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 #include <capsicum_helpers.h> 61 62 #define BPF_FORMAT "/dev/bpf%d" 63 64 /* 65 * Called by get_interface_list for each interface that's discovered. 66 * Opens a packet filter for each interface and adds it to the select 67 * mask. 68 */ 69 int 70 if_register_bpf(struct interface_info *info, int flags) 71 { 72 char filename[50]; 73 int sock, b; 74 75 /* Open a BPF device */ 76 for (b = 0;; b++) { 77 snprintf(filename, sizeof(filename), BPF_FORMAT, b); 78 sock = open(filename, flags); 79 if (sock < 0) { 80 if (errno == EBUSY) 81 continue; 82 else 83 error("Can't find free bpf: %m"); 84 } else 85 break; 86 } 87 88 /* Set the BPF device to point at this interface. */ 89 if (ioctl(sock, BIOCSETIF, info->ifp) < 0) 90 error("Can't attach interface %s to bpf device %s: %m", 91 info->name, filename); 92 93 return (sock); 94 } 95 96 /* 97 * Packet write filter program: 98 * 'ip and udp and src port bootps and dst port (bootps or bootpc)' 99 */ 100 static struct bpf_insn dhcp_bpf_wfilter[] = { 101 BPF_STMT(BPF_LD + BPF_B + BPF_IND, 14), 102 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, (IPVERSION << 4) + 5, 0, 12), 103 104 /* Make sure this is an IP packet... */ 105 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12), 106 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 10), 107 108 /* Make sure it's a UDP packet... */ 109 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23), 110 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 8), 111 112 /* Make sure this isn't a fragment... */ 113 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20), 114 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 6, 0), /* patched */ 115 116 /* Get the IP header length... */ 117 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14), 118 119 /* Make sure it's from the right port... */ 120 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 14), 121 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 68, 0, 3), 122 123 /* Make sure it is to the right ports ... */ 124 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16), 125 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1), 126 127 /* If we passed all the tests, ask for the whole packet. */ 128 BPF_STMT(BPF_RET+BPF_K, (u_int)-1), 129 130 /* Otherwise, drop it. */ 131 BPF_STMT(BPF_RET+BPF_K, 0), 132 }; 133 134 static int dhcp_bpf_wfilter_len = nitems(dhcp_bpf_wfilter); 135 136 void 137 if_register_send(struct interface_info *info) 138 { 139 cap_rights_t rights; 140 struct bpf_version v; 141 struct bpf_program p; 142 int sock, on = 1; 143 144 /* Open a BPF device and hang it on this interface... */ 145 info->wfdesc = if_register_bpf(info, O_WRONLY); 146 147 /* Make sure the BPF version is in range... */ 148 if (ioctl(info->wfdesc, BIOCVERSION, &v) < 0) 149 error("Can't get BPF version: %m"); 150 151 if (v.bv_major != BPF_MAJOR_VERSION || 152 v.bv_minor < BPF_MINOR_VERSION) 153 error("Kernel BPF version out of range - recompile dhcpd!"); 154 155 /* Set up the bpf write filter program structure. */ 156 p.bf_len = dhcp_bpf_wfilter_len; 157 p.bf_insns = dhcp_bpf_wfilter; 158 159 if (dhcp_bpf_wfilter[7].k == 0x1fff) 160 dhcp_bpf_wfilter[7].k = htons(IP_MF|IP_OFFMASK); 161 162 if (ioctl(info->wfdesc, BIOCSETWF, &p) < 0) 163 error("Can't install write filter program: %m"); 164 165 if (ioctl(info->wfdesc, BIOCLOCK, NULL) < 0) 166 error("Cannot lock bpf"); 167 168 cap_rights_init(&rights, CAP_WRITE); 169 if (caph_rights_limit(info->wfdesc, &rights) < 0) 170 error("Can't limit bpf descriptor: %m"); 171 172 /* 173 * Use raw socket for unicast send. 174 */ 175 if ((sock = socket(AF_INET, SOCK_RAW, IPPROTO_UDP)) == -1) 176 error("socket(SOCK_RAW): %m"); 177 if (setsockopt(sock, IPPROTO_IP, IP_HDRINCL, &on, 178 sizeof(on)) == -1) 179 error("setsockopt(IP_HDRINCL): %m"); 180 info->ufdesc = sock; 181 } 182 183 /* 184 * Packet filter program... 185 * 186 * XXX: Changes to the filter program may require changes to the 187 * constant offsets used in if_register_send to patch the BPF program! 188 */ 189 static struct bpf_insn dhcp_bpf_filter[] = { 190 /* Make sure this is an IP packet... */ 191 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12), 192 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 8), 193 194 /* Make sure it's a UDP packet... */ 195 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23), 196 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 6), 197 198 /* Make sure this isn't a fragment... */ 199 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20), 200 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 4, 0), 201 202 /* Get the IP header length... */ 203 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14), 204 205 /* Make sure it's to the right port... */ 206 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16), 207 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1), /* patch */ 208 209 /* If we passed all the tests, ask for the whole packet. */ 210 BPF_STMT(BPF_RET+BPF_K, (u_int)-1), 211 212 /* Otherwise, drop it. */ 213 BPF_STMT(BPF_RET+BPF_K, 0), 214 }; 215 216 static int dhcp_bpf_filter_len = nitems(dhcp_bpf_filter); 217 218 void 219 if_register_receive(struct interface_info *info) 220 { 221 static const unsigned long cmds[2] = { SIOCGIFFLAGS, SIOCGIFMEDIA }; 222 cap_rights_t rights; 223 struct bpf_version v; 224 struct bpf_program p; 225 int flag = 1, sz; 226 227 /* Open a BPF device and hang it on this interface... */ 228 info->rfdesc = if_register_bpf(info, O_RDONLY); 229 230 /* Make sure the BPF version is in range... */ 231 if (ioctl(info->rfdesc, BIOCVERSION, &v) < 0) 232 error("Can't get BPF version: %m"); 233 234 if (v.bv_major != BPF_MAJOR_VERSION || 235 v.bv_minor < BPF_MINOR_VERSION) 236 error("Kernel BPF version out of range - recompile dhcpd!"); 237 238 /* 239 * Set immediate mode so that reads return as soon as a packet 240 * comes in, rather than waiting for the input buffer to fill 241 * with packets. 242 */ 243 if (ioctl(info->rfdesc, BIOCIMMEDIATE, &flag) < 0) 244 error("Can't set immediate mode on bpf device: %m"); 245 246 /* Get the required BPF buffer length from the kernel. */ 247 if (ioctl(info->rfdesc, BIOCGBLEN, &sz) < 0) 248 error("Can't get bpf buffer length: %m"); 249 info->rbuf_max = sz; 250 info->rbuf = malloc(info->rbuf_max); 251 if (!info->rbuf) 252 error("Can't allocate %lu bytes for bpf input buffer.", 253 (unsigned long)info->rbuf_max); 254 info->rbuf_offset = 0; 255 info->rbuf_len = 0; 256 257 /* Set up the bpf filter program structure. */ 258 p.bf_len = dhcp_bpf_filter_len; 259 p.bf_insns = dhcp_bpf_filter; 260 261 /* Patch the server port into the BPF program... 262 * 263 * XXX: changes to filter program may require changes to the 264 * insn number(s) used below! 265 */ 266 dhcp_bpf_filter[8].k = LOCAL_PORT; 267 268 if (ioctl(info->rfdesc, BIOCSETF, &p) < 0) 269 error("Can't install packet filter program: %m"); 270 271 if (ioctl(info->rfdesc, BIOCLOCK, NULL) < 0) 272 error("Cannot lock bpf"); 273 274 cap_rights_init(&rights, CAP_IOCTL, CAP_EVENT, CAP_READ); 275 if (caph_rights_limit(info->rfdesc, &rights) < 0) 276 error("Can't limit bpf descriptor: %m"); 277 if (caph_ioctls_limit(info->rfdesc, cmds, 2) < 0) 278 error("Can't limit ioctls for bpf descriptor: %m"); 279 } 280 281 void 282 send_packet_unpriv(int privfd, struct dhcp_packet *raw, size_t len, 283 struct in_addr from, struct in_addr to) 284 { 285 struct imsg_hdr hdr; 286 struct buf *buf; 287 int errs; 288 289 hdr.code = IMSG_SEND_PACKET; 290 hdr.len = sizeof(hdr) + 291 sizeof(size_t) + len + 292 sizeof(from) + sizeof(to); 293 294 if ((buf = buf_open(hdr.len)) == NULL) 295 error("buf_open: %m"); 296 297 errs = 0; 298 errs += buf_add(buf, &hdr, sizeof(hdr)); 299 errs += buf_add(buf, &len, sizeof(len)); 300 errs += buf_add(buf, raw, len); 301 errs += buf_add(buf, &from, sizeof(from)); 302 errs += buf_add(buf, &to, sizeof(to)); 303 if (errs) 304 error("buf_add: %m"); 305 306 if (buf_close(privfd, buf) == -1) 307 error("buf_close: %m"); 308 } 309 310 void 311 send_packet_priv(struct interface_info *interface, struct imsg_hdr *hdr, int fd) 312 { 313 unsigned char buf[256]; 314 struct iovec iov[2]; 315 struct msghdr msg; 316 struct dhcp_packet raw; 317 size_t len; 318 struct in_addr from, to; 319 int result, bufp = 0; 320 321 if (hdr->len < sizeof(*hdr) + sizeof(size_t)) 322 error("corrupted message received"); 323 buf_read(fd, &len, sizeof(len)); 324 if (hdr->len != sizeof(*hdr) + sizeof(size_t) + len + 325 sizeof(from) + sizeof(to)) { 326 error("corrupted message received"); 327 } 328 if (len > sizeof(raw)) 329 error("corrupted message received"); 330 buf_read(fd, &raw, len); 331 buf_read(fd, &from, sizeof(from)); 332 buf_read(fd, &to, sizeof(to)); 333 334 /* Assemble the headers... */ 335 if (to.s_addr == INADDR_BROADCAST) 336 assemble_hw_header(interface, buf, &bufp); 337 assemble_udp_ip_header(buf, &bufp, from.s_addr, to.s_addr, 338 htons(REMOTE_PORT), (unsigned char *)&raw, len); 339 340 iov[0].iov_base = buf; 341 iov[0].iov_len = bufp; 342 iov[1].iov_base = &raw; 343 iov[1].iov_len = len; 344 345 /* Fire it off */ 346 if (to.s_addr == INADDR_BROADCAST) 347 result = writev(interface->wfdesc, iov, 2); 348 else { 349 struct sockaddr_in sato; 350 351 sato.sin_addr = to; 352 sato.sin_port = htons(REMOTE_PORT); 353 sato.sin_family = AF_INET; 354 sato.sin_len = sizeof(sato); 355 356 memset(&msg, 0, sizeof(msg)); 357 msg.msg_name = (struct sockaddr *)&sato; 358 msg.msg_namelen = sizeof(sato); 359 msg.msg_iov = iov; 360 msg.msg_iovlen = 2; 361 result = sendmsg(interface->ufdesc, &msg, 0); 362 } 363 364 if (result < 0) 365 warning("send_packet: %m"); 366 } 367 368 ssize_t 369 receive_packet(struct interface_info *interface, unsigned char *buf, 370 size_t len, struct sockaddr_in *from, struct hardware *hfrom) 371 { 372 int length = 0, offset = 0; 373 struct bpf_hdr hdr; 374 375 /* 376 * All this complexity is because BPF doesn't guarantee that 377 * only one packet will be returned at a time. We're getting 378 * what we deserve, though - this is a terrible abuse of the BPF 379 * interface. Sigh. 380 */ 381 382 /* Process packets until we get one we can return or until we've 383 * done a read and gotten nothing we can return... 384 */ 385 do { 386 /* If the buffer is empty, fill it. */ 387 if (interface->rbuf_offset >= interface->rbuf_len) { 388 length = read(interface->rfdesc, interface->rbuf, 389 interface->rbuf_max); 390 if (length <= 0) 391 return (length); 392 interface->rbuf_offset = 0; 393 interface->rbuf_len = length; 394 } 395 396 /* 397 * If there isn't room for a whole bpf header, something 398 * went wrong, but we'll ignore it and hope it goes 399 * away... XXX 400 */ 401 if (interface->rbuf_len - interface->rbuf_offset < 402 sizeof(hdr)) { 403 interface->rbuf_offset = interface->rbuf_len; 404 continue; 405 } 406 407 /* Copy out a bpf header... */ 408 memcpy(&hdr, &interface->rbuf[interface->rbuf_offset], 409 sizeof(hdr)); 410 411 /* 412 * If the bpf header plus data doesn't fit in what's 413 * left of the buffer, stick head in sand yet again... 414 */ 415 if (interface->rbuf_offset + hdr.bh_hdrlen + hdr.bh_caplen > 416 interface->rbuf_len) { 417 interface->rbuf_offset = interface->rbuf_len; 418 continue; 419 } 420 421 /* Skip over the BPF header... */ 422 interface->rbuf_offset += hdr.bh_hdrlen; 423 424 /* 425 * If the captured data wasn't the whole packet, or if 426 * the packet won't fit in the input buffer, all we can 427 * do is drop it. 428 */ 429 if (hdr.bh_caplen != hdr.bh_datalen) { 430 interface->rbuf_offset = 431 BPF_WORDALIGN(interface->rbuf_offset + 432 hdr.bh_caplen); 433 continue; 434 } 435 436 /* Decode the physical header... */ 437 offset = decode_hw_header(interface->rbuf, 438 interface->rbuf_offset, hfrom); 439 440 /* 441 * If a physical layer checksum failed (dunno of any 442 * physical layer that supports this, but WTH), skip 443 * this packet. 444 */ 445 if (offset < 0) { 446 interface->rbuf_offset = 447 BPF_WORDALIGN(interface->rbuf_offset + 448 hdr.bh_caplen); 449 continue; 450 } 451 interface->rbuf_offset += offset; 452 hdr.bh_caplen -= offset; 453 454 /* Decode the IP and UDP headers... */ 455 offset = decode_udp_ip_header(interface->rbuf, 456 interface->rbuf_offset, from, NULL, hdr.bh_caplen); 457 458 /* If the IP or UDP checksum was bad, skip the packet... */ 459 if (offset < 0) { 460 interface->rbuf_offset = 461 BPF_WORDALIGN(interface->rbuf_offset + 462 hdr.bh_caplen); 463 continue; 464 } 465 interface->rbuf_offset += offset; 466 hdr.bh_caplen -= offset; 467 468 /* 469 * If there's not enough room to stash the packet data, 470 * we have to skip it (this shouldn't happen in real 471 * life, though). 472 */ 473 if (hdr.bh_caplen > len) { 474 interface->rbuf_offset = 475 BPF_WORDALIGN(interface->rbuf_offset + 476 hdr.bh_caplen); 477 continue; 478 } 479 480 /* Copy out the data in the packet... */ 481 memcpy(buf, interface->rbuf + interface->rbuf_offset, 482 hdr.bh_caplen); 483 interface->rbuf_offset = 484 BPF_WORDALIGN(interface->rbuf_offset + 485 hdr.bh_caplen); 486 return (hdr.bh_caplen); 487 } while (!length); 488 return (0); 489 } 490