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 /* Tag the packets with the proper VLAN PCP setting. */ 94 if (info->client->config->vlan_pcp != 0) { 95 if (ioctl(sock, BIOCSETVLANPCP, 96 &info->client->config->vlan_pcp) < 0) 97 error( "Can't set the VLAN PCP tag on interface %s: %m", 98 info->name); 99 } 100 101 return (sock); 102 } 103 104 /* 105 * Packet write filter program: 106 * 'ip and udp and src port bootps and dst port (bootps or bootpc)' 107 */ 108 static struct bpf_insn dhcp_bpf_wfilter[] = { 109 BPF_STMT(BPF_LD + BPF_B + BPF_IND, 14), 110 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, (IPVERSION << 4) + 5, 0, 12), 111 112 /* Make sure this is an IP packet... */ 113 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12), 114 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 10), 115 116 /* Make sure it's a UDP packet... */ 117 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23), 118 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 8), 119 120 /* Make sure this isn't a fragment... */ 121 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20), 122 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 6, 0), /* patched */ 123 124 /* Get the IP header length... */ 125 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14), 126 127 /* Make sure it's from the right port... */ 128 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 14), 129 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 68, 0, 3), 130 131 /* Make sure it is to the right ports ... */ 132 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16), 133 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1), 134 135 /* If we passed all the tests, ask for the whole packet. */ 136 BPF_STMT(BPF_RET+BPF_K, (u_int)-1), 137 138 /* Otherwise, drop it. */ 139 BPF_STMT(BPF_RET+BPF_K, 0), 140 }; 141 142 static int dhcp_bpf_wfilter_len = nitems(dhcp_bpf_wfilter); 143 144 void 145 if_register_send(struct interface_info *info) 146 { 147 cap_rights_t rights; 148 struct bpf_version v; 149 struct bpf_program p; 150 int sock, on = 1; 151 152 /* Open a BPF device and hang it on this interface... */ 153 info->wfdesc = if_register_bpf(info, O_WRONLY); 154 155 /* Make sure the BPF version is in range... */ 156 if (ioctl(info->wfdesc, BIOCVERSION, &v) < 0) 157 error("Can't get BPF version: %m"); 158 159 if (v.bv_major != BPF_MAJOR_VERSION || 160 v.bv_minor < BPF_MINOR_VERSION) 161 error("Kernel BPF version out of range - recompile dhcpd!"); 162 163 /* Set up the bpf write filter program structure. */ 164 p.bf_len = dhcp_bpf_wfilter_len; 165 p.bf_insns = dhcp_bpf_wfilter; 166 167 if (dhcp_bpf_wfilter[7].k == 0x1fff) 168 dhcp_bpf_wfilter[7].k = htons(IP_MF|IP_OFFMASK); 169 170 if (ioctl(info->wfdesc, BIOCSETWF, &p) < 0) 171 error("Can't install write filter program: %m"); 172 173 if (ioctl(info->wfdesc, BIOCLOCK, NULL) < 0) 174 error("Cannot lock bpf"); 175 176 cap_rights_init(&rights, CAP_WRITE); 177 if (caph_rights_limit(info->wfdesc, &rights) < 0) 178 error("Can't limit bpf descriptor: %m"); 179 180 /* 181 * Use raw socket for unicast send. 182 */ 183 if ((sock = socket(AF_INET, SOCK_RAW, IPPROTO_UDP)) == -1) 184 error("socket(SOCK_RAW): %m"); 185 if (setsockopt(sock, IPPROTO_IP, IP_HDRINCL, &on, 186 sizeof(on)) == -1) 187 error("setsockopt(IP_HDRINCL): %m"); 188 info->ufdesc = sock; 189 } 190 191 /* 192 * Packet filter program... 193 * 194 * XXX: Changes to the filter program may require changes to the 195 * constant offsets used in if_register_send to patch the BPF program! 196 */ 197 static struct bpf_insn dhcp_bpf_filter[] = { 198 /* Make sure this is an IP packet... */ 199 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12), 200 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 8), 201 202 /* Make sure it's a UDP packet... */ 203 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23), 204 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 6), 205 206 /* Make sure this isn't a fragment... */ 207 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20), 208 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 4, 0), 209 210 /* Get the IP header length... */ 211 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14), 212 213 /* Make sure it's to the right port... */ 214 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16), 215 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1), /* patch */ 216 217 /* If we passed all the tests, ask for the whole packet. */ 218 BPF_STMT(BPF_RET+BPF_K, (u_int)-1), 219 220 /* Otherwise, drop it. */ 221 BPF_STMT(BPF_RET+BPF_K, 0), 222 }; 223 224 static int dhcp_bpf_filter_len = nitems(dhcp_bpf_filter); 225 226 void 227 if_register_receive(struct interface_info *info) 228 { 229 static const unsigned long cmds[2] = { SIOCGIFFLAGS, SIOCGIFMEDIA }; 230 cap_rights_t rights; 231 struct bpf_version v; 232 struct bpf_program p; 233 int flag = 1, sz; 234 235 /* Open a BPF device and hang it on this interface... */ 236 info->rfdesc = if_register_bpf(info, O_RDONLY); 237 238 /* Make sure the BPF version is in range... */ 239 if (ioctl(info->rfdesc, BIOCVERSION, &v) < 0) 240 error("Can't get BPF version: %m"); 241 242 if (v.bv_major != BPF_MAJOR_VERSION || 243 v.bv_minor < BPF_MINOR_VERSION) 244 error("Kernel BPF version out of range - recompile dhcpd!"); 245 246 /* 247 * Set immediate mode so that reads return as soon as a packet 248 * comes in, rather than waiting for the input buffer to fill 249 * with packets. 250 */ 251 if (ioctl(info->rfdesc, BIOCIMMEDIATE, &flag) < 0) 252 error("Can't set immediate mode on bpf device: %m"); 253 254 /* Get the required BPF buffer length from the kernel. */ 255 if (ioctl(info->rfdesc, BIOCGBLEN, &sz) < 0) 256 error("Can't get bpf buffer length: %m"); 257 info->rbuf_max = sz; 258 info->rbuf = malloc(info->rbuf_max); 259 if (!info->rbuf) 260 error("Can't allocate %lu bytes for bpf input buffer.", 261 (unsigned long)info->rbuf_max); 262 info->rbuf_offset = 0; 263 info->rbuf_len = 0; 264 265 /* Set up the bpf filter program structure. */ 266 p.bf_len = dhcp_bpf_filter_len; 267 p.bf_insns = dhcp_bpf_filter; 268 269 /* Patch the server port into the BPF program... 270 * 271 * XXX: changes to filter program may require changes to the 272 * insn number(s) used below! 273 */ 274 dhcp_bpf_filter[8].k = LOCAL_PORT; 275 276 if (ioctl(info->rfdesc, BIOCSETF, &p) < 0) 277 error("Can't install packet filter program: %m"); 278 279 if (ioctl(info->rfdesc, BIOCLOCK, NULL) < 0) 280 error("Cannot lock bpf"); 281 282 cap_rights_init(&rights, CAP_IOCTL, CAP_EVENT, CAP_READ); 283 if (caph_rights_limit(info->rfdesc, &rights) < 0) 284 error("Can't limit bpf descriptor: %m"); 285 if (caph_ioctls_limit(info->rfdesc, cmds, 2) < 0) 286 error("Can't limit ioctls for bpf descriptor: %m"); 287 } 288 289 void 290 send_packet_unpriv(int privfd, struct dhcp_packet *raw, size_t len, 291 struct in_addr from, struct in_addr to) 292 { 293 struct imsg_hdr hdr; 294 struct buf *buf; 295 int errs; 296 297 hdr.code = IMSG_SEND_PACKET; 298 hdr.len = sizeof(hdr) + 299 sizeof(size_t) + len + 300 sizeof(from) + sizeof(to); 301 302 if ((buf = buf_open(hdr.len)) == NULL) 303 error("buf_open: %m"); 304 305 errs = 0; 306 errs += buf_add(buf, &hdr, sizeof(hdr)); 307 errs += buf_add(buf, &len, sizeof(len)); 308 errs += buf_add(buf, raw, len); 309 errs += buf_add(buf, &from, sizeof(from)); 310 errs += buf_add(buf, &to, sizeof(to)); 311 if (errs) 312 error("buf_add: %m"); 313 314 if (buf_close(privfd, buf) == -1) 315 error("buf_close: %m"); 316 } 317 318 void 319 send_packet_priv(struct interface_info *interface, struct imsg_hdr *hdr, int fd) 320 { 321 unsigned char buf[256]; 322 struct iovec iov[2]; 323 struct msghdr msg; 324 struct dhcp_packet raw; 325 size_t len; 326 struct in_addr from, to; 327 int result, bufp = 0; 328 329 if (hdr->len < sizeof(*hdr) + sizeof(size_t)) 330 error("corrupted message received"); 331 buf_read(fd, &len, sizeof(len)); 332 if (hdr->len != sizeof(*hdr) + sizeof(size_t) + len + 333 sizeof(from) + sizeof(to)) { 334 error("corrupted message received"); 335 } 336 if (len > sizeof(raw)) 337 error("corrupted message received"); 338 buf_read(fd, &raw, len); 339 buf_read(fd, &from, sizeof(from)); 340 buf_read(fd, &to, sizeof(to)); 341 342 /* Assemble the headers... */ 343 if (to.s_addr == INADDR_BROADCAST) 344 assemble_hw_header(interface, buf, &bufp); 345 assemble_udp_ip_header(buf, &bufp, from.s_addr, to.s_addr, 346 htons(REMOTE_PORT), (unsigned char *)&raw, len); 347 348 iov[0].iov_base = buf; 349 iov[0].iov_len = bufp; 350 iov[1].iov_base = &raw; 351 iov[1].iov_len = len; 352 353 /* Fire it off */ 354 if (to.s_addr == INADDR_BROADCAST) 355 result = writev(interface->wfdesc, iov, 2); 356 else { 357 struct sockaddr_in sato; 358 359 sato.sin_addr = to; 360 sato.sin_port = htons(REMOTE_PORT); 361 sato.sin_family = AF_INET; 362 sato.sin_len = sizeof(sato); 363 364 memset(&msg, 0, sizeof(msg)); 365 msg.msg_name = (struct sockaddr *)&sato; 366 msg.msg_namelen = sizeof(sato); 367 msg.msg_iov = iov; 368 msg.msg_iovlen = 2; 369 result = sendmsg(interface->ufdesc, &msg, 0); 370 } 371 372 if (result < 0) 373 warning("send_packet: %m"); 374 } 375 376 ssize_t 377 receive_packet(struct interface_info *interface, unsigned char *buf, 378 size_t len, struct sockaddr_in *from, struct hardware *hfrom) 379 { 380 int length = 0, offset = 0; 381 struct bpf_hdr hdr; 382 383 /* 384 * All this complexity is because BPF doesn't guarantee that 385 * only one packet will be returned at a time. We're getting 386 * what we deserve, though - this is a terrible abuse of the BPF 387 * interface. Sigh. 388 */ 389 390 /* Process packets until we get one we can return or until we've 391 * done a read and gotten nothing we can return... 392 */ 393 do { 394 /* If the buffer is empty, fill it. */ 395 if (interface->rbuf_offset >= interface->rbuf_len) { 396 length = read(interface->rfdesc, interface->rbuf, 397 interface->rbuf_max); 398 if (length <= 0) 399 return (length); 400 interface->rbuf_offset = 0; 401 interface->rbuf_len = length; 402 } 403 404 /* 405 * If there isn't room for a whole bpf header, something 406 * went wrong, but we'll ignore it and hope it goes 407 * away... XXX 408 */ 409 if (interface->rbuf_len - interface->rbuf_offset < 410 sizeof(hdr)) { 411 interface->rbuf_offset = interface->rbuf_len; 412 continue; 413 } 414 415 /* Copy out a bpf header... */ 416 memcpy(&hdr, &interface->rbuf[interface->rbuf_offset], 417 sizeof(hdr)); 418 419 /* 420 * If the bpf header plus data doesn't fit in what's 421 * left of the buffer, stick head in sand yet again... 422 */ 423 if (interface->rbuf_offset + hdr.bh_hdrlen + hdr.bh_caplen > 424 interface->rbuf_len) { 425 interface->rbuf_offset = interface->rbuf_len; 426 continue; 427 } 428 429 /* Skip over the BPF header... */ 430 interface->rbuf_offset += hdr.bh_hdrlen; 431 432 /* 433 * If the captured data wasn't the whole packet, or if 434 * the packet won't fit in the input buffer, all we can 435 * do is drop it. 436 */ 437 if (hdr.bh_caplen != hdr.bh_datalen) { 438 interface->rbuf_offset = 439 BPF_WORDALIGN(interface->rbuf_offset + 440 hdr.bh_caplen); 441 continue; 442 } 443 444 /* Decode the physical header... */ 445 offset = decode_hw_header(interface->rbuf, 446 interface->rbuf_offset, hfrom); 447 448 /* 449 * If a physical layer checksum failed (dunno of any 450 * physical layer that supports this, but WTH), skip 451 * this packet. 452 */ 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 /* Decode the IP and UDP headers... */ 463 offset = decode_udp_ip_header(interface->rbuf, 464 interface->rbuf_offset, from, NULL, hdr.bh_caplen); 465 466 /* If the IP or UDP checksum was bad, skip the packet... */ 467 if (offset < 0) { 468 interface->rbuf_offset = 469 BPF_WORDALIGN(interface->rbuf_offset + 470 hdr.bh_caplen); 471 continue; 472 } 473 interface->rbuf_offset += offset; 474 hdr.bh_caplen -= offset; 475 476 /* 477 * If there's not enough room to stash the packet data, 478 * we have to skip it (this shouldn't happen in real 479 * life, though). 480 */ 481 if (hdr.bh_caplen > len) { 482 interface->rbuf_offset = 483 BPF_WORDALIGN(interface->rbuf_offset + 484 hdr.bh_caplen); 485 continue; 486 } 487 488 /* Copy out the data in the packet... */ 489 memcpy(buf, interface->rbuf + interface->rbuf_offset, 490 hdr.bh_caplen); 491 interface->rbuf_offset = 492 BPF_WORDALIGN(interface->rbuf_offset + 493 hdr.bh_caplen); 494 return (hdr.bh_caplen); 495 } while (!length); 496 return (0); 497 } 498