1 /* 2 * common UDP/RAW code 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Pedro Roque <roque@di.fc.ul.pt> 7 * 8 * This program is free software; you can redistribute it and/or 9 * modify it under the terms of the GNU General Public License 10 * as published by the Free Software Foundation; either version 11 * 2 of the License, or (at your option) any later version. 12 */ 13 14 #include <linux/capability.h> 15 #include <linux/errno.h> 16 #include <linux/types.h> 17 #include <linux/kernel.h> 18 #include <linux/interrupt.h> 19 #include <linux/socket.h> 20 #include <linux/sockios.h> 21 #include <linux/in6.h> 22 #include <linux/ipv6.h> 23 #include <linux/route.h> 24 #include <linux/slab.h> 25 #include <linux/export.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 #include <net/dsfield.h> 34 35 #include <linux/errqueue.h> 36 #include <linux/uaccess.h> 37 38 static bool ipv6_mapped_addr_any(const struct in6_addr *a) 39 { 40 return ipv6_addr_v4mapped(a) && (a->s6_addr32[3] == 0); 41 } 42 43 static void ip6_datagram_flow_key_init(struct flowi6 *fl6, struct sock *sk) 44 { 45 struct inet_sock *inet = inet_sk(sk); 46 struct ipv6_pinfo *np = inet6_sk(sk); 47 48 memset(fl6, 0, sizeof(*fl6)); 49 fl6->flowi6_proto = sk->sk_protocol; 50 fl6->daddr = sk->sk_v6_daddr; 51 fl6->saddr = np->saddr; 52 fl6->flowi6_oif = sk->sk_bound_dev_if; 53 fl6->flowi6_mark = sk->sk_mark; 54 fl6->fl6_dport = inet->inet_dport; 55 fl6->fl6_sport = inet->inet_sport; 56 fl6->flowlabel = np->flow_label; 57 fl6->flowi6_uid = sk->sk_uid; 58 59 if (!fl6->flowi6_oif) 60 fl6->flowi6_oif = np->sticky_pktinfo.ipi6_ifindex; 61 62 if (!fl6->flowi6_oif && ipv6_addr_is_multicast(&fl6->daddr)) 63 fl6->flowi6_oif = np->mcast_oif; 64 65 security_sk_classify_flow(sk, flowi6_to_flowi(fl6)); 66 } 67 68 int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr) 69 { 70 struct ip6_flowlabel *flowlabel = NULL; 71 struct in6_addr *final_p, final; 72 struct ipv6_txoptions *opt; 73 struct dst_entry *dst; 74 struct inet_sock *inet = inet_sk(sk); 75 struct ipv6_pinfo *np = inet6_sk(sk); 76 struct flowi6 fl6; 77 int err = 0; 78 79 if (np->sndflow && (np->flow_label & IPV6_FLOWLABEL_MASK)) { 80 flowlabel = fl6_sock_lookup(sk, np->flow_label); 81 if (!flowlabel) 82 return -EINVAL; 83 } 84 ip6_datagram_flow_key_init(&fl6, sk); 85 86 rcu_read_lock(); 87 opt = flowlabel ? flowlabel->opt : rcu_dereference(np->opt); 88 final_p = fl6_update_dst(&fl6, opt, &final); 89 rcu_read_unlock(); 90 91 dst = ip6_dst_lookup_flow(sk, &fl6, final_p); 92 if (IS_ERR(dst)) { 93 err = PTR_ERR(dst); 94 goto out; 95 } 96 97 if (fix_sk_saddr) { 98 if (ipv6_addr_any(&np->saddr)) 99 np->saddr = fl6.saddr; 100 101 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) { 102 sk->sk_v6_rcv_saddr = fl6.saddr; 103 inet->inet_rcv_saddr = LOOPBACK4_IPV6; 104 if (sk->sk_prot->rehash) 105 sk->sk_prot->rehash(sk); 106 } 107 } 108 109 ip6_dst_store(sk, dst, 110 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ? 111 &sk->sk_v6_daddr : NULL, 112 #ifdef CONFIG_IPV6_SUBTREES 113 ipv6_addr_equal(&fl6.saddr, &np->saddr) ? 114 &np->saddr : 115 #endif 116 NULL); 117 118 out: 119 fl6_sock_release(flowlabel); 120 return err; 121 } 122 123 void ip6_datagram_release_cb(struct sock *sk) 124 { 125 struct dst_entry *dst; 126 127 if (ipv6_addr_v4mapped(&sk->sk_v6_daddr)) 128 return; 129 130 rcu_read_lock(); 131 dst = __sk_dst_get(sk); 132 if (!dst || !dst->obsolete || 133 dst->ops->check(dst, inet6_sk(sk)->dst_cookie)) { 134 rcu_read_unlock(); 135 return; 136 } 137 rcu_read_unlock(); 138 139 ip6_datagram_dst_update(sk, false); 140 } 141 EXPORT_SYMBOL_GPL(ip6_datagram_release_cb); 142 143 int __ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, 144 int addr_len) 145 { 146 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr; 147 struct inet_sock *inet = inet_sk(sk); 148 struct ipv6_pinfo *np = inet6_sk(sk); 149 struct in6_addr *daddr; 150 int addr_type; 151 int err; 152 __be32 fl6_flowlabel = 0; 153 154 if (usin->sin6_family == AF_INET) { 155 if (__ipv6_only_sock(sk)) 156 return -EAFNOSUPPORT; 157 err = __ip4_datagram_connect(sk, uaddr, addr_len); 158 goto ipv4_connected; 159 } 160 161 if (addr_len < SIN6_LEN_RFC2133) 162 return -EINVAL; 163 164 if (usin->sin6_family != AF_INET6) 165 return -EAFNOSUPPORT; 166 167 if (np->sndflow) 168 fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK; 169 170 if (ipv6_addr_any(&usin->sin6_addr)) { 171 /* 172 * connect to self 173 */ 174 if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr)) 175 ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK), 176 &usin->sin6_addr); 177 else 178 usin->sin6_addr = in6addr_loopback; 179 } 180 181 addr_type = ipv6_addr_type(&usin->sin6_addr); 182 183 daddr = &usin->sin6_addr; 184 185 if (addr_type & IPV6_ADDR_MAPPED) { 186 struct sockaddr_in sin; 187 188 if (__ipv6_only_sock(sk)) { 189 err = -ENETUNREACH; 190 goto out; 191 } 192 sin.sin_family = AF_INET; 193 sin.sin_addr.s_addr = daddr->s6_addr32[3]; 194 sin.sin_port = usin->sin6_port; 195 196 err = __ip4_datagram_connect(sk, 197 (struct sockaddr *) &sin, 198 sizeof(sin)); 199 200 ipv4_connected: 201 if (err) 202 goto out; 203 204 ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr); 205 206 if (ipv6_addr_any(&np->saddr) || 207 ipv6_mapped_addr_any(&np->saddr)) 208 ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr); 209 210 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) || 211 ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) { 212 ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, 213 &sk->sk_v6_rcv_saddr); 214 if (sk->sk_prot->rehash) 215 sk->sk_prot->rehash(sk); 216 } 217 218 goto out; 219 } 220 221 if (__ipv6_addr_needs_scope_id(addr_type)) { 222 if (addr_len >= sizeof(struct sockaddr_in6) && 223 usin->sin6_scope_id) { 224 if (sk->sk_bound_dev_if && 225 sk->sk_bound_dev_if != usin->sin6_scope_id) { 226 err = -EINVAL; 227 goto out; 228 } 229 sk->sk_bound_dev_if = usin->sin6_scope_id; 230 } 231 232 if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST)) 233 sk->sk_bound_dev_if = np->mcast_oif; 234 235 /* Connect to link-local address requires an interface */ 236 if (!sk->sk_bound_dev_if) { 237 err = -EINVAL; 238 goto out; 239 } 240 } 241 242 sk->sk_v6_daddr = *daddr; 243 np->flow_label = fl6_flowlabel; 244 245 inet->inet_dport = usin->sin6_port; 246 247 /* 248 * Check for a route to destination an obtain the 249 * destination cache for it. 250 */ 251 252 err = ip6_datagram_dst_update(sk, true); 253 if (err) 254 goto out; 255 256 sk->sk_state = TCP_ESTABLISHED; 257 sk_set_txhash(sk); 258 out: 259 return err; 260 } 261 EXPORT_SYMBOL_GPL(__ip6_datagram_connect); 262 263 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) 264 { 265 int res; 266 267 lock_sock(sk); 268 res = __ip6_datagram_connect(sk, uaddr, addr_len); 269 release_sock(sk); 270 return res; 271 } 272 EXPORT_SYMBOL_GPL(ip6_datagram_connect); 273 274 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr, 275 int addr_len) 276 { 277 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr); 278 if (sin6->sin6_family != AF_INET6) 279 return -EAFNOSUPPORT; 280 return ip6_datagram_connect(sk, uaddr, addr_len); 281 } 282 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only); 283 284 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, 285 __be16 port, u32 info, u8 *payload) 286 { 287 struct ipv6_pinfo *np = inet6_sk(sk); 288 struct icmp6hdr *icmph = icmp6_hdr(skb); 289 struct sock_exterr_skb *serr; 290 291 if (!np->recverr) 292 return; 293 294 skb = skb_clone(skb, GFP_ATOMIC); 295 if (!skb) 296 return; 297 298 skb->protocol = htons(ETH_P_IPV6); 299 300 serr = SKB_EXT_ERR(skb); 301 serr->ee.ee_errno = err; 302 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6; 303 serr->ee.ee_type = icmph->icmp6_type; 304 serr->ee.ee_code = icmph->icmp6_code; 305 serr->ee.ee_pad = 0; 306 serr->ee.ee_info = info; 307 serr->ee.ee_data = 0; 308 serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) - 309 skb_network_header(skb); 310 serr->port = port; 311 312 __skb_pull(skb, payload - skb->data); 313 skb_reset_transport_header(skb); 314 315 if (sock_queue_err_skb(sk, skb)) 316 kfree_skb(skb); 317 } 318 319 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info) 320 { 321 const struct ipv6_pinfo *np = inet6_sk(sk); 322 struct sock_exterr_skb *serr; 323 struct ipv6hdr *iph; 324 struct sk_buff *skb; 325 326 if (!np->recverr) 327 return; 328 329 skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 330 if (!skb) 331 return; 332 333 skb->protocol = htons(ETH_P_IPV6); 334 335 skb_put(skb, sizeof(struct ipv6hdr)); 336 skb_reset_network_header(skb); 337 iph = ipv6_hdr(skb); 338 iph->daddr = fl6->daddr; 339 340 serr = SKB_EXT_ERR(skb); 341 serr->ee.ee_errno = err; 342 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL; 343 serr->ee.ee_type = 0; 344 serr->ee.ee_code = 0; 345 serr->ee.ee_pad = 0; 346 serr->ee.ee_info = info; 347 serr->ee.ee_data = 0; 348 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb); 349 serr->port = fl6->fl6_dport; 350 351 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 352 skb_reset_transport_header(skb); 353 354 if (sock_queue_err_skb(sk, skb)) 355 kfree_skb(skb); 356 } 357 358 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu) 359 { 360 struct ipv6_pinfo *np = inet6_sk(sk); 361 struct ipv6hdr *iph; 362 struct sk_buff *skb; 363 struct ip6_mtuinfo *mtu_info; 364 365 if (!np->rxopt.bits.rxpmtu) 366 return; 367 368 skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 369 if (!skb) 370 return; 371 372 skb_put(skb, sizeof(struct ipv6hdr)); 373 skb_reset_network_header(skb); 374 iph = ipv6_hdr(skb); 375 iph->daddr = fl6->daddr; 376 377 mtu_info = IP6CBMTU(skb); 378 379 mtu_info->ip6m_mtu = mtu; 380 mtu_info->ip6m_addr.sin6_family = AF_INET6; 381 mtu_info->ip6m_addr.sin6_port = 0; 382 mtu_info->ip6m_addr.sin6_flowinfo = 0; 383 mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif; 384 mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr; 385 386 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 387 skb_reset_transport_header(skb); 388 389 skb = xchg(&np->rxpmtu, skb); 390 kfree_skb(skb); 391 } 392 393 /* For some errors we have valid addr_offset even with zero payload and 394 * zero port. Also, addr_offset should be supported if port is set. 395 */ 396 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr) 397 { 398 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 || 399 serr->ee.ee_origin == SO_EE_ORIGIN_ICMP || 400 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port; 401 } 402 403 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL. 404 * 405 * At one point, excluding local errors was a quick test to identify icmp/icmp6 406 * errors. This is no longer true, but the test remained, so the v6 stack, 407 * unlike v4, also honors cmsg requests on all wifi and timestamp errors. 408 * 409 * Timestamp code paths do not initialize the fields expected by cmsg: 410 * the PKTINFO fields in skb->cb[]. Fill those in here. 411 */ 412 static bool ip6_datagram_support_cmsg(struct sk_buff *skb, 413 struct sock_exterr_skb *serr) 414 { 415 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP || 416 serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6) 417 return true; 418 419 if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL) 420 return false; 421 422 if (!skb->dev) 423 return false; 424 425 if (skb->protocol == htons(ETH_P_IPV6)) 426 IP6CB(skb)->iif = skb->dev->ifindex; 427 else 428 PKTINFO_SKB_CB(skb)->ipi_ifindex = skb->dev->ifindex; 429 430 return true; 431 } 432 433 /* 434 * Handle MSG_ERRQUEUE 435 */ 436 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) 437 { 438 struct ipv6_pinfo *np = inet6_sk(sk); 439 struct sock_exterr_skb *serr; 440 struct sk_buff *skb; 441 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name); 442 struct { 443 struct sock_extended_err ee; 444 struct sockaddr_in6 offender; 445 } errhdr; 446 int err; 447 int copied; 448 449 err = -EAGAIN; 450 skb = sock_dequeue_err_skb(sk); 451 if (!skb) 452 goto out; 453 454 copied = skb->len; 455 if (copied > len) { 456 msg->msg_flags |= MSG_TRUNC; 457 copied = len; 458 } 459 err = skb_copy_datagram_msg(skb, 0, msg, copied); 460 if (unlikely(err)) { 461 kfree_skb(skb); 462 return err; 463 } 464 sock_recv_timestamp(msg, sk, skb); 465 466 serr = SKB_EXT_ERR(skb); 467 468 if (sin && ipv6_datagram_support_addr(serr)) { 469 const unsigned char *nh = skb_network_header(skb); 470 sin->sin6_family = AF_INET6; 471 sin->sin6_flowinfo = 0; 472 sin->sin6_port = serr->port; 473 if (skb->protocol == htons(ETH_P_IPV6)) { 474 const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset), 475 struct ipv6hdr, daddr); 476 sin->sin6_addr = ip6h->daddr; 477 if (np->sndflow) 478 sin->sin6_flowinfo = ip6_flowinfo(ip6h); 479 sin->sin6_scope_id = 480 ipv6_iface_scope_id(&sin->sin6_addr, 481 IP6CB(skb)->iif); 482 } else { 483 ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset), 484 &sin->sin6_addr); 485 sin->sin6_scope_id = 0; 486 } 487 *addr_len = sizeof(*sin); 488 } 489 490 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err)); 491 sin = &errhdr.offender; 492 memset(sin, 0, sizeof(*sin)); 493 494 if (ip6_datagram_support_cmsg(skb, serr)) { 495 sin->sin6_family = AF_INET6; 496 if (np->rxopt.all) 497 ip6_datagram_recv_common_ctl(sk, msg, skb); 498 if (skb->protocol == htons(ETH_P_IPV6)) { 499 sin->sin6_addr = ipv6_hdr(skb)->saddr; 500 if (np->rxopt.all) 501 ip6_datagram_recv_specific_ctl(sk, msg, skb); 502 sin->sin6_scope_id = 503 ipv6_iface_scope_id(&sin->sin6_addr, 504 IP6CB(skb)->iif); 505 } else { 506 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, 507 &sin->sin6_addr); 508 if (inet_sk(sk)->cmsg_flags) 509 ip_cmsg_recv(msg, skb); 510 } 511 } 512 513 put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr); 514 515 /* Now we could try to dump offended packet options */ 516 517 msg->msg_flags |= MSG_ERRQUEUE; 518 err = copied; 519 520 consume_skb(skb); 521 out: 522 return err; 523 } 524 EXPORT_SYMBOL_GPL(ipv6_recv_error); 525 526 /* 527 * Handle IPV6_RECVPATHMTU 528 */ 529 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len, 530 int *addr_len) 531 { 532 struct ipv6_pinfo *np = inet6_sk(sk); 533 struct sk_buff *skb; 534 struct ip6_mtuinfo mtu_info; 535 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name); 536 int err; 537 int copied; 538 539 err = -EAGAIN; 540 skb = xchg(&np->rxpmtu, NULL); 541 if (!skb) 542 goto out; 543 544 copied = skb->len; 545 if (copied > len) { 546 msg->msg_flags |= MSG_TRUNC; 547 copied = len; 548 } 549 err = skb_copy_datagram_msg(skb, 0, msg, copied); 550 if (err) 551 goto out_free_skb; 552 553 sock_recv_timestamp(msg, sk, skb); 554 555 memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info)); 556 557 if (sin) { 558 sin->sin6_family = AF_INET6; 559 sin->sin6_flowinfo = 0; 560 sin->sin6_port = 0; 561 sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id; 562 sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr; 563 *addr_len = sizeof(*sin); 564 } 565 566 put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info); 567 568 err = copied; 569 570 out_free_skb: 571 kfree_skb(skb); 572 out: 573 return err; 574 } 575 576 577 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg, 578 struct sk_buff *skb) 579 { 580 struct ipv6_pinfo *np = inet6_sk(sk); 581 bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6); 582 583 if (np->rxopt.bits.rxinfo) { 584 struct in6_pktinfo src_info; 585 586 if (is_ipv6) { 587 src_info.ipi6_ifindex = IP6CB(skb)->iif; 588 src_info.ipi6_addr = ipv6_hdr(skb)->daddr; 589 } else { 590 src_info.ipi6_ifindex = 591 PKTINFO_SKB_CB(skb)->ipi_ifindex; 592 ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr, 593 &src_info.ipi6_addr); 594 } 595 596 if (src_info.ipi6_ifindex >= 0) 597 put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO, 598 sizeof(src_info), &src_info); 599 } 600 } 601 602 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg, 603 struct sk_buff *skb) 604 { 605 struct ipv6_pinfo *np = inet6_sk(sk); 606 struct inet6_skb_parm *opt = IP6CB(skb); 607 unsigned char *nh = skb_network_header(skb); 608 609 if (np->rxopt.bits.rxhlim) { 610 int hlim = ipv6_hdr(skb)->hop_limit; 611 put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim); 612 } 613 614 if (np->rxopt.bits.rxtclass) { 615 int tclass = ipv6_get_dsfield(ipv6_hdr(skb)); 616 put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass); 617 } 618 619 if (np->rxopt.bits.rxflow) { 620 __be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh); 621 if (flowinfo) 622 put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo); 623 } 624 625 /* HbH is allowed only once */ 626 if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) { 627 u8 *ptr = nh + sizeof(struct ipv6hdr); 628 put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr); 629 } 630 631 if (opt->lastopt && 632 (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) { 633 /* 634 * Silly enough, but we need to reparse in order to 635 * report extension headers (except for HbH) 636 * in order. 637 * 638 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT 639 * (and WILL NOT be) defined because 640 * IPV6_RECVDSTOPTS is more generic. --yoshfuji 641 */ 642 unsigned int off = sizeof(struct ipv6hdr); 643 u8 nexthdr = ipv6_hdr(skb)->nexthdr; 644 645 while (off <= opt->lastopt) { 646 unsigned int len; 647 u8 *ptr = nh + off; 648 649 switch (nexthdr) { 650 case IPPROTO_DSTOPTS: 651 nexthdr = ptr[0]; 652 len = (ptr[1] + 1) << 3; 653 if (np->rxopt.bits.dstopts) 654 put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr); 655 break; 656 case IPPROTO_ROUTING: 657 nexthdr = ptr[0]; 658 len = (ptr[1] + 1) << 3; 659 if (np->rxopt.bits.srcrt) 660 put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr); 661 break; 662 case IPPROTO_AH: 663 nexthdr = ptr[0]; 664 len = (ptr[1] + 2) << 2; 665 break; 666 default: 667 nexthdr = ptr[0]; 668 len = (ptr[1] + 1) << 3; 669 break; 670 } 671 672 off += len; 673 } 674 } 675 676 /* socket options in old style */ 677 if (np->rxopt.bits.rxoinfo) { 678 struct in6_pktinfo src_info; 679 680 src_info.ipi6_ifindex = opt->iif; 681 src_info.ipi6_addr = ipv6_hdr(skb)->daddr; 682 put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info); 683 } 684 if (np->rxopt.bits.rxohlim) { 685 int hlim = ipv6_hdr(skb)->hop_limit; 686 put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim); 687 } 688 if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) { 689 u8 *ptr = nh + sizeof(struct ipv6hdr); 690 put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr); 691 } 692 if (np->rxopt.bits.odstopts && opt->dst0) { 693 u8 *ptr = nh + opt->dst0; 694 put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr); 695 } 696 if (np->rxopt.bits.osrcrt && opt->srcrt) { 697 struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt); 698 put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr); 699 } 700 if (np->rxopt.bits.odstopts && opt->dst1) { 701 u8 *ptr = nh + opt->dst1; 702 put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr); 703 } 704 if (np->rxopt.bits.rxorigdstaddr) { 705 struct sockaddr_in6 sin6; 706 __be16 *ports = (__be16 *) skb_transport_header(skb); 707 708 if (skb_transport_offset(skb) + 4 <= (int)skb->len) { 709 /* All current transport protocols have the port numbers in the 710 * first four bytes of the transport header and this function is 711 * written with this assumption in mind. 712 */ 713 714 sin6.sin6_family = AF_INET6; 715 sin6.sin6_addr = ipv6_hdr(skb)->daddr; 716 sin6.sin6_port = ports[1]; 717 sin6.sin6_flowinfo = 0; 718 sin6.sin6_scope_id = 719 ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr, 720 opt->iif); 721 722 put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6); 723 } 724 } 725 if (np->rxopt.bits.recvfragsize && opt->frag_max_size) { 726 int val = opt->frag_max_size; 727 728 put_cmsg(msg, SOL_IPV6, IPV6_RECVFRAGSIZE, sizeof(val), &val); 729 } 730 } 731 732 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg, 733 struct sk_buff *skb) 734 { 735 ip6_datagram_recv_common_ctl(sk, msg, skb); 736 ip6_datagram_recv_specific_ctl(sk, msg, skb); 737 } 738 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl); 739 740 int ip6_datagram_send_ctl(struct net *net, struct sock *sk, 741 struct msghdr *msg, struct flowi6 *fl6, 742 struct ipcm6_cookie *ipc6, struct sockcm_cookie *sockc) 743 { 744 struct in6_pktinfo *src_info; 745 struct cmsghdr *cmsg; 746 struct ipv6_rt_hdr *rthdr; 747 struct ipv6_opt_hdr *hdr; 748 struct ipv6_txoptions *opt = ipc6->opt; 749 int len; 750 int err = 0; 751 752 for_each_cmsghdr(cmsg, msg) { 753 int addr_type; 754 755 if (!CMSG_OK(msg, cmsg)) { 756 err = -EINVAL; 757 goto exit_f; 758 } 759 760 if (cmsg->cmsg_level == SOL_SOCKET) { 761 err = __sock_cmsg_send(sk, msg, cmsg, sockc); 762 if (err) 763 return err; 764 continue; 765 } 766 767 if (cmsg->cmsg_level != SOL_IPV6) 768 continue; 769 770 switch (cmsg->cmsg_type) { 771 case IPV6_PKTINFO: 772 case IPV6_2292PKTINFO: 773 { 774 struct net_device *dev = NULL; 775 776 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) { 777 err = -EINVAL; 778 goto exit_f; 779 } 780 781 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg); 782 783 if (src_info->ipi6_ifindex) { 784 if (fl6->flowi6_oif && 785 src_info->ipi6_ifindex != fl6->flowi6_oif) 786 return -EINVAL; 787 fl6->flowi6_oif = src_info->ipi6_ifindex; 788 } 789 790 addr_type = __ipv6_addr_type(&src_info->ipi6_addr); 791 792 rcu_read_lock(); 793 if (fl6->flowi6_oif) { 794 dev = dev_get_by_index_rcu(net, fl6->flowi6_oif); 795 if (!dev) { 796 rcu_read_unlock(); 797 return -ENODEV; 798 } 799 } else if (addr_type & IPV6_ADDR_LINKLOCAL) { 800 rcu_read_unlock(); 801 return -EINVAL; 802 } 803 804 if (addr_type != IPV6_ADDR_ANY) { 805 int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL; 806 if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) && 807 !ipv6_chk_addr(net, &src_info->ipi6_addr, 808 strict ? dev : NULL, 0) && 809 !ipv6_chk_acast_addr_src(net, dev, 810 &src_info->ipi6_addr)) 811 err = -EINVAL; 812 else 813 fl6->saddr = src_info->ipi6_addr; 814 } 815 816 rcu_read_unlock(); 817 818 if (err) 819 goto exit_f; 820 821 break; 822 } 823 824 case IPV6_FLOWINFO: 825 if (cmsg->cmsg_len < CMSG_LEN(4)) { 826 err = -EINVAL; 827 goto exit_f; 828 } 829 830 if (fl6->flowlabel&IPV6_FLOWINFO_MASK) { 831 if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) { 832 err = -EINVAL; 833 goto exit_f; 834 } 835 } 836 fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg); 837 break; 838 839 case IPV6_2292HOPOPTS: 840 case IPV6_HOPOPTS: 841 if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) { 842 err = -EINVAL; 843 goto exit_f; 844 } 845 846 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg); 847 len = ((hdr->hdrlen + 1) << 3); 848 if (cmsg->cmsg_len < CMSG_LEN(len)) { 849 err = -EINVAL; 850 goto exit_f; 851 } 852 if (!ns_capable(net->user_ns, CAP_NET_RAW)) { 853 err = -EPERM; 854 goto exit_f; 855 } 856 opt->opt_nflen += len; 857 opt->hopopt = hdr; 858 break; 859 860 case IPV6_2292DSTOPTS: 861 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) { 862 err = -EINVAL; 863 goto exit_f; 864 } 865 866 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg); 867 len = ((hdr->hdrlen + 1) << 3); 868 if (cmsg->cmsg_len < CMSG_LEN(len)) { 869 err = -EINVAL; 870 goto exit_f; 871 } 872 if (!ns_capable(net->user_ns, CAP_NET_RAW)) { 873 err = -EPERM; 874 goto exit_f; 875 } 876 if (opt->dst1opt) { 877 err = -EINVAL; 878 goto exit_f; 879 } 880 opt->opt_flen += len; 881 opt->dst1opt = hdr; 882 break; 883 884 case IPV6_DSTOPTS: 885 case IPV6_RTHDRDSTOPTS: 886 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) { 887 err = -EINVAL; 888 goto exit_f; 889 } 890 891 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg); 892 len = ((hdr->hdrlen + 1) << 3); 893 if (cmsg->cmsg_len < CMSG_LEN(len)) { 894 err = -EINVAL; 895 goto exit_f; 896 } 897 if (!ns_capable(net->user_ns, CAP_NET_RAW)) { 898 err = -EPERM; 899 goto exit_f; 900 } 901 if (cmsg->cmsg_type == IPV6_DSTOPTS) { 902 opt->opt_flen += len; 903 opt->dst1opt = hdr; 904 } else { 905 opt->opt_nflen += len; 906 opt->dst0opt = hdr; 907 } 908 break; 909 910 case IPV6_2292RTHDR: 911 case IPV6_RTHDR: 912 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) { 913 err = -EINVAL; 914 goto exit_f; 915 } 916 917 rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg); 918 919 switch (rthdr->type) { 920 #if IS_ENABLED(CONFIG_IPV6_MIP6) 921 case IPV6_SRCRT_TYPE_2: 922 if (rthdr->hdrlen != 2 || 923 rthdr->segments_left != 1) { 924 err = -EINVAL; 925 goto exit_f; 926 } 927 break; 928 #endif 929 default: 930 err = -EINVAL; 931 goto exit_f; 932 } 933 934 len = ((rthdr->hdrlen + 1) << 3); 935 936 if (cmsg->cmsg_len < CMSG_LEN(len)) { 937 err = -EINVAL; 938 goto exit_f; 939 } 940 941 /* segments left must also match */ 942 if ((rthdr->hdrlen >> 1) != rthdr->segments_left) { 943 err = -EINVAL; 944 goto exit_f; 945 } 946 947 opt->opt_nflen += len; 948 opt->srcrt = rthdr; 949 950 if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) { 951 int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3); 952 953 opt->opt_nflen += dsthdrlen; 954 opt->dst0opt = opt->dst1opt; 955 opt->dst1opt = NULL; 956 opt->opt_flen -= dsthdrlen; 957 } 958 959 break; 960 961 case IPV6_2292HOPLIMIT: 962 case IPV6_HOPLIMIT: 963 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) { 964 err = -EINVAL; 965 goto exit_f; 966 } 967 968 ipc6->hlimit = *(int *)CMSG_DATA(cmsg); 969 if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) { 970 err = -EINVAL; 971 goto exit_f; 972 } 973 974 break; 975 976 case IPV6_TCLASS: 977 { 978 int tc; 979 980 err = -EINVAL; 981 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) 982 goto exit_f; 983 984 tc = *(int *)CMSG_DATA(cmsg); 985 if (tc < -1 || tc > 0xff) 986 goto exit_f; 987 988 err = 0; 989 ipc6->tclass = tc; 990 991 break; 992 } 993 994 case IPV6_DONTFRAG: 995 { 996 int df; 997 998 err = -EINVAL; 999 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) 1000 goto exit_f; 1001 1002 df = *(int *)CMSG_DATA(cmsg); 1003 if (df < 0 || df > 1) 1004 goto exit_f; 1005 1006 err = 0; 1007 ipc6->dontfrag = df; 1008 1009 break; 1010 } 1011 default: 1012 net_dbg_ratelimited("invalid cmsg type: %d\n", 1013 cmsg->cmsg_type); 1014 err = -EINVAL; 1015 goto exit_f; 1016 } 1017 } 1018 1019 exit_f: 1020 return err; 1021 } 1022 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl); 1023 1024 void ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp, 1025 __u16 srcp, __u16 destp, int bucket) 1026 { 1027 const struct in6_addr *dest, *src; 1028 1029 dest = &sp->sk_v6_daddr; 1030 src = &sp->sk_v6_rcv_saddr; 1031 seq_printf(seq, 1032 "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " 1033 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n", 1034 bucket, 1035 src->s6_addr32[0], src->s6_addr32[1], 1036 src->s6_addr32[2], src->s6_addr32[3], srcp, 1037 dest->s6_addr32[0], dest->s6_addr32[1], 1038 dest->s6_addr32[2], dest->s6_addr32[3], destp, 1039 sp->sk_state, 1040 sk_wmem_alloc_get(sp), 1041 sk_rmem_alloc_get(sp), 1042 0, 0L, 0, 1043 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), 1044 0, 1045 sock_i_ino(sp), 1046 atomic_read(&sp->sk_refcnt), sp, 1047 atomic_read(&sp->sk_drops)); 1048 } 1049