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 { 257 __be32 spec_dst = htonl(INADDR_ANY); 258 unsigned char *pp_ptr = NULL; 259 struct rtable *rt = NULL; 260 unsigned char *optptr; 261 unsigned char *iph; 262 int optlen, l; 263 264 if (skb) { 265 rt = skb_rtable(skb); 266 optptr = (unsigned char *)&(ip_hdr(skb)[1]); 267 } else 268 optptr = opt->__data; 269 iph = optptr - sizeof(struct iphdr); 270 271 for (l = opt->optlen; l > 0; ) { 272 switch (*optptr) { 273 case IPOPT_END: 274 for (optptr++, l--; l > 0; optptr++, l--) { 275 if (*optptr != IPOPT_END) { 276 *optptr = IPOPT_END; 277 opt->is_changed = 1; 278 } 279 } 280 goto eol; 281 case IPOPT_NOOP: 282 l--; 283 optptr++; 284 continue; 285 } 286 if (unlikely(l < 2)) { 287 pp_ptr = optptr; 288 goto error; 289 } 290 optlen = optptr[1]; 291 if (optlen < 2 || optlen > l) { 292 pp_ptr = optptr; 293 goto error; 294 } 295 switch (*optptr) { 296 case IPOPT_SSRR: 297 case IPOPT_LSRR: 298 if (optlen < 3) { 299 pp_ptr = optptr + 1; 300 goto error; 301 } 302 if (optptr[2] < 4) { 303 pp_ptr = optptr + 2; 304 goto error; 305 } 306 /* NB: cf RFC-1812 5.2.4.1 */ 307 if (opt->srr) { 308 pp_ptr = optptr; 309 goto error; 310 } 311 if (!skb) { 312 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) { 313 pp_ptr = optptr + 1; 314 goto error; 315 } 316 memcpy(&opt->faddr, &optptr[3], 4); 317 if (optlen > 7) 318 memmove(&optptr[3], &optptr[7], optlen-7); 319 } 320 opt->is_strictroute = (optptr[0] == IPOPT_SSRR); 321 opt->srr = optptr - iph; 322 break; 323 case IPOPT_RR: 324 if (opt->rr) { 325 pp_ptr = optptr; 326 goto error; 327 } 328 if (optlen < 3) { 329 pp_ptr = optptr + 1; 330 goto error; 331 } 332 if (optptr[2] < 4) { 333 pp_ptr = optptr + 2; 334 goto error; 335 } 336 if (optptr[2] <= optlen) { 337 if (optptr[2]+3 > optlen) { 338 pp_ptr = optptr + 2; 339 goto error; 340 } 341 if (rt) { 342 spec_dst_fill(&spec_dst, skb); 343 memcpy(&optptr[optptr[2]-1], &spec_dst, 4); 344 opt->is_changed = 1; 345 } 346 optptr[2] += 4; 347 opt->rr_needaddr = 1; 348 } 349 opt->rr = optptr - iph; 350 break; 351 case IPOPT_TIMESTAMP: 352 if (opt->ts) { 353 pp_ptr = optptr; 354 goto error; 355 } 356 if (optlen < 4) { 357 pp_ptr = optptr + 1; 358 goto error; 359 } 360 if (optptr[2] < 5) { 361 pp_ptr = optptr + 2; 362 goto error; 363 } 364 if (optptr[2] <= optlen) { 365 unsigned char *timeptr = NULL; 366 if (optptr[2]+3 > optlen) { 367 pp_ptr = optptr + 2; 368 goto error; 369 } 370 switch (optptr[3]&0xF) { 371 case IPOPT_TS_TSONLY: 372 if (skb) 373 timeptr = &optptr[optptr[2]-1]; 374 opt->ts_needtime = 1; 375 optptr[2] += 4; 376 break; 377 case IPOPT_TS_TSANDADDR: 378 if (optptr[2]+7 > optlen) { 379 pp_ptr = optptr + 2; 380 goto error; 381 } 382 if (rt) { 383 spec_dst_fill(&spec_dst, skb); 384 memcpy(&optptr[optptr[2]-1], &spec_dst, 4); 385 timeptr = &optptr[optptr[2]+3]; 386 } 387 opt->ts_needaddr = 1; 388 opt->ts_needtime = 1; 389 optptr[2] += 8; 390 break; 391 case IPOPT_TS_PRESPEC: 392 if (optptr[2]+7 > optlen) { 393 pp_ptr = optptr + 2; 394 goto error; 395 } 396 { 397 __be32 addr; 398 memcpy(&addr, &optptr[optptr[2]-1], 4); 399 if (inet_addr_type(net, addr) == RTN_UNICAST) 400 break; 401 if (skb) 402 timeptr = &optptr[optptr[2]+3]; 403 } 404 opt->ts_needtime = 1; 405 optptr[2] += 8; 406 break; 407 default: 408 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) { 409 pp_ptr = optptr + 3; 410 goto error; 411 } 412 break; 413 } 414 if (timeptr) { 415 __be32 midtime; 416 417 midtime = inet_current_timestamp(); 418 memcpy(timeptr, &midtime, 4); 419 opt->is_changed = 1; 420 } 421 } else if ((optptr[3]&0xF) != IPOPT_TS_PRESPEC) { 422 unsigned int overflow = optptr[3]>>4; 423 if (overflow == 15) { 424 pp_ptr = optptr + 3; 425 goto error; 426 } 427 if (skb) { 428 optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4); 429 opt->is_changed = 1; 430 } 431 } 432 opt->ts = optptr - iph; 433 break; 434 case IPOPT_RA: 435 if (optlen < 4) { 436 pp_ptr = optptr + 1; 437 goto error; 438 } 439 if (optptr[2] == 0 && optptr[3] == 0) 440 opt->router_alert = optptr - iph; 441 break; 442 case IPOPT_CIPSO: 443 if ((!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) || opt->cipso) { 444 pp_ptr = optptr; 445 goto error; 446 } 447 opt->cipso = optptr - iph; 448 if (cipso_v4_validate(skb, &optptr)) { 449 pp_ptr = optptr; 450 goto error; 451 } 452 break; 453 case IPOPT_SEC: 454 case IPOPT_SID: 455 default: 456 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) { 457 pp_ptr = optptr; 458 goto error; 459 } 460 break; 461 } 462 l -= optlen; 463 optptr += optlen; 464 } 465 466 eol: 467 if (!pp_ptr) 468 return 0; 469 470 error: 471 if (skb) { 472 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24)); 473 } 474 return -EINVAL; 475 } 476 EXPORT_SYMBOL(ip_options_compile); 477 478 /* 479 * Undo all the changes done by ip_options_compile(). 480 */ 481 482 void ip_options_undo(struct ip_options *opt) 483 { 484 if (opt->srr) { 485 unsigned char *optptr = opt->__data+opt->srr-sizeof(struct iphdr); 486 memmove(optptr+7, optptr+3, optptr[1]-7); 487 memcpy(optptr+3, &opt->faddr, 4); 488 } 489 if (opt->rr_needaddr) { 490 unsigned char *optptr = opt->__data+opt->rr-sizeof(struct iphdr); 491 optptr[2] -= 4; 492 memset(&optptr[optptr[2]-1], 0, 4); 493 } 494 if (opt->ts) { 495 unsigned char *optptr = opt->__data+opt->ts-sizeof(struct iphdr); 496 if (opt->ts_needtime) { 497 optptr[2] -= 4; 498 memset(&optptr[optptr[2]-1], 0, 4); 499 if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC) 500 optptr[2] -= 4; 501 } 502 if (opt->ts_needaddr) { 503 optptr[2] -= 4; 504 memset(&optptr[optptr[2]-1], 0, 4); 505 } 506 } 507 } 508 509 static struct ip_options_rcu *ip_options_get_alloc(const int optlen) 510 { 511 return kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3), 512 GFP_KERNEL); 513 } 514 515 static int ip_options_get_finish(struct net *net, struct ip_options_rcu **optp, 516 struct ip_options_rcu *opt, int optlen) 517 { 518 while (optlen & 3) 519 opt->opt.__data[optlen++] = IPOPT_END; 520 opt->opt.optlen = optlen; 521 if (optlen && ip_options_compile(net, &opt->opt, NULL)) { 522 kfree(opt); 523 return -EINVAL; 524 } 525 kfree(*optp); 526 *optp = opt; 527 return 0; 528 } 529 530 int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp, 531 unsigned char __user *data, int optlen) 532 { 533 struct ip_options_rcu *opt = ip_options_get_alloc(optlen); 534 535 if (!opt) 536 return -ENOMEM; 537 if (optlen && copy_from_user(opt->opt.__data, data, optlen)) { 538 kfree(opt); 539 return -EFAULT; 540 } 541 return ip_options_get_finish(net, optp, opt, optlen); 542 } 543 544 int ip_options_get(struct net *net, struct ip_options_rcu **optp, 545 unsigned char *data, int optlen) 546 { 547 struct ip_options_rcu *opt = ip_options_get_alloc(optlen); 548 549 if (!opt) 550 return -ENOMEM; 551 if (optlen) 552 memcpy(opt->opt.__data, data, optlen); 553 return ip_options_get_finish(net, optp, opt, optlen); 554 } 555 556 void ip_forward_options(struct sk_buff *skb) 557 { 558 struct ip_options *opt = &(IPCB(skb)->opt); 559 unsigned char *optptr; 560 struct rtable *rt = skb_rtable(skb); 561 unsigned char *raw = skb_network_header(skb); 562 563 if (opt->rr_needaddr) { 564 optptr = (unsigned char *)raw + opt->rr; 565 ip_rt_get_source(&optptr[optptr[2]-5], skb, rt); 566 opt->is_changed = 1; 567 } 568 if (opt->srr_is_hit) { 569 int srrptr, srrspace; 570 571 optptr = raw + opt->srr; 572 573 for ( srrptr = optptr[2], srrspace = optptr[1]; 574 srrptr <= srrspace; 575 srrptr += 4 576 ) { 577 if (srrptr + 3 > srrspace) 578 break; 579 if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0) 580 break; 581 } 582 if (srrptr + 3 <= srrspace) { 583 opt->is_changed = 1; 584 ip_hdr(skb)->daddr = opt->nexthop; 585 ip_rt_get_source(&optptr[srrptr-1], skb, rt); 586 optptr[2] = srrptr+4; 587 } else { 588 net_crit_ratelimited("%s(): Argh! Destination lost!\n", 589 __func__); 590 } 591 if (opt->ts_needaddr) { 592 optptr = raw + opt->ts; 593 ip_rt_get_source(&optptr[optptr[2]-9], skb, rt); 594 opt->is_changed = 1; 595 } 596 } 597 if (opt->is_changed) { 598 opt->is_changed = 0; 599 ip_send_check(ip_hdr(skb)); 600 } 601 } 602 603 int ip_options_rcv_srr(struct sk_buff *skb) 604 { 605 struct ip_options *opt = &(IPCB(skb)->opt); 606 int srrspace, srrptr; 607 __be32 nexthop; 608 struct iphdr *iph = ip_hdr(skb); 609 unsigned char *optptr = skb_network_header(skb) + opt->srr; 610 struct rtable *rt = skb_rtable(skb); 611 struct rtable *rt2; 612 unsigned long orefdst; 613 int err; 614 615 if (!rt) 616 return 0; 617 618 if (skb->pkt_type != PACKET_HOST) 619 return -EINVAL; 620 if (rt->rt_type == RTN_UNICAST) { 621 if (!opt->is_strictroute) 622 return 0; 623 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24)); 624 return -EINVAL; 625 } 626 if (rt->rt_type != RTN_LOCAL) 627 return -EINVAL; 628 629 for (srrptr = optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) { 630 if (srrptr + 3 > srrspace) { 631 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24)); 632 return -EINVAL; 633 } 634 memcpy(&nexthop, &optptr[srrptr-1], 4); 635 636 orefdst = skb->_skb_refdst; 637 skb_dst_set(skb, NULL); 638 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev); 639 rt2 = skb_rtable(skb); 640 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) { 641 skb_dst_drop(skb); 642 skb->_skb_refdst = orefdst; 643 return -EINVAL; 644 } 645 refdst_drop(orefdst); 646 if (rt2->rt_type != RTN_LOCAL) 647 break; 648 /* Superfast 8) loopback forward */ 649 iph->daddr = nexthop; 650 opt->is_changed = 1; 651 } 652 if (srrptr <= srrspace) { 653 opt->srr_is_hit = 1; 654 opt->nexthop = nexthop; 655 opt->is_changed = 1; 656 } 657 return 0; 658 } 659 EXPORT_SYMBOL(ip_options_rcv_srr); 660