1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * The options processing module for ip.c 8 * 9 * Authors: A.N.Kuznetsov 10 * 11 */ 12 13 #define pr_fmt(fmt) "IPv4: " fmt 14 15 #include <linux/capability.h> 16 #include <linux/module.h> 17 #include <linux/slab.h> 18 #include <linux/types.h> 19 #include <linux/uaccess.h> 20 #include <asm/unaligned.h> 21 #include <linux/skbuff.h> 22 #include <linux/ip.h> 23 #include <linux/icmp.h> 24 #include <linux/netdevice.h> 25 #include <linux/rtnetlink.h> 26 #include <net/sock.h> 27 #include <net/ip.h> 28 #include <net/icmp.h> 29 #include <net/route.h> 30 #include <net/cipso_ipv4.h> 31 #include <net/ip_fib.h> 32 33 /* 34 * Write options to IP header, record destination address to 35 * source route option, address of outgoing interface 36 * (we should already know it, so that this function is allowed be 37 * called only after routing decision) and timestamp, 38 * if we originate this datagram. 39 * 40 * daddr is real destination address, next hop is recorded in IP header. 41 * saddr is address of outgoing interface. 42 */ 43 44 void ip_options_build(struct sk_buff *skb, struct ip_options *opt, 45 __be32 daddr, struct rtable *rt, int is_frag) 46 { 47 unsigned char *iph = skb_network_header(skb); 48 49 memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options)); 50 memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen); 51 opt = &(IPCB(skb)->opt); 52 53 if (opt->srr) 54 memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4); 55 56 if (!is_frag) { 57 if (opt->rr_needaddr) 58 ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, skb, rt); 59 if (opt->ts_needaddr) 60 ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, skb, rt); 61 if (opt->ts_needtime) { 62 __be32 midtime; 63 64 midtime = inet_current_timestamp(); 65 memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4); 66 } 67 return; 68 } 69 if (opt->rr) { 70 memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]); 71 opt->rr = 0; 72 opt->rr_needaddr = 0; 73 } 74 if (opt->ts) { 75 memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]); 76 opt->ts = 0; 77 opt->ts_needaddr = opt->ts_needtime = 0; 78 } 79 } 80 81 /* 82 * Provided (sopt, skb) points to received options, 83 * build in dopt compiled option set appropriate for answering. 84 * i.e. invert SRR option, copy anothers, 85 * and grab room in RR/TS options. 86 * 87 * NOTE: dopt cannot point to skb. 88 */ 89 90 int __ip_options_echo(struct net *net, struct ip_options *dopt, 91 struct sk_buff *skb, const struct ip_options *sopt) 92 { 93 unsigned char *sptr, *dptr; 94 int soffset, doffset; 95 int optlen; 96 97 memset(dopt, 0, sizeof(struct ip_options)); 98 99 if (sopt->optlen == 0) 100 return 0; 101 102 sptr = skb_network_header(skb); 103 dptr = dopt->__data; 104 105 if (sopt->rr) { 106 optlen = sptr[sopt->rr+1]; 107 soffset = sptr[sopt->rr+2]; 108 dopt->rr = dopt->optlen + sizeof(struct iphdr); 109 memcpy(dptr, sptr+sopt->rr, optlen); 110 if (sopt->rr_needaddr && soffset <= optlen) { 111 if (soffset + 3 > optlen) 112 return -EINVAL; 113 dptr[2] = soffset + 4; 114 dopt->rr_needaddr = 1; 115 } 116 dptr += optlen; 117 dopt->optlen += optlen; 118 } 119 if (sopt->ts) { 120 optlen = sptr[sopt->ts+1]; 121 soffset = sptr[sopt->ts+2]; 122 dopt->ts = dopt->optlen + sizeof(struct iphdr); 123 memcpy(dptr, sptr+sopt->ts, optlen); 124 if (soffset <= optlen) { 125 if (sopt->ts_needaddr) { 126 if (soffset + 3 > optlen) 127 return -EINVAL; 128 dopt->ts_needaddr = 1; 129 soffset += 4; 130 } 131 if (sopt->ts_needtime) { 132 if (soffset + 3 > optlen) 133 return -EINVAL; 134 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) { 135 dopt->ts_needtime = 1; 136 soffset += 4; 137 } else { 138 dopt->ts_needtime = 0; 139 140 if (soffset + 7 <= optlen) { 141 __be32 addr; 142 143 memcpy(&addr, dptr+soffset-1, 4); 144 if (inet_addr_type(net, addr) != RTN_UNICAST) { 145 dopt->ts_needtime = 1; 146 soffset += 8; 147 } 148 } 149 } 150 } 151 dptr[2] = soffset; 152 } 153 dptr += optlen; 154 dopt->optlen += optlen; 155 } 156 if (sopt->srr) { 157 unsigned char *start = sptr+sopt->srr; 158 __be32 faddr; 159 160 optlen = start[1]; 161 soffset = start[2]; 162 doffset = 0; 163 if (soffset > optlen) 164 soffset = optlen + 1; 165 soffset -= 4; 166 if (soffset > 3) { 167 memcpy(&faddr, &start[soffset-1], 4); 168 for (soffset -= 4, doffset = 4; soffset > 3; soffset -= 4, doffset += 4) 169 memcpy(&dptr[doffset-1], &start[soffset-1], 4); 170 /* 171 * RFC1812 requires to fix illegal source routes. 172 */ 173 if (memcmp(&ip_hdr(skb)->saddr, 174 &start[soffset + 3], 4) == 0) 175 doffset -= 4; 176 } 177 if (doffset > 3) { 178 dopt->faddr = faddr; 179 dptr[0] = start[0]; 180 dptr[1] = doffset+3; 181 dptr[2] = 4; 182 dptr += doffset+3; 183 dopt->srr = dopt->optlen + sizeof(struct iphdr); 184 dopt->optlen += doffset+3; 185 dopt->is_strictroute = sopt->is_strictroute; 186 } 187 } 188 if (sopt->cipso) { 189 optlen = sptr[sopt->cipso+1]; 190 dopt->cipso = dopt->optlen+sizeof(struct iphdr); 191 memcpy(dptr, sptr+sopt->cipso, optlen); 192 dptr += optlen; 193 dopt->optlen += optlen; 194 } 195 while (dopt->optlen & 3) { 196 *dptr++ = IPOPT_END; 197 dopt->optlen++; 198 } 199 return 0; 200 } 201 202 /* 203 * Options "fragmenting", just fill options not 204 * allowed in fragments with NOOPs. 205 * Simple and stupid 8), but the most efficient way. 206 */ 207 208 void ip_options_fragment(struct sk_buff *skb) 209 { 210 unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr); 211 struct ip_options *opt = &(IPCB(skb)->opt); 212 int l = opt->optlen; 213 int optlen; 214 215 while (l > 0) { 216 switch (*optptr) { 217 case IPOPT_END: 218 return; 219 case IPOPT_NOOP: 220 l--; 221 optptr++; 222 continue; 223 } 224 optlen = optptr[1]; 225 if (optlen < 2 || optlen > l) 226 return; 227 if (!IPOPT_COPIED(*optptr)) 228 memset(optptr, IPOPT_NOOP, optlen); 229 l -= optlen; 230 optptr += optlen; 231 } 232 opt->ts = 0; 233 opt->rr = 0; 234 opt->rr_needaddr = 0; 235 opt->ts_needaddr = 0; 236 opt->ts_needtime = 0; 237 } 238 239 /* helper used by ip_options_compile() to call fib_compute_spec_dst() 240 * at most one time. 241 */ 242 static void spec_dst_fill(__be32 *spec_dst, struct sk_buff *skb) 243 { 244 if (*spec_dst == htonl(INADDR_ANY)) 245 *spec_dst = fib_compute_spec_dst(skb); 246 } 247 248 /* 249 * Verify options and fill pointers in struct options. 250 * Caller should clear *opt, and set opt->data. 251 * If opt == NULL, then skb->data should point to IP header. 252 */ 253 254 int __ip_options_compile(struct net *net, 255 struct ip_options *opt, struct sk_buff *skb, 256 __be32 *info) 257 { 258 __be32 spec_dst = htonl(INADDR_ANY); 259 unsigned char *pp_ptr = NULL; 260 struct rtable *rt = NULL; 261 unsigned char *optptr; 262 unsigned char *iph; 263 int optlen, l; 264 265 if (skb) { 266 rt = skb_rtable(skb); 267 optptr = (unsigned char *)&(ip_hdr(skb)[1]); 268 } else 269 optptr = opt->__data; 270 iph = optptr - sizeof(struct iphdr); 271 272 for (l = opt->optlen; l > 0; ) { 273 switch (*optptr) { 274 case IPOPT_END: 275 for (optptr++, l--; l > 0; optptr++, l--) { 276 if (*optptr != IPOPT_END) { 277 *optptr = IPOPT_END; 278 opt->is_changed = 1; 279 } 280 } 281 goto eol; 282 case IPOPT_NOOP: 283 l--; 284 optptr++; 285 continue; 286 } 287 if (unlikely(l < 2)) { 288 pp_ptr = optptr; 289 goto error; 290 } 291 optlen = optptr[1]; 292 if (optlen < 2 || optlen > l) { 293 pp_ptr = optptr; 294 goto error; 295 } 296 switch (*optptr) { 297 case IPOPT_SSRR: 298 case IPOPT_LSRR: 299 if (optlen < 3) { 300 pp_ptr = optptr + 1; 301 goto error; 302 } 303 if (optptr[2] < 4) { 304 pp_ptr = optptr + 2; 305 goto error; 306 } 307 /* NB: cf RFC-1812 5.2.4.1 */ 308 if (opt->srr) { 309 pp_ptr = optptr; 310 goto error; 311 } 312 if (!skb) { 313 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) { 314 pp_ptr = optptr + 1; 315 goto error; 316 } 317 memcpy(&opt->faddr, &optptr[3], 4); 318 if (optlen > 7) 319 memmove(&optptr[3], &optptr[7], optlen-7); 320 } 321 opt->is_strictroute = (optptr[0] == IPOPT_SSRR); 322 opt->srr = optptr - iph; 323 break; 324 case IPOPT_RR: 325 if (opt->rr) { 326 pp_ptr = optptr; 327 goto error; 328 } 329 if (optlen < 3) { 330 pp_ptr = optptr + 1; 331 goto error; 332 } 333 if (optptr[2] < 4) { 334 pp_ptr = optptr + 2; 335 goto error; 336 } 337 if (optptr[2] <= optlen) { 338 if (optptr[2]+3 > optlen) { 339 pp_ptr = optptr + 2; 340 goto error; 341 } 342 if (rt) { 343 spec_dst_fill(&spec_dst, skb); 344 memcpy(&optptr[optptr[2]-1], &spec_dst, 4); 345 opt->is_changed = 1; 346 } 347 optptr[2] += 4; 348 opt->rr_needaddr = 1; 349 } 350 opt->rr = optptr - iph; 351 break; 352 case IPOPT_TIMESTAMP: 353 if (opt->ts) { 354 pp_ptr = optptr; 355 goto error; 356 } 357 if (optlen < 4) { 358 pp_ptr = optptr + 1; 359 goto error; 360 } 361 if (optptr[2] < 5) { 362 pp_ptr = optptr + 2; 363 goto error; 364 } 365 if (optptr[2] <= optlen) { 366 unsigned char *timeptr = NULL; 367 if (optptr[2]+3 > optlen) { 368 pp_ptr = optptr + 2; 369 goto error; 370 } 371 switch (optptr[3]&0xF) { 372 case IPOPT_TS_TSONLY: 373 if (skb) 374 timeptr = &optptr[optptr[2]-1]; 375 opt->ts_needtime = 1; 376 optptr[2] += 4; 377 break; 378 case IPOPT_TS_TSANDADDR: 379 if (optptr[2]+7 > optlen) { 380 pp_ptr = optptr + 2; 381 goto error; 382 } 383 if (rt) { 384 spec_dst_fill(&spec_dst, skb); 385 memcpy(&optptr[optptr[2]-1], &spec_dst, 4); 386 timeptr = &optptr[optptr[2]+3]; 387 } 388 opt->ts_needaddr = 1; 389 opt->ts_needtime = 1; 390 optptr[2] += 8; 391 break; 392 case IPOPT_TS_PRESPEC: 393 if (optptr[2]+7 > optlen) { 394 pp_ptr = optptr + 2; 395 goto error; 396 } 397 { 398 __be32 addr; 399 memcpy(&addr, &optptr[optptr[2]-1], 4); 400 if (inet_addr_type(net, addr) == RTN_UNICAST) 401 break; 402 if (skb) 403 timeptr = &optptr[optptr[2]+3]; 404 } 405 opt->ts_needtime = 1; 406 optptr[2] += 8; 407 break; 408 default: 409 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) { 410 pp_ptr = optptr + 3; 411 goto error; 412 } 413 break; 414 } 415 if (timeptr) { 416 __be32 midtime; 417 418 midtime = inet_current_timestamp(); 419 memcpy(timeptr, &midtime, 4); 420 opt->is_changed = 1; 421 } 422 } else if ((optptr[3]&0xF) != IPOPT_TS_PRESPEC) { 423 unsigned int overflow = optptr[3]>>4; 424 if (overflow == 15) { 425 pp_ptr = optptr + 3; 426 goto error; 427 } 428 if (skb) { 429 optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4); 430 opt->is_changed = 1; 431 } 432 } 433 opt->ts = optptr - iph; 434 break; 435 case IPOPT_RA: 436 if (optlen < 4) { 437 pp_ptr = optptr + 1; 438 goto error; 439 } 440 if (optptr[2] == 0 && optptr[3] == 0) 441 opt->router_alert = optptr - iph; 442 break; 443 case IPOPT_CIPSO: 444 if ((!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) || opt->cipso) { 445 pp_ptr = optptr; 446 goto error; 447 } 448 opt->cipso = optptr - iph; 449 if (cipso_v4_validate(skb, &optptr)) { 450 pp_ptr = optptr; 451 goto error; 452 } 453 break; 454 case IPOPT_SEC: 455 case IPOPT_SID: 456 default: 457 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) { 458 pp_ptr = optptr; 459 goto error; 460 } 461 break; 462 } 463 l -= optlen; 464 optptr += optlen; 465 } 466 467 eol: 468 if (!pp_ptr) 469 return 0; 470 471 error: 472 if (info) 473 *info = htonl((pp_ptr-iph)<<24); 474 return -EINVAL; 475 } 476 EXPORT_SYMBOL(__ip_options_compile); 477 478 int ip_options_compile(struct net *net, 479 struct ip_options *opt, struct sk_buff *skb) 480 { 481 int ret; 482 __be32 info; 483 484 ret = __ip_options_compile(net, opt, skb, &info); 485 if (ret != 0 && skb) 486 icmp_send(skb, ICMP_PARAMETERPROB, 0, info); 487 return ret; 488 } 489 EXPORT_SYMBOL(ip_options_compile); 490 491 /* 492 * Undo all the changes done by ip_options_compile(). 493 */ 494 495 void ip_options_undo(struct ip_options *opt) 496 { 497 if (opt->srr) { 498 unsigned char *optptr = opt->__data+opt->srr-sizeof(struct iphdr); 499 memmove(optptr+7, optptr+3, optptr[1]-7); 500 memcpy(optptr+3, &opt->faddr, 4); 501 } 502 if (opt->rr_needaddr) { 503 unsigned char *optptr = opt->__data+opt->rr-sizeof(struct iphdr); 504 optptr[2] -= 4; 505 memset(&optptr[optptr[2]-1], 0, 4); 506 } 507 if (opt->ts) { 508 unsigned char *optptr = opt->__data+opt->ts-sizeof(struct iphdr); 509 if (opt->ts_needtime) { 510 optptr[2] -= 4; 511 memset(&optptr[optptr[2]-1], 0, 4); 512 if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC) 513 optptr[2] -= 4; 514 } 515 if (opt->ts_needaddr) { 516 optptr[2] -= 4; 517 memset(&optptr[optptr[2]-1], 0, 4); 518 } 519 } 520 } 521 522 int ip_options_get(struct net *net, struct ip_options_rcu **optp, 523 sockptr_t data, int optlen) 524 { 525 struct ip_options_rcu *opt; 526 527 opt = kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3), 528 GFP_KERNEL); 529 if (!opt) 530 return -ENOMEM; 531 if (optlen && copy_from_sockptr(opt->opt.__data, data, optlen)) { 532 kfree(opt); 533 return -EFAULT; 534 } 535 536 while (optlen & 3) 537 opt->opt.__data[optlen++] = IPOPT_END; 538 opt->opt.optlen = optlen; 539 if (optlen && ip_options_compile(net, &opt->opt, NULL)) { 540 kfree(opt); 541 return -EINVAL; 542 } 543 kfree(*optp); 544 *optp = opt; 545 return 0; 546 } 547 548 void ip_forward_options(struct sk_buff *skb) 549 { 550 struct ip_options *opt = &(IPCB(skb)->opt); 551 unsigned char *optptr; 552 struct rtable *rt = skb_rtable(skb); 553 unsigned char *raw = skb_network_header(skb); 554 555 if (opt->rr_needaddr) { 556 optptr = (unsigned char *)raw + opt->rr; 557 ip_rt_get_source(&optptr[optptr[2]-5], skb, rt); 558 opt->is_changed = 1; 559 } 560 if (opt->srr_is_hit) { 561 int srrptr, srrspace; 562 563 optptr = raw + opt->srr; 564 565 for ( srrptr = optptr[2], srrspace = optptr[1]; 566 srrptr <= srrspace; 567 srrptr += 4 568 ) { 569 if (srrptr + 3 > srrspace) 570 break; 571 if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0) 572 break; 573 } 574 if (srrptr + 3 <= srrspace) { 575 opt->is_changed = 1; 576 ip_hdr(skb)->daddr = opt->nexthop; 577 ip_rt_get_source(&optptr[srrptr-1], skb, rt); 578 optptr[2] = srrptr+4; 579 } else { 580 net_crit_ratelimited("%s(): Argh! Destination lost!\n", 581 __func__); 582 } 583 if (opt->ts_needaddr) { 584 optptr = raw + opt->ts; 585 ip_rt_get_source(&optptr[optptr[2]-9], skb, rt); 586 opt->is_changed = 1; 587 } 588 } 589 if (opt->is_changed) { 590 opt->is_changed = 0; 591 ip_send_check(ip_hdr(skb)); 592 } 593 } 594 595 int ip_options_rcv_srr(struct sk_buff *skb, struct net_device *dev) 596 { 597 struct ip_options *opt = &(IPCB(skb)->opt); 598 int srrspace, srrptr; 599 __be32 nexthop; 600 struct iphdr *iph = ip_hdr(skb); 601 unsigned char *optptr = skb_network_header(skb) + opt->srr; 602 struct rtable *rt = skb_rtable(skb); 603 struct rtable *rt2; 604 unsigned long orefdst; 605 int err; 606 607 if (!rt) 608 return 0; 609 610 if (skb->pkt_type != PACKET_HOST) 611 return -EINVAL; 612 if (rt->rt_type == RTN_UNICAST) { 613 if (!opt->is_strictroute) 614 return 0; 615 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24)); 616 return -EINVAL; 617 } 618 if (rt->rt_type != RTN_LOCAL) 619 return -EINVAL; 620 621 for (srrptr = optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) { 622 if (srrptr + 3 > srrspace) { 623 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24)); 624 return -EINVAL; 625 } 626 memcpy(&nexthop, &optptr[srrptr-1], 4); 627 628 orefdst = skb->_skb_refdst; 629 skb_dst_set(skb, NULL); 630 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, dev); 631 rt2 = skb_rtable(skb); 632 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) { 633 skb_dst_drop(skb); 634 skb->_skb_refdst = orefdst; 635 return -EINVAL; 636 } 637 refdst_drop(orefdst); 638 if (rt2->rt_type != RTN_LOCAL) 639 break; 640 /* Superfast 8) loopback forward */ 641 iph->daddr = nexthop; 642 opt->is_changed = 1; 643 } 644 if (srrptr <= srrspace) { 645 opt->srr_is_hit = 1; 646 opt->nexthop = nexthop; 647 opt->is_changed = 1; 648 } 649 return 0; 650 } 651 EXPORT_SYMBOL(ip_options_rcv_srr); 652