1 /* $OpenBSD: bpf.c,v 1.13 2004/05/05 14:28:58 deraadt Exp $ */ 2 3 /* BPF socket interface code, originally contributed by Archie Cobbs. */ 4 5 /*- 6 * SPDX-License-Identifier: BSD-3-Clause 7 * 8 * Copyright (c) 1995, 1996, 1998, 1999 9 * The Internet Software Consortium. All rights reserved. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 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 Internet Software Consortium nor the names 21 * of its contributors may be used to endorse or promote products derived 22 * from this software without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE INTERNET SOFTWARE CONSORTIUM AND 25 * CONTRIBUTORS ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, 26 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 27 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 28 * DISCLAIMED. IN NO EVENT SHALL THE INTERNET SOFTWARE CONSORTIUM OR 29 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 30 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 31 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF 32 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 33 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 34 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 35 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * This software has been written for the Internet Software Consortium 39 * by Ted Lemon <mellon@fugue.com> in cooperation with Vixie 40 * Enterprises. To learn more about the Internet Software Consortium, 41 * see ``http://www.vix.com/isc''. To learn more about Vixie 42 * Enterprises, see ``http://www.vix.com''. 43 */ 44 45 #include <sys/cdefs.h> 46 __FBSDID("$FreeBSD$"); 47 48 #include "dhcpd.h" 49 #include "privsep.h" 50 #include <sys/capsicum.h> 51 #include <sys/ioctl.h> 52 #include <sys/uio.h> 53 54 #include <net/bpf.h> 55 #include <netinet/in_systm.h> 56 #include <netinet/ip.h> 57 #include <netinet/udp.h> 58 #include <netinet/if_ether.h> 59 60 #define BPF_FORMAT "/dev/bpf%d" 61 62 /* 63 * Called by get_interface_list for each interface that's discovered. 64 * Opens a packet filter for each interface and adds it to the select 65 * mask. 66 */ 67 int 68 if_register_bpf(struct interface_info *info, int flags) 69 { 70 char filename[50]; 71 int sock, b; 72 73 /* Open a BPF device */ 74 for (b = 0;; b++) { 75 snprintf(filename, sizeof(filename), BPF_FORMAT, b); 76 sock = open(filename, flags); 77 if (sock < 0) { 78 if (errno == EBUSY) 79 continue; 80 else 81 error("Can't find free bpf: %m"); 82 } else 83 break; 84 } 85 86 /* Set the BPF device to point at this interface. */ 87 if (ioctl(sock, BIOCSETIF, info->ifp) < 0) 88 error("Can't attach interface %s to bpf device %s: %m", 89 info->name, filename); 90 91 return (sock); 92 } 93 94 /* 95 * Packet write filter program: 96 * 'ip and udp and src port bootps and dst port (bootps or bootpc)' 97 */ 98 struct bpf_insn dhcp_bpf_wfilter[] = { 99 BPF_STMT(BPF_LD + BPF_B + BPF_IND, 14), 100 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, (IPVERSION << 4) + 5, 0, 12), 101 102 /* Make sure this is an IP packet... */ 103 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12), 104 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 10), 105 106 /* Make sure it's a UDP packet... */ 107 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23), 108 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 8), 109 110 /* Make sure this isn't a fragment... */ 111 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20), 112 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 6, 0), /* patched */ 113 114 /* Get the IP header length... */ 115 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14), 116 117 /* Make sure it's from the right port... */ 118 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 14), 119 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 68, 0, 3), 120 121 /* Make sure it is to the right ports ... */ 122 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16), 123 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1), 124 125 /* If we passed all the tests, ask for the whole packet. */ 126 BPF_STMT(BPF_RET+BPF_K, (u_int)-1), 127 128 /* Otherwise, drop it. */ 129 BPF_STMT(BPF_RET+BPF_K, 0), 130 }; 131 132 int dhcp_bpf_wfilter_len = sizeof(dhcp_bpf_wfilter) / sizeof(struct bpf_insn); 133 134 void 135 if_register_send(struct interface_info *info) 136 { 137 cap_rights_t rights; 138 struct bpf_version v; 139 struct bpf_program p; 140 int sock, on = 1; 141 142 /* Open a BPF device and hang it on this interface... */ 143 info->wfdesc = if_register_bpf(info, O_WRONLY); 144 145 /* Make sure the BPF version is in range... */ 146 if (ioctl(info->wfdesc, BIOCVERSION, &v) < 0) 147 error("Can't get BPF version: %m"); 148 149 if (v.bv_major != BPF_MAJOR_VERSION || 150 v.bv_minor < BPF_MINOR_VERSION) 151 error("Kernel BPF version out of range - recompile dhcpd!"); 152 153 /* Set up the bpf write filter program structure. */ 154 p.bf_len = dhcp_bpf_wfilter_len; 155 p.bf_insns = dhcp_bpf_wfilter; 156 157 if (dhcp_bpf_wfilter[7].k == 0x1fff) 158 dhcp_bpf_wfilter[7].k = htons(IP_MF|IP_OFFMASK); 159 160 if (ioctl(info->wfdesc, BIOCSETWF, &p) < 0) 161 error("Can't install write filter program: %m"); 162 163 if (ioctl(info->wfdesc, BIOCLOCK, NULL) < 0) 164 error("Cannot lock bpf"); 165 166 cap_rights_init(&rights, CAP_WRITE); 167 if (cap_rights_limit(info->wfdesc, &rights) < 0 && errno != ENOSYS) 168 error("Can't limit bpf descriptor: %m"); 169 170 /* 171 * Use raw socket for unicast send. 172 */ 173 if ((sock = socket(AF_INET, SOCK_RAW, IPPROTO_UDP)) == -1) 174 error("socket(SOCK_RAW): %m"); 175 if (setsockopt(sock, IPPROTO_IP, IP_HDRINCL, &on, 176 sizeof(on)) == -1) 177 error("setsockopt(IP_HDRINCL): %m"); 178 info->ufdesc = sock; 179 } 180 181 /* 182 * Packet filter program... 183 * 184 * XXX: Changes to the filter program may require changes to the 185 * constant offsets used in if_register_send to patch the BPF program! 186 */ 187 struct bpf_insn dhcp_bpf_filter[] = { 188 /* Make sure this is an IP packet... */ 189 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12), 190 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 8), 191 192 /* Make sure it's a UDP packet... */ 193 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23), 194 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 6), 195 196 /* Make sure this isn't a fragment... */ 197 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20), 198 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 4, 0), 199 200 /* Get the IP header length... */ 201 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14), 202 203 /* Make sure it's to the right port... */ 204 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16), 205 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1), /* patch */ 206 207 /* If we passed all the tests, ask for the whole packet. */ 208 BPF_STMT(BPF_RET+BPF_K, (u_int)-1), 209 210 /* Otherwise, drop it. */ 211 BPF_STMT(BPF_RET+BPF_K, 0), 212 }; 213 214 int dhcp_bpf_filter_len = sizeof(dhcp_bpf_filter) / sizeof(struct bpf_insn); 215 216 void 217 if_register_receive(struct interface_info *info) 218 { 219 static const unsigned long cmds[2] = { SIOCGIFFLAGS, SIOCGIFMEDIA }; 220 cap_rights_t rights; 221 struct bpf_version v; 222 struct bpf_program p; 223 int flag = 1, sz; 224 225 /* Open a BPF device and hang it on this interface... */ 226 info->rfdesc = if_register_bpf(info, O_RDONLY); 227 228 /* Make sure the BPF version is in range... */ 229 if (ioctl(info->rfdesc, BIOCVERSION, &v) < 0) 230 error("Can't get BPF version: %m"); 231 232 if (v.bv_major != BPF_MAJOR_VERSION || 233 v.bv_minor < BPF_MINOR_VERSION) 234 error("Kernel BPF version out of range - recompile dhcpd!"); 235 236 /* 237 * Set immediate mode so that reads return as soon as a packet 238 * comes in, rather than waiting for the input buffer to fill 239 * with packets. 240 */ 241 if (ioctl(info->rfdesc, BIOCIMMEDIATE, &flag) < 0) 242 error("Can't set immediate mode on bpf device: %m"); 243 244 /* Get the required BPF buffer length from the kernel. */ 245 if (ioctl(info->rfdesc, BIOCGBLEN, &sz) < 0) 246 error("Can't get bpf buffer length: %m"); 247 info->rbuf_max = sz; 248 info->rbuf = malloc(info->rbuf_max); 249 if (!info->rbuf) 250 error("Can't allocate %lu bytes for bpf input buffer.", 251 (unsigned long)info->rbuf_max); 252 info->rbuf_offset = 0; 253 info->rbuf_len = 0; 254 255 /* Set up the bpf filter program structure. */ 256 p.bf_len = dhcp_bpf_filter_len; 257 p.bf_insns = dhcp_bpf_filter; 258 259 /* Patch the server port into the BPF program... 260 * 261 * XXX: changes to filter program may require changes to the 262 * insn number(s) used below! 263 */ 264 dhcp_bpf_filter[8].k = LOCAL_PORT; 265 266 if (ioctl(info->rfdesc, BIOCSETF, &p) < 0) 267 error("Can't install packet filter program: %m"); 268 269 if (ioctl(info->rfdesc, BIOCLOCK, NULL) < 0) 270 error("Cannot lock bpf"); 271 272 cap_rights_init(&rights, CAP_IOCTL, CAP_EVENT, CAP_READ); 273 if (cap_rights_limit(info->rfdesc, &rights) < 0 && errno != ENOSYS) 274 error("Can't limit bpf descriptor: %m"); 275 if (cap_ioctls_limit(info->rfdesc, cmds, 2) < 0 && errno != ENOSYS) 276 error("Can't limit ioctls for bpf descriptor: %m"); 277 } 278 279 void 280 send_packet_unpriv(int privfd, struct dhcp_packet *raw, size_t len, 281 struct in_addr from, struct in_addr to) 282 { 283 struct imsg_hdr hdr; 284 struct buf *buf; 285 int errs; 286 287 hdr.code = IMSG_SEND_PACKET; 288 hdr.len = sizeof(hdr) + 289 sizeof(size_t) + len + 290 sizeof(from) + sizeof(to); 291 292 if ((buf = buf_open(hdr.len)) == NULL) 293 error("buf_open: %m"); 294 295 errs = 0; 296 errs += buf_add(buf, &hdr, sizeof(hdr)); 297 errs += buf_add(buf, &len, sizeof(len)); 298 errs += buf_add(buf, raw, len); 299 errs += buf_add(buf, &from, sizeof(from)); 300 errs += buf_add(buf, &to, sizeof(to)); 301 if (errs) 302 error("buf_add: %m"); 303 304 if (buf_close(privfd, buf) == -1) 305 error("buf_close: %m"); 306 } 307 308 void 309 send_packet_priv(struct interface_info *interface, struct imsg_hdr *hdr, int fd) 310 { 311 unsigned char buf[256]; 312 struct iovec iov[2]; 313 struct msghdr msg; 314 struct dhcp_packet raw; 315 size_t len; 316 struct in_addr from, to; 317 int result, bufp = 0; 318 319 if (hdr->len < sizeof(*hdr) + sizeof(size_t)) 320 error("corrupted message received"); 321 buf_read(fd, &len, sizeof(len)); 322 if (hdr->len != sizeof(*hdr) + sizeof(size_t) + len + 323 sizeof(from) + sizeof(to)) { 324 error("corrupted message received"); 325 } 326 if (len > sizeof(raw)) 327 error("corrupted message received"); 328 buf_read(fd, &raw, len); 329 buf_read(fd, &from, sizeof(from)); 330 buf_read(fd, &to, sizeof(to)); 331 332 /* Assemble the headers... */ 333 if (to.s_addr == INADDR_BROADCAST) 334 assemble_hw_header(interface, buf, &bufp); 335 assemble_udp_ip_header(buf, &bufp, from.s_addr, to.s_addr, 336 htons(REMOTE_PORT), (unsigned char *)&raw, len); 337 338 iov[0].iov_base = buf; 339 iov[0].iov_len = bufp; 340 iov[1].iov_base = &raw; 341 iov[1].iov_len = len; 342 343 /* Fire it off */ 344 if (to.s_addr == INADDR_BROADCAST) 345 result = writev(interface->wfdesc, iov, 2); 346 else { 347 struct sockaddr_in sato; 348 349 sato.sin_addr = to; 350 sato.sin_port = htons(REMOTE_PORT); 351 sato.sin_family = AF_INET; 352 sato.sin_len = sizeof(sato); 353 354 memset(&msg, 0, sizeof(msg)); 355 msg.msg_name = (struct sockaddr *)&sato; 356 msg.msg_namelen = sizeof(sato); 357 msg.msg_iov = iov; 358 msg.msg_iovlen = 2; 359 result = sendmsg(interface->ufdesc, &msg, 0); 360 } 361 362 if (result < 0) 363 warning("send_packet: %m"); 364 } 365 366 ssize_t 367 receive_packet(struct interface_info *interface, unsigned char *buf, 368 size_t len, struct sockaddr_in *from, struct hardware *hfrom) 369 { 370 int length = 0, offset = 0; 371 struct bpf_hdr hdr; 372 373 /* 374 * All this complexity is because BPF doesn't guarantee that 375 * only one packet will be returned at a time. We're getting 376 * what we deserve, though - this is a terrible abuse of the BPF 377 * interface. Sigh. 378 */ 379 380 /* Process packets until we get one we can return or until we've 381 * done a read and gotten nothing we can return... 382 */ 383 do { 384 /* If the buffer is empty, fill it. */ 385 if (interface->rbuf_offset >= interface->rbuf_len) { 386 length = read(interface->rfdesc, interface->rbuf, 387 interface->rbuf_max); 388 if (length <= 0) 389 return (length); 390 interface->rbuf_offset = 0; 391 interface->rbuf_len = length; 392 } 393 394 /* 395 * If there isn't room for a whole bpf header, something 396 * went wrong, but we'll ignore it and hope it goes 397 * away... XXX 398 */ 399 if (interface->rbuf_len - interface->rbuf_offset < 400 sizeof(hdr)) { 401 interface->rbuf_offset = interface->rbuf_len; 402 continue; 403 } 404 405 /* Copy out a bpf header... */ 406 memcpy(&hdr, &interface->rbuf[interface->rbuf_offset], 407 sizeof(hdr)); 408 409 /* 410 * If the bpf header plus data doesn't fit in what's 411 * left of the buffer, stick head in sand yet again... 412 */ 413 if (interface->rbuf_offset + hdr.bh_hdrlen + hdr.bh_caplen > 414 interface->rbuf_len) { 415 interface->rbuf_offset = interface->rbuf_len; 416 continue; 417 } 418 419 /* Skip over the BPF header... */ 420 interface->rbuf_offset += hdr.bh_hdrlen; 421 422 /* 423 * If the captured data wasn't the whole packet, or if 424 * the packet won't fit in the input buffer, all we can 425 * do is drop it. 426 */ 427 if (hdr.bh_caplen != hdr.bh_datalen) { 428 interface->rbuf_offset = 429 BPF_WORDALIGN(interface->rbuf_offset + 430 hdr.bh_caplen); 431 continue; 432 } 433 434 /* Decode the physical header... */ 435 offset = decode_hw_header(interface->rbuf, 436 interface->rbuf_offset, hfrom); 437 438 /* 439 * If a physical layer checksum failed (dunno of any 440 * physical layer that supports this, but WTH), skip 441 * this packet. 442 */ 443 if (offset < 0) { 444 interface->rbuf_offset = 445 BPF_WORDALIGN(interface->rbuf_offset + 446 hdr.bh_caplen); 447 continue; 448 } 449 interface->rbuf_offset += offset; 450 hdr.bh_caplen -= offset; 451 452 /* Decode the IP and UDP headers... */ 453 offset = decode_udp_ip_header(interface->rbuf, 454 interface->rbuf_offset, from, NULL, hdr.bh_caplen); 455 456 /* If the IP or UDP checksum was bad, skip the packet... */ 457 if (offset < 0) { 458 interface->rbuf_offset = 459 BPF_WORDALIGN(interface->rbuf_offset + 460 hdr.bh_caplen); 461 continue; 462 } 463 interface->rbuf_offset += offset; 464 hdr.bh_caplen -= offset; 465 466 /* 467 * If there's not enough room to stash the packet data, 468 * we have to skip it (this shouldn't happen in real 469 * life, though). 470 */ 471 if (hdr.bh_caplen > len) { 472 interface->rbuf_offset = 473 BPF_WORDALIGN(interface->rbuf_offset + 474 hdr.bh_caplen); 475 continue; 476 } 477 478 /* Copy out the data in the packet... */ 479 memcpy(buf, interface->rbuf + interface->rbuf_offset, 480 hdr.bh_caplen); 481 interface->rbuf_offset = 482 BPF_WORDALIGN(interface->rbuf_offset + 483 hdr.bh_caplen); 484 return (hdr.bh_caplen); 485 } while (!length); 486 return (0); 487 } 488