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 * Copyright (c) 1995, 1996, 1998, 1999 7 * The Internet Software Consortium. All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 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. Neither the name of The Internet Software Consortium nor the names 19 * of its contributors may be used to endorse or promote products derived 20 * from this software without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE INTERNET SOFTWARE CONSORTIUM AND 23 * CONTRIBUTORS ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, 24 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 26 * DISCLAIMED. IN NO EVENT SHALL THE INTERNET SOFTWARE CONSORTIUM OR 27 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 28 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 29 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF 30 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 31 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 32 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 33 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * This software has been written for the Internet Software Consortium 37 * by Ted Lemon <mellon@fugue.com> in cooperation with Vixie 38 * Enterprises. To learn more about the Internet Software Consortium, 39 * see ``http://www.vix.com/isc''. To learn more about Vixie 40 * Enterprises, see ``http://www.vix.com''. 41 */ 42 43 #include <sys/cdefs.h> 44 __FBSDID("$FreeBSD$"); 45 46 #include "dhcpd.h" 47 #include "privsep.h" 48 #include <sys/capsicum.h> 49 #include <sys/ioctl.h> 50 #include <sys/uio.h> 51 52 #include <net/bpf.h> 53 #include <netinet/in_systm.h> 54 #include <netinet/ip.h> 55 #include <netinet/udp.h> 56 #include <netinet/if_ether.h> 57 58 #define BPF_FORMAT "/dev/bpf%d" 59 60 /* 61 * Called by get_interface_list for each interface that's discovered. 62 * Opens a packet filter for each interface and adds it to the select 63 * mask. 64 */ 65 int 66 if_register_bpf(struct interface_info *info, int flags) 67 { 68 char filename[50]; 69 int sock, b; 70 71 /* Open a BPF device */ 72 for (b = 0;; b++) { 73 snprintf(filename, sizeof(filename), BPF_FORMAT, b); 74 sock = open(filename, flags); 75 if (sock < 0) { 76 if (errno == EBUSY) 77 continue; 78 else 79 error("Can't find free bpf: %m"); 80 } else 81 break; 82 } 83 84 /* Set the BPF device to point at this interface. */ 85 if (ioctl(sock, BIOCSETIF, info->ifp) < 0) 86 error("Can't attach interface %s to bpf device %s: %m", 87 info->name, filename); 88 89 return (sock); 90 } 91 92 /* 93 * Packet write filter program: 94 * 'ip and udp and src port bootps and dst port (bootps or bootpc)' 95 */ 96 struct bpf_insn dhcp_bpf_wfilter[] = { 97 BPF_STMT(BPF_LD + BPF_B + BPF_IND, 14), 98 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, (IPVERSION << 4) + 5, 0, 12), 99 100 /* Make sure this is an IP packet... */ 101 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12), 102 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 10), 103 104 /* Make sure it's a UDP packet... */ 105 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23), 106 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 8), 107 108 /* Make sure this isn't a fragment... */ 109 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20), 110 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 6, 0), /* patched */ 111 112 /* Get the IP header length... */ 113 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14), 114 115 /* Make sure it's from the right port... */ 116 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 14), 117 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 68, 0, 3), 118 119 /* Make sure it is to the right ports ... */ 120 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16), 121 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1), 122 123 /* If we passed all the tests, ask for the whole packet. */ 124 BPF_STMT(BPF_RET+BPF_K, (u_int)-1), 125 126 /* Otherwise, drop it. */ 127 BPF_STMT(BPF_RET+BPF_K, 0), 128 }; 129 130 int dhcp_bpf_wfilter_len = sizeof(dhcp_bpf_wfilter) / sizeof(struct bpf_insn); 131 132 void 133 if_register_send(struct interface_info *info) 134 { 135 cap_rights_t rights; 136 struct bpf_version v; 137 struct bpf_program p; 138 int sock, on = 1; 139 140 /* Open a BPF device and hang it on this interface... */ 141 info->wfdesc = if_register_bpf(info, O_WRONLY); 142 143 /* Make sure the BPF version is in range... */ 144 if (ioctl(info->wfdesc, BIOCVERSION, &v) < 0) 145 error("Can't get BPF version: %m"); 146 147 if (v.bv_major != BPF_MAJOR_VERSION || 148 v.bv_minor < BPF_MINOR_VERSION) 149 error("Kernel BPF version out of range - recompile dhcpd!"); 150 151 /* Set up the bpf write filter program structure. */ 152 p.bf_len = dhcp_bpf_wfilter_len; 153 p.bf_insns = dhcp_bpf_wfilter; 154 155 if (dhcp_bpf_wfilter[7].k == 0x1fff) 156 dhcp_bpf_wfilter[7].k = htons(IP_MF|IP_OFFMASK); 157 158 if (ioctl(info->wfdesc, BIOCSETWF, &p) < 0) 159 error("Can't install write filter program: %m"); 160 161 if (ioctl(info->wfdesc, BIOCLOCK, NULL) < 0) 162 error("Cannot lock bpf"); 163 164 cap_rights_init(&rights, CAP_WRITE); 165 if (cap_rights_limit(info->wfdesc, &rights) < 0 && errno != ENOSYS) 166 error("Can't limit bpf descriptor: %m"); 167 168 /* 169 * Use raw socket for unicast send. 170 */ 171 if ((sock = socket(AF_INET, SOCK_RAW, IPPROTO_UDP)) == -1) 172 error("socket(SOCK_RAW): %m"); 173 if (setsockopt(sock, IPPROTO_IP, IP_HDRINCL, &on, 174 sizeof(on)) == -1) 175 error("setsockopt(IP_HDRINCL): %m"); 176 info->ufdesc = sock; 177 } 178 179 /* 180 * Packet filter program... 181 * 182 * XXX: Changes to the filter program may require changes to the 183 * constant offsets used in if_register_send to patch the BPF program! 184 */ 185 struct bpf_insn dhcp_bpf_filter[] = { 186 /* Make sure this is an IP packet... */ 187 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12), 188 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 8), 189 190 /* Make sure it's a UDP packet... */ 191 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23), 192 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 6), 193 194 /* Make sure this isn't a fragment... */ 195 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20), 196 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 4, 0), 197 198 /* Get the IP header length... */ 199 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14), 200 201 /* Make sure it's to the right port... */ 202 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16), 203 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1), /* patch */ 204 205 /* If we passed all the tests, ask for the whole packet. */ 206 BPF_STMT(BPF_RET+BPF_K, (u_int)-1), 207 208 /* Otherwise, drop it. */ 209 BPF_STMT(BPF_RET+BPF_K, 0), 210 }; 211 212 int dhcp_bpf_filter_len = sizeof(dhcp_bpf_filter) / sizeof(struct bpf_insn); 213 214 void 215 if_register_receive(struct interface_info *info) 216 { 217 static const unsigned long cmds[2] = { SIOCGIFFLAGS, SIOCGIFMEDIA }; 218 cap_rights_t rights; 219 struct bpf_version v; 220 struct bpf_program p; 221 int flag = 1, sz; 222 223 /* Open a BPF device and hang it on this interface... */ 224 info->rfdesc = if_register_bpf(info, O_RDONLY); 225 226 /* Make sure the BPF version is in range... */ 227 if (ioctl(info->rfdesc, BIOCVERSION, &v) < 0) 228 error("Can't get BPF version: %m"); 229 230 if (v.bv_major != BPF_MAJOR_VERSION || 231 v.bv_minor < BPF_MINOR_VERSION) 232 error("Kernel BPF version out of range - recompile dhcpd!"); 233 234 /* 235 * Set immediate mode so that reads return as soon as a packet 236 * comes in, rather than waiting for the input buffer to fill 237 * with packets. 238 */ 239 if (ioctl(info->rfdesc, BIOCIMMEDIATE, &flag) < 0) 240 error("Can't set immediate mode on bpf device: %m"); 241 242 /* Get the required BPF buffer length from the kernel. */ 243 if (ioctl(info->rfdesc, BIOCGBLEN, &sz) < 0) 244 error("Can't get bpf buffer length: %m"); 245 info->rbuf_max = sz; 246 info->rbuf = malloc(info->rbuf_max); 247 if (!info->rbuf) 248 error("Can't allocate %lu bytes for bpf input buffer.", 249 (unsigned long)info->rbuf_max); 250 info->rbuf_offset = 0; 251 info->rbuf_len = 0; 252 253 /* Set up the bpf filter program structure. */ 254 p.bf_len = dhcp_bpf_filter_len; 255 p.bf_insns = dhcp_bpf_filter; 256 257 /* Patch the server port into the BPF program... 258 * 259 * XXX: changes to filter program may require changes to the 260 * insn number(s) used below! 261 */ 262 dhcp_bpf_filter[8].k = LOCAL_PORT; 263 264 if (ioctl(info->rfdesc, BIOCSETF, &p) < 0) 265 error("Can't install packet filter program: %m"); 266 267 if (ioctl(info->rfdesc, BIOCLOCK, NULL) < 0) 268 error("Cannot lock bpf"); 269 270 cap_rights_init(&rights, CAP_IOCTL, CAP_EVENT, CAP_READ); 271 if (cap_rights_limit(info->rfdesc, &rights) < 0 && errno != ENOSYS) 272 error("Can't limit bpf descriptor: %m"); 273 if (cap_ioctls_limit(info->rfdesc, cmds, 2) < 0 && errno != ENOSYS) 274 error("Can't limit ioctls for bpf descriptor: %m"); 275 } 276 277 void 278 send_packet_unpriv(int privfd, struct dhcp_packet *raw, size_t len, 279 struct in_addr from, struct in_addr to) 280 { 281 struct imsg_hdr hdr; 282 struct buf *buf; 283 int errs; 284 285 hdr.code = IMSG_SEND_PACKET; 286 hdr.len = sizeof(hdr) + 287 sizeof(size_t) + len + 288 sizeof(from) + sizeof(to); 289 290 if ((buf = buf_open(hdr.len)) == NULL) 291 error("buf_open: %m"); 292 293 errs = 0; 294 errs += buf_add(buf, &hdr, sizeof(hdr)); 295 errs += buf_add(buf, &len, sizeof(len)); 296 errs += buf_add(buf, raw, len); 297 errs += buf_add(buf, &from, sizeof(from)); 298 errs += buf_add(buf, &to, sizeof(to)); 299 if (errs) 300 error("buf_add: %m"); 301 302 if (buf_close(privfd, buf) == -1) 303 error("buf_close: %m"); 304 } 305 306 void 307 send_packet_priv(struct interface_info *interface, struct imsg_hdr *hdr, int fd) 308 { 309 unsigned char buf[256]; 310 struct iovec iov[2]; 311 struct msghdr msg; 312 struct dhcp_packet raw; 313 size_t len; 314 struct in_addr from, to; 315 int result, bufp = 0; 316 317 if (hdr->len < sizeof(*hdr) + sizeof(size_t)) 318 error("corrupted message received"); 319 buf_read(fd, &len, sizeof(len)); 320 if (hdr->len != sizeof(*hdr) + sizeof(size_t) + len + 321 sizeof(from) + sizeof(to)) { 322 error("corrupted message received"); 323 } 324 if (len > sizeof(raw)) 325 error("corrupted message received"); 326 buf_read(fd, &raw, len); 327 buf_read(fd, &from, sizeof(from)); 328 buf_read(fd, &to, sizeof(to)); 329 330 /* Assemble the headers... */ 331 if (to.s_addr == INADDR_BROADCAST) 332 assemble_hw_header(interface, buf, &bufp); 333 assemble_udp_ip_header(buf, &bufp, from.s_addr, to.s_addr, 334 htons(REMOTE_PORT), (unsigned char *)&raw, len); 335 336 iov[0].iov_base = buf; 337 iov[0].iov_len = bufp; 338 iov[1].iov_base = &raw; 339 iov[1].iov_len = len; 340 341 /* Fire it off */ 342 if (to.s_addr == INADDR_BROADCAST) 343 result = writev(interface->wfdesc, iov, 2); 344 else { 345 struct sockaddr_in sato; 346 347 sato.sin_addr = to; 348 sato.sin_port = htons(REMOTE_PORT); 349 sato.sin_family = AF_INET; 350 sato.sin_len = sizeof(sato); 351 352 memset(&msg, 0, sizeof(msg)); 353 msg.msg_name = (struct sockaddr *)&sato; 354 msg.msg_namelen = sizeof(sato); 355 msg.msg_iov = iov; 356 msg.msg_iovlen = 2; 357 result = sendmsg(interface->ufdesc, &msg, 0); 358 } 359 360 if (result < 0) 361 warning("send_packet: %m"); 362 } 363 364 ssize_t 365 receive_packet(struct interface_info *interface, unsigned char *buf, 366 size_t len, struct sockaddr_in *from, struct hardware *hfrom) 367 { 368 int length = 0, offset = 0; 369 struct bpf_hdr hdr; 370 371 /* 372 * All this complexity is because BPF doesn't guarantee that 373 * only one packet will be returned at a time. We're getting 374 * what we deserve, though - this is a terrible abuse of the BPF 375 * interface. Sigh. 376 */ 377 378 /* Process packets until we get one we can return or until we've 379 * done a read and gotten nothing we can return... 380 */ 381 do { 382 /* If the buffer is empty, fill it. */ 383 if (interface->rbuf_offset >= interface->rbuf_len) { 384 length = read(interface->rfdesc, interface->rbuf, 385 interface->rbuf_max); 386 if (length <= 0) 387 return (length); 388 interface->rbuf_offset = 0; 389 interface->rbuf_len = length; 390 } 391 392 /* 393 * If there isn't room for a whole bpf header, something 394 * went wrong, but we'll ignore it and hope it goes 395 * away... XXX 396 */ 397 if (interface->rbuf_len - interface->rbuf_offset < 398 sizeof(hdr)) { 399 interface->rbuf_offset = interface->rbuf_len; 400 continue; 401 } 402 403 /* Copy out a bpf header... */ 404 memcpy(&hdr, &interface->rbuf[interface->rbuf_offset], 405 sizeof(hdr)); 406 407 /* 408 * If the bpf header plus data doesn't fit in what's 409 * left of the buffer, stick head in sand yet again... 410 */ 411 if (interface->rbuf_offset + hdr.bh_hdrlen + hdr.bh_caplen > 412 interface->rbuf_len) { 413 interface->rbuf_offset = interface->rbuf_len; 414 continue; 415 } 416 417 /* Skip over the BPF header... */ 418 interface->rbuf_offset += hdr.bh_hdrlen; 419 420 /* 421 * If the captured data wasn't the whole packet, or if 422 * the packet won't fit in the input buffer, all we can 423 * do is drop it. 424 */ 425 if (hdr.bh_caplen != hdr.bh_datalen) { 426 interface->rbuf_offset = 427 BPF_WORDALIGN(interface->rbuf_offset + 428 hdr.bh_caplen); 429 continue; 430 } 431 432 /* Decode the physical header... */ 433 offset = decode_hw_header(interface->rbuf, 434 interface->rbuf_offset, hfrom); 435 436 /* 437 * If a physical layer checksum failed (dunno of any 438 * physical layer that supports this, but WTH), skip 439 * this packet. 440 */ 441 if (offset < 0) { 442 interface->rbuf_offset = 443 BPF_WORDALIGN(interface->rbuf_offset + 444 hdr.bh_caplen); 445 continue; 446 } 447 interface->rbuf_offset += offset; 448 hdr.bh_caplen -= offset; 449 450 /* Decode the IP and UDP headers... */ 451 offset = decode_udp_ip_header(interface->rbuf, 452 interface->rbuf_offset, from, NULL, hdr.bh_caplen); 453 454 /* If the IP or UDP checksum was bad, skip the packet... */ 455 if (offset < 0) { 456 interface->rbuf_offset = 457 BPF_WORDALIGN(interface->rbuf_offset + 458 hdr.bh_caplen); 459 continue; 460 } 461 interface->rbuf_offset += offset; 462 hdr.bh_caplen -= offset; 463 464 /* 465 * If there's not enough room to stash the packet data, 466 * we have to skip it (this shouldn't happen in real 467 * life, though). 468 */ 469 if (hdr.bh_caplen > len) { 470 interface->rbuf_offset = 471 BPF_WORDALIGN(interface->rbuf_offset + 472 hdr.bh_caplen); 473 continue; 474 } 475 476 /* Copy out the data in the packet... */ 477 memcpy(buf, interface->rbuf + interface->rbuf_offset, 478 hdr.bh_caplen); 479 interface->rbuf_offset = 480 BPF_WORDALIGN(interface->rbuf_offset + 481 hdr.bh_caplen); 482 return (hdr.bh_caplen); 483 } while (!length); 484 return (0); 485 } 486