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/capability.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 struct bpf_version v; 136 struct bpf_program p; 137 int sock, on = 1; 138 139 /* Open a BPF device and hang it on this interface... */ 140 info->wfdesc = if_register_bpf(info, O_WRONLY); 141 142 /* Make sure the BPF version is in range... */ 143 if (ioctl(info->wfdesc, BIOCVERSION, &v) < 0) 144 error("Can't get BPF version: %m"); 145 146 if (v.bv_major != BPF_MAJOR_VERSION || 147 v.bv_minor < BPF_MINOR_VERSION) 148 error("Kernel BPF version out of range - recompile dhcpd!"); 149 150 /* Set up the bpf write filter program structure. */ 151 p.bf_len = dhcp_bpf_wfilter_len; 152 p.bf_insns = dhcp_bpf_wfilter; 153 154 if (dhcp_bpf_wfilter[7].k == 0x1fff) 155 dhcp_bpf_wfilter[7].k = htons(IP_MF|IP_OFFMASK); 156 157 if (ioctl(info->wfdesc, BIOCSETWF, &p) < 0) 158 error("Can't install write filter program: %m"); 159 160 if (ioctl(info->wfdesc, BIOCLOCK, NULL) < 0) 161 error("Cannot lock bpf"); 162 163 if (cap_rights_limit(info->wfdesc, CAP_WRITE) < 0 && errno != ENOSYS) 164 error("Can't limit bpf descriptor: %m"); 165 166 /* 167 * Use raw socket for unicast send. 168 */ 169 if ((sock = socket(AF_INET, SOCK_RAW, IPPROTO_UDP)) == -1) 170 error("socket(SOCK_RAW): %m"); 171 if (setsockopt(sock, IPPROTO_IP, IP_HDRINCL, &on, 172 sizeof(on)) == -1) 173 error("setsockopt(IP_HDRINCL): %m"); 174 info->ufdesc = sock; 175 } 176 177 /* 178 * Packet filter program... 179 * 180 * XXX: Changes to the filter program may require changes to the 181 * constant offsets used in if_register_send to patch the BPF program! 182 */ 183 struct bpf_insn dhcp_bpf_filter[] = { 184 /* Make sure this is an IP packet... */ 185 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12), 186 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 8), 187 188 /* Make sure it's a UDP packet... */ 189 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23), 190 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 6), 191 192 /* Make sure this isn't a fragment... */ 193 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20), 194 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 4, 0), 195 196 /* Get the IP header length... */ 197 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14), 198 199 /* Make sure it's to the right port... */ 200 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16), 201 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1), /* patch */ 202 203 /* If we passed all the tests, ask for the whole packet. */ 204 BPF_STMT(BPF_RET+BPF_K, (u_int)-1), 205 206 /* Otherwise, drop it. */ 207 BPF_STMT(BPF_RET+BPF_K, 0), 208 }; 209 210 int dhcp_bpf_filter_len = sizeof(dhcp_bpf_filter) / sizeof(struct bpf_insn); 211 212 void 213 if_register_receive(struct interface_info *info) 214 { 215 static const unsigned long cmds[2] = { SIOCGIFFLAGS, SIOCGIFMEDIA }; 216 struct bpf_version v; 217 struct bpf_program p; 218 int flag = 1, sz; 219 220 /* Open a BPF device and hang it on this interface... */ 221 info->rfdesc = if_register_bpf(info, O_RDONLY); 222 223 /* Make sure the BPF version is in range... */ 224 if (ioctl(info->rfdesc, BIOCVERSION, &v) < 0) 225 error("Can't get BPF version: %m"); 226 227 if (v.bv_major != BPF_MAJOR_VERSION || 228 v.bv_minor < BPF_MINOR_VERSION) 229 error("Kernel BPF version out of range - recompile dhcpd!"); 230 231 /* 232 * Set immediate mode so that reads return as soon as a packet 233 * comes in, rather than waiting for the input buffer to fill 234 * with packets. 235 */ 236 if (ioctl(info->rfdesc, BIOCIMMEDIATE, &flag) < 0) 237 error("Can't set immediate mode on bpf device: %m"); 238 239 /* Get the required BPF buffer length from the kernel. */ 240 if (ioctl(info->rfdesc, BIOCGBLEN, &sz) < 0) 241 error("Can't get bpf buffer length: %m"); 242 info->rbuf_max = sz; 243 info->rbuf = malloc(info->rbuf_max); 244 if (!info->rbuf) 245 error("Can't allocate %lu bytes for bpf input buffer.", 246 (unsigned long)info->rbuf_max); 247 info->rbuf_offset = 0; 248 info->rbuf_len = 0; 249 250 /* Set up the bpf filter program structure. */ 251 p.bf_len = dhcp_bpf_filter_len; 252 p.bf_insns = dhcp_bpf_filter; 253 254 /* Patch the server port into the BPF program... 255 * 256 * XXX: changes to filter program may require changes to the 257 * insn number(s) used below! 258 */ 259 dhcp_bpf_filter[8].k = LOCAL_PORT; 260 261 if (ioctl(info->rfdesc, BIOCSETF, &p) < 0) 262 error("Can't install packet filter program: %m"); 263 264 if (ioctl(info->rfdesc, BIOCLOCK, NULL) < 0) 265 error("Cannot lock bpf"); 266 267 if (cap_rights_limit(info->rfdesc, 268 CAP_IOCTL | CAP_POLL_EVENT | CAP_READ) < 0 && errno != ENOSYS) { 269 error("Can't limit bpf descriptor: %m"); 270 } 271 if (cap_ioctls_limit(info->rfdesc, cmds, 2) < 0 && errno != ENOSYS) 272 error("Can't limit ioctls for bpf descriptor: %m"); 273 } 274 275 void 276 send_packet_unpriv(int privfd, struct dhcp_packet *raw, size_t len, 277 struct in_addr from, struct in_addr to) 278 { 279 struct imsg_hdr hdr; 280 struct buf *buf; 281 int errs; 282 283 hdr.code = IMSG_SEND_PACKET; 284 hdr.len = sizeof(hdr) + 285 sizeof(size_t) + len + 286 sizeof(from) + sizeof(to); 287 288 if ((buf = buf_open(hdr.len)) == NULL) 289 error("buf_open: %m"); 290 291 errs = 0; 292 errs += buf_add(buf, &hdr, sizeof(hdr)); 293 errs += buf_add(buf, &len, sizeof(len)); 294 errs += buf_add(buf, raw, len); 295 errs += buf_add(buf, &from, sizeof(from)); 296 errs += buf_add(buf, &to, sizeof(to)); 297 if (errs) 298 error("buf_add: %m"); 299 300 if (buf_close(privfd, buf) == -1) 301 error("buf_close: %m"); 302 } 303 304 void 305 send_packet_priv(struct interface_info *interface, struct imsg_hdr *hdr, int fd) 306 { 307 unsigned char buf[256]; 308 struct iovec iov[2]; 309 struct msghdr msg; 310 struct dhcp_packet raw; 311 size_t len; 312 struct in_addr from, to; 313 int result, bufp = 0; 314 315 if (hdr->len < sizeof(*hdr) + sizeof(size_t)) 316 error("corrupted message received"); 317 buf_read(fd, &len, sizeof(len)); 318 if (hdr->len != sizeof(*hdr) + sizeof(size_t) + len + 319 sizeof(from) + sizeof(to)) { 320 error("corrupted message received"); 321 } 322 if (len > sizeof(raw)) 323 error("corrupted message received"); 324 buf_read(fd, &raw, len); 325 buf_read(fd, &from, sizeof(from)); 326 buf_read(fd, &to, sizeof(to)); 327 328 /* Assemble the headers... */ 329 if (to.s_addr == INADDR_BROADCAST) 330 assemble_hw_header(interface, buf, &bufp); 331 assemble_udp_ip_header(buf, &bufp, from.s_addr, to.s_addr, 332 htons(REMOTE_PORT), (unsigned char *)&raw, len); 333 334 iov[0].iov_base = buf; 335 iov[0].iov_len = bufp; 336 iov[1].iov_base = &raw; 337 iov[1].iov_len = len; 338 339 /* Fire it off */ 340 if (to.s_addr == INADDR_BROADCAST) 341 result = writev(interface->wfdesc, iov, 2); 342 else { 343 struct sockaddr_in sato; 344 345 sato.sin_addr = to; 346 sato.sin_port = htons(REMOTE_PORT); 347 sato.sin_family = AF_INET; 348 sato.sin_len = sizeof(sato); 349 350 memset(&msg, 0, sizeof(msg)); 351 msg.msg_name = (struct sockaddr *)&sato; 352 msg.msg_namelen = sizeof(sato); 353 msg.msg_iov = iov; 354 msg.msg_iovlen = 2; 355 result = sendmsg(interface->ufdesc, &msg, 0); 356 } 357 358 if (result < 0) 359 warning("send_packet: %m"); 360 } 361 362 ssize_t 363 receive_packet(struct interface_info *interface, unsigned char *buf, 364 size_t len, struct sockaddr_in *from, struct hardware *hfrom) 365 { 366 int length = 0, offset = 0; 367 struct bpf_hdr hdr; 368 369 /* 370 * All this complexity is because BPF doesn't guarantee that 371 * only one packet will be returned at a time. We're getting 372 * what we deserve, though - this is a terrible abuse of the BPF 373 * interface. Sigh. 374 */ 375 376 /* Process packets until we get one we can return or until we've 377 * done a read and gotten nothing we can return... 378 */ 379 do { 380 /* If the buffer is empty, fill it. */ 381 if (interface->rbuf_offset >= interface->rbuf_len) { 382 length = read(interface->rfdesc, interface->rbuf, 383 interface->rbuf_max); 384 if (length <= 0) 385 return (length); 386 interface->rbuf_offset = 0; 387 interface->rbuf_len = length; 388 } 389 390 /* 391 * If there isn't room for a whole bpf header, something 392 * went wrong, but we'll ignore it and hope it goes 393 * away... XXX 394 */ 395 if (interface->rbuf_len - interface->rbuf_offset < 396 sizeof(hdr)) { 397 interface->rbuf_offset = interface->rbuf_len; 398 continue; 399 } 400 401 /* Copy out a bpf header... */ 402 memcpy(&hdr, &interface->rbuf[interface->rbuf_offset], 403 sizeof(hdr)); 404 405 /* 406 * If the bpf header plus data doesn't fit in what's 407 * left of the buffer, stick head in sand yet again... 408 */ 409 if (interface->rbuf_offset + hdr.bh_hdrlen + hdr.bh_caplen > 410 interface->rbuf_len) { 411 interface->rbuf_offset = interface->rbuf_len; 412 continue; 413 } 414 415 /* Skip over the BPF header... */ 416 interface->rbuf_offset += hdr.bh_hdrlen; 417 418 /* 419 * If the captured data wasn't the whole packet, or if 420 * the packet won't fit in the input buffer, all we can 421 * do is drop it. 422 */ 423 if (hdr.bh_caplen != hdr.bh_datalen) { 424 interface->rbuf_offset = 425 BPF_WORDALIGN(interface->rbuf_offset + 426 hdr.bh_caplen); 427 continue; 428 } 429 430 /* Decode the physical header... */ 431 offset = decode_hw_header(interface->rbuf, 432 interface->rbuf_offset, hfrom); 433 434 /* 435 * If a physical layer checksum failed (dunno of any 436 * physical layer that supports this, but WTH), skip 437 * this packet. 438 */ 439 if (offset < 0) { 440 interface->rbuf_offset = 441 BPF_WORDALIGN(interface->rbuf_offset + 442 hdr.bh_caplen); 443 continue; 444 } 445 interface->rbuf_offset += offset; 446 hdr.bh_caplen -= offset; 447 448 /* Decode the IP and UDP headers... */ 449 offset = decode_udp_ip_header(interface->rbuf, 450 interface->rbuf_offset, from, NULL, hdr.bh_caplen); 451 452 /* If the IP or UDP checksum was bad, skip the packet... */ 453 if (offset < 0) { 454 interface->rbuf_offset = 455 BPF_WORDALIGN(interface->rbuf_offset + 456 hdr.bh_caplen); 457 continue; 458 } 459 interface->rbuf_offset += offset; 460 hdr.bh_caplen -= offset; 461 462 /* 463 * If there's not enough room to stash the packet data, 464 * we have to skip it (this shouldn't happen in real 465 * life, though). 466 */ 467 if (hdr.bh_caplen > len) { 468 interface->rbuf_offset = 469 BPF_WORDALIGN(interface->rbuf_offset + 470 hdr.bh_caplen); 471 continue; 472 } 473 474 /* Copy out the data in the packet... */ 475 memcpy(buf, interface->rbuf + interface->rbuf_offset, 476 hdr.bh_caplen); 477 interface->rbuf_offset = 478 BPF_WORDALIGN(interface->rbuf_offset + 479 hdr.bh_caplen); 480 return (hdr.bh_caplen); 481 } while (!length); 482 return (0); 483 } 484