1 /* 2 * common UDP/RAW code 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Pedro Roque <roque@di.fc.ul.pt> 7 * 8 * $Id: datagram.c,v 1.24 2002/02/01 22:01:04 davem Exp $ 9 * 10 * This program is free software; you can redistribute it and/or 11 * modify it under the terms of the GNU General Public License 12 * as published by the Free Software Foundation; either version 13 * 2 of the License, or (at your option) any later version. 14 */ 15 16 #include <linux/errno.h> 17 #include <linux/types.h> 18 #include <linux/kernel.h> 19 #include <linux/sched.h> 20 #include <linux/interrupt.h> 21 #include <linux/socket.h> 22 #include <linux/sockios.h> 23 #include <linux/in6.h> 24 #include <linux/ipv6.h> 25 #include <linux/route.h> 26 27 #include <net/ipv6.h> 28 #include <net/ndisc.h> 29 #include <net/addrconf.h> 30 #include <net/transp_v6.h> 31 #include <net/ip6_route.h> 32 #include <net/tcp_states.h> 33 34 #include <linux/errqueue.h> 35 #include <asm/uaccess.h> 36 37 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) 38 { 39 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr; 40 struct inet_sock *inet = inet_sk(sk); 41 struct ipv6_pinfo *np = inet6_sk(sk); 42 struct in6_addr *daddr, *final_p = NULL, final; 43 struct dst_entry *dst; 44 struct flowi fl; 45 struct ip6_flowlabel *flowlabel = NULL; 46 int addr_type; 47 int err; 48 49 if (usin->sin6_family == AF_INET) { 50 if (__ipv6_only_sock(sk)) 51 return -EAFNOSUPPORT; 52 err = ip4_datagram_connect(sk, uaddr, addr_len); 53 goto ipv4_connected; 54 } 55 56 if (addr_len < SIN6_LEN_RFC2133) 57 return -EINVAL; 58 59 if (usin->sin6_family != AF_INET6) 60 return -EAFNOSUPPORT; 61 62 memset(&fl, 0, sizeof(fl)); 63 if (np->sndflow) { 64 fl.fl6_flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK; 65 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) { 66 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 67 if (flowlabel == NULL) 68 return -EINVAL; 69 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst); 70 } 71 } 72 73 addr_type = ipv6_addr_type(&usin->sin6_addr); 74 75 if (addr_type == IPV6_ADDR_ANY) { 76 /* 77 * connect to self 78 */ 79 usin->sin6_addr.s6_addr[15] = 0x01; 80 } 81 82 daddr = &usin->sin6_addr; 83 84 if (addr_type == IPV6_ADDR_MAPPED) { 85 struct sockaddr_in sin; 86 87 if (__ipv6_only_sock(sk)) { 88 err = -ENETUNREACH; 89 goto out; 90 } 91 sin.sin_family = AF_INET; 92 sin.sin_addr.s_addr = daddr->s6_addr32[3]; 93 sin.sin_port = usin->sin6_port; 94 95 err = ip4_datagram_connect(sk, 96 (struct sockaddr*) &sin, 97 sizeof(sin)); 98 99 ipv4_connected: 100 if (err) 101 goto out; 102 103 ipv6_addr_set(&np->daddr, 0, 0, htonl(0x0000ffff), inet->daddr); 104 105 if (ipv6_addr_any(&np->saddr)) { 106 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000ffff), 107 inet->saddr); 108 } 109 110 if (ipv6_addr_any(&np->rcv_saddr)) { 111 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000ffff), 112 inet->rcv_saddr); 113 } 114 goto out; 115 } 116 117 if (addr_type&IPV6_ADDR_LINKLOCAL) { 118 if (addr_len >= sizeof(struct sockaddr_in6) && 119 usin->sin6_scope_id) { 120 if (sk->sk_bound_dev_if && 121 sk->sk_bound_dev_if != usin->sin6_scope_id) { 122 err = -EINVAL; 123 goto out; 124 } 125 sk->sk_bound_dev_if = usin->sin6_scope_id; 126 if (!sk->sk_bound_dev_if && 127 (addr_type & IPV6_ADDR_MULTICAST)) 128 fl.oif = np->mcast_oif; 129 } 130 131 /* Connect to link-local address requires an interface */ 132 if (!sk->sk_bound_dev_if) { 133 err = -EINVAL; 134 goto out; 135 } 136 } 137 138 ipv6_addr_copy(&np->daddr, daddr); 139 np->flow_label = fl.fl6_flowlabel; 140 141 inet->dport = usin->sin6_port; 142 143 /* 144 * Check for a route to destination an obtain the 145 * destination cache for it. 146 */ 147 148 fl.proto = sk->sk_protocol; 149 ipv6_addr_copy(&fl.fl6_dst, &np->daddr); 150 ipv6_addr_copy(&fl.fl6_src, &np->saddr); 151 fl.oif = sk->sk_bound_dev_if; 152 fl.fl_ip_dport = inet->dport; 153 fl.fl_ip_sport = inet->sport; 154 155 if (!fl.oif && (addr_type&IPV6_ADDR_MULTICAST)) 156 fl.oif = np->mcast_oif; 157 158 if (flowlabel) { 159 if (flowlabel->opt && flowlabel->opt->srcrt) { 160 struct rt0_hdr *rt0 = (struct rt0_hdr *) flowlabel->opt->srcrt; 161 ipv6_addr_copy(&final, &fl.fl6_dst); 162 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 163 final_p = &final; 164 } 165 } else if (np->opt && np->opt->srcrt) { 166 struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt; 167 ipv6_addr_copy(&final, &fl.fl6_dst); 168 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 169 final_p = &final; 170 } 171 172 err = ip6_dst_lookup(sk, &dst, &fl); 173 if (err) 174 goto out; 175 if (final_p) 176 ipv6_addr_copy(&fl.fl6_dst, final_p); 177 178 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) { 179 dst_release(dst); 180 goto out; 181 } 182 183 /* source address lookup done in ip6_dst_lookup */ 184 185 if (ipv6_addr_any(&np->saddr)) 186 ipv6_addr_copy(&np->saddr, &fl.fl6_src); 187 188 if (ipv6_addr_any(&np->rcv_saddr)) { 189 ipv6_addr_copy(&np->rcv_saddr, &fl.fl6_src); 190 inet->rcv_saddr = LOOPBACK4_IPV6; 191 } 192 193 ip6_dst_store(sk, dst, 194 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ? 195 &np->daddr : NULL); 196 197 sk->sk_state = TCP_ESTABLISHED; 198 out: 199 fl6_sock_release(flowlabel); 200 return err; 201 } 202 203 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, 204 u16 port, u32 info, u8 *payload) 205 { 206 struct ipv6_pinfo *np = inet6_sk(sk); 207 struct icmp6hdr *icmph = (struct icmp6hdr *)skb->h.raw; 208 struct sock_exterr_skb *serr; 209 210 if (!np->recverr) 211 return; 212 213 skb = skb_clone(skb, GFP_ATOMIC); 214 if (!skb) 215 return; 216 217 serr = SKB_EXT_ERR(skb); 218 serr->ee.ee_errno = err; 219 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6; 220 serr->ee.ee_type = icmph->icmp6_type; 221 serr->ee.ee_code = icmph->icmp6_code; 222 serr->ee.ee_pad = 0; 223 serr->ee.ee_info = info; 224 serr->ee.ee_data = 0; 225 serr->addr_offset = (u8*)&(((struct ipv6hdr*)(icmph+1))->daddr) - skb->nh.raw; 226 serr->port = port; 227 228 skb->h.raw = payload; 229 __skb_pull(skb, payload - skb->data); 230 231 if (sock_queue_err_skb(sk, skb)) 232 kfree_skb(skb); 233 } 234 235 void ipv6_local_error(struct sock *sk, int err, struct flowi *fl, u32 info) 236 { 237 struct ipv6_pinfo *np = inet6_sk(sk); 238 struct sock_exterr_skb *serr; 239 struct ipv6hdr *iph; 240 struct sk_buff *skb; 241 242 if (!np->recverr) 243 return; 244 245 skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 246 if (!skb) 247 return; 248 249 iph = (struct ipv6hdr*)skb_put(skb, sizeof(struct ipv6hdr)); 250 skb->nh.ipv6h = iph; 251 ipv6_addr_copy(&iph->daddr, &fl->fl6_dst); 252 253 serr = SKB_EXT_ERR(skb); 254 serr->ee.ee_errno = err; 255 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL; 256 serr->ee.ee_type = 0; 257 serr->ee.ee_code = 0; 258 serr->ee.ee_pad = 0; 259 serr->ee.ee_info = info; 260 serr->ee.ee_data = 0; 261 serr->addr_offset = (u8*)&iph->daddr - skb->nh.raw; 262 serr->port = fl->fl_ip_dport; 263 264 skb->h.raw = skb->tail; 265 __skb_pull(skb, skb->tail - skb->data); 266 267 if (sock_queue_err_skb(sk, skb)) 268 kfree_skb(skb); 269 } 270 271 /* 272 * Handle MSG_ERRQUEUE 273 */ 274 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len) 275 { 276 struct ipv6_pinfo *np = inet6_sk(sk); 277 struct sock_exterr_skb *serr; 278 struct sk_buff *skb, *skb2; 279 struct sockaddr_in6 *sin; 280 struct { 281 struct sock_extended_err ee; 282 struct sockaddr_in6 offender; 283 } errhdr; 284 int err; 285 int copied; 286 287 err = -EAGAIN; 288 skb = skb_dequeue(&sk->sk_error_queue); 289 if (skb == NULL) 290 goto out; 291 292 copied = skb->len; 293 if (copied > len) { 294 msg->msg_flags |= MSG_TRUNC; 295 copied = len; 296 } 297 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 298 if (err) 299 goto out_free_skb; 300 301 sock_recv_timestamp(msg, sk, skb); 302 303 serr = SKB_EXT_ERR(skb); 304 305 sin = (struct sockaddr_in6 *)msg->msg_name; 306 if (sin) { 307 sin->sin6_family = AF_INET6; 308 sin->sin6_flowinfo = 0; 309 sin->sin6_port = serr->port; 310 sin->sin6_scope_id = 0; 311 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6) { 312 ipv6_addr_copy(&sin->sin6_addr, 313 (struct in6_addr *)(skb->nh.raw + serr->addr_offset)); 314 if (np->sndflow) 315 sin->sin6_flowinfo = *(u32*)(skb->nh.raw + serr->addr_offset - 24) & IPV6_FLOWINFO_MASK; 316 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL) 317 sin->sin6_scope_id = IP6CB(skb)->iif; 318 } else { 319 ipv6_addr_set(&sin->sin6_addr, 0, 0, 320 htonl(0xffff), 321 *(u32*)(skb->nh.raw + serr->addr_offset)); 322 } 323 } 324 325 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err)); 326 sin = &errhdr.offender; 327 sin->sin6_family = AF_UNSPEC; 328 if (serr->ee.ee_origin != SO_EE_ORIGIN_LOCAL) { 329 sin->sin6_family = AF_INET6; 330 sin->sin6_flowinfo = 0; 331 sin->sin6_scope_id = 0; 332 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6) { 333 ipv6_addr_copy(&sin->sin6_addr, &skb->nh.ipv6h->saddr); 334 if (np->rxopt.all) 335 datagram_recv_ctl(sk, msg, skb); 336 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL) 337 sin->sin6_scope_id = IP6CB(skb)->iif; 338 } else { 339 struct inet_sock *inet = inet_sk(sk); 340 341 ipv6_addr_set(&sin->sin6_addr, 0, 0, 342 htonl(0xffff), 343 skb->nh.iph->saddr); 344 if (inet->cmsg_flags) 345 ip_cmsg_recv(msg, skb); 346 } 347 } 348 349 put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr); 350 351 /* Now we could try to dump offended packet options */ 352 353 msg->msg_flags |= MSG_ERRQUEUE; 354 err = copied; 355 356 /* Reset and regenerate socket error */ 357 spin_lock_bh(&sk->sk_error_queue.lock); 358 sk->sk_err = 0; 359 if ((skb2 = skb_peek(&sk->sk_error_queue)) != NULL) { 360 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno; 361 spin_unlock_bh(&sk->sk_error_queue.lock); 362 sk->sk_error_report(sk); 363 } else { 364 spin_unlock_bh(&sk->sk_error_queue.lock); 365 } 366 367 out_free_skb: 368 kfree_skb(skb); 369 out: 370 return err; 371 } 372 373 374 375 int datagram_recv_ctl(struct sock *sk, struct msghdr *msg, struct sk_buff *skb) 376 { 377 struct ipv6_pinfo *np = inet6_sk(sk); 378 struct inet6_skb_parm *opt = IP6CB(skb); 379 380 if (np->rxopt.bits.rxinfo) { 381 struct in6_pktinfo src_info; 382 383 src_info.ipi6_ifindex = opt->iif; 384 ipv6_addr_copy(&src_info.ipi6_addr, &skb->nh.ipv6h->daddr); 385 put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info); 386 } 387 388 if (np->rxopt.bits.rxhlim) { 389 int hlim = skb->nh.ipv6h->hop_limit; 390 put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim); 391 } 392 393 if (np->rxopt.bits.rxflow && (*(u32*)skb->nh.raw & IPV6_FLOWINFO_MASK)) { 394 u32 flowinfo = *(u32*)skb->nh.raw & IPV6_FLOWINFO_MASK; 395 put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo); 396 } 397 if (np->rxopt.bits.hopopts && opt->hop) { 398 u8 *ptr = skb->nh.raw + opt->hop; 399 put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr); 400 } 401 if (np->rxopt.bits.dstopts && opt->dst0) { 402 u8 *ptr = skb->nh.raw + opt->dst0; 403 put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, (ptr[1]+1)<<3, ptr); 404 } 405 if (np->rxopt.bits.srcrt && opt->srcrt) { 406 struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(skb->nh.raw + opt->srcrt); 407 put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, (rthdr->hdrlen+1) << 3, rthdr); 408 } 409 if (np->rxopt.bits.dstopts && opt->dst1) { 410 u8 *ptr = skb->nh.raw + opt->dst1; 411 put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, (ptr[1]+1)<<3, ptr); 412 } 413 return 0; 414 } 415 416 int datagram_send_ctl(struct msghdr *msg, struct flowi *fl, 417 struct ipv6_txoptions *opt, 418 int *hlimit) 419 { 420 struct in6_pktinfo *src_info; 421 struct cmsghdr *cmsg; 422 struct ipv6_rt_hdr *rthdr; 423 struct ipv6_opt_hdr *hdr; 424 int len; 425 int err = 0; 426 427 for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) { 428 int addr_type; 429 struct net_device *dev = NULL; 430 431 if (!CMSG_OK(msg, cmsg)) { 432 err = -EINVAL; 433 goto exit_f; 434 } 435 436 if (cmsg->cmsg_level != SOL_IPV6) 437 continue; 438 439 switch (cmsg->cmsg_type) { 440 case IPV6_PKTINFO: 441 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) { 442 err = -EINVAL; 443 goto exit_f; 444 } 445 446 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg); 447 448 if (src_info->ipi6_ifindex) { 449 if (fl->oif && src_info->ipi6_ifindex != fl->oif) 450 return -EINVAL; 451 fl->oif = src_info->ipi6_ifindex; 452 } 453 454 addr_type = ipv6_addr_type(&src_info->ipi6_addr); 455 456 if (addr_type == IPV6_ADDR_ANY) 457 break; 458 459 if (addr_type & IPV6_ADDR_LINKLOCAL) { 460 if (!src_info->ipi6_ifindex) 461 return -EINVAL; 462 else { 463 dev = dev_get_by_index(src_info->ipi6_ifindex); 464 if (!dev) 465 return -ENODEV; 466 } 467 } 468 if (!ipv6_chk_addr(&src_info->ipi6_addr, dev, 0)) { 469 if (dev) 470 dev_put(dev); 471 err = -EINVAL; 472 goto exit_f; 473 } 474 if (dev) 475 dev_put(dev); 476 477 ipv6_addr_copy(&fl->fl6_src, &src_info->ipi6_addr); 478 break; 479 480 case IPV6_FLOWINFO: 481 if (cmsg->cmsg_len < CMSG_LEN(4)) { 482 err = -EINVAL; 483 goto exit_f; 484 } 485 486 if (fl->fl6_flowlabel&IPV6_FLOWINFO_MASK) { 487 if ((fl->fl6_flowlabel^*(u32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) { 488 err = -EINVAL; 489 goto exit_f; 490 } 491 } 492 fl->fl6_flowlabel = IPV6_FLOWINFO_MASK & *(u32 *)CMSG_DATA(cmsg); 493 break; 494 495 case IPV6_HOPOPTS: 496 if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) { 497 err = -EINVAL; 498 goto exit_f; 499 } 500 501 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg); 502 len = ((hdr->hdrlen + 1) << 3); 503 if (cmsg->cmsg_len < CMSG_LEN(len)) { 504 err = -EINVAL; 505 goto exit_f; 506 } 507 if (!capable(CAP_NET_RAW)) { 508 err = -EPERM; 509 goto exit_f; 510 } 511 opt->opt_nflen += len; 512 opt->hopopt = hdr; 513 break; 514 515 case IPV6_DSTOPTS: 516 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) { 517 err = -EINVAL; 518 goto exit_f; 519 } 520 521 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg); 522 len = ((hdr->hdrlen + 1) << 3); 523 if (cmsg->cmsg_len < CMSG_LEN(len)) { 524 err = -EINVAL; 525 goto exit_f; 526 } 527 if (!capable(CAP_NET_RAW)) { 528 err = -EPERM; 529 goto exit_f; 530 } 531 if (opt->dst1opt) { 532 err = -EINVAL; 533 goto exit_f; 534 } 535 opt->opt_flen += len; 536 opt->dst1opt = hdr; 537 break; 538 539 case IPV6_RTHDR: 540 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) { 541 err = -EINVAL; 542 goto exit_f; 543 } 544 545 rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg); 546 547 /* 548 * TYPE 0 549 */ 550 if (rthdr->type) { 551 err = -EINVAL; 552 goto exit_f; 553 } 554 555 len = ((rthdr->hdrlen + 1) << 3); 556 557 if (cmsg->cmsg_len < CMSG_LEN(len)) { 558 err = -EINVAL; 559 goto exit_f; 560 } 561 562 /* segments left must also match */ 563 if ((rthdr->hdrlen >> 1) != rthdr->segments_left) { 564 err = -EINVAL; 565 goto exit_f; 566 } 567 568 opt->opt_nflen += len; 569 opt->srcrt = rthdr; 570 571 if (opt->dst1opt) { 572 int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3); 573 574 opt->opt_nflen += dsthdrlen; 575 opt->dst0opt = opt->dst1opt; 576 opt->dst1opt = NULL; 577 opt->opt_flen -= dsthdrlen; 578 } 579 580 break; 581 582 case IPV6_HOPLIMIT: 583 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) { 584 err = -EINVAL; 585 goto exit_f; 586 } 587 588 *hlimit = *(int *)CMSG_DATA(cmsg); 589 break; 590 591 default: 592 LIMIT_NETDEBUG(KERN_DEBUG "invalid cmsg type: %d\n", 593 cmsg->cmsg_type); 594 err = -EINVAL; 595 break; 596 }; 597 } 598 599 exit_f: 600 return err; 601 } 602