* 235 * Attach file to the bpf interface, i.e. make d listen on bp. 236 * Must be called at splimp. 237 */ 238 static void 239 bpf_attachd(struct bpf_d *d, struct bpf_if *bp) 240 { 241 /* 242 * Point d at bp, and add d to the interface's list of listeners. 243 * Finally, point the driver's bpf cookie at the interface so 244 * it will divert packets to bpf. 245 */ 246 d->bd_bif = bp; 247 SLIST_INSERT_HEAD(&bp->bif_dlist, d, bd_next); 248 *bp->bif_driverp = bp; 249 } 250 251 /* 252 * Detach a file from its interface. 253 */ 254 static void 255 bpf_detachd(struct bpf_d *d) 256 { 257 int error; 258 struct bpf_if *bp; 259 struct ifnet *ifp; 260 261 bp = d->bd_bif; 262 ifp = bp->bif_ifp; 263 264 /* Remove d from the interface's descriptor list. */ 265 SLIST_REMOVE(&bp->bif_dlist, d, bpf_d, bd_next); 266 267 if (SLIST_EMPTY(&bp->bif_dlist)) { 268 /* 269 * Let the driver know that there are no more listeners. 270 */ 271 *bp->bif_driverp = NULL; 272 } 273 d->bd_bif = NULL; 274 /* 275 * Check if this descriptor had requested promiscuous mode. 276 * If so, turn it off. 277 */ 278 if (d->bd_promisc) { 279 d->bd_promisc = 0; 280 error = ifpromisc(ifp, 0); 281 if (error != 0 && error != ENXIO) { 282 /* 283 * ENXIO can happen if a pccard is unplugged, 284 * Something is really wrong if we were able to put 285 * the driver into promiscuous mode, but can't 286 * take it out. 287 */ 288 if_printf(ifp, "bpf_detach: ifpromisc failed(%d)\n", 289 error); 290 } 291 } 292 } 293 294 /* 295 * Open ethernet device. Returns ENXIO for illegal minor device number, 296 * EBUSY if file is open by another process. 297 */ 298 /* ARGSUSED */ 299 static int 300 bpfopen(struct dev_open_args *ap) 301 { 302 cdev_t dev = ap->a_head.a_dev; 303 struct bpf_d *d; 304 305 if (ap->a_cred->cr_prison) 306 return(EPERM); 307 308 d = dev->si_drv1; 309 /* 310 * Each minor can be opened by only one process. If the requested 311 * minor is in use, return EBUSY. 312 */ 313 if (d != NULL) 314 return(EBUSY); 315 make_dev(&bpf_ops, minor(dev), 0, 0, 0600, "bpf%d", lminor(dev)); 316 MALLOC(d, struct bpf_d *, sizeof *d, M_BPF, M_WAITOK | M_ZERO); 317 dev->si_drv1 = d; 318 d->bd_bufsize = bpf_bufsize; 319 d->bd_sig = SIGIO; 320 d->bd_seesent = 1; 321 callout_init(&d->bd_callout); 322 return(0); 323 } 324 325 /* 326 * Close the descriptor by detaching it from its interface, 327 * deallocating its buffers, and marking it free. 328 */ 329 /* ARGSUSED */ 330 static int 331 bpfclose(struct dev_close_args *ap) 332 { 333 cdev_t dev = ap->a_head.a_dev; 334 struct bpf_d *d = dev->si_drv1; 335 336 funsetown(d->bd_sigio); 337 crit_enter(); 338 if (d->bd_state == BPF_WAITING) 339 callout_stop(&d->bd_callout); 340 d->bd_state = BPF_IDLE; 341 if (d->bd_bif != NULL) 342 bpf_detachd(d); 343 crit_exit(); 344 bpf_freed(d); 345 dev->si_drv1 = NULL; 346 kfree(d, M_BPF); 347 348 return(0); 349 } 350 351 /* 352 * Rotate the packet buffers in descriptor d. Move the store buffer 353 * into the hold slot, and the free buffer into the store slot. 354 * Zero the length of the new store buffer. 355 */ 356 #define ROTATE_BUFFERS(d) \ 357 (d)->bd_hbuf = (d)->bd_sbuf; \ 358 (d)->bd_hlen = (d)->bd_slen; \ 359 (d)->bd_sbuf = (d)->bd_fbuf; \ 360 (d)->bd_slen = 0; \ 361 (d)->bd_fbuf = NULL; 362 /* 363 * bpfread - read next chunk of packets from buffers 364 */ 365 static int 366 bpfread(struct dev_read_args *ap) 367 { 368 cdev_t dev = ap->a_head.a_dev; 369 struct bpf_d *d = dev->si_drv1; 370 int timed_out; 371 int error; 372 373 /* 374 * Restrict application to use a buffer the same size as 375 * as kernel buffers. 376 */ 377 if (ap->a_uio->uio_resid != d->bd_bufsize) 378 return(EINVAL); 379 380 crit_enter(); 381 if (d->bd_state == BPF_WAITING) 382 callout_stop(&d->bd_callout); 383 timed_out = (d->bd_state == BPF_TIMED_OUT); 384 d->bd_state = BPF_IDLE; 385 /* 386 * If the hold buffer is empty, then do a timed sleep, which 387 * ends when the timeout expires or when enough packets 388 * have arrived to fill the store buffer. 389 */ 390 while (d->bd_hbuf == NULL) { 391 if ((d->bd_immediate || timed_out) && d->bd_slen != 0) { 392 /* 393 * A packet(s) either arrived since the previous 394 * read or arrived while we were asleep. 395 * Rotate the buffers and return what's here. 396 */ 397 ROTATE_BUFFERS(d); 398 break; 399 } 400 401 /* 402 * No data is available, check to see if the bpf device 403 * is still pointed at a real interface. If not, return 404 * ENXIO so that the userland process knows to rebind 405 * it before using it again. 406 */ 407 if (d->bd_bif == NULL) { 408 crit_exit(); 409 return(ENXIO); 410 } 411 412 if (ap->a_ioflag & IO_NDELAY) { 413 crit_exit(); 414 return(EWOULDBLOCK); 415 } 416 error = tsleep(d, PCATCH, "bpf", d->bd_rtout); 417 if (error == EINTR || error == ERESTART) { 418 crit_exit(); 419 return(error); 420 } 421 if (error == EWOULDBLOCK) { 422 /* 423 * On a timeout, return what's in the buffer, 424 * which may be nothing. If there is something 425 * in the store buffer, we can rotate the buffers. 426 */ 427 if (d->bd_hbuf) 428 /* 429 * We filled up the buffer in between 430 * getting the timeout and arriving 431 * here, so we don't need to rotate. 432 */ 433 break; 434 435 if (d->bd_slen == 0) { 436 crit_exit(); 437 return(0); 438 } 439 ROTATE_BUFFERS(d); 440 break; 441 } 442 } 443 /* 444 * At this point, we know we have something in the hold slot. 445 */ 446 crit_exit(); 447 448 /* 449 * Move data from hold buffer into user space. 450 * We know the entire buffer is transferred since 451 * we checked above that the read buffer is bpf_bufsize bytes. 452 */ 453 error = uiomove(d->bd_hbuf, d->bd_hlen, ap->a_uio); 454 455 crit_enter(); 456 d->bd_fbuf = d->bd_hbuf; 457 d->bd_hbuf = NULL; 458 d->bd_hlen = 0; 459 crit_exit(); 460 461 return(error); 462 } 463 464 465 /* 466 * If there are processes sleeping on this descriptor, wake them up. 467 */ 468 static void 469 bpf_wakeup(struct bpf_d *d) 470 { 471 if (d->bd_state == BPF_WAITING) { 472 callout_stop(&d->bd_callout); 473 d->bd_state = BPF_IDLE; 474 } 475 wakeup(d); 476 if (d->bd_async && d->bd_sig && d->bd_sigio) 477 pgsigio(d->bd_sigio, d->bd_sig, 0); 478 479 get_mplock(); 480 selwakeup(&d->bd_sel); 481 rel_mplock(); 482 /* XXX */ 483 d->bd_sel.si_pid = 0; 484 } 485 486 static void 487 bpf_timed_out(void *arg) 488 { 489 struct bpf_d *d = (struct bpf_d *)arg; 490 491 crit_enter(); 492 if (d->bd_state == BPF_WAITING) { 493 d->bd_state = BPF_TIMED_OUT; 494 if (d->bd_slen != 0) 495 bpf_wakeup(d); 496 } 497 crit_exit(); 498 } 499 500 static int 501 bpfwrite(struct dev_write_args *ap) 502 { 503 cdev_t dev = ap->a_head.a_dev; 504 struct bpf_d *d = dev->si_drv1; 505 struct ifnet *ifp; 506 struct mbuf *m; 507 int error; 508 static struct sockaddr dst; 509 int datlen; 510 511 if (d->bd_bif == NULL) 512 return(ENXIO); 513 514 ifp = d->bd_bif->bif_ifp; 515 516 if (ap->a_uio->uio_resid == 0) 517 return(0); 518 519 error = bpf_movein(ap->a_uio, (int)d->bd_bif->bif_dlt, &m, 520 &dst, &datlen, d->bd_wfilter); 521 if (error) 522 return(error); 523 524 if (datlen > ifp->if_mtu) { 525 m_freem(m); 526 return(EMSGSIZE); 527 } 528 529 if (d->bd_hdrcmplt) 530 dst.sa_family = pseudo_AF_HDRCMPLT; 531 532 crit_enter(); 533 error = ifp->if_output(ifp, m, &dst, NULL); 534 crit_exit(); 535 /* 536 * The driver frees the mbuf. 537 */ 538 return(error); 539 } 540 541 /* 542 * Reset a descriptor by flushing its packet buffer and clearing the 543 * receive and drop counts. Should be called at splimp. 544 */ 545 static void 546 bpf_resetd(struct bpf_d *d) 547 { 548 if (d->bd_hbuf) { 549 /* Free the hold buffer. */ 550 d->bd_fbuf = d->bd_hbuf; 551 d->bd_hbuf = NULL; 552 } 553 d->bd_slen = 0; 554 d->bd_hlen = 0; 555 d->bd_rcount = 0; 556 d->bd_dcount = 0; 557 } 558 559 /* 560 * FIONREAD Check for read packet available. 561 * SIOCGIFADDR Get interface address - convenient hook to driver. 562 * BIOCGBLEN Get buffer len [for read()]. 563 * BIOCSETF Set ethernet read filter. 564 * BIOCSETWF Set ethernet write filter. 565 * BIOCFLUSH Flush read packet buffer. 566 * BIOCPROMISC Put interface into promiscuous mode. 567 * BIOCGDLT Get link layer type. 568 * BIOCGETIF Get interface name. 569 * BIOCSETIF Set interface. 570 * BIOCSRTIMEOUT Set read timeout. 571 * BIOCGRTIMEOUT Get read timeout. 572 * BIOCGSTATS Get packet stats. 573 * BIOCIMMEDIATE Set immediate mode. 574 * BIOCVERSION Get filter language version. 575 * BIOCGHDRCMPLT Get "header already complete" flag 576 * BIOCSHDRCMPLT Set "header already complete" flag 577 * BIOCGSEESENT Get "see packets sent" flag 578 * BIOCSSEESENT Set "see packets sent" flag 579 * BIOCLOCK Set "locked" flag 580 */ 581 /* ARGSUSED */ 582 static int 583 bpfioctl(struct dev_ioctl_args *ap) 584 { 585 cdev_t dev = ap->a_head.a_dev; 586 struct bpf_d *d = dev->si_drv1; 587 int error = 0; 588 589 crit_enter(); 590 if (d->bd_state == BPF_WAITING) 591 callout_stop(&d->bd_callout); 592 d->bd_state = BPF_IDLE; 593 crit_exit(); 594 595 if (d->bd_locked == 1) { 596 switch (ap->a_cmd) { 597 case BIOCGBLEN: 598 case BIOCFLUSH: 599 case BIOCGDLT: 600 case BIOCGDLTLIST: 601 case BIOCGETIF: 602 case BIOCGRTIMEOUT: 603 case BIOCGSTATS: 604 case BIOCVERSION: 605 case BIOCGRSIG: 606 case BIOCGHDRCMPLT: 607 case FIONREAD: 608 case BIOCLOCK: 609 case BIOCSRTIMEOUT: 610 case BIOCIMMEDIATE: 611 case TIOCGPGRP: 612 break; 613 default: 614 return (EPERM); 615 } 616 } 617 switch (ap->a_cmd) { 618 default: 619 error = EINVAL; 620 break; 621 622 /* 623 * Check for read packet available. 624 */ 625 case FIONREAD: 626 { 627 int n; 628 629 crit_enter(); 630 n = d->bd_slen; 631 if (d->bd_hbuf) 632 n += d->bd_hlen; 633 crit_exit(); 634 635 *(int *)ap->a_data = n; 636 break; 637 } 638 639 case SIOCGIFADDR: 640 { 641 struct ifnet *ifp; 642 643 if (d->bd_bif == NULL) { 644 error = EINVAL; 645 } else { 646 ifp = d->bd_bif->bif_ifp; 647 lwkt_serialize_enter(ifp->if_serializer); 648 error = ifp->if_ioctl(ifp, ap->a_cmd, 649 ap->a_data, ap->a_cred); 650 lwkt_serialize_exit(ifp->if_serializer); 651 } 652 break; 653 } 654 655 /* 656 * Get buffer len [for read()]. 657 */ 658 case BIOCGBLEN: 659 *(u_int *)ap->a_data = d->bd_bufsize; 660 break; 661 662 /* 663 * Set buffer length. 664 */ 665 case BIOCSBLEN: 666 if (d->bd_bif != NULL) { 667 error = EINVAL; 668 } else { 669 u_int size = *(u_int *)ap->a_data; 670 671 if (size > bpf_maxbufsize) 672 *(u_int *)ap->a_data = size = bpf_maxbufsize; 673 else if (size < BPF_MINBUFSIZE) 674 *(u_int *)ap->a_data = size = BPF_MINBUFSIZE; 675 d->bd_bufsize = size; 676 } 677 break; 678 679 /* 680 * Set link layer read filter. 681 */ 682 case BIOCSETF: 683 case BIOCSETWF: 684 error = bpf_setf(d, (struct bpf_program *)ap->a_data, 685 ap->a_cmd); 686 break; 687 688 /* 689 * Flush read packet buffer. 690 */ 691 case BIOCFLUSH: 692 crit_enter(); 693 bpf_resetd(d); 694 crit_exit(); 695 break; 696 697 /* 698 * Put interface into promiscuous mode. 699 */ 700 case BIOCPROMISC: 701 if (d->bd_bif == NULL) { 702 /* 703 * No interface attached yet. 704 */ 705 error = EINVAL; 706 break; 707 } 708 crit_enter(); 709 if (d->bd_promisc == 0) { 710 error = ifpromisc(d->bd_bif->bif_ifp, 1); 711 if (error == 0) 712 d->bd_promisc = 1; 713 } 714 crit_exit(); 715 break; 716 717 /* 718 * Get device parameters. 719 */ 720 case BIOCGDLT: 721 if (d->bd_bif == NULL) 722 error = EINVAL; 723 else 724 *(u_int *)ap->a_data = d->bd_bif->bif_dlt; 725 break; 726 727 /* 728 * Get a list of supported data link types. 729 */ 730 case BIOCGDLTLIST: 731 if (d->bd_bif == NULL) { 732 error = EINVAL; 733 } else { 734 error = bpf_getdltlist(d, 735 (struct bpf_dltlist *)ap->a_data); 736 } 737 break; 738 739 /* 740 * Set data link type. 741 */ 742 case BIOCSDLT: 743 if (d->bd_bif == NULL) 744 error = EINVAL; 745 else 746 error = bpf_setdlt(d, *(u_int *)ap->a_data); 747 break; 748 749 /* 750 * Get interface name. 751 */ 752 case BIOCGETIF: 753 if (d->bd_bif == NULL) { 754 error = EINVAL; 755 } else { 756 struct ifnet *const ifp = d->bd_bif->bif_ifp; 757 struct ifreq *const ifr = (struct ifreq *)ap->a_data; 758 759 strlcpy(ifr->ifr_name, ifp->if_xname, 760 sizeof ifr->ifr_name); 761 } 762 break; 763 764 /* 765 * Set interface. 766 */ 767 case BIOCSETIF: 768 error = bpf_setif(d, (struct ifreq *)ap->a_data); 769 break; 770 771 /* 772 * Set read timeout. 773 */ 774 case BIOCSRTIMEOUT: 775 { 776 struct timeval *tv = (struct timeval *)ap->a_data; 777 778 /* 779 * Subtract 1 tick from tvtohz() since this isn't 780 * a one-shot timer. 781 */ 782 if ((error = itimerfix(tv)) == 0) 783 d->bd_rtout = tvtohz_low(tv); 784 break; 785 } 786 787 /* 788 * Get read timeout. 789 */ 790 case BIOCGRTIMEOUT: 791 { 792 struct timeval *tv = (struct timeval *)ap->a_data; 793 794 tv->tv_sec = d->bd_rtout / hz; 795 tv->tv_usec = (d->bd_rtout % hz) * tick; 796 break; 797 } 798 799 /* 800 * Get packet stats. 801 */ 802 case BIOCGSTATS: 803 { 804 struct bpf_stat *bs = (struct bpf_stat *)ap->a_data; 805 806 bs->bs_recv = d->bd_rcount; 807 bs->bs_drop = d->bd_dcount; 808 break; 809 } 810 811 /* 812 * Set immediate mode. 813 */ 814 case BIOCIMMEDIATE: 815 d->bd_immediate = *(u_int *)ap->a_data; 816 break; 817 818 case BIOCVERSION: 819 { 820 struct bpf_version *bv = (struct bpf_version *)ap->a_data; 821 822 bv->bv_major = BPF_MAJOR_VERSION; 823 bv->bv_minor = BPF_MINOR_VERSION; 824 break; 825 } 826 827 /* 828 * Get "header already complete" flag 829 */ 830 case BIOCGHDRCMPLT: 831 *(u_int *)ap->a_data = d->bd_hdrcmplt; 832 break; 833 834 /* 835 * Set "header already complete" flag 836 */ 837 case BIOCSHDRCMPLT: 838 d->bd_hdrcmplt = *(u_int *)ap->a_data ? 1 : 0; 839 break; 840 841 /* 842 * Get "see sent packets" flag 843 */ 844 case BIOCGSEESENT: 845 *(u_int *)ap->a_data = d->bd_seesent; 846 break; 847 848 /* 849 * Set "see sent packets" flag 850 */ 851 case BIOCSSEESENT: 852 d->bd_seesent = *(u_int *)ap->a_data; 853 break; 854 855 case FIOASYNC: /* Send signal on receive packets */ 856 d->bd_async = *(int *)ap->a_data; 857 break; 858 859 case FIOSETOWN: 860 error = fsetown(*(int *)ap->a_data, &d->bd_sigio); 861 break; 862 863 case FIOGETOWN: 864 *(int *)ap->a_data = fgetown(d->bd_sigio); 865 break; 866 867 /* This is deprecated, FIOSETOWN should be used instead. */ 868 case TIOCSPGRP: 869 error = fsetown(-(*(int *)ap->a_data), &d->bd_sigio); 870 break; 871 872 /* This is deprecated, FIOGETOWN should be used instead. */ 873 case TIOCGPGRP: 874 *(int *)ap->a_data = -fgetown(d->bd_sigio); 875 break; 876 877 case BIOCSRSIG: /* Set receive signal */ 878 { 879 u_int sig; 880 881 sig = *(u_int *)ap->a_data; 882 883 if (sig >= NSIG) 884 error = EINVAL; 885 else 886 d->bd_sig = sig; 887 break; 888 } 889 case BIOCGRSIG: 890 *(u_int *)ap->a_data = d->bd_sig; 891 break; 892 case BIOCLOCK: 893 d->bd_locked = 1; 894 break; 895 } 896 return(error); 897 } 898 899 /* 900 * Set d's packet filter program to fp. If this file already has a filter, 901 * free it and replace it. Returns EINVAL for bogus requests. 902 */ 903 static int 904 bpf_setf(struct bpf_d *d, struct bpf_program *fp, u_long cmd) 905 { 906 struct bpf_insn *fcode, *old; 907 u_int wfilter, flen, size; 908 909 if (cmd == BIOCSETWF) { 910 old = d->bd_wfilter; 911 wfilter = 1; 912 } else { 913 wfilter = 0; 914 old = d->bd_rfilter; 915 } 916 if (fp->bf_insns == NULL) { 917 if (fp->bf_len != 0) 918 return(EINVAL); 919 crit_enter(); 920 if (wfilter) 921 d->bd_wfilter = NULL; 922 else 923 d->bd_rfilter = NULL; 924 bpf_resetd(d); 925 crit_exit(); 926 if (old != NULL) 927 kfree(old, M_BPF); 928 return(0); 929 } 930 flen = fp->bf_len; 931 if (flen > BPF_MAXINSNS) 932 return(EINVAL); 933 934 size = flen * sizeof *fp->bf_insns; 935 fcode = (struct bpf_insn *)kmalloc(size, M_BPF, M_WAITOK); 936 if (copyin(fp->bf_insns, fcode, size) == 0 && 937 bpf_validate(fcode, (int)flen)) { 938 crit_enter(); 939 if (wfilter) 940 d->bd_wfilter = fcode; 941 else 942 d->bd_rfilter = fcode; 943 bpf_resetd(d); 944 crit_exit(); 945 if (old != NULL) 946 kfree(old, M_BPF); 947 948 return(0); 949 } 950 kfree(fcode, M_BPF); 951 return(EINVAL); 952 } 953 954 /* 955 * Detach a file from its current interface (if attached at all) and attach 956 * to the interface indicated by the name stored in ifr. 957 * Return an errno or 0. 958 */ 959 static int 960 bpf_setif(struct bpf_d *d, struct ifreq *ifr) 961 { 962 struct bpf_if *bp; 963 int error; 964 struct ifnet *theywant; 965 966 theywant = ifunit(ifr->ifr_name); 967 if (theywant == NULL) 968 return(ENXIO); 969 970 /* 971 * Look through attached interfaces for the named one. 972 */ 973 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) { 974 struct ifnet *ifp = bp->bif_ifp; 975 976 if (ifp == NULL || ifp != theywant) 977 continue; 978 /* skip additional entry */ 979 if (bp->bif_driverp != &ifp->if_bpf) 980 continue; 981 /* 982 * We found the requested interface. 983 * If it's not up, return an error. 984 * Allocate the packet buffers if we need to. 985 * If we're already attached to requested interface, 986 * just flush the buffer. 987 */ 988 if (!(ifp->if_flags & IFF_UP)) 989 return(ENETDOWN); 990 991 if (d->bd_sbuf == NULL) { 992 error = bpf_allocbufs(d); 993 if (error != 0) 994 return(error); 995 } 996 crit_enter(); 997 if (bp != d->bd_bif) { 998 if (d->bd_bif != NULL) { 999 /* 1000 * Detach if attached to something else. 1001 */ 1002 bpf_detachd(d); 1003 } 1004 1005 bpf_attachd(d, bp); 1006 } 1007 bpf_resetd(d); 1008 crit_exit(); 1009 return(0); 1010 } 1011 1012 /* Not found. */ 1013 return(ENXIO); 1014 } 1015 1016 /* 1017 * Support for select() and poll() system calls 1018 * 1019 * Return true iff the specific operation will not block indefinitely. 1020 * Otherwise, return false but make a note that a selwakeup() must be done. 1021 */ 1022 int 1023 bpfpoll(struct dev_poll_args *ap) 1024 { 1025 cdev_t dev = ap->a_head.a_dev; 1026 struct bpf_d *d; 1027 int revents; 1028 1029 d = dev->si_drv1; 1030 if (d->bd_bif == NULL) 1031 return(ENXIO); 1032 1033 revents = ap->a_events & (POLLOUT | POLLWRNORM); 1034 crit_enter(); 1035 if (ap->a_events & (POLLIN | POLLRDNORM)) { 1036 /* 1037 * An imitation of the FIONREAD ioctl code. 1038 * XXX not quite. An exact imitation: 1039 * if (d->b_slen != 0 || 1040 * (d->bd_hbuf != NULL && d->bd_hlen != 0) 1041 */ 1042 if (d->bd_hlen != 0 || 1043 ((d->bd_immediate || d->bd_state == BPF_TIMED_OUT) && 1044 d->bd_slen != 0)) { 1045 revents |= ap->a_events & (POLLIN | POLLRDNORM); 1046 } else { 1047 selrecord(curthread, &d->bd_sel); 1048 /* Start the read timeout if necessary. */ 1049 if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) { 1050 callout_reset(&d->bd_callout, d->bd_rtout, 1051 bpf_timed_out, d); 1052 d->bd_state = BPF_WAITING; 1053 } 1054 } 1055 } 1056 crit_exit(); 1057 ap->a_events = revents; 1058 return(0); 1059 } 1060 1061 /* 1062 * Process the packet pkt of length pktlen. The packet is parsed 1063 * by each listener's filter, and if accepted, stashed into the 1064 * corresponding buffer. 1065 */ 1066 void 1067 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen) 1068 { 1069 struct bpf_d *d; 1070 struct timeval tv; 1071 int gottime = 0; 1072 u_int slen; 1073 1074 /* 1075 * Note that the ipl does not have to be raised at this point. 1076 * The only problem that could arise here is that if two different 1077 * interfaces shared any data. This is not the case. 1078 */ 1079 SLIST_FOREACH(d, &bp->bif_dlist, bd_next) { 1080 ++d->bd_rcount; 1081 slen = bpf_filter(d->bd_rfilter, pkt, pktlen, pktlen); 1082 if (slen != 0) { 1083 if (!gottime) { 1084 microtime(&tv); 1085 gottime = 1; 1086 } 1087 catchpacket(d, pkt, pktlen, slen, ovbcopy, &tv); 1088 } 1089 } 1090 } 1091 1092 /* 1093 * Copy data from an mbuf chain into a buffer. This code is derived 1094 * from m_copydata in sys/uipc_mbuf.c. 1095 */ 1096 static void 1097 bpf_mcopy(const void *src_arg, void *dst_arg, size_t len) 1098 { 1099 const struct mbuf *m; 1100 u_int count; 1101 u_char *dst; 1102 1103 m = src_arg; 1104 dst = dst_arg; 1105 while (len > 0) { 1106 if (m == NULL) 1107 panic("bpf_mcopy"); 1108 count = min(m->m_len, len); 1109 bcopy(mtod(m, void *), dst, count); 1110 m = m->m_next; 1111 dst += count; 1112 len -= count; 1113 } 1114 } 1115 1116 /* 1117 * Process the packet in the mbuf chain m. The packet is parsed by each 1118 * listener's filter, and if accepted, stashed into the corresponding 1119 * buffer. 1120 */ 1121 void 1122 bpf_mtap(struct bpf_if *bp, struct mbuf *m) 1123 { 1124 struct bpf_d *d; 1125 u_int pktlen, slen; 1126 struct timeval tv; 1127 int gottime = 0; 1128 1129 /* Don't compute pktlen, if no descriptor is attached. */ 1130 if (SLIST_EMPTY(&bp->bif_dlist)) 1131 return; 1132 1133 pktlen = m_lengthm(m, NULL); 1134 1135 SLIST_FOREACH(d, &bp->bif_dlist, bd_next) { 1136 if (!d->bd_seesent && (m->m_pkthdr.rcvif == NULL)) 1137 continue; 1138 ++d->bd_rcount; 1139 slen = bpf_filter(d->bd_rfilter, (u_char *)m, pktlen, 0); 1140 if (slen != 0) { 1141 if (!gottime) { 1142 microtime(&tv); 1143 gottime = 1; 1144 } 1145 catchpacket(d, (u_char *)m, pktlen, slen, bpf_mcopy, 1146 &tv); 1147 } 1148 } 1149 } 1150 1151 void 1152 bpf_mtap_family(struct bpf_if *bp, struct mbuf *m, sa_family_t family) 1153 { 1154 u_int family4; 1155 1156 KKASSERT(family != AF_UNSPEC); 1157 1158 family4 = (u_int)family; 1159 bpf_ptap(bp, m, &family4, sizeof(family4)); 1160 } 1161 1162 /* 1163 * Process the packet in the mbuf chain m with the header in m prepended. 1164 * The packet is parsed by each listener's filter, and if accepted, 1165 * stashed into the corresponding buffer. 1166 */ 1167 void 1168 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen) 1169 { 1170 struct mbuf mb; 1171 1172 /* 1173 * Craft on-stack mbuf suitable for passing to bpf_mtap. 1174 * Note that we cut corners here; we only setup what's 1175 * absolutely needed--this mbuf should never go anywhere else. 1176 */ 1177 mb.m_next = m; 1178 mb.m_data = __DECONST(void *, data); /* LINTED */ 1179 mb.m_len = dlen; 1180 mb.m_pkthdr.rcvif = m->m_pkthdr.rcvif; 1181 1182 bpf_mtap(bp, &mb); 1183 } 1184 1185 /* 1186 * Move the packet data from interface memory (pkt) into the 1187 * store buffer. Return 1 if it's time to wakeup a listener (buffer full), 1188 * otherwise 0. "copy" is the routine called to do the actual data 1189 * transfer. bcopy is passed in to copy contiguous chunks, while 1190 * bpf_mcopy is passed in to copy mbuf chains. In the latter case, 1191 * pkt is really an mbuf. 1192 */ 1193 static void 1194 catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen, 1195 void (*cpfn)(const void *, void *, size_t), 1196 const struct timeval *tv) 1197 { 1198 struct bpf_hdr *hp; 1199 int totlen, curlen; 1200 int hdrlen = d->bd_bif->bif_hdrlen; 1201 /* 1202 * Figure out how many bytes to move. If the packet is 1203 * greater or equal to the snapshot length, transfer that 1204 * much. Otherwise, transfer the whole packet (unless 1205 * we hit the buffer size limit). 1206 */ 1207 totlen = hdrlen + min(snaplen, pktlen); 1208 if (totlen > d->bd_bufsize) 1209 totlen = d->bd_bufsize; 1210 1211 /* 1212 * Round up the end of the previous packet to the next longword. 1213 */ 1214 curlen = BPF_WORDALIGN(d->bd_slen); 1215 if (curlen + totlen > d->bd_bufsize) { 1216 /* 1217 * This packet will overflow the storage buffer. 1218 * Rotate the buffers if we can, then wakeup any 1219 * pending reads. 1220 */ 1221 if (d->bd_fbuf == NULL) { 1222 /* 1223 * We haven't completed the previous read yet, 1224 * so drop the packet. 1225 */ 1226 ++d->bd_dcount; 1227 return; 1228 } 1229 ROTATE_BUFFERS(d); 1230 bpf_wakeup(d); 1231 curlen = 0; 1232 } else if (d->bd_immediate || d->bd_state == BPF_TIMED_OUT) { 1233 /* 1234 * Immediate mode is set, or the read timeout has 1235 * already expired during a select call. A packet 1236 * arrived, so the reader should be woken up. 1237 */ 1238 bpf_wakeup(d); 1239 } 1240 1241 /* 1242 * Append the bpf header. 1243 */ 1244 hp = (struct bpf_hdr *)(d->bd_sbuf + curlen); 1245 hp->bh_tstamp = *tv; 1246 hp->bh_datalen = pktlen; 1247 hp->bh_hdrlen = hdrlen; 1248 /* 1249 * Copy the packet data into the store buffer and update its length. 1250 */ 1251 (*cpfn)(pkt, (u_char *)hp + hdrlen, (hp->bh_caplen = totlen - hdrlen)); 1252 d->bd_slen = curlen + totlen; 1253 } 1254 1255 /* 1256 * Initialize all nonzero fields of a descriptor. 1257 */ 1258 static int 1259 bpf_allocbufs(struct bpf_d *d) 1260 { 1261 d->bd_fbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK); 1262 d->bd_sbuf = kmalloc(d->bd_bufsize, M_BPF, M_WAITOK); 1263 d->bd_slen = 0; 1264 d->bd_hlen = 0; 1265 return(0); 1266 } 1267 1268 /* 1269 * Free buffers and packet filter program currently in use by a descriptor. 1270 * Called on close. 1271 */ 1272 static void 1273 bpf_freed(struct bpf_d *d) 1274 { 1275 /* 1276 * We don't need to lock out interrupts since this descriptor has 1277 * been detached from its interface and it yet hasn't been marked 1278 * free. 1279 */ 1280 if (d->bd_sbuf != NULL) { 1281 kfree(d->bd_sbuf, M_BPF); 1282 if (d->bd_hbuf != NULL) 1283 kfree(d->bd_hbuf, M_BPF); 1284 if (d->bd_fbuf != NULL) 1285 kfree(d->bd_fbuf, M_BPF); 1286 } 1287 if (d->bd_rfilter) 1288 kfree(d->bd_rfilter, M_BPF); 1289 if (d->bd_wfilter) 1290 kfree(d->bd_wfilter, M_BPF); 1291 } 1292 1293 /* 1294 * Attach an interface to bpf. ifp is a pointer to the structure 1295 * defining the interface to be attached, dlt is the link layer type, 1296 * and hdrlen is the fixed size of the link header (variable length 1297 * headers are not yet supported). 1298 */ 1299 void 1300 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen) 1301 { 1302 bpfattach_dlt(ifp, dlt, hdrlen, &ifp->if_bpf); 1303 } 1304 1305 void 1306 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp) 1307 { 1308 struct bpf_if *bp; 1309 1310 bp = kmalloc(sizeof *bp, M_BPF, M_WAITOK | M_ZERO); 1311 1312 SLIST_INIT(&bp->bif_dlist); 1313 bp->bif_ifp = ifp; 1314 bp->bif_dlt = dlt; 1315 bp->bif_driverp = driverp; 1316 *bp->bif_driverp = NULL; 1317 1318 bp->bif_next = bpf_iflist; 1319 bpf_iflist = bp; 1320 1321 /* 1322 * Compute the length of the bpf header. This is not necessarily 1323 * equal to SIZEOF_BPF_HDR because we want to insert spacing such 1324 * that the network layer header begins on a longword boundary (for 1325 * performance reasons and to alleviate alignment restrictions). 1326 */ 1327 bp->bif_hdrlen = BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen; 1328 1329 if (bootverbose) 1330 if_printf(ifp, "bpf attached\n"); 1331 } 1332 1333 /* 1334 * Detach bpf from an interface. This involves detaching each descriptor 1335 * associated with the interface, and leaving bd_bif NULL. Notify each 1336 * descriptor as it's detached so that any sleepers wake up and get 1337 * ENXIO. 1338 */ 1339 void 1340 bpfdetach(struct ifnet *ifp) 1341 { 1342 struct bpf_if *bp, *bp_prev; 1343 struct bpf_d *d; 1344 1345 crit_enter(); 1346 1347 /* Locate BPF interface information */ 1348 bp_prev = NULL; 1349 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) { 1350 if (ifp == bp->bif_ifp) 1351 break; 1352 bp_prev = bp; 1353 } 1354 1355 /* Interface wasn't attached */ 1356 if (bp->bif_ifp == NULL) { 1357 crit_exit(); 1358 kprintf("bpfdetach: %s was not attached\n", ifp->if_xname); 1359 return; 1360 } 1361 1362 while ((d = SLIST_FIRST(&bp->bif_dlist)) != NULL) { 1363 bpf_detachd(d); 1364 bpf_wakeup(d); 1365 } 1366 1367 if (bp_prev != NULL) 1368 bp_prev->bif_next = bp->bif_next; 1369 else 1370 bpf_iflist = bp->bif_next; 1371 1372 kfree(bp, M_BPF); 1373 1374 crit_exit(); 1375 } 1376 1377 /* 1378 * Get a list of available data link type of the interface. 1379 */ 1380 static int 1381 bpf_getdltlist(struct bpf_d *d, struct bpf_dltlist *bfl) 1382 { 1383 int n, error; 1384 struct ifnet *ifp; 1385 struct bpf_if *bp; 1386 1387 ifp = d->bd_bif->bif_ifp; 1388 n = 0; 1389 error = 0; 1390 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) { 1391 if (bp->bif_ifp != ifp) 1392 continue; 1393 if (bfl->bfl_list != NULL) { 1394 if (n >= bfl->bfl_len) { 1395 return (ENOMEM); 1396 } 1397 error = copyout(&bp->bif_dlt, 1398 bfl->bfl_list + n, sizeof(u_int)); 1399 } 1400 n++; 1401 } 1402 bfl->bfl_len = n; 1403 return(error); 1404 } 1405 1406 /* 1407 * Set the data link type of a BPF instance. 1408 */ 1409 static int 1410 bpf_setdlt(struct bpf_d *d, u_int dlt) 1411 { 1412 int error, opromisc; 1413 struct ifnet *ifp; 1414 struct bpf_if *bp; 1415 1416 if (d->bd_bif->bif_dlt == dlt) 1417 return (0); 1418 ifp = d->bd_bif->bif_ifp; 1419 for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) { 1420 if (bp->bif_ifp == ifp && bp->bif_dlt == dlt) 1421 break; 1422 } 1423 if (bp != NULL) { 1424 opromisc = d->bd_promisc; 1425 crit_enter(); 1426 bpf_detachd(d); 1427 bpf_attachd(d, bp); 1428 bpf_resetd(d); 1429 if (opromisc) { 1430 error = ifpromisc(bp->bif_ifp, 1); 1431 if (error) { 1432 if_printf(bp->bif_ifp, 1433 "bpf_setdlt: ifpromisc failed (%d)\n", 1434 error); 1435 } else { 1436 d->bd_promisc = 1; 1437 } 1438 } 1439 crit_exit(); 1440 } 1441 return(bp == NULL ? EINVAL : 0); 1442 } 1443 1444 static void 1445 bpf_drvinit(void *unused) 1446 { 1447 dev_ops_add(&bpf_ops, 0, 0); 1448 } 1449 1450 SYSINIT(bpfdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,bpf_drvinit,NULL) 1451 1452 #else /* !BPF */ 1453 /* 1454 * NOP stubs to allow bpf-using drivers to load and function. 1455 * 1456 * A 'better' implementation would allow the core bpf functionality 1457 * to be loaded at runtime. 1458 */ 1459 1460 void 1461 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen) 1462 { 1463 } 1464 1465 void 1466 bpf_mtap(struct bpf_if *bp, struct mbuf *m) 1467 { 1468 } 1469 1470 void 1471 bpf_ptap(struct bpf_if *bp, struct mbuf *m, const void *data, u_int dlen) 1472 { 1473 } 1474 1475 void 1476 bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen) 1477 { 1478 } 1479 1480 void 1481 bpfattach_dlt(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp) 1482 { 1483 } 1484 1485 void 1486 bpfdetach(struct ifnet *ifp) 1487 { 1488 } 1489 1490 u_int 1491 bpf_filter(const struct bpf_insn *pc, u_char *p, u_int wirelen, u_int buflen) 1492 { 1493 return -1; /* "no filter" behaviour */ 1494 } 1495 1496 #endif /* !BPF */ 1497