| 1 | /* $NetBSD: bpf.c,v 1.203 2016/07/19 02:47:45 pgoyette Exp $ */ |
| 2 | |
| 3 | /* |
| 4 | * Copyright (c) 1990, 1991, 1993 |
| 5 | * The Regents of the University of California. All rights reserved. |
| 6 | * |
| 7 | * This code is derived from the Stanford/CMU enet packet filter, |
| 8 | * (net/enet.c) distributed as part of 4.3BSD, and code contributed |
| 9 | * to Berkeley by Steven McCanne and Van Jacobson both of Lawrence |
| 10 | * Berkeley Laboratory. |
| 11 | * |
| 12 | * Redistribution and use in source and binary forms, with or without |
| 13 | * modification, are permitted provided that the following conditions |
| 14 | * are met: |
| 15 | * 1. Redistributions of source code must retain the above copyright |
| 16 | * notice, this list of conditions and the following disclaimer. |
| 17 | * 2. Redistributions in binary form must reproduce the above copyright |
| 18 | * notice, this list of conditions and the following disclaimer in the |
| 19 | * documentation and/or other materials provided with the distribution. |
| 20 | * 3. Neither the name of the University nor the names of its contributors |
| 21 | * may be used to endorse or promote products derived from this software |
| 22 | * without specific prior written permission. |
| 23 | * |
| 24 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
| 25 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 26 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 27 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
| 28 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 29 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 30 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 31 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 32 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 33 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 34 | * SUCH DAMAGE. |
| 35 | * |
| 36 | * @(#)bpf.c 8.4 (Berkeley) 1/9/95 |
| 37 | * static char rcsid[] = |
| 38 | * "Header: bpf.c,v 1.67 96/09/26 22:00:52 leres Exp "; |
| 39 | */ |
| 40 | |
| 41 | #include <sys/cdefs.h> |
| 42 | __KERNEL_RCSID(0, "$NetBSD: bpf.c,v 1.203 2016/07/19 02:47:45 pgoyette Exp $" ); |
| 43 | |
| 44 | #if defined(_KERNEL_OPT) |
| 45 | #include "opt_bpf.h" |
| 46 | #include "sl.h" |
| 47 | #include "strip.h" |
| 48 | #endif |
| 49 | |
| 50 | #include <sys/param.h> |
| 51 | #include <sys/systm.h> |
| 52 | #include <sys/mbuf.h> |
| 53 | #include <sys/buf.h> |
| 54 | #include <sys/time.h> |
| 55 | #include <sys/proc.h> |
| 56 | #include <sys/ioctl.h> |
| 57 | #include <sys/conf.h> |
| 58 | #include <sys/vnode.h> |
| 59 | #include <sys/queue.h> |
| 60 | #include <sys/stat.h> |
| 61 | #include <sys/module.h> |
| 62 | #include <sys/atomic.h> |
| 63 | |
| 64 | #include <sys/file.h> |
| 65 | #include <sys/filedesc.h> |
| 66 | #include <sys/tty.h> |
| 67 | #include <sys/uio.h> |
| 68 | |
| 69 | #include <sys/protosw.h> |
| 70 | #include <sys/socket.h> |
| 71 | #include <sys/errno.h> |
| 72 | #include <sys/kernel.h> |
| 73 | #include <sys/poll.h> |
| 74 | #include <sys/sysctl.h> |
| 75 | #include <sys/kauth.h> |
| 76 | |
| 77 | #include <net/if.h> |
| 78 | #include <net/slip.h> |
| 79 | |
| 80 | #include <net/bpf.h> |
| 81 | #include <net/bpfdesc.h> |
| 82 | #include <net/bpfjit.h> |
| 83 | |
| 84 | #include <net/if_arc.h> |
| 85 | #include <net/if_ether.h> |
| 86 | |
| 87 | #include <netinet/in.h> |
| 88 | #include <netinet/if_inarp.h> |
| 89 | |
| 90 | |
| 91 | #include <compat/sys/sockio.h> |
| 92 | |
| 93 | #ifndef BPF_BUFSIZE |
| 94 | /* |
| 95 | * 4096 is too small for FDDI frames. 8192 is too small for gigabit Ethernet |
| 96 | * jumbos (circa 9k), ATM, or Intel gig/10gig ethernet jumbos (16k). |
| 97 | */ |
| 98 | # define BPF_BUFSIZE 32768 |
| 99 | #endif |
| 100 | |
| 101 | #define PRINET 26 /* interruptible */ |
| 102 | |
| 103 | /* |
| 104 | * The default read buffer size, and limit for BIOCSBLEN, is sysctl'able. |
| 105 | * XXX the default values should be computed dynamically based |
| 106 | * on available memory size and available mbuf clusters. |
| 107 | */ |
| 108 | int bpf_bufsize = BPF_BUFSIZE; |
| 109 | int bpf_maxbufsize = BPF_DFLTBUFSIZE; /* XXX set dynamically, see above */ |
| 110 | bool bpf_jit = false; |
| 111 | |
| 112 | struct bpfjit_ops bpfjit_module_ops = { |
| 113 | .bj_generate_code = NULL, |
| 114 | .bj_free_code = NULL |
| 115 | }; |
| 116 | |
| 117 | /* |
| 118 | * Global BPF statistics returned by net.bpf.stats sysctl. |
| 119 | */ |
| 120 | struct bpf_stat bpf_gstats; |
| 121 | |
| 122 | /* |
| 123 | * Use a mutex to avoid a race condition between gathering the stats/peers |
| 124 | * and opening/closing the device. |
| 125 | */ |
| 126 | static kmutex_t bpf_mtx; |
| 127 | |
| 128 | /* |
| 129 | * bpf_iflist is the list of interfaces; each corresponds to an ifnet |
| 130 | * bpf_dtab holds the descriptors, indexed by minor device # |
| 131 | */ |
| 132 | struct bpf_if *bpf_iflist; |
| 133 | LIST_HEAD(, bpf_d) bpf_list; |
| 134 | |
| 135 | static int bpf_allocbufs(struct bpf_d *); |
| 136 | static void bpf_deliver(struct bpf_if *, |
| 137 | void *(*cpfn)(void *, const void *, size_t), |
| 138 | void *, u_int, u_int, const bool); |
| 139 | static void bpf_freed(struct bpf_d *); |
| 140 | static void bpf_ifname(struct ifnet *, struct ifreq *); |
| 141 | static void *bpf_mcpy(void *, const void *, size_t); |
| 142 | static int bpf_movein(struct uio *, int, uint64_t, |
| 143 | struct mbuf **, struct sockaddr *); |
| 144 | static void bpf_attachd(struct bpf_d *, struct bpf_if *); |
| 145 | static void bpf_detachd(struct bpf_d *); |
| 146 | static int bpf_setif(struct bpf_d *, struct ifreq *); |
| 147 | static void bpf_timed_out(void *); |
| 148 | static inline void |
| 149 | bpf_wakeup(struct bpf_d *); |
| 150 | static int bpf_hdrlen(struct bpf_d *); |
| 151 | static void catchpacket(struct bpf_d *, u_char *, u_int, u_int, |
| 152 | void *(*)(void *, const void *, size_t), struct timespec *); |
| 153 | static void reset_d(struct bpf_d *); |
| 154 | static int bpf_getdltlist(struct bpf_d *, struct bpf_dltlist *); |
| 155 | static int bpf_setdlt(struct bpf_d *, u_int); |
| 156 | |
| 157 | static int bpf_read(struct file *, off_t *, struct uio *, kauth_cred_t, |
| 158 | int); |
| 159 | static int bpf_write(struct file *, off_t *, struct uio *, kauth_cred_t, |
| 160 | int); |
| 161 | static int bpf_ioctl(struct file *, u_long, void *); |
| 162 | static int bpf_poll(struct file *, int); |
| 163 | static int bpf_stat(struct file *, struct stat *); |
| 164 | static int bpf_close(struct file *); |
| 165 | static int bpf_kqfilter(struct file *, struct knote *); |
| 166 | static void bpf_softintr(void *); |
| 167 | |
| 168 | static const struct fileops bpf_fileops = { |
| 169 | .fo_read = bpf_read, |
| 170 | .fo_write = bpf_write, |
| 171 | .fo_ioctl = bpf_ioctl, |
| 172 | .fo_fcntl = fnullop_fcntl, |
| 173 | .fo_poll = bpf_poll, |
| 174 | .fo_stat = bpf_stat, |
| 175 | .fo_close = bpf_close, |
| 176 | .fo_kqfilter = bpf_kqfilter, |
| 177 | .fo_restart = fnullop_restart, |
| 178 | }; |
| 179 | |
| 180 | dev_type_open(bpfopen); |
| 181 | |
| 182 | const struct cdevsw bpf_cdevsw = { |
| 183 | .d_open = bpfopen, |
| 184 | .d_close = noclose, |
| 185 | .d_read = noread, |
| 186 | .d_write = nowrite, |
| 187 | .d_ioctl = noioctl, |
| 188 | .d_stop = nostop, |
| 189 | .d_tty = notty, |
| 190 | .d_poll = nopoll, |
| 191 | .d_mmap = nommap, |
| 192 | .d_kqfilter = nokqfilter, |
| 193 | .d_discard = nodiscard, |
| 194 | .d_flag = D_OTHER |
| 195 | }; |
| 196 | |
| 197 | bpfjit_func_t |
| 198 | bpf_jit_generate(bpf_ctx_t *bc, void *code, size_t size) |
| 199 | { |
| 200 | |
| 201 | membar_consumer(); |
| 202 | if (bpfjit_module_ops.bj_generate_code != NULL) { |
| 203 | return bpfjit_module_ops.bj_generate_code(bc, code, size); |
| 204 | } |
| 205 | return NULL; |
| 206 | } |
| 207 | |
| 208 | void |
| 209 | bpf_jit_freecode(bpfjit_func_t jcode) |
| 210 | { |
| 211 | KASSERT(bpfjit_module_ops.bj_free_code != NULL); |
| 212 | bpfjit_module_ops.bj_free_code(jcode); |
| 213 | } |
| 214 | |
| 215 | static int |
| 216 | bpf_movein(struct uio *uio, int linktype, uint64_t mtu, struct mbuf **mp, |
| 217 | struct sockaddr *sockp) |
| 218 | { |
| 219 | struct mbuf *m; |
| 220 | int error; |
| 221 | size_t len; |
| 222 | size_t hlen; |
| 223 | size_t align; |
| 224 | |
| 225 | /* |
| 226 | * Build a sockaddr based on the data link layer type. |
| 227 | * We do this at this level because the ethernet header |
| 228 | * is copied directly into the data field of the sockaddr. |
| 229 | * In the case of SLIP, there is no header and the packet |
| 230 | * is forwarded as is. |
| 231 | * Also, we are careful to leave room at the front of the mbuf |
| 232 | * for the link level header. |
| 233 | */ |
| 234 | switch (linktype) { |
| 235 | |
| 236 | case DLT_SLIP: |
| 237 | sockp->sa_family = AF_INET; |
| 238 | hlen = 0; |
| 239 | align = 0; |
| 240 | break; |
| 241 | |
| 242 | case DLT_PPP: |
| 243 | sockp->sa_family = AF_UNSPEC; |
| 244 | hlen = 0; |
| 245 | align = 0; |
| 246 | break; |
| 247 | |
| 248 | case DLT_EN10MB: |
| 249 | sockp->sa_family = AF_UNSPEC; |
| 250 | /* XXX Would MAXLINKHDR be better? */ |
| 251 | /* 6(dst)+6(src)+2(type) */ |
| 252 | hlen = sizeof(struct ether_header); |
| 253 | align = 2; |
| 254 | break; |
| 255 | |
| 256 | case DLT_ARCNET: |
| 257 | sockp->sa_family = AF_UNSPEC; |
| 258 | hlen = ARC_HDRLEN; |
| 259 | align = 5; |
| 260 | break; |
| 261 | |
| 262 | case DLT_FDDI: |
| 263 | sockp->sa_family = AF_LINK; |
| 264 | /* XXX 4(FORMAC)+6(dst)+6(src) */ |
| 265 | hlen = 16; |
| 266 | align = 0; |
| 267 | break; |
| 268 | |
| 269 | case DLT_ECONET: |
| 270 | sockp->sa_family = AF_UNSPEC; |
| 271 | hlen = 6; |
| 272 | align = 2; |
| 273 | break; |
| 274 | |
| 275 | case DLT_NULL: |
| 276 | sockp->sa_family = AF_UNSPEC; |
| 277 | hlen = 0; |
| 278 | align = 0; |
| 279 | break; |
| 280 | |
| 281 | default: |
| 282 | return (EIO); |
| 283 | } |
| 284 | |
| 285 | len = uio->uio_resid; |
| 286 | /* |
| 287 | * If there aren't enough bytes for a link level header or the |
| 288 | * packet length exceeds the interface mtu, return an error. |
| 289 | */ |
| 290 | if (len - hlen > mtu) |
| 291 | return (EMSGSIZE); |
| 292 | |
| 293 | /* |
| 294 | * XXX Avoid complicated buffer chaining --- |
| 295 | * bail if it won't fit in a single mbuf. |
| 296 | * (Take into account possible alignment bytes) |
| 297 | */ |
| 298 | if (len + align > MCLBYTES) |
| 299 | return (EIO); |
| 300 | |
| 301 | m = m_gethdr(M_WAIT, MT_DATA); |
| 302 | m_reset_rcvif(m); |
| 303 | m->m_pkthdr.len = (int)(len - hlen); |
| 304 | if (len + align > MHLEN) { |
| 305 | m_clget(m, M_WAIT); |
| 306 | if ((m->m_flags & M_EXT) == 0) { |
| 307 | error = ENOBUFS; |
| 308 | goto bad; |
| 309 | } |
| 310 | } |
| 311 | |
| 312 | /* Insure the data is properly aligned */ |
| 313 | if (align > 0) { |
| 314 | m->m_data += align; |
| 315 | m->m_len -= (int)align; |
| 316 | } |
| 317 | |
| 318 | error = uiomove(mtod(m, void *), len, uio); |
| 319 | if (error) |
| 320 | goto bad; |
| 321 | if (hlen != 0) { |
| 322 | memcpy(sockp->sa_data, mtod(m, void *), hlen); |
| 323 | m->m_data += hlen; /* XXX */ |
| 324 | len -= hlen; |
| 325 | } |
| 326 | m->m_len = (int)len; |
| 327 | *mp = m; |
| 328 | return (0); |
| 329 | |
| 330 | bad: |
| 331 | m_freem(m); |
| 332 | return (error); |
| 333 | } |
| 334 | |
| 335 | /* |
| 336 | * Attach file to the bpf interface, i.e. make d listen on bp. |
| 337 | * Must be called at splnet. |
| 338 | */ |
| 339 | static void |
| 340 | bpf_attachd(struct bpf_d *d, struct bpf_if *bp) |
| 341 | { |
| 342 | KASSERT(mutex_owned(&bpf_mtx)); |
| 343 | /* |
| 344 | * Point d at bp, and add d to the interface's list of listeners. |
| 345 | * Finally, point the driver's bpf cookie at the interface so |
| 346 | * it will divert packets to bpf. |
| 347 | */ |
| 348 | d->bd_bif = bp; |
| 349 | d->bd_next = bp->bif_dlist; |
| 350 | bp->bif_dlist = d; |
| 351 | |
| 352 | *bp->bif_driverp = bp; |
| 353 | } |
| 354 | |
| 355 | /* |
| 356 | * Detach a file from its interface. |
| 357 | */ |
| 358 | static void |
| 359 | bpf_detachd(struct bpf_d *d) |
| 360 | { |
| 361 | struct bpf_d **p; |
| 362 | struct bpf_if *bp; |
| 363 | |
| 364 | KASSERT(mutex_owned(&bpf_mtx)); |
| 365 | |
| 366 | bp = d->bd_bif; |
| 367 | /* |
| 368 | * Check if this descriptor had requested promiscuous mode. |
| 369 | * If so, turn it off. |
| 370 | */ |
| 371 | if (d->bd_promisc) { |
| 372 | int error __diagused; |
| 373 | |
| 374 | d->bd_promisc = 0; |
| 375 | /* |
| 376 | * Take device out of promiscuous mode. Since we were |
| 377 | * able to enter promiscuous mode, we should be able |
| 378 | * to turn it off. But we can get an error if |
| 379 | * the interface was configured down, so only panic |
| 380 | * if we don't get an unexpected error. |
| 381 | */ |
| 382 | error = ifpromisc(bp->bif_ifp, 0); |
| 383 | #ifdef DIAGNOSTIC |
| 384 | if (error) |
| 385 | printf("%s: ifpromisc failed: %d" , __func__, error); |
| 386 | #endif |
| 387 | } |
| 388 | /* Remove d from the interface's descriptor list. */ |
| 389 | p = &bp->bif_dlist; |
| 390 | while (*p != d) { |
| 391 | p = &(*p)->bd_next; |
| 392 | if (*p == NULL) |
| 393 | panic("%s: descriptor not in list" , __func__); |
| 394 | } |
| 395 | *p = (*p)->bd_next; |
| 396 | if (bp->bif_dlist == NULL) |
| 397 | /* |
| 398 | * Let the driver know that there are no more listeners. |
| 399 | */ |
| 400 | *d->bd_bif->bif_driverp = NULL; |
| 401 | d->bd_bif = NULL; |
| 402 | } |
| 403 | |
| 404 | static void |
| 405 | bpf_init(void) |
| 406 | { |
| 407 | |
| 408 | mutex_init(&bpf_mtx, MUTEX_DEFAULT, IPL_NONE); |
| 409 | |
| 410 | LIST_INIT(&bpf_list); |
| 411 | |
| 412 | bpf_gstats.bs_recv = 0; |
| 413 | bpf_gstats.bs_drop = 0; |
| 414 | bpf_gstats.bs_capt = 0; |
| 415 | |
| 416 | return; |
| 417 | } |
| 418 | |
| 419 | /* |
| 420 | * bpfilterattach() is called at boot time. We don't need to do anything |
| 421 | * here, since any initialization will happen as part of module init code. |
| 422 | */ |
| 423 | /* ARGSUSED */ |
| 424 | void |
| 425 | bpfilterattach(int n) |
| 426 | { |
| 427 | |
| 428 | } |
| 429 | |
| 430 | /* |
| 431 | * Open ethernet device. Clones. |
| 432 | */ |
| 433 | /* ARGSUSED */ |
| 434 | int |
| 435 | bpfopen(dev_t dev, int flag, int mode, struct lwp *l) |
| 436 | { |
| 437 | struct bpf_d *d; |
| 438 | struct file *fp; |
| 439 | int error, fd; |
| 440 | |
| 441 | /* falloc() will fill in the descriptor for us. */ |
| 442 | if ((error = fd_allocfile(&fp, &fd)) != 0) |
| 443 | return error; |
| 444 | |
| 445 | d = malloc(sizeof(*d), M_DEVBUF, M_WAITOK|M_ZERO); |
| 446 | d->bd_bufsize = bpf_bufsize; |
| 447 | d->bd_seesent = 1; |
| 448 | d->bd_feedback = 0; |
| 449 | d->bd_pid = l->l_proc->p_pid; |
| 450 | #ifdef _LP64 |
| 451 | if (curproc->p_flag & PK_32) |
| 452 | d->bd_compat32 = 1; |
| 453 | #endif |
| 454 | getnanotime(&d->bd_btime); |
| 455 | d->bd_atime = d->bd_mtime = d->bd_btime; |
| 456 | callout_init(&d->bd_callout, 0); |
| 457 | selinit(&d->bd_sel); |
| 458 | d->bd_sih = softint_establish(SOFTINT_CLOCK, bpf_softintr, d); |
| 459 | d->bd_jitcode = NULL; |
| 460 | |
| 461 | mutex_enter(&bpf_mtx); |
| 462 | LIST_INSERT_HEAD(&bpf_list, d, bd_list); |
| 463 | mutex_exit(&bpf_mtx); |
| 464 | |
| 465 | return fd_clone(fp, fd, flag, &bpf_fileops, d); |
| 466 | } |
| 467 | |
| 468 | /* |
| 469 | * Close the descriptor by detaching it from its interface, |
| 470 | * deallocating its buffers, and marking it free. |
| 471 | */ |
| 472 | /* ARGSUSED */ |
| 473 | static int |
| 474 | bpf_close(struct file *fp) |
| 475 | { |
| 476 | struct bpf_d *d; |
| 477 | int s; |
| 478 | |
| 479 | KERNEL_LOCK(1, NULL); |
| 480 | mutex_enter(&bpf_mtx); |
| 481 | |
| 482 | if ((d = fp->f_bpf) == NULL) { |
| 483 | mutex_exit(&bpf_mtx); |
| 484 | KERNEL_UNLOCK_ONE(NULL); |
| 485 | return 0; |
| 486 | } |
| 487 | |
| 488 | /* |
| 489 | * Refresh the PID associated with this bpf file. |
| 490 | */ |
| 491 | d->bd_pid = curproc->p_pid; |
| 492 | |
| 493 | s = splnet(); |
| 494 | if (d->bd_state == BPF_WAITING) |
| 495 | callout_stop(&d->bd_callout); |
| 496 | d->bd_state = BPF_IDLE; |
| 497 | if (d->bd_bif) |
| 498 | bpf_detachd(d); |
| 499 | splx(s); |
| 500 | bpf_freed(d); |
| 501 | LIST_REMOVE(d, bd_list); |
| 502 | fp->f_bpf = NULL; |
| 503 | |
| 504 | mutex_exit(&bpf_mtx); |
| 505 | KERNEL_UNLOCK_ONE(NULL); |
| 506 | |
| 507 | callout_destroy(&d->bd_callout); |
| 508 | seldestroy(&d->bd_sel); |
| 509 | softint_disestablish(d->bd_sih); |
| 510 | free(d, M_DEVBUF); |
| 511 | |
| 512 | return (0); |
| 513 | } |
| 514 | |
| 515 | /* |
| 516 | * Rotate the packet buffers in descriptor d. Move the store buffer |
| 517 | * into the hold slot, and the free buffer into the store slot. |
| 518 | * Zero the length of the new store buffer. |
| 519 | */ |
| 520 | #define ROTATE_BUFFERS(d) \ |
| 521 | (d)->bd_hbuf = (d)->bd_sbuf; \ |
| 522 | (d)->bd_hlen = (d)->bd_slen; \ |
| 523 | (d)->bd_sbuf = (d)->bd_fbuf; \ |
| 524 | (d)->bd_slen = 0; \ |
| 525 | (d)->bd_fbuf = NULL; |
| 526 | /* |
| 527 | * bpfread - read next chunk of packets from buffers |
| 528 | */ |
| 529 | static int |
| 530 | bpf_read(struct file *fp, off_t *offp, struct uio *uio, |
| 531 | kauth_cred_t cred, int flags) |
| 532 | { |
| 533 | struct bpf_d *d = fp->f_bpf; |
| 534 | int timed_out; |
| 535 | int error; |
| 536 | int s; |
| 537 | |
| 538 | getnanotime(&d->bd_atime); |
| 539 | /* |
| 540 | * Restrict application to use a buffer the same size as |
| 541 | * the kernel buffers. |
| 542 | */ |
| 543 | if (uio->uio_resid != d->bd_bufsize) |
| 544 | return (EINVAL); |
| 545 | |
| 546 | KERNEL_LOCK(1, NULL); |
| 547 | s = splnet(); |
| 548 | if (d->bd_state == BPF_WAITING) |
| 549 | callout_stop(&d->bd_callout); |
| 550 | timed_out = (d->bd_state == BPF_TIMED_OUT); |
| 551 | d->bd_state = BPF_IDLE; |
| 552 | /* |
| 553 | * If the hold buffer is empty, then do a timed sleep, which |
| 554 | * ends when the timeout expires or when enough packets |
| 555 | * have arrived to fill the store buffer. |
| 556 | */ |
| 557 | while (d->bd_hbuf == NULL) { |
| 558 | if (fp->f_flag & FNONBLOCK) { |
| 559 | if (d->bd_slen == 0) { |
| 560 | splx(s); |
| 561 | KERNEL_UNLOCK_ONE(NULL); |
| 562 | return (EWOULDBLOCK); |
| 563 | } |
| 564 | ROTATE_BUFFERS(d); |
| 565 | break; |
| 566 | } |
| 567 | |
| 568 | if ((d->bd_immediate || timed_out) && d->bd_slen != 0) { |
| 569 | /* |
| 570 | * A packet(s) either arrived since the previous |
| 571 | * read or arrived while we were asleep. |
| 572 | * Rotate the buffers and return what's here. |
| 573 | */ |
| 574 | ROTATE_BUFFERS(d); |
| 575 | break; |
| 576 | } |
| 577 | error = tsleep(d, PRINET|PCATCH, "bpf" , |
| 578 | d->bd_rtout); |
| 579 | if (error == EINTR || error == ERESTART) { |
| 580 | splx(s); |
| 581 | KERNEL_UNLOCK_ONE(NULL); |
| 582 | return (error); |
| 583 | } |
| 584 | if (error == EWOULDBLOCK) { |
| 585 | /* |
| 586 | * On a timeout, return what's in the buffer, |
| 587 | * which may be nothing. If there is something |
| 588 | * in the store buffer, we can rotate the buffers. |
| 589 | */ |
| 590 | if (d->bd_hbuf) |
| 591 | /* |
| 592 | * We filled up the buffer in between |
| 593 | * getting the timeout and arriving |
| 594 | * here, so we don't need to rotate. |
| 595 | */ |
| 596 | break; |
| 597 | |
| 598 | if (d->bd_slen == 0) { |
| 599 | splx(s); |
| 600 | KERNEL_UNLOCK_ONE(NULL); |
| 601 | return (0); |
| 602 | } |
| 603 | ROTATE_BUFFERS(d); |
| 604 | break; |
| 605 | } |
| 606 | if (error != 0) |
| 607 | goto done; |
| 608 | } |
| 609 | /* |
| 610 | * At this point, we know we have something in the hold slot. |
| 611 | */ |
| 612 | splx(s); |
| 613 | |
| 614 | /* |
| 615 | * Move data from hold buffer into user space. |
| 616 | * We know the entire buffer is transferred since |
| 617 | * we checked above that the read buffer is bpf_bufsize bytes. |
| 618 | */ |
| 619 | error = uiomove(d->bd_hbuf, d->bd_hlen, uio); |
| 620 | |
| 621 | s = splnet(); |
| 622 | d->bd_fbuf = d->bd_hbuf; |
| 623 | d->bd_hbuf = NULL; |
| 624 | d->bd_hlen = 0; |
| 625 | done: |
| 626 | splx(s); |
| 627 | KERNEL_UNLOCK_ONE(NULL); |
| 628 | return (error); |
| 629 | } |
| 630 | |
| 631 | |
| 632 | /* |
| 633 | * If there are processes sleeping on this descriptor, wake them up. |
| 634 | */ |
| 635 | static inline void |
| 636 | bpf_wakeup(struct bpf_d *d) |
| 637 | { |
| 638 | wakeup(d); |
| 639 | if (d->bd_async) |
| 640 | softint_schedule(d->bd_sih); |
| 641 | selnotify(&d->bd_sel, 0, 0); |
| 642 | } |
| 643 | |
| 644 | static void |
| 645 | bpf_softintr(void *cookie) |
| 646 | { |
| 647 | struct bpf_d *d; |
| 648 | |
| 649 | d = cookie; |
| 650 | if (d->bd_async) |
| 651 | fownsignal(d->bd_pgid, SIGIO, 0, 0, NULL); |
| 652 | } |
| 653 | |
| 654 | static void |
| 655 | bpf_timed_out(void *arg) |
| 656 | { |
| 657 | struct bpf_d *d = arg; |
| 658 | int s; |
| 659 | |
| 660 | s = splnet(); |
| 661 | if (d->bd_state == BPF_WAITING) { |
| 662 | d->bd_state = BPF_TIMED_OUT; |
| 663 | if (d->bd_slen != 0) |
| 664 | bpf_wakeup(d); |
| 665 | } |
| 666 | splx(s); |
| 667 | } |
| 668 | |
| 669 | |
| 670 | static int |
| 671 | bpf_write(struct file *fp, off_t *offp, struct uio *uio, |
| 672 | kauth_cred_t cred, int flags) |
| 673 | { |
| 674 | struct bpf_d *d = fp->f_bpf; |
| 675 | struct ifnet *ifp; |
| 676 | struct mbuf *m, *mc; |
| 677 | int error, s; |
| 678 | static struct sockaddr_storage dst; |
| 679 | |
| 680 | m = NULL; /* XXX gcc */ |
| 681 | |
| 682 | KERNEL_LOCK(1, NULL); |
| 683 | |
| 684 | if (d->bd_bif == NULL) { |
| 685 | KERNEL_UNLOCK_ONE(NULL); |
| 686 | return (ENXIO); |
| 687 | } |
| 688 | getnanotime(&d->bd_mtime); |
| 689 | |
| 690 | ifp = d->bd_bif->bif_ifp; |
| 691 | |
| 692 | if (uio->uio_resid == 0) { |
| 693 | KERNEL_UNLOCK_ONE(NULL); |
| 694 | return (0); |
| 695 | } |
| 696 | |
| 697 | error = bpf_movein(uio, (int)d->bd_bif->bif_dlt, ifp->if_mtu, &m, |
| 698 | (struct sockaddr *) &dst); |
| 699 | if (error) { |
| 700 | KERNEL_UNLOCK_ONE(NULL); |
| 701 | return (error); |
| 702 | } |
| 703 | |
| 704 | if (m->m_pkthdr.len > ifp->if_mtu) { |
| 705 | KERNEL_UNLOCK_ONE(NULL); |
| 706 | m_freem(m); |
| 707 | return (EMSGSIZE); |
| 708 | } |
| 709 | |
| 710 | if (d->bd_hdrcmplt) |
| 711 | dst.ss_family = pseudo_AF_HDRCMPLT; |
| 712 | |
| 713 | if (d->bd_feedback) { |
| 714 | mc = m_dup(m, 0, M_COPYALL, M_NOWAIT); |
| 715 | if (mc != NULL) |
| 716 | m_set_rcvif(mc, ifp); |
| 717 | /* Set M_PROMISC for outgoing packets to be discarded. */ |
| 718 | if (1 /*d->bd_direction == BPF_D_INOUT*/) |
| 719 | m->m_flags |= M_PROMISC; |
| 720 | } else |
| 721 | mc = NULL; |
| 722 | |
| 723 | s = splsoftnet(); |
| 724 | error = if_output_lock(ifp, ifp, m, (struct sockaddr *) &dst, NULL); |
| 725 | |
| 726 | if (mc != NULL) { |
| 727 | if (error == 0) |
| 728 | ifp->_if_input(ifp, mc); |
| 729 | else |
| 730 | m_freem(mc); |
| 731 | } |
| 732 | splx(s); |
| 733 | KERNEL_UNLOCK_ONE(NULL); |
| 734 | /* |
| 735 | * The driver frees the mbuf. |
| 736 | */ |
| 737 | return (error); |
| 738 | } |
| 739 | |
| 740 | /* |
| 741 | * Reset a descriptor by flushing its packet buffer and clearing the |
| 742 | * receive and drop counts. Should be called at splnet. |
| 743 | */ |
| 744 | static void |
| 745 | reset_d(struct bpf_d *d) |
| 746 | { |
| 747 | if (d->bd_hbuf) { |
| 748 | /* Free the hold buffer. */ |
| 749 | d->bd_fbuf = d->bd_hbuf; |
| 750 | d->bd_hbuf = NULL; |
| 751 | } |
| 752 | d->bd_slen = 0; |
| 753 | d->bd_hlen = 0; |
| 754 | d->bd_rcount = 0; |
| 755 | d->bd_dcount = 0; |
| 756 | d->bd_ccount = 0; |
| 757 | } |
| 758 | |
| 759 | /* |
| 760 | * FIONREAD Check for read packet available. |
| 761 | * BIOCGBLEN Get buffer len [for read()]. |
| 762 | * BIOCSETF Set ethernet read filter. |
| 763 | * BIOCFLUSH Flush read packet buffer. |
| 764 | * BIOCPROMISC Put interface into promiscuous mode. |
| 765 | * BIOCGDLT Get link layer type. |
| 766 | * BIOCGETIF Get interface name. |
| 767 | * BIOCSETIF Set interface. |
| 768 | * BIOCSRTIMEOUT Set read timeout. |
| 769 | * BIOCGRTIMEOUT Get read timeout. |
| 770 | * BIOCGSTATS Get packet stats. |
| 771 | * BIOCIMMEDIATE Set immediate mode. |
| 772 | * BIOCVERSION Get filter language version. |
| 773 | * BIOCGHDRCMPLT Get "header already complete" flag. |
| 774 | * BIOCSHDRCMPLT Set "header already complete" flag. |
| 775 | * BIOCSFEEDBACK Set packet feedback mode. |
| 776 | * BIOCGFEEDBACK Get packet feedback mode. |
| 777 | * BIOCGSEESENT Get "see sent packets" mode. |
| 778 | * BIOCSSEESENT Set "see sent packets" mode. |
| 779 | */ |
| 780 | /* ARGSUSED */ |
| 781 | static int |
| 782 | bpf_ioctl(struct file *fp, u_long cmd, void *addr) |
| 783 | { |
| 784 | struct bpf_d *d = fp->f_bpf; |
| 785 | int s, error = 0; |
| 786 | |
| 787 | /* |
| 788 | * Refresh the PID associated with this bpf file. |
| 789 | */ |
| 790 | KERNEL_LOCK(1, NULL); |
| 791 | d->bd_pid = curproc->p_pid; |
| 792 | #ifdef _LP64 |
| 793 | if (curproc->p_flag & PK_32) |
| 794 | d->bd_compat32 = 1; |
| 795 | else |
| 796 | d->bd_compat32 = 0; |
| 797 | #endif |
| 798 | |
| 799 | s = splnet(); |
| 800 | if (d->bd_state == BPF_WAITING) |
| 801 | callout_stop(&d->bd_callout); |
| 802 | d->bd_state = BPF_IDLE; |
| 803 | splx(s); |
| 804 | |
| 805 | switch (cmd) { |
| 806 | |
| 807 | default: |
| 808 | error = EINVAL; |
| 809 | break; |
| 810 | |
| 811 | /* |
| 812 | * Check for read packet available. |
| 813 | */ |
| 814 | case FIONREAD: |
| 815 | { |
| 816 | int n; |
| 817 | |
| 818 | s = splnet(); |
| 819 | n = d->bd_slen; |
| 820 | if (d->bd_hbuf) |
| 821 | n += d->bd_hlen; |
| 822 | splx(s); |
| 823 | |
| 824 | *(int *)addr = n; |
| 825 | break; |
| 826 | } |
| 827 | |
| 828 | /* |
| 829 | * Get buffer len [for read()]. |
| 830 | */ |
| 831 | case BIOCGBLEN: |
| 832 | *(u_int *)addr = d->bd_bufsize; |
| 833 | break; |
| 834 | |
| 835 | /* |
| 836 | * Set buffer length. |
| 837 | */ |
| 838 | case BIOCSBLEN: |
| 839 | if (d->bd_bif != NULL) |
| 840 | error = EINVAL; |
| 841 | else { |
| 842 | u_int size = *(u_int *)addr; |
| 843 | |
| 844 | if (size > bpf_maxbufsize) |
| 845 | *(u_int *)addr = size = bpf_maxbufsize; |
| 846 | else if (size < BPF_MINBUFSIZE) |
| 847 | *(u_int *)addr = size = BPF_MINBUFSIZE; |
| 848 | d->bd_bufsize = size; |
| 849 | } |
| 850 | break; |
| 851 | |
| 852 | /* |
| 853 | * Set link layer read filter. |
| 854 | */ |
| 855 | case BIOCSETF: |
| 856 | error = bpf_setf(d, addr); |
| 857 | break; |
| 858 | |
| 859 | /* |
| 860 | * Flush read packet buffer. |
| 861 | */ |
| 862 | case BIOCFLUSH: |
| 863 | s = splnet(); |
| 864 | reset_d(d); |
| 865 | splx(s); |
| 866 | break; |
| 867 | |
| 868 | /* |
| 869 | * Put interface into promiscuous mode. |
| 870 | */ |
| 871 | case BIOCPROMISC: |
| 872 | if (d->bd_bif == NULL) { |
| 873 | /* |
| 874 | * No interface attached yet. |
| 875 | */ |
| 876 | error = EINVAL; |
| 877 | break; |
| 878 | } |
| 879 | s = splnet(); |
| 880 | if (d->bd_promisc == 0) { |
| 881 | error = ifpromisc(d->bd_bif->bif_ifp, 1); |
| 882 | if (error == 0) |
| 883 | d->bd_promisc = 1; |
| 884 | } |
| 885 | splx(s); |
| 886 | break; |
| 887 | |
| 888 | /* |
| 889 | * Get device parameters. |
| 890 | */ |
| 891 | case BIOCGDLT: |
| 892 | if (d->bd_bif == NULL) |
| 893 | error = EINVAL; |
| 894 | else |
| 895 | *(u_int *)addr = d->bd_bif->bif_dlt; |
| 896 | break; |
| 897 | |
| 898 | /* |
| 899 | * Get a list of supported device parameters. |
| 900 | */ |
| 901 | case BIOCGDLTLIST: |
| 902 | if (d->bd_bif == NULL) |
| 903 | error = EINVAL; |
| 904 | else |
| 905 | error = bpf_getdltlist(d, addr); |
| 906 | break; |
| 907 | |
| 908 | /* |
| 909 | * Set device parameters. |
| 910 | */ |
| 911 | case BIOCSDLT: |
| 912 | mutex_enter(&bpf_mtx); |
| 913 | if (d->bd_bif == NULL) |
| 914 | error = EINVAL; |
| 915 | else |
| 916 | error = bpf_setdlt(d, *(u_int *)addr); |
| 917 | mutex_exit(&bpf_mtx); |
| 918 | break; |
| 919 | |
| 920 | /* |
| 921 | * Set interface name. |
| 922 | */ |
| 923 | #ifdef OBIOCGETIF |
| 924 | case OBIOCGETIF: |
| 925 | #endif |
| 926 | case BIOCGETIF: |
| 927 | if (d->bd_bif == NULL) |
| 928 | error = EINVAL; |
| 929 | else |
| 930 | bpf_ifname(d->bd_bif->bif_ifp, addr); |
| 931 | break; |
| 932 | |
| 933 | /* |
| 934 | * Set interface. |
| 935 | */ |
| 936 | #ifdef OBIOCSETIF |
| 937 | case OBIOCSETIF: |
| 938 | #endif |
| 939 | case BIOCSETIF: |
| 940 | mutex_enter(&bpf_mtx); |
| 941 | error = bpf_setif(d, addr); |
| 942 | mutex_exit(&bpf_mtx); |
| 943 | break; |
| 944 | |
| 945 | /* |
| 946 | * Set read timeout. |
| 947 | */ |
| 948 | case BIOCSRTIMEOUT: |
| 949 | { |
| 950 | struct timeval *tv = addr; |
| 951 | |
| 952 | /* Compute number of ticks. */ |
| 953 | d->bd_rtout = tv->tv_sec * hz + tv->tv_usec / tick; |
| 954 | if ((d->bd_rtout == 0) && (tv->tv_usec != 0)) |
| 955 | d->bd_rtout = 1; |
| 956 | break; |
| 957 | } |
| 958 | |
| 959 | #ifdef BIOCGORTIMEOUT |
| 960 | /* |
| 961 | * Get read timeout. |
| 962 | */ |
| 963 | case BIOCGORTIMEOUT: |
| 964 | { |
| 965 | struct timeval50 *tv = addr; |
| 966 | |
| 967 | tv->tv_sec = d->bd_rtout / hz; |
| 968 | tv->tv_usec = (d->bd_rtout % hz) * tick; |
| 969 | break; |
| 970 | } |
| 971 | #endif |
| 972 | |
| 973 | #ifdef BIOCSORTIMEOUT |
| 974 | /* |
| 975 | * Set read timeout. |
| 976 | */ |
| 977 | case BIOCSORTIMEOUT: |
| 978 | { |
| 979 | struct timeval50 *tv = addr; |
| 980 | |
| 981 | /* Compute number of ticks. */ |
| 982 | d->bd_rtout = tv->tv_sec * hz + tv->tv_usec / tick; |
| 983 | if ((d->bd_rtout == 0) && (tv->tv_usec != 0)) |
| 984 | d->bd_rtout = 1; |
| 985 | break; |
| 986 | } |
| 987 | #endif |
| 988 | |
| 989 | /* |
| 990 | * Get read timeout. |
| 991 | */ |
| 992 | case BIOCGRTIMEOUT: |
| 993 | { |
| 994 | struct timeval *tv = addr; |
| 995 | |
| 996 | tv->tv_sec = d->bd_rtout / hz; |
| 997 | tv->tv_usec = (d->bd_rtout % hz) * tick; |
| 998 | break; |
| 999 | } |
| 1000 | /* |
| 1001 | * Get packet stats. |
| 1002 | */ |
| 1003 | case BIOCGSTATS: |
| 1004 | { |
| 1005 | struct bpf_stat *bs = addr; |
| 1006 | |
| 1007 | bs->bs_recv = d->bd_rcount; |
| 1008 | bs->bs_drop = d->bd_dcount; |
| 1009 | bs->bs_capt = d->bd_ccount; |
| 1010 | break; |
| 1011 | } |
| 1012 | |
| 1013 | case BIOCGSTATSOLD: |
| 1014 | { |
| 1015 | struct bpf_stat_old *bs = addr; |
| 1016 | |
| 1017 | bs->bs_recv = d->bd_rcount; |
| 1018 | bs->bs_drop = d->bd_dcount; |
| 1019 | break; |
| 1020 | } |
| 1021 | |
| 1022 | /* |
| 1023 | * Set immediate mode. |
| 1024 | */ |
| 1025 | case BIOCIMMEDIATE: |
| 1026 | d->bd_immediate = *(u_int *)addr; |
| 1027 | break; |
| 1028 | |
| 1029 | case BIOCVERSION: |
| 1030 | { |
| 1031 | struct bpf_version *bv = addr; |
| 1032 | |
| 1033 | bv->bv_major = BPF_MAJOR_VERSION; |
| 1034 | bv->bv_minor = BPF_MINOR_VERSION; |
| 1035 | break; |
| 1036 | } |
| 1037 | |
| 1038 | case BIOCGHDRCMPLT: /* get "header already complete" flag */ |
| 1039 | *(u_int *)addr = d->bd_hdrcmplt; |
| 1040 | break; |
| 1041 | |
| 1042 | case BIOCSHDRCMPLT: /* set "header already complete" flag */ |
| 1043 | d->bd_hdrcmplt = *(u_int *)addr ? 1 : 0; |
| 1044 | break; |
| 1045 | |
| 1046 | /* |
| 1047 | * Get "see sent packets" flag |
| 1048 | */ |
| 1049 | case BIOCGSEESENT: |
| 1050 | *(u_int *)addr = d->bd_seesent; |
| 1051 | break; |
| 1052 | |
| 1053 | /* |
| 1054 | * Set "see sent" packets flag |
| 1055 | */ |
| 1056 | case BIOCSSEESENT: |
| 1057 | d->bd_seesent = *(u_int *)addr; |
| 1058 | break; |
| 1059 | |
| 1060 | /* |
| 1061 | * Set "feed packets from bpf back to input" mode |
| 1062 | */ |
| 1063 | case BIOCSFEEDBACK: |
| 1064 | d->bd_feedback = *(u_int *)addr; |
| 1065 | break; |
| 1066 | |
| 1067 | /* |
| 1068 | * Get "feed packets from bpf back to input" mode |
| 1069 | */ |
| 1070 | case BIOCGFEEDBACK: |
| 1071 | *(u_int *)addr = d->bd_feedback; |
| 1072 | break; |
| 1073 | |
| 1074 | case FIONBIO: /* Non-blocking I/O */ |
| 1075 | /* |
| 1076 | * No need to do anything special as we use IO_NDELAY in |
| 1077 | * bpfread() as an indication of whether or not to block |
| 1078 | * the read. |
| 1079 | */ |
| 1080 | break; |
| 1081 | |
| 1082 | case FIOASYNC: /* Send signal on receive packets */ |
| 1083 | d->bd_async = *(int *)addr; |
| 1084 | break; |
| 1085 | |
| 1086 | case TIOCSPGRP: /* Process or group to send signals to */ |
| 1087 | case FIOSETOWN: |
| 1088 | error = fsetown(&d->bd_pgid, cmd, addr); |
| 1089 | break; |
| 1090 | |
| 1091 | case TIOCGPGRP: |
| 1092 | case FIOGETOWN: |
| 1093 | error = fgetown(d->bd_pgid, cmd, addr); |
| 1094 | break; |
| 1095 | } |
| 1096 | KERNEL_UNLOCK_ONE(NULL); |
| 1097 | return (error); |
| 1098 | } |
| 1099 | |
| 1100 | /* |
| 1101 | * Set d's packet filter program to fp. If this file already has a filter, |
| 1102 | * free it and replace it. Returns EINVAL for bogus requests. |
| 1103 | */ |
| 1104 | int |
| 1105 | bpf_setf(struct bpf_d *d, struct bpf_program *fp) |
| 1106 | { |
| 1107 | struct bpf_insn *fcode, *old; |
| 1108 | bpfjit_func_t jcode, oldj; |
| 1109 | size_t flen, size; |
| 1110 | int s; |
| 1111 | |
| 1112 | jcode = NULL; |
| 1113 | flen = fp->bf_len; |
| 1114 | |
| 1115 | if ((fp->bf_insns == NULL && flen) || flen > BPF_MAXINSNS) { |
| 1116 | return EINVAL; |
| 1117 | } |
| 1118 | |
| 1119 | if (flen) { |
| 1120 | /* |
| 1121 | * Allocate the buffer, copy the byte-code from |
| 1122 | * userspace and validate it. |
| 1123 | */ |
| 1124 | size = flen * sizeof(*fp->bf_insns); |
| 1125 | fcode = malloc(size, M_DEVBUF, M_WAITOK); |
| 1126 | if (copyin(fp->bf_insns, fcode, size) != 0 || |
| 1127 | !bpf_validate(fcode, (int)flen)) { |
| 1128 | free(fcode, M_DEVBUF); |
| 1129 | return EINVAL; |
| 1130 | } |
| 1131 | membar_consumer(); |
| 1132 | if (bpf_jit) |
| 1133 | jcode = bpf_jit_generate(NULL, fcode, flen); |
| 1134 | } else { |
| 1135 | fcode = NULL; |
| 1136 | } |
| 1137 | |
| 1138 | s = splnet(); |
| 1139 | old = d->bd_filter; |
| 1140 | d->bd_filter = fcode; |
| 1141 | oldj = d->bd_jitcode; |
| 1142 | d->bd_jitcode = jcode; |
| 1143 | reset_d(d); |
| 1144 | splx(s); |
| 1145 | |
| 1146 | if (old) { |
| 1147 | free(old, M_DEVBUF); |
| 1148 | } |
| 1149 | if (oldj) { |
| 1150 | bpf_jit_freecode(oldj); |
| 1151 | } |
| 1152 | |
| 1153 | return 0; |
| 1154 | } |
| 1155 | |
| 1156 | /* |
| 1157 | * Detach a file from its current interface (if attached at all) and attach |
| 1158 | * to the interface indicated by the name stored in ifr. |
| 1159 | * Return an errno or 0. |
| 1160 | */ |
| 1161 | static int |
| 1162 | bpf_setif(struct bpf_d *d, struct ifreq *ifr) |
| 1163 | { |
| 1164 | struct bpf_if *bp; |
| 1165 | char *cp; |
| 1166 | int unit_seen, i, s, error; |
| 1167 | |
| 1168 | KASSERT(mutex_owned(&bpf_mtx)); |
| 1169 | /* |
| 1170 | * Make sure the provided name has a unit number, and default |
| 1171 | * it to '0' if not specified. |
| 1172 | * XXX This is ugly ... do this differently? |
| 1173 | */ |
| 1174 | unit_seen = 0; |
| 1175 | cp = ifr->ifr_name; |
| 1176 | cp[sizeof(ifr->ifr_name) - 1] = '\0'; /* sanity */ |
| 1177 | while (*cp++) |
| 1178 | if (*cp >= '0' && *cp <= '9') |
| 1179 | unit_seen = 1; |
| 1180 | if (!unit_seen) { |
| 1181 | /* Make sure to leave room for the '\0'. */ |
| 1182 | for (i = 0; i < (IFNAMSIZ - 1); ++i) { |
| 1183 | if ((ifr->ifr_name[i] >= 'a' && |
| 1184 | ifr->ifr_name[i] <= 'z') || |
| 1185 | (ifr->ifr_name[i] >= 'A' && |
| 1186 | ifr->ifr_name[i] <= 'Z')) |
| 1187 | continue; |
| 1188 | ifr->ifr_name[i] = '0'; |
| 1189 | } |
| 1190 | } |
| 1191 | |
| 1192 | /* |
| 1193 | * Look through attached interfaces for the named one. |
| 1194 | */ |
| 1195 | for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) { |
| 1196 | struct ifnet *ifp = bp->bif_ifp; |
| 1197 | |
| 1198 | if (ifp == NULL || |
| 1199 | strcmp(ifp->if_xname, ifr->ifr_name) != 0) |
| 1200 | continue; |
| 1201 | /* skip additional entry */ |
| 1202 | if (bp->bif_driverp != &ifp->if_bpf) |
| 1203 | continue; |
| 1204 | /* |
| 1205 | * We found the requested interface. |
| 1206 | * Allocate the packet buffers if we need to. |
| 1207 | * If we're already attached to requested interface, |
| 1208 | * just flush the buffer. |
| 1209 | */ |
| 1210 | if (d->bd_sbuf == NULL) { |
| 1211 | error = bpf_allocbufs(d); |
| 1212 | if (error != 0) |
| 1213 | return (error); |
| 1214 | } |
| 1215 | s = splnet(); |
| 1216 | if (bp != d->bd_bif) { |
| 1217 | if (d->bd_bif) |
| 1218 | /* |
| 1219 | * Detach if attached to something else. |
| 1220 | */ |
| 1221 | bpf_detachd(d); |
| 1222 | |
| 1223 | bpf_attachd(d, bp); |
| 1224 | } |
| 1225 | reset_d(d); |
| 1226 | splx(s); |
| 1227 | return (0); |
| 1228 | } |
| 1229 | /* Not found. */ |
| 1230 | return (ENXIO); |
| 1231 | } |
| 1232 | |
| 1233 | /* |
| 1234 | * Copy the interface name to the ifreq. |
| 1235 | */ |
| 1236 | static void |
| 1237 | bpf_ifname(struct ifnet *ifp, struct ifreq *ifr) |
| 1238 | { |
| 1239 | memcpy(ifr->ifr_name, ifp->if_xname, IFNAMSIZ); |
| 1240 | } |
| 1241 | |
| 1242 | static int |
| 1243 | bpf_stat(struct file *fp, struct stat *st) |
| 1244 | { |
| 1245 | struct bpf_d *d = fp->f_bpf; |
| 1246 | |
| 1247 | (void)memset(st, 0, sizeof(*st)); |
| 1248 | KERNEL_LOCK(1, NULL); |
| 1249 | st->st_dev = makedev(cdevsw_lookup_major(&bpf_cdevsw), d->bd_pid); |
| 1250 | st->st_atimespec = d->bd_atime; |
| 1251 | st->st_mtimespec = d->bd_mtime; |
| 1252 | st->st_ctimespec = st->st_birthtimespec = d->bd_btime; |
| 1253 | st->st_uid = kauth_cred_geteuid(fp->f_cred); |
| 1254 | st->st_gid = kauth_cred_getegid(fp->f_cred); |
| 1255 | st->st_mode = S_IFCHR; |
| 1256 | KERNEL_UNLOCK_ONE(NULL); |
| 1257 | return 0; |
| 1258 | } |
| 1259 | |
| 1260 | /* |
| 1261 | * Support for poll() system call |
| 1262 | * |
| 1263 | * Return true iff the specific operation will not block indefinitely - with |
| 1264 | * the assumption that it is safe to positively acknowledge a request for the |
| 1265 | * ability to write to the BPF device. |
| 1266 | * Otherwise, return false but make a note that a selnotify() must be done. |
| 1267 | */ |
| 1268 | static int |
| 1269 | bpf_poll(struct file *fp, int events) |
| 1270 | { |
| 1271 | struct bpf_d *d = fp->f_bpf; |
| 1272 | int s = splnet(); |
| 1273 | int revents; |
| 1274 | |
| 1275 | /* |
| 1276 | * Refresh the PID associated with this bpf file. |
| 1277 | */ |
| 1278 | KERNEL_LOCK(1, NULL); |
| 1279 | d->bd_pid = curproc->p_pid; |
| 1280 | |
| 1281 | revents = events & (POLLOUT | POLLWRNORM); |
| 1282 | if (events & (POLLIN | POLLRDNORM)) { |
| 1283 | /* |
| 1284 | * An imitation of the FIONREAD ioctl code. |
| 1285 | */ |
| 1286 | if (d->bd_hlen != 0 || |
| 1287 | ((d->bd_immediate || d->bd_state == BPF_TIMED_OUT) && |
| 1288 | d->bd_slen != 0)) { |
| 1289 | revents |= events & (POLLIN | POLLRDNORM); |
| 1290 | } else { |
| 1291 | selrecord(curlwp, &d->bd_sel); |
| 1292 | /* Start the read timeout if necessary */ |
| 1293 | if (d->bd_rtout > 0 && d->bd_state == BPF_IDLE) { |
| 1294 | callout_reset(&d->bd_callout, d->bd_rtout, |
| 1295 | bpf_timed_out, d); |
| 1296 | d->bd_state = BPF_WAITING; |
| 1297 | } |
| 1298 | } |
| 1299 | } |
| 1300 | |
| 1301 | KERNEL_UNLOCK_ONE(NULL); |
| 1302 | splx(s); |
| 1303 | return (revents); |
| 1304 | } |
| 1305 | |
| 1306 | static void |
| 1307 | filt_bpfrdetach(struct knote *kn) |
| 1308 | { |
| 1309 | struct bpf_d *d = kn->kn_hook; |
| 1310 | int s; |
| 1311 | |
| 1312 | KERNEL_LOCK(1, NULL); |
| 1313 | s = splnet(); |
| 1314 | SLIST_REMOVE(&d->bd_sel.sel_klist, kn, knote, kn_selnext); |
| 1315 | splx(s); |
| 1316 | KERNEL_UNLOCK_ONE(NULL); |
| 1317 | } |
| 1318 | |
| 1319 | static int |
| 1320 | filt_bpfread(struct knote *kn, long hint) |
| 1321 | { |
| 1322 | struct bpf_d *d = kn->kn_hook; |
| 1323 | int rv; |
| 1324 | |
| 1325 | KERNEL_LOCK(1, NULL); |
| 1326 | kn->kn_data = d->bd_hlen; |
| 1327 | if (d->bd_immediate) |
| 1328 | kn->kn_data += d->bd_slen; |
| 1329 | rv = (kn->kn_data > 0); |
| 1330 | KERNEL_UNLOCK_ONE(NULL); |
| 1331 | return rv; |
| 1332 | } |
| 1333 | |
| 1334 | static const struct filterops bpfread_filtops = |
| 1335 | { 1, NULL, filt_bpfrdetach, filt_bpfread }; |
| 1336 | |
| 1337 | static int |
| 1338 | bpf_kqfilter(struct file *fp, struct knote *kn) |
| 1339 | { |
| 1340 | struct bpf_d *d = fp->f_bpf; |
| 1341 | struct klist *klist; |
| 1342 | int s; |
| 1343 | |
| 1344 | KERNEL_LOCK(1, NULL); |
| 1345 | |
| 1346 | switch (kn->kn_filter) { |
| 1347 | case EVFILT_READ: |
| 1348 | klist = &d->bd_sel.sel_klist; |
| 1349 | kn->kn_fop = &bpfread_filtops; |
| 1350 | break; |
| 1351 | |
| 1352 | default: |
| 1353 | KERNEL_UNLOCK_ONE(NULL); |
| 1354 | return (EINVAL); |
| 1355 | } |
| 1356 | |
| 1357 | kn->kn_hook = d; |
| 1358 | |
| 1359 | s = splnet(); |
| 1360 | SLIST_INSERT_HEAD(klist, kn, kn_selnext); |
| 1361 | splx(s); |
| 1362 | KERNEL_UNLOCK_ONE(NULL); |
| 1363 | |
| 1364 | return (0); |
| 1365 | } |
| 1366 | |
| 1367 | /* |
| 1368 | * Copy data from an mbuf chain into a buffer. This code is derived |
| 1369 | * from m_copydata in sys/uipc_mbuf.c. |
| 1370 | */ |
| 1371 | static void * |
| 1372 | bpf_mcpy(void *dst_arg, const void *src_arg, size_t len) |
| 1373 | { |
| 1374 | const struct mbuf *m; |
| 1375 | u_int count; |
| 1376 | u_char *dst; |
| 1377 | |
| 1378 | m = src_arg; |
| 1379 | dst = dst_arg; |
| 1380 | while (len > 0) { |
| 1381 | if (m == NULL) |
| 1382 | panic("bpf_mcpy" ); |
| 1383 | count = min(m->m_len, len); |
| 1384 | memcpy(dst, mtod(m, const void *), count); |
| 1385 | m = m->m_next; |
| 1386 | dst += count; |
| 1387 | len -= count; |
| 1388 | } |
| 1389 | return dst_arg; |
| 1390 | } |
| 1391 | |
| 1392 | /* |
| 1393 | * Dispatch a packet to all the listeners on interface bp. |
| 1394 | * |
| 1395 | * pkt pointer to the packet, either a data buffer or an mbuf chain |
| 1396 | * buflen buffer length, if pkt is a data buffer |
| 1397 | * cpfn a function that can copy pkt into the listener's buffer |
| 1398 | * pktlen length of the packet |
| 1399 | * rcv true if packet came in |
| 1400 | */ |
| 1401 | static inline void |
| 1402 | bpf_deliver(struct bpf_if *bp, void *(*cpfn)(void *, const void *, size_t), |
| 1403 | void *pkt, u_int pktlen, u_int buflen, const bool rcv) |
| 1404 | { |
| 1405 | uint32_t mem[BPF_MEMWORDS]; |
| 1406 | bpf_args_t args = { |
| 1407 | .pkt = (const uint8_t *)pkt, |
| 1408 | .wirelen = pktlen, |
| 1409 | .buflen = buflen, |
| 1410 | .mem = mem, |
| 1411 | .arg = NULL |
| 1412 | }; |
| 1413 | bool gottime = false; |
| 1414 | struct timespec ts; |
| 1415 | |
| 1416 | /* |
| 1417 | * Note that the IPL does not have to be raised at this point. |
| 1418 | * The only problem that could arise here is that if two different |
| 1419 | * interfaces shared any data. This is not the case. |
| 1420 | */ |
| 1421 | for (struct bpf_d *d = bp->bif_dlist; d != NULL; d = d->bd_next) { |
| 1422 | u_int slen; |
| 1423 | |
| 1424 | if (!d->bd_seesent && !rcv) { |
| 1425 | continue; |
| 1426 | } |
| 1427 | d->bd_rcount++; |
| 1428 | bpf_gstats.bs_recv++; |
| 1429 | |
| 1430 | if (d->bd_jitcode) |
| 1431 | slen = d->bd_jitcode(NULL, &args); |
| 1432 | else |
| 1433 | slen = bpf_filter_ext(NULL, d->bd_filter, &args); |
| 1434 | |
| 1435 | if (!slen) { |
| 1436 | continue; |
| 1437 | } |
| 1438 | if (!gottime) { |
| 1439 | gottime = true; |
| 1440 | nanotime(&ts); |
| 1441 | } |
| 1442 | catchpacket(d, pkt, pktlen, slen, cpfn, &ts); |
| 1443 | } |
| 1444 | } |
| 1445 | |
| 1446 | /* |
| 1447 | * Incoming linkage from device drivers. Process the packet pkt, of length |
| 1448 | * pktlen, which is stored in a contiguous buffer. The packet is parsed |
| 1449 | * by each process' filter, and if accepted, stashed into the corresponding |
| 1450 | * buffer. |
| 1451 | */ |
| 1452 | static void |
| 1453 | _bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen) |
| 1454 | { |
| 1455 | |
| 1456 | bpf_deliver(bp, memcpy, pkt, pktlen, pktlen, true); |
| 1457 | } |
| 1458 | |
| 1459 | /* |
| 1460 | * Incoming linkage from device drivers, when the head of the packet is in |
| 1461 | * a buffer, and the tail is in an mbuf chain. |
| 1462 | */ |
| 1463 | static void |
| 1464 | _bpf_mtap2(struct bpf_if *bp, void *data, u_int dlen, struct mbuf *m) |
| 1465 | { |
| 1466 | u_int pktlen; |
| 1467 | struct mbuf mb; |
| 1468 | |
| 1469 | /* Skip outgoing duplicate packets. */ |
| 1470 | if ((m->m_flags & M_PROMISC) != 0 && m->m_pkthdr.rcvif_index == 0) { |
| 1471 | m->m_flags &= ~M_PROMISC; |
| 1472 | return; |
| 1473 | } |
| 1474 | |
| 1475 | pktlen = m_length(m) + dlen; |
| 1476 | |
| 1477 | /* |
| 1478 | * Craft on-stack mbuf suitable for passing to bpf_filter. |
| 1479 | * Note that we cut corners here; we only setup what's |
| 1480 | * absolutely needed--this mbuf should never go anywhere else. |
| 1481 | */ |
| 1482 | (void)memset(&mb, 0, sizeof(mb)); |
| 1483 | mb.m_next = m; |
| 1484 | mb.m_data = data; |
| 1485 | mb.m_len = dlen; |
| 1486 | |
| 1487 | bpf_deliver(bp, bpf_mcpy, &mb, pktlen, 0, m->m_pkthdr.rcvif_index != 0); |
| 1488 | } |
| 1489 | |
| 1490 | /* |
| 1491 | * Incoming linkage from device drivers, when packet is in an mbuf chain. |
| 1492 | */ |
| 1493 | static void |
| 1494 | _bpf_mtap(struct bpf_if *bp, struct mbuf *m) |
| 1495 | { |
| 1496 | void *(*cpfn)(void *, const void *, size_t); |
| 1497 | u_int pktlen, buflen; |
| 1498 | void *marg; |
| 1499 | |
| 1500 | /* Skip outgoing duplicate packets. */ |
| 1501 | if ((m->m_flags & M_PROMISC) != 0 && m->m_pkthdr.rcvif_index == 0) { |
| 1502 | m->m_flags &= ~M_PROMISC; |
| 1503 | return; |
| 1504 | } |
| 1505 | |
| 1506 | pktlen = m_length(m); |
| 1507 | |
| 1508 | if (pktlen == m->m_len) { |
| 1509 | cpfn = (void *)memcpy; |
| 1510 | marg = mtod(m, void *); |
| 1511 | buflen = pktlen; |
| 1512 | } else { |
| 1513 | cpfn = bpf_mcpy; |
| 1514 | marg = m; |
| 1515 | buflen = 0; |
| 1516 | } |
| 1517 | |
| 1518 | bpf_deliver(bp, cpfn, marg, pktlen, buflen, m->m_pkthdr.rcvif_index != 0); |
| 1519 | } |
| 1520 | |
| 1521 | /* |
| 1522 | * We need to prepend the address family as |
| 1523 | * a four byte field. Cons up a dummy header |
| 1524 | * to pacify bpf. This is safe because bpf |
| 1525 | * will only read from the mbuf (i.e., it won't |
| 1526 | * try to free it or keep a pointer a to it). |
| 1527 | */ |
| 1528 | static void |
| 1529 | _bpf_mtap_af(struct bpf_if *bp, uint32_t af, struct mbuf *m) |
| 1530 | { |
| 1531 | struct mbuf m0; |
| 1532 | |
| 1533 | m0.m_flags = 0; |
| 1534 | m0.m_next = m; |
| 1535 | m0.m_len = 4; |
| 1536 | m0.m_data = (char *)⁡ |
| 1537 | |
| 1538 | _bpf_mtap(bp, &m0); |
| 1539 | } |
| 1540 | |
| 1541 | /* |
| 1542 | * Put the SLIP pseudo-"link header" in place. |
| 1543 | * Note this M_PREPEND() should never fail, |
| 1544 | * swince we know we always have enough space |
| 1545 | * in the input buffer. |
| 1546 | */ |
| 1547 | static void |
| 1548 | _bpf_mtap_sl_in(struct bpf_if *bp, u_char *chdr, struct mbuf **m) |
| 1549 | { |
| 1550 | int s; |
| 1551 | u_char *hp; |
| 1552 | |
| 1553 | M_PREPEND(*m, SLIP_HDRLEN, M_DONTWAIT); |
| 1554 | if (*m == NULL) |
| 1555 | return; |
| 1556 | |
| 1557 | hp = mtod(*m, u_char *); |
| 1558 | hp[SLX_DIR] = SLIPDIR_IN; |
| 1559 | (void)memcpy(&hp[SLX_CHDR], chdr, CHDR_LEN); |
| 1560 | |
| 1561 | s = splnet(); |
| 1562 | _bpf_mtap(bp, *m); |
| 1563 | splx(s); |
| 1564 | |
| 1565 | m_adj(*m, SLIP_HDRLEN); |
| 1566 | } |
| 1567 | |
| 1568 | /* |
| 1569 | * Put the SLIP pseudo-"link header" in |
| 1570 | * place. The compressed header is now |
| 1571 | * at the beginning of the mbuf. |
| 1572 | */ |
| 1573 | static void |
| 1574 | _bpf_mtap_sl_out(struct bpf_if *bp, u_char *chdr, struct mbuf *m) |
| 1575 | { |
| 1576 | struct mbuf m0; |
| 1577 | u_char *hp; |
| 1578 | int s; |
| 1579 | |
| 1580 | m0.m_flags = 0; |
| 1581 | m0.m_next = m; |
| 1582 | m0.m_data = m0.m_dat; |
| 1583 | m0.m_len = SLIP_HDRLEN; |
| 1584 | |
| 1585 | hp = mtod(&m0, u_char *); |
| 1586 | |
| 1587 | hp[SLX_DIR] = SLIPDIR_OUT; |
| 1588 | (void)memcpy(&hp[SLX_CHDR], chdr, CHDR_LEN); |
| 1589 | |
| 1590 | s = splnet(); |
| 1591 | _bpf_mtap(bp, &m0); |
| 1592 | splx(s); |
| 1593 | m_freem(m); |
| 1594 | } |
| 1595 | |
| 1596 | static int |
| 1597 | bpf_hdrlen(struct bpf_d *d) |
| 1598 | { |
| 1599 | int hdrlen = d->bd_bif->bif_hdrlen; |
| 1600 | /* |
| 1601 | * Compute the length of the bpf header. This is not necessarily |
| 1602 | * equal to SIZEOF_BPF_HDR because we want to insert spacing such |
| 1603 | * that the network layer header begins on a longword boundary (for |
| 1604 | * performance reasons and to alleviate alignment restrictions). |
| 1605 | */ |
| 1606 | #ifdef _LP64 |
| 1607 | if (d->bd_compat32) |
| 1608 | return (BPF_WORDALIGN32(hdrlen + SIZEOF_BPF_HDR32) - hdrlen); |
| 1609 | else |
| 1610 | #endif |
| 1611 | return (BPF_WORDALIGN(hdrlen + SIZEOF_BPF_HDR) - hdrlen); |
| 1612 | } |
| 1613 | |
| 1614 | /* |
| 1615 | * Move the packet data from interface memory (pkt) into the |
| 1616 | * store buffer. Call the wakeup functions if it's time to wakeup |
| 1617 | * a listener (buffer full), "cpfn" is the routine called to do the |
| 1618 | * actual data transfer. memcpy is passed in to copy contiguous chunks, |
| 1619 | * while bpf_mcpy is passed in to copy mbuf chains. In the latter case, |
| 1620 | * pkt is really an mbuf. |
| 1621 | */ |
| 1622 | static void |
| 1623 | catchpacket(struct bpf_d *d, u_char *pkt, u_int pktlen, u_int snaplen, |
| 1624 | void *(*cpfn)(void *, const void *, size_t), struct timespec *ts) |
| 1625 | { |
| 1626 | char *h; |
| 1627 | int totlen, curlen, caplen; |
| 1628 | int hdrlen = bpf_hdrlen(d); |
| 1629 | int do_wakeup = 0; |
| 1630 | |
| 1631 | ++d->bd_ccount; |
| 1632 | ++bpf_gstats.bs_capt; |
| 1633 | /* |
| 1634 | * Figure out how many bytes to move. If the packet is |
| 1635 | * greater or equal to the snapshot length, transfer that |
| 1636 | * much. Otherwise, transfer the whole packet (unless |
| 1637 | * we hit the buffer size limit). |
| 1638 | */ |
| 1639 | totlen = hdrlen + min(snaplen, pktlen); |
| 1640 | if (totlen > d->bd_bufsize) |
| 1641 | totlen = d->bd_bufsize; |
| 1642 | /* |
| 1643 | * If we adjusted totlen to fit the bufsize, it could be that |
| 1644 | * totlen is smaller than hdrlen because of the link layer header. |
| 1645 | */ |
| 1646 | caplen = totlen - hdrlen; |
| 1647 | if (caplen < 0) |
| 1648 | caplen = 0; |
| 1649 | |
| 1650 | /* |
| 1651 | * Round up the end of the previous packet to the next longword. |
| 1652 | */ |
| 1653 | #ifdef _LP64 |
| 1654 | if (d->bd_compat32) |
| 1655 | curlen = BPF_WORDALIGN32(d->bd_slen); |
| 1656 | else |
| 1657 | #endif |
| 1658 | curlen = BPF_WORDALIGN(d->bd_slen); |
| 1659 | if (curlen + totlen > d->bd_bufsize) { |
| 1660 | /* |
| 1661 | * This packet will overflow the storage buffer. |
| 1662 | * Rotate the buffers if we can, then wakeup any |
| 1663 | * pending reads. |
| 1664 | */ |
| 1665 | if (d->bd_fbuf == NULL) { |
| 1666 | /* |
| 1667 | * We haven't completed the previous read yet, |
| 1668 | * so drop the packet. |
| 1669 | */ |
| 1670 | ++d->bd_dcount; |
| 1671 | ++bpf_gstats.bs_drop; |
| 1672 | return; |
| 1673 | } |
| 1674 | ROTATE_BUFFERS(d); |
| 1675 | do_wakeup = 1; |
| 1676 | curlen = 0; |
| 1677 | } else if (d->bd_immediate || d->bd_state == BPF_TIMED_OUT) { |
| 1678 | /* |
| 1679 | * Immediate mode is set, or the read timeout has |
| 1680 | * already expired during a select call. A packet |
| 1681 | * arrived, so the reader should be woken up. |
| 1682 | */ |
| 1683 | do_wakeup = 1; |
| 1684 | } |
| 1685 | |
| 1686 | /* |
| 1687 | * Append the bpf header. |
| 1688 | */ |
| 1689 | h = (char *)d->bd_sbuf + curlen; |
| 1690 | #ifdef _LP64 |
| 1691 | if (d->bd_compat32) { |
| 1692 | struct bpf_hdr32 *hp32; |
| 1693 | |
| 1694 | hp32 = (struct bpf_hdr32 *)h; |
| 1695 | hp32->bh_tstamp.tv_sec = ts->tv_sec; |
| 1696 | hp32->bh_tstamp.tv_usec = ts->tv_nsec / 1000; |
| 1697 | hp32->bh_datalen = pktlen; |
| 1698 | hp32->bh_hdrlen = hdrlen; |
| 1699 | hp32->bh_caplen = caplen; |
| 1700 | } else |
| 1701 | #endif |
| 1702 | { |
| 1703 | struct bpf_hdr *hp; |
| 1704 | |
| 1705 | hp = (struct bpf_hdr *)h; |
| 1706 | hp->bh_tstamp.tv_sec = ts->tv_sec; |
| 1707 | hp->bh_tstamp.tv_usec = ts->tv_nsec / 1000; |
| 1708 | hp->bh_datalen = pktlen; |
| 1709 | hp->bh_hdrlen = hdrlen; |
| 1710 | hp->bh_caplen = caplen; |
| 1711 | } |
| 1712 | |
| 1713 | /* |
| 1714 | * Copy the packet data into the store buffer and update its length. |
| 1715 | */ |
| 1716 | (*cpfn)(h + hdrlen, pkt, caplen); |
| 1717 | d->bd_slen = curlen + totlen; |
| 1718 | |
| 1719 | /* |
| 1720 | * Call bpf_wakeup after bd_slen has been updated so that kevent(2) |
| 1721 | * will cause filt_bpfread() to be called with it adjusted. |
| 1722 | */ |
| 1723 | if (do_wakeup) |
| 1724 | bpf_wakeup(d); |
| 1725 | } |
| 1726 | |
| 1727 | /* |
| 1728 | * Initialize all nonzero fields of a descriptor. |
| 1729 | */ |
| 1730 | static int |
| 1731 | bpf_allocbufs(struct bpf_d *d) |
| 1732 | { |
| 1733 | |
| 1734 | d->bd_fbuf = malloc(d->bd_bufsize, M_DEVBUF, M_NOWAIT); |
| 1735 | if (!d->bd_fbuf) |
| 1736 | return (ENOBUFS); |
| 1737 | d->bd_sbuf = malloc(d->bd_bufsize, M_DEVBUF, M_NOWAIT); |
| 1738 | if (!d->bd_sbuf) { |
| 1739 | free(d->bd_fbuf, M_DEVBUF); |
| 1740 | return (ENOBUFS); |
| 1741 | } |
| 1742 | d->bd_slen = 0; |
| 1743 | d->bd_hlen = 0; |
| 1744 | return (0); |
| 1745 | } |
| 1746 | |
| 1747 | /* |
| 1748 | * Free buffers currently in use by a descriptor. |
| 1749 | * Called on close. |
| 1750 | */ |
| 1751 | static void |
| 1752 | bpf_freed(struct bpf_d *d) |
| 1753 | { |
| 1754 | /* |
| 1755 | * We don't need to lock out interrupts since this descriptor has |
| 1756 | * been detached from its interface and it yet hasn't been marked |
| 1757 | * free. |
| 1758 | */ |
| 1759 | if (d->bd_sbuf != NULL) { |
| 1760 | free(d->bd_sbuf, M_DEVBUF); |
| 1761 | if (d->bd_hbuf != NULL) |
| 1762 | free(d->bd_hbuf, M_DEVBUF); |
| 1763 | if (d->bd_fbuf != NULL) |
| 1764 | free(d->bd_fbuf, M_DEVBUF); |
| 1765 | } |
| 1766 | if (d->bd_filter) |
| 1767 | free(d->bd_filter, M_DEVBUF); |
| 1768 | |
| 1769 | if (d->bd_jitcode != NULL) { |
| 1770 | bpf_jit_freecode(d->bd_jitcode); |
| 1771 | } |
| 1772 | } |
| 1773 | |
| 1774 | /* |
| 1775 | * Attach an interface to bpf. dlt is the link layer type; |
| 1776 | * hdrlen is the fixed size of the link header for the specified dlt |
| 1777 | * (variable length headers not yet supported). |
| 1778 | */ |
| 1779 | static void |
| 1780 | _bpfattach(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp) |
| 1781 | { |
| 1782 | struct bpf_if *bp; |
| 1783 | bp = malloc(sizeof(*bp), M_DEVBUF, M_DONTWAIT); |
| 1784 | if (bp == NULL) |
| 1785 | panic("bpfattach" ); |
| 1786 | |
| 1787 | mutex_enter(&bpf_mtx); |
| 1788 | bp->bif_dlist = NULL; |
| 1789 | bp->bif_driverp = driverp; |
| 1790 | bp->bif_ifp = ifp; |
| 1791 | bp->bif_dlt = dlt; |
| 1792 | |
| 1793 | bp->bif_next = bpf_iflist; |
| 1794 | bpf_iflist = bp; |
| 1795 | |
| 1796 | *bp->bif_driverp = NULL; |
| 1797 | |
| 1798 | bp->bif_hdrlen = hdrlen; |
| 1799 | mutex_exit(&bpf_mtx); |
| 1800 | #if 0 |
| 1801 | printf("bpf: %s attached\n" , ifp->if_xname); |
| 1802 | #endif |
| 1803 | } |
| 1804 | |
| 1805 | /* |
| 1806 | * Remove an interface from bpf. |
| 1807 | */ |
| 1808 | static void |
| 1809 | _bpfdetach(struct ifnet *ifp) |
| 1810 | { |
| 1811 | struct bpf_if *bp, **pbp; |
| 1812 | struct bpf_d *d; |
| 1813 | int s; |
| 1814 | |
| 1815 | mutex_enter(&bpf_mtx); |
| 1816 | /* Nuke the vnodes for any open instances */ |
| 1817 | LIST_FOREACH(d, &bpf_list, bd_list) { |
| 1818 | if (d->bd_bif != NULL && d->bd_bif->bif_ifp == ifp) { |
| 1819 | /* |
| 1820 | * Detach the descriptor from an interface now. |
| 1821 | * It will be free'ed later by close routine. |
| 1822 | */ |
| 1823 | s = splnet(); |
| 1824 | d->bd_promisc = 0; /* we can't touch device. */ |
| 1825 | bpf_detachd(d); |
| 1826 | splx(s); |
| 1827 | } |
| 1828 | } |
| 1829 | |
| 1830 | again: |
| 1831 | for (bp = bpf_iflist, pbp = &bpf_iflist; |
| 1832 | bp != NULL; pbp = &bp->bif_next, bp = bp->bif_next) { |
| 1833 | if (bp->bif_ifp == ifp) { |
| 1834 | *pbp = bp->bif_next; |
| 1835 | free(bp, M_DEVBUF); |
| 1836 | goto again; |
| 1837 | } |
| 1838 | } |
| 1839 | mutex_exit(&bpf_mtx); |
| 1840 | } |
| 1841 | |
| 1842 | /* |
| 1843 | * Change the data link type of a interface. |
| 1844 | */ |
| 1845 | static void |
| 1846 | _bpf_change_type(struct ifnet *ifp, u_int dlt, u_int hdrlen) |
| 1847 | { |
| 1848 | struct bpf_if *bp; |
| 1849 | |
| 1850 | for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) { |
| 1851 | if (bp->bif_driverp == &ifp->if_bpf) |
| 1852 | break; |
| 1853 | } |
| 1854 | if (bp == NULL) |
| 1855 | panic("bpf_change_type" ); |
| 1856 | |
| 1857 | bp->bif_dlt = dlt; |
| 1858 | |
| 1859 | bp->bif_hdrlen = hdrlen; |
| 1860 | } |
| 1861 | |
| 1862 | /* |
| 1863 | * Get a list of available data link type of the interface. |
| 1864 | */ |
| 1865 | static int |
| 1866 | bpf_getdltlist(struct bpf_d *d, struct bpf_dltlist *bfl) |
| 1867 | { |
| 1868 | int n, error; |
| 1869 | struct ifnet *ifp; |
| 1870 | struct bpf_if *bp; |
| 1871 | |
| 1872 | ifp = d->bd_bif->bif_ifp; |
| 1873 | n = 0; |
| 1874 | error = 0; |
| 1875 | for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) { |
| 1876 | if (bp->bif_ifp != ifp) |
| 1877 | continue; |
| 1878 | if (bfl->bfl_list != NULL) { |
| 1879 | if (n >= bfl->bfl_len) |
| 1880 | return ENOMEM; |
| 1881 | error = copyout(&bp->bif_dlt, |
| 1882 | bfl->bfl_list + n, sizeof(u_int)); |
| 1883 | } |
| 1884 | n++; |
| 1885 | } |
| 1886 | bfl->bfl_len = n; |
| 1887 | return error; |
| 1888 | } |
| 1889 | |
| 1890 | /* |
| 1891 | * Set the data link type of a BPF instance. |
| 1892 | */ |
| 1893 | static int |
| 1894 | bpf_setdlt(struct bpf_d *d, u_int dlt) |
| 1895 | { |
| 1896 | int s, error, opromisc; |
| 1897 | struct ifnet *ifp; |
| 1898 | struct bpf_if *bp; |
| 1899 | |
| 1900 | KASSERT(mutex_owned(&bpf_mtx)); |
| 1901 | |
| 1902 | if (d->bd_bif->bif_dlt == dlt) |
| 1903 | return 0; |
| 1904 | ifp = d->bd_bif->bif_ifp; |
| 1905 | for (bp = bpf_iflist; bp != NULL; bp = bp->bif_next) { |
| 1906 | if (bp->bif_ifp == ifp && bp->bif_dlt == dlt) |
| 1907 | break; |
| 1908 | } |
| 1909 | if (bp == NULL) |
| 1910 | return EINVAL; |
| 1911 | s = splnet(); |
| 1912 | opromisc = d->bd_promisc; |
| 1913 | bpf_detachd(d); |
| 1914 | bpf_attachd(d, bp); |
| 1915 | reset_d(d); |
| 1916 | if (opromisc) { |
| 1917 | error = ifpromisc(bp->bif_ifp, 1); |
| 1918 | if (error) |
| 1919 | printf("%s: bpf_setdlt: ifpromisc failed (%d)\n" , |
| 1920 | bp->bif_ifp->if_xname, error); |
| 1921 | else |
| 1922 | d->bd_promisc = 1; |
| 1923 | } |
| 1924 | splx(s); |
| 1925 | return 0; |
| 1926 | } |
| 1927 | |
| 1928 | static int |
| 1929 | sysctl_net_bpf_maxbufsize(SYSCTLFN_ARGS) |
| 1930 | { |
| 1931 | int newsize, error; |
| 1932 | struct sysctlnode node; |
| 1933 | |
| 1934 | node = *rnode; |
| 1935 | node.sysctl_data = &newsize; |
| 1936 | newsize = bpf_maxbufsize; |
| 1937 | error = sysctl_lookup(SYSCTLFN_CALL(&node)); |
| 1938 | if (error || newp == NULL) |
| 1939 | return (error); |
| 1940 | |
| 1941 | if (newsize < BPF_MINBUFSIZE || newsize > BPF_MAXBUFSIZE) |
| 1942 | return (EINVAL); |
| 1943 | |
| 1944 | bpf_maxbufsize = newsize; |
| 1945 | |
| 1946 | return (0); |
| 1947 | } |
| 1948 | |
| 1949 | #if defined(MODULAR) || defined(BPFJIT) |
| 1950 | static int |
| 1951 | sysctl_net_bpf_jit(SYSCTLFN_ARGS) |
| 1952 | { |
| 1953 | bool newval; |
| 1954 | int error; |
| 1955 | struct sysctlnode node; |
| 1956 | |
| 1957 | node = *rnode; |
| 1958 | node.sysctl_data = &newval; |
| 1959 | newval = bpf_jit; |
| 1960 | error = sysctl_lookup(SYSCTLFN_CALL(&node)); |
| 1961 | if (error != 0 || newp == NULL) |
| 1962 | return error; |
| 1963 | |
| 1964 | bpf_jit = newval; |
| 1965 | |
| 1966 | /* |
| 1967 | * Do a full sync to publish new bpf_jit value and |
| 1968 | * update bpfjit_module_ops.bj_generate_code variable. |
| 1969 | */ |
| 1970 | membar_sync(); |
| 1971 | |
| 1972 | if (newval && bpfjit_module_ops.bj_generate_code == NULL) { |
| 1973 | printf("JIT compilation is postponed " |
| 1974 | "until after bpfjit module is loaded\n" ); |
| 1975 | } |
| 1976 | |
| 1977 | return 0; |
| 1978 | } |
| 1979 | #endif |
| 1980 | |
| 1981 | static int |
| 1982 | sysctl_net_bpf_peers(SYSCTLFN_ARGS) |
| 1983 | { |
| 1984 | int error, elem_count; |
| 1985 | struct bpf_d *dp; |
| 1986 | struct bpf_d_ext dpe; |
| 1987 | size_t len, needed, elem_size, out_size; |
| 1988 | char *sp; |
| 1989 | |
| 1990 | if (namelen == 1 && name[0] == CTL_QUERY) |
| 1991 | return (sysctl_query(SYSCTLFN_CALL(rnode))); |
| 1992 | |
| 1993 | if (namelen != 2) |
| 1994 | return (EINVAL); |
| 1995 | |
| 1996 | /* BPF peers is privileged information. */ |
| 1997 | error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_INTERFACE, |
| 1998 | KAUTH_REQ_NETWORK_INTERFACE_GETPRIV, NULL, NULL, NULL); |
| 1999 | if (error) |
| 2000 | return (EPERM); |
| 2001 | |
| 2002 | len = (oldp != NULL) ? *oldlenp : 0; |
| 2003 | sp = oldp; |
| 2004 | elem_size = name[0]; |
| 2005 | elem_count = name[1]; |
| 2006 | out_size = MIN(sizeof(dpe), elem_size); |
| 2007 | needed = 0; |
| 2008 | |
| 2009 | if (elem_size < 1 || elem_count < 0) |
| 2010 | return (EINVAL); |
| 2011 | |
| 2012 | mutex_enter(&bpf_mtx); |
| 2013 | LIST_FOREACH(dp, &bpf_list, bd_list) { |
| 2014 | if (len >= elem_size && elem_count > 0) { |
| 2015 | #define BPF_EXT(field) dpe.bde_ ## field = dp->bd_ ## field |
| 2016 | BPF_EXT(bufsize); |
| 2017 | BPF_EXT(promisc); |
| 2018 | BPF_EXT(state); |
| 2019 | BPF_EXT(immediate); |
| 2020 | BPF_EXT(hdrcmplt); |
| 2021 | BPF_EXT(seesent); |
| 2022 | BPF_EXT(pid); |
| 2023 | BPF_EXT(rcount); |
| 2024 | BPF_EXT(dcount); |
| 2025 | BPF_EXT(ccount); |
| 2026 | #undef BPF_EXT |
| 2027 | if (dp->bd_bif) |
| 2028 | (void)strlcpy(dpe.bde_ifname, |
| 2029 | dp->bd_bif->bif_ifp->if_xname, |
| 2030 | IFNAMSIZ - 1); |
| 2031 | else |
| 2032 | dpe.bde_ifname[0] = '\0'; |
| 2033 | |
| 2034 | error = copyout(&dpe, sp, out_size); |
| 2035 | if (error) |
| 2036 | break; |
| 2037 | sp += elem_size; |
| 2038 | len -= elem_size; |
| 2039 | } |
| 2040 | needed += elem_size; |
| 2041 | if (elem_count > 0 && elem_count != INT_MAX) |
| 2042 | elem_count--; |
| 2043 | } |
| 2044 | mutex_exit(&bpf_mtx); |
| 2045 | |
| 2046 | *oldlenp = needed; |
| 2047 | |
| 2048 | return (error); |
| 2049 | } |
| 2050 | |
| 2051 | static struct sysctllog *bpf_sysctllog; |
| 2052 | static void |
| 2053 | sysctl_net_bpf_setup(void) |
| 2054 | { |
| 2055 | const struct sysctlnode *node; |
| 2056 | |
| 2057 | node = NULL; |
| 2058 | sysctl_createv(&bpf_sysctllog, 0, NULL, &node, |
| 2059 | CTLFLAG_PERMANENT, |
| 2060 | CTLTYPE_NODE, "bpf" , |
| 2061 | SYSCTL_DESCR("BPF options" ), |
| 2062 | NULL, 0, NULL, 0, |
| 2063 | CTL_NET, CTL_CREATE, CTL_EOL); |
| 2064 | if (node != NULL) { |
| 2065 | #if defined(MODULAR) || defined(BPFJIT) |
| 2066 | sysctl_createv(&bpf_sysctllog, 0, NULL, NULL, |
| 2067 | CTLFLAG_PERMANENT|CTLFLAG_READWRITE, |
| 2068 | CTLTYPE_BOOL, "jit" , |
| 2069 | SYSCTL_DESCR("Toggle Just-In-Time compilation" ), |
| 2070 | sysctl_net_bpf_jit, 0, &bpf_jit, 0, |
| 2071 | CTL_NET, node->sysctl_num, CTL_CREATE, CTL_EOL); |
| 2072 | #endif |
| 2073 | sysctl_createv(&bpf_sysctllog, 0, NULL, NULL, |
| 2074 | CTLFLAG_PERMANENT|CTLFLAG_READWRITE, |
| 2075 | CTLTYPE_INT, "maxbufsize" , |
| 2076 | SYSCTL_DESCR("Maximum size for data capture buffer" ), |
| 2077 | sysctl_net_bpf_maxbufsize, 0, &bpf_maxbufsize, 0, |
| 2078 | CTL_NET, node->sysctl_num, CTL_CREATE, CTL_EOL); |
| 2079 | sysctl_createv(&bpf_sysctllog, 0, NULL, NULL, |
| 2080 | CTLFLAG_PERMANENT, |
| 2081 | CTLTYPE_STRUCT, "stats" , |
| 2082 | SYSCTL_DESCR("BPF stats" ), |
| 2083 | NULL, 0, &bpf_gstats, sizeof(bpf_gstats), |
| 2084 | CTL_NET, node->sysctl_num, CTL_CREATE, CTL_EOL); |
| 2085 | sysctl_createv(&bpf_sysctllog, 0, NULL, NULL, |
| 2086 | CTLFLAG_PERMANENT, |
| 2087 | CTLTYPE_STRUCT, "peers" , |
| 2088 | SYSCTL_DESCR("BPF peers" ), |
| 2089 | sysctl_net_bpf_peers, 0, NULL, 0, |
| 2090 | CTL_NET, node->sysctl_num, CTL_CREATE, CTL_EOL); |
| 2091 | } |
| 2092 | |
| 2093 | } |
| 2094 | |
| 2095 | struct bpf_ops bpf_ops_kernel = { |
| 2096 | .bpf_attach = _bpfattach, |
| 2097 | .bpf_detach = _bpfdetach, |
| 2098 | .bpf_change_type = _bpf_change_type, |
| 2099 | |
| 2100 | .bpf_tap = _bpf_tap, |
| 2101 | .bpf_mtap = _bpf_mtap, |
| 2102 | .bpf_mtap2 = _bpf_mtap2, |
| 2103 | .bpf_mtap_af = _bpf_mtap_af, |
| 2104 | .bpf_mtap_sl_in = _bpf_mtap_sl_in, |
| 2105 | .bpf_mtap_sl_out = _bpf_mtap_sl_out, |
| 2106 | }; |
| 2107 | |
| 2108 | MODULE(MODULE_CLASS_DRIVER, bpf, "bpf_filter" ); |
| 2109 | |
| 2110 | static int |
| 2111 | bpf_modcmd(modcmd_t cmd, void *arg) |
| 2112 | { |
| 2113 | #ifdef _MODULE |
| 2114 | devmajor_t bmajor, cmajor; |
| 2115 | #endif |
| 2116 | int error = 0; |
| 2117 | |
| 2118 | switch (cmd) { |
| 2119 | case MODULE_CMD_INIT: |
| 2120 | bpf_init(); |
| 2121 | #ifdef _MODULE |
| 2122 | bmajor = cmajor = NODEVMAJOR; |
| 2123 | error = devsw_attach("bpf" , NULL, &bmajor, |
| 2124 | &bpf_cdevsw, &cmajor); |
| 2125 | if (error) |
| 2126 | break; |
| 2127 | #endif |
| 2128 | |
| 2129 | bpf_ops_handover_enter(&bpf_ops_kernel); |
| 2130 | atomic_swap_ptr(&bpf_ops, &bpf_ops_kernel); |
| 2131 | bpf_ops_handover_exit(); |
| 2132 | sysctl_net_bpf_setup(); |
| 2133 | break; |
| 2134 | |
| 2135 | case MODULE_CMD_FINI: |
| 2136 | /* |
| 2137 | * While there is no reference counting for bpf callers, |
| 2138 | * unload could at least in theory be done similarly to |
| 2139 | * system call disestablishment. This should even be |
| 2140 | * a little simpler: |
| 2141 | * |
| 2142 | * 1) replace op vector with stubs |
| 2143 | * 2) post update to all cpus with xc |
| 2144 | * 3) check that nobody is in bpf anymore |
| 2145 | * (it's doubtful we'd want something like l_sysent, |
| 2146 | * but we could do something like *signed* percpu |
| 2147 | * counters. if the sum is 0, we're good). |
| 2148 | * 4) if fail, unroll changes |
| 2149 | * |
| 2150 | * NOTE: change won't be atomic to the outside. some |
| 2151 | * packets may be not captured even if unload is |
| 2152 | * not succesful. I think packet capture not working |
| 2153 | * is a perfectly logical consequence of trying to |
| 2154 | * disable packet capture. |
| 2155 | */ |
| 2156 | error = EOPNOTSUPP; |
| 2157 | /* insert sysctl teardown */ |
| 2158 | break; |
| 2159 | |
| 2160 | default: |
| 2161 | error = ENOTTY; |
| 2162 | break; |
| 2163 | } |
| 2164 | |
| 2165 | return error; |
| 2166 | } |
| 2167 | |