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, old_daddr; 150 __be32 fl6_flowlabel = 0; 151 __be32 old_fl6_flowlabel; 152 __be16 old_dport; 153 int addr_type; 154 int err; 155 156 if (usin->sin6_family == AF_INET) { 157 if (__ipv6_only_sock(sk)) 158 return -EAFNOSUPPORT; 159 err = __ip4_datagram_connect(sk, uaddr, addr_len); 160 goto ipv4_connected; 161 } 162 163 if (addr_len < SIN6_LEN_RFC2133) 164 return -EINVAL; 165 166 if (usin->sin6_family != AF_INET6) 167 return -EAFNOSUPPORT; 168 169 if (np->sndflow) 170 fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK; 171 172 if (ipv6_addr_any(&usin->sin6_addr)) { 173 /* 174 * connect to self 175 */ 176 if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr)) 177 ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK), 178 &usin->sin6_addr); 179 else 180 usin->sin6_addr = in6addr_loopback; 181 } 182 183 addr_type = ipv6_addr_type(&usin->sin6_addr); 184 185 daddr = &usin->sin6_addr; 186 187 if (addr_type & IPV6_ADDR_MAPPED) { 188 struct sockaddr_in sin; 189 190 if (__ipv6_only_sock(sk)) { 191 err = -ENETUNREACH; 192 goto out; 193 } 194 sin.sin_family = AF_INET; 195 sin.sin_addr.s_addr = daddr->s6_addr32[3]; 196 sin.sin_port = usin->sin6_port; 197 198 err = __ip4_datagram_connect(sk, 199 (struct sockaddr *) &sin, 200 sizeof(sin)); 201 202 ipv4_connected: 203 if (err) 204 goto out; 205 206 ipv6_addr_set_v4mapped(inet->inet_daddr, &sk->sk_v6_daddr); 207 208 if (ipv6_addr_any(&np->saddr) || 209 ipv6_mapped_addr_any(&np->saddr)) 210 ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr); 211 212 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) || 213 ipv6_mapped_addr_any(&sk->sk_v6_rcv_saddr)) { 214 ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, 215 &sk->sk_v6_rcv_saddr); 216 if (sk->sk_prot->rehash) 217 sk->sk_prot->rehash(sk); 218 } 219 220 goto out; 221 } 222 223 if (__ipv6_addr_needs_scope_id(addr_type)) { 224 if (addr_len >= sizeof(struct sockaddr_in6) && 225 usin->sin6_scope_id) { 226 if (!sk_dev_equal_l3scope(sk, usin->sin6_scope_id)) { 227 err = -EINVAL; 228 goto out; 229 } 230 sk->sk_bound_dev_if = usin->sin6_scope_id; 231 } 232 233 if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST)) 234 sk->sk_bound_dev_if = np->mcast_oif; 235 236 /* Connect to link-local address requires an interface */ 237 if (!sk->sk_bound_dev_if) { 238 err = -EINVAL; 239 goto out; 240 } 241 } 242 243 /* save the current peer information before updating it */ 244 old_daddr = sk->sk_v6_daddr; 245 old_fl6_flowlabel = np->flow_label; 246 old_dport = inet->inet_dport; 247 248 sk->sk_v6_daddr = *daddr; 249 np->flow_label = fl6_flowlabel; 250 inet->inet_dport = usin->sin6_port; 251 252 /* 253 * Check for a route to destination an obtain the 254 * destination cache for it. 255 */ 256 257 err = ip6_datagram_dst_update(sk, true); 258 if (err) { 259 /* Restore the socket peer info, to keep it consistent with 260 * the old socket state 261 */ 262 sk->sk_v6_daddr = old_daddr; 263 np->flow_label = old_fl6_flowlabel; 264 inet->inet_dport = old_dport; 265 goto out; 266 } 267 268 sk->sk_state = TCP_ESTABLISHED; 269 sk_set_txhash(sk); 270 out: 271 return err; 272 } 273 EXPORT_SYMBOL_GPL(__ip6_datagram_connect); 274 275 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) 276 { 277 int res; 278 279 lock_sock(sk); 280 res = __ip6_datagram_connect(sk, uaddr, addr_len); 281 release_sock(sk); 282 return res; 283 } 284 EXPORT_SYMBOL_GPL(ip6_datagram_connect); 285 286 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *uaddr, 287 int addr_len) 288 { 289 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, uaddr); 290 if (sin6->sin6_family != AF_INET6) 291 return -EAFNOSUPPORT; 292 return ip6_datagram_connect(sk, uaddr, addr_len); 293 } 294 EXPORT_SYMBOL_GPL(ip6_datagram_connect_v6_only); 295 296 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, 297 __be16 port, u32 info, u8 *payload) 298 { 299 struct ipv6_pinfo *np = inet6_sk(sk); 300 struct icmp6hdr *icmph = icmp6_hdr(skb); 301 struct sock_exterr_skb *serr; 302 303 if (!np->recverr) 304 return; 305 306 skb = skb_clone(skb, GFP_ATOMIC); 307 if (!skb) 308 return; 309 310 skb->protocol = htons(ETH_P_IPV6); 311 312 serr = SKB_EXT_ERR(skb); 313 serr->ee.ee_errno = err; 314 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6; 315 serr->ee.ee_type = icmph->icmp6_type; 316 serr->ee.ee_code = icmph->icmp6_code; 317 serr->ee.ee_pad = 0; 318 serr->ee.ee_info = info; 319 serr->ee.ee_data = 0; 320 serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) - 321 skb_network_header(skb); 322 serr->port = port; 323 324 __skb_pull(skb, payload - skb->data); 325 skb_reset_transport_header(skb); 326 327 if (sock_queue_err_skb(sk, skb)) 328 kfree_skb(skb); 329 } 330 331 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info) 332 { 333 const struct ipv6_pinfo *np = inet6_sk(sk); 334 struct sock_exterr_skb *serr; 335 struct ipv6hdr *iph; 336 struct sk_buff *skb; 337 338 if (!np->recverr) 339 return; 340 341 skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 342 if (!skb) 343 return; 344 345 skb->protocol = htons(ETH_P_IPV6); 346 347 skb_put(skb, sizeof(struct ipv6hdr)); 348 skb_reset_network_header(skb); 349 iph = ipv6_hdr(skb); 350 iph->daddr = fl6->daddr; 351 352 serr = SKB_EXT_ERR(skb); 353 serr->ee.ee_errno = err; 354 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL; 355 serr->ee.ee_type = 0; 356 serr->ee.ee_code = 0; 357 serr->ee.ee_pad = 0; 358 serr->ee.ee_info = info; 359 serr->ee.ee_data = 0; 360 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb); 361 serr->port = fl6->fl6_dport; 362 363 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 364 skb_reset_transport_header(skb); 365 366 if (sock_queue_err_skb(sk, skb)) 367 kfree_skb(skb); 368 } 369 370 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu) 371 { 372 struct ipv6_pinfo *np = inet6_sk(sk); 373 struct ipv6hdr *iph; 374 struct sk_buff *skb; 375 struct ip6_mtuinfo *mtu_info; 376 377 if (!np->rxopt.bits.rxpmtu) 378 return; 379 380 skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC); 381 if (!skb) 382 return; 383 384 skb_put(skb, sizeof(struct ipv6hdr)); 385 skb_reset_network_header(skb); 386 iph = ipv6_hdr(skb); 387 iph->daddr = fl6->daddr; 388 389 mtu_info = IP6CBMTU(skb); 390 391 mtu_info->ip6m_mtu = mtu; 392 mtu_info->ip6m_addr.sin6_family = AF_INET6; 393 mtu_info->ip6m_addr.sin6_port = 0; 394 mtu_info->ip6m_addr.sin6_flowinfo = 0; 395 mtu_info->ip6m_addr.sin6_scope_id = fl6->flowi6_oif; 396 mtu_info->ip6m_addr.sin6_addr = ipv6_hdr(skb)->daddr; 397 398 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 399 skb_reset_transport_header(skb); 400 401 skb = xchg(&np->rxpmtu, skb); 402 kfree_skb(skb); 403 } 404 405 /* For some errors we have valid addr_offset even with zero payload and 406 * zero port. Also, addr_offset should be supported if port is set. 407 */ 408 static inline bool ipv6_datagram_support_addr(struct sock_exterr_skb *serr) 409 { 410 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6 || 411 serr->ee.ee_origin == SO_EE_ORIGIN_ICMP || 412 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port; 413 } 414 415 /* IPv6 supports cmsg on all origins aside from SO_EE_ORIGIN_LOCAL. 416 * 417 * At one point, excluding local errors was a quick test to identify icmp/icmp6 418 * errors. This is no longer true, but the test remained, so the v6 stack, 419 * unlike v4, also honors cmsg requests on all wifi and timestamp errors. 420 */ 421 static bool ip6_datagram_support_cmsg(struct sk_buff *skb, 422 struct sock_exterr_skb *serr) 423 { 424 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP || 425 serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6) 426 return true; 427 428 if (serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL) 429 return false; 430 431 if (!IP6CB(skb)->iif) 432 return false; 433 434 return true; 435 } 436 437 /* 438 * Handle MSG_ERRQUEUE 439 */ 440 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) 441 { 442 struct ipv6_pinfo *np = inet6_sk(sk); 443 struct sock_exterr_skb *serr; 444 struct sk_buff *skb; 445 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name); 446 struct { 447 struct sock_extended_err ee; 448 struct sockaddr_in6 offender; 449 } errhdr; 450 int err; 451 int copied; 452 453 err = -EAGAIN; 454 skb = sock_dequeue_err_skb(sk); 455 if (!skb) 456 goto out; 457 458 copied = skb->len; 459 if (copied > len) { 460 msg->msg_flags |= MSG_TRUNC; 461 copied = len; 462 } 463 err = skb_copy_datagram_msg(skb, 0, msg, copied); 464 if (unlikely(err)) { 465 kfree_skb(skb); 466 return err; 467 } 468 sock_recv_timestamp(msg, sk, skb); 469 470 serr = SKB_EXT_ERR(skb); 471 472 if (sin && ipv6_datagram_support_addr(serr)) { 473 const unsigned char *nh = skb_network_header(skb); 474 sin->sin6_family = AF_INET6; 475 sin->sin6_flowinfo = 0; 476 sin->sin6_port = serr->port; 477 if (skb->protocol == htons(ETH_P_IPV6)) { 478 const struct ipv6hdr *ip6h = container_of((struct in6_addr *)(nh + serr->addr_offset), 479 struct ipv6hdr, daddr); 480 sin->sin6_addr = ip6h->daddr; 481 if (np->sndflow) 482 sin->sin6_flowinfo = ip6_flowinfo(ip6h); 483 sin->sin6_scope_id = 484 ipv6_iface_scope_id(&sin->sin6_addr, 485 IP6CB(skb)->iif); 486 } else { 487 ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset), 488 &sin->sin6_addr); 489 sin->sin6_scope_id = 0; 490 } 491 *addr_len = sizeof(*sin); 492 } 493 494 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err)); 495 sin = &errhdr.offender; 496 memset(sin, 0, sizeof(*sin)); 497 498 if (ip6_datagram_support_cmsg(skb, serr)) { 499 sin->sin6_family = AF_INET6; 500 if (np->rxopt.all) 501 ip6_datagram_recv_common_ctl(sk, msg, skb); 502 if (skb->protocol == htons(ETH_P_IPV6)) { 503 sin->sin6_addr = ipv6_hdr(skb)->saddr; 504 if (np->rxopt.all) 505 ip6_datagram_recv_specific_ctl(sk, msg, skb); 506 sin->sin6_scope_id = 507 ipv6_iface_scope_id(&sin->sin6_addr, 508 IP6CB(skb)->iif); 509 } else { 510 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, 511 &sin->sin6_addr); 512 if (inet_sk(sk)->cmsg_flags) 513 ip_cmsg_recv(msg, skb); 514 } 515 } 516 517 put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr); 518 519 /* Now we could try to dump offended packet options */ 520 521 msg->msg_flags |= MSG_ERRQUEUE; 522 err = copied; 523 524 consume_skb(skb); 525 out: 526 return err; 527 } 528 EXPORT_SYMBOL_GPL(ipv6_recv_error); 529 530 /* 531 * Handle IPV6_RECVPATHMTU 532 */ 533 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len, 534 int *addr_len) 535 { 536 struct ipv6_pinfo *np = inet6_sk(sk); 537 struct sk_buff *skb; 538 struct ip6_mtuinfo mtu_info; 539 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin, msg->msg_name); 540 int err; 541 int copied; 542 543 err = -EAGAIN; 544 skb = xchg(&np->rxpmtu, NULL); 545 if (!skb) 546 goto out; 547 548 copied = skb->len; 549 if (copied > len) { 550 msg->msg_flags |= MSG_TRUNC; 551 copied = len; 552 } 553 err = skb_copy_datagram_msg(skb, 0, msg, copied); 554 if (err) 555 goto out_free_skb; 556 557 sock_recv_timestamp(msg, sk, skb); 558 559 memcpy(&mtu_info, IP6CBMTU(skb), sizeof(mtu_info)); 560 561 if (sin) { 562 sin->sin6_family = AF_INET6; 563 sin->sin6_flowinfo = 0; 564 sin->sin6_port = 0; 565 sin->sin6_scope_id = mtu_info.ip6m_addr.sin6_scope_id; 566 sin->sin6_addr = mtu_info.ip6m_addr.sin6_addr; 567 *addr_len = sizeof(*sin); 568 } 569 570 put_cmsg(msg, SOL_IPV6, IPV6_PATHMTU, sizeof(mtu_info), &mtu_info); 571 572 err = copied; 573 574 out_free_skb: 575 kfree_skb(skb); 576 out: 577 return err; 578 } 579 580 581 void ip6_datagram_recv_common_ctl(struct sock *sk, struct msghdr *msg, 582 struct sk_buff *skb) 583 { 584 struct ipv6_pinfo *np = inet6_sk(sk); 585 bool is_ipv6 = skb->protocol == htons(ETH_P_IPV6); 586 587 if (np->rxopt.bits.rxinfo) { 588 struct in6_pktinfo src_info; 589 590 if (is_ipv6) { 591 src_info.ipi6_ifindex = IP6CB(skb)->iif; 592 src_info.ipi6_addr = ipv6_hdr(skb)->daddr; 593 } else { 594 src_info.ipi6_ifindex = 595 PKTINFO_SKB_CB(skb)->ipi_ifindex; 596 ipv6_addr_set_v4mapped(ip_hdr(skb)->daddr, 597 &src_info.ipi6_addr); 598 } 599 600 if (src_info.ipi6_ifindex >= 0) 601 put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO, 602 sizeof(src_info), &src_info); 603 } 604 } 605 606 void ip6_datagram_recv_specific_ctl(struct sock *sk, struct msghdr *msg, 607 struct sk_buff *skb) 608 { 609 struct ipv6_pinfo *np = inet6_sk(sk); 610 struct inet6_skb_parm *opt = IP6CB(skb); 611 unsigned char *nh = skb_network_header(skb); 612 613 if (np->rxopt.bits.rxhlim) { 614 int hlim = ipv6_hdr(skb)->hop_limit; 615 put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim); 616 } 617 618 if (np->rxopt.bits.rxtclass) { 619 int tclass = ipv6_get_dsfield(ipv6_hdr(skb)); 620 put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass); 621 } 622 623 if (np->rxopt.bits.rxflow) { 624 __be32 flowinfo = ip6_flowinfo((struct ipv6hdr *)nh); 625 if (flowinfo) 626 put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo); 627 } 628 629 /* HbH is allowed only once */ 630 if (np->rxopt.bits.hopopts && (opt->flags & IP6SKB_HOPBYHOP)) { 631 u8 *ptr = nh + sizeof(struct ipv6hdr); 632 put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr); 633 } 634 635 if (opt->lastopt && 636 (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) { 637 /* 638 * Silly enough, but we need to reparse in order to 639 * report extension headers (except for HbH) 640 * in order. 641 * 642 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT 643 * (and WILL NOT be) defined because 644 * IPV6_RECVDSTOPTS is more generic. --yoshfuji 645 */ 646 unsigned int off = sizeof(struct ipv6hdr); 647 u8 nexthdr = ipv6_hdr(skb)->nexthdr; 648 649 while (off <= opt->lastopt) { 650 unsigned int len; 651 u8 *ptr = nh + off; 652 653 switch (nexthdr) { 654 case IPPROTO_DSTOPTS: 655 nexthdr = ptr[0]; 656 len = (ptr[1] + 1) << 3; 657 if (np->rxopt.bits.dstopts) 658 put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr); 659 break; 660 case IPPROTO_ROUTING: 661 nexthdr = ptr[0]; 662 len = (ptr[1] + 1) << 3; 663 if (np->rxopt.bits.srcrt) 664 put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr); 665 break; 666 case IPPROTO_AH: 667 nexthdr = ptr[0]; 668 len = (ptr[1] + 2) << 2; 669 break; 670 default: 671 nexthdr = ptr[0]; 672 len = (ptr[1] + 1) << 3; 673 break; 674 } 675 676 off += len; 677 } 678 } 679 680 /* socket options in old style */ 681 if (np->rxopt.bits.rxoinfo) { 682 struct in6_pktinfo src_info; 683 684 src_info.ipi6_ifindex = opt->iif; 685 src_info.ipi6_addr = ipv6_hdr(skb)->daddr; 686 put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info); 687 } 688 if (np->rxopt.bits.rxohlim) { 689 int hlim = ipv6_hdr(skb)->hop_limit; 690 put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim); 691 } 692 if (np->rxopt.bits.ohopopts && (opt->flags & IP6SKB_HOPBYHOP)) { 693 u8 *ptr = nh + sizeof(struct ipv6hdr); 694 put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr); 695 } 696 if (np->rxopt.bits.odstopts && opt->dst0) { 697 u8 *ptr = nh + opt->dst0; 698 put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr); 699 } 700 if (np->rxopt.bits.osrcrt && opt->srcrt) { 701 struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt); 702 put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr); 703 } 704 if (np->rxopt.bits.odstopts && opt->dst1) { 705 u8 *ptr = nh + opt->dst1; 706 put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr); 707 } 708 if (np->rxopt.bits.rxorigdstaddr) { 709 struct sockaddr_in6 sin6; 710 __be16 *ports = (__be16 *) skb_transport_header(skb); 711 712 if (skb_transport_offset(skb) + 4 <= (int)skb->len) { 713 /* All current transport protocols have the port numbers in the 714 * first four bytes of the transport header and this function is 715 * written with this assumption in mind. 716 */ 717 718 sin6.sin6_family = AF_INET6; 719 sin6.sin6_addr = ipv6_hdr(skb)->daddr; 720 sin6.sin6_port = ports[1]; 721 sin6.sin6_flowinfo = 0; 722 sin6.sin6_scope_id = 723 ipv6_iface_scope_id(&ipv6_hdr(skb)->daddr, 724 opt->iif); 725 726 put_cmsg(msg, SOL_IPV6, IPV6_ORIGDSTADDR, sizeof(sin6), &sin6); 727 } 728 } 729 if (np->rxopt.bits.recvfragsize && opt->frag_max_size) { 730 int val = opt->frag_max_size; 731 732 put_cmsg(msg, SOL_IPV6, IPV6_RECVFRAGSIZE, sizeof(val), &val); 733 } 734 } 735 736 void ip6_datagram_recv_ctl(struct sock *sk, struct msghdr *msg, 737 struct sk_buff *skb) 738 { 739 ip6_datagram_recv_common_ctl(sk, msg, skb); 740 ip6_datagram_recv_specific_ctl(sk, msg, skb); 741 } 742 EXPORT_SYMBOL_GPL(ip6_datagram_recv_ctl); 743 744 int ip6_datagram_send_ctl(struct net *net, struct sock *sk, 745 struct msghdr *msg, struct flowi6 *fl6, 746 struct ipcm6_cookie *ipc6, struct sockcm_cookie *sockc) 747 { 748 struct in6_pktinfo *src_info; 749 struct cmsghdr *cmsg; 750 struct ipv6_rt_hdr *rthdr; 751 struct ipv6_opt_hdr *hdr; 752 struct ipv6_txoptions *opt = ipc6->opt; 753 int len; 754 int err = 0; 755 756 for_each_cmsghdr(cmsg, msg) { 757 int addr_type; 758 759 if (!CMSG_OK(msg, cmsg)) { 760 err = -EINVAL; 761 goto exit_f; 762 } 763 764 if (cmsg->cmsg_level == SOL_SOCKET) { 765 err = __sock_cmsg_send(sk, msg, cmsg, sockc); 766 if (err) 767 return err; 768 continue; 769 } 770 771 if (cmsg->cmsg_level != SOL_IPV6) 772 continue; 773 774 switch (cmsg->cmsg_type) { 775 case IPV6_PKTINFO: 776 case IPV6_2292PKTINFO: 777 { 778 struct net_device *dev = NULL; 779 780 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) { 781 err = -EINVAL; 782 goto exit_f; 783 } 784 785 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg); 786 787 if (src_info->ipi6_ifindex) { 788 if (fl6->flowi6_oif && 789 src_info->ipi6_ifindex != fl6->flowi6_oif) 790 return -EINVAL; 791 fl6->flowi6_oif = src_info->ipi6_ifindex; 792 } 793 794 addr_type = __ipv6_addr_type(&src_info->ipi6_addr); 795 796 rcu_read_lock(); 797 if (fl6->flowi6_oif) { 798 dev = dev_get_by_index_rcu(net, fl6->flowi6_oif); 799 if (!dev) { 800 rcu_read_unlock(); 801 return -ENODEV; 802 } 803 } else if (addr_type & IPV6_ADDR_LINKLOCAL) { 804 rcu_read_unlock(); 805 return -EINVAL; 806 } 807 808 if (addr_type != IPV6_ADDR_ANY) { 809 int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL; 810 if (!(inet_sk(sk)->freebind || inet_sk(sk)->transparent) && 811 !ipv6_chk_addr_and_flags(net, &src_info->ipi6_addr, 812 dev, !strict, 0, 813 IFA_F_TENTATIVE) && 814 !ipv6_chk_acast_addr_src(net, dev, 815 &src_info->ipi6_addr)) 816 err = -EINVAL; 817 else 818 fl6->saddr = src_info->ipi6_addr; 819 } 820 821 rcu_read_unlock(); 822 823 if (err) 824 goto exit_f; 825 826 break; 827 } 828 829 case IPV6_FLOWINFO: 830 if (cmsg->cmsg_len < CMSG_LEN(4)) { 831 err = -EINVAL; 832 goto exit_f; 833 } 834 835 if (fl6->flowlabel&IPV6_FLOWINFO_MASK) { 836 if ((fl6->flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) { 837 err = -EINVAL; 838 goto exit_f; 839 } 840 } 841 fl6->flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg); 842 break; 843 844 case IPV6_2292HOPOPTS: 845 case IPV6_HOPOPTS: 846 if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) { 847 err = -EINVAL; 848 goto exit_f; 849 } 850 851 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg); 852 len = ((hdr->hdrlen + 1) << 3); 853 if (cmsg->cmsg_len < CMSG_LEN(len)) { 854 err = -EINVAL; 855 goto exit_f; 856 } 857 if (!ns_capable(net->user_ns, CAP_NET_RAW)) { 858 err = -EPERM; 859 goto exit_f; 860 } 861 opt->opt_nflen += len; 862 opt->hopopt = hdr; 863 break; 864 865 case IPV6_2292DSTOPTS: 866 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) { 867 err = -EINVAL; 868 goto exit_f; 869 } 870 871 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg); 872 len = ((hdr->hdrlen + 1) << 3); 873 if (cmsg->cmsg_len < CMSG_LEN(len)) { 874 err = -EINVAL; 875 goto exit_f; 876 } 877 if (!ns_capable(net->user_ns, CAP_NET_RAW)) { 878 err = -EPERM; 879 goto exit_f; 880 } 881 if (opt->dst1opt) { 882 err = -EINVAL; 883 goto exit_f; 884 } 885 opt->opt_flen += len; 886 opt->dst1opt = hdr; 887 break; 888 889 case IPV6_DSTOPTS: 890 case IPV6_RTHDRDSTOPTS: 891 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) { 892 err = -EINVAL; 893 goto exit_f; 894 } 895 896 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg); 897 len = ((hdr->hdrlen + 1) << 3); 898 if (cmsg->cmsg_len < CMSG_LEN(len)) { 899 err = -EINVAL; 900 goto exit_f; 901 } 902 if (!ns_capable(net->user_ns, CAP_NET_RAW)) { 903 err = -EPERM; 904 goto exit_f; 905 } 906 if (cmsg->cmsg_type == IPV6_DSTOPTS) { 907 opt->opt_flen += len; 908 opt->dst1opt = hdr; 909 } else { 910 opt->opt_nflen += len; 911 opt->dst0opt = hdr; 912 } 913 break; 914 915 case IPV6_2292RTHDR: 916 case IPV6_RTHDR: 917 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) { 918 err = -EINVAL; 919 goto exit_f; 920 } 921 922 rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg); 923 924 switch (rthdr->type) { 925 #if IS_ENABLED(CONFIG_IPV6_MIP6) 926 case IPV6_SRCRT_TYPE_2: 927 if (rthdr->hdrlen != 2 || 928 rthdr->segments_left != 1) { 929 err = -EINVAL; 930 goto exit_f; 931 } 932 break; 933 #endif 934 default: 935 err = -EINVAL; 936 goto exit_f; 937 } 938 939 len = ((rthdr->hdrlen + 1) << 3); 940 941 if (cmsg->cmsg_len < CMSG_LEN(len)) { 942 err = -EINVAL; 943 goto exit_f; 944 } 945 946 /* segments left must also match */ 947 if ((rthdr->hdrlen >> 1) != rthdr->segments_left) { 948 err = -EINVAL; 949 goto exit_f; 950 } 951 952 opt->opt_nflen += len; 953 opt->srcrt = rthdr; 954 955 if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) { 956 int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3); 957 958 opt->opt_nflen += dsthdrlen; 959 opt->dst0opt = opt->dst1opt; 960 opt->dst1opt = NULL; 961 opt->opt_flen -= dsthdrlen; 962 } 963 964 break; 965 966 case IPV6_2292HOPLIMIT: 967 case IPV6_HOPLIMIT: 968 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) { 969 err = -EINVAL; 970 goto exit_f; 971 } 972 973 ipc6->hlimit = *(int *)CMSG_DATA(cmsg); 974 if (ipc6->hlimit < -1 || ipc6->hlimit > 0xff) { 975 err = -EINVAL; 976 goto exit_f; 977 } 978 979 break; 980 981 case IPV6_TCLASS: 982 { 983 int tc; 984 985 err = -EINVAL; 986 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) 987 goto exit_f; 988 989 tc = *(int *)CMSG_DATA(cmsg); 990 if (tc < -1 || tc > 0xff) 991 goto exit_f; 992 993 err = 0; 994 ipc6->tclass = tc; 995 996 break; 997 } 998 999 case IPV6_DONTFRAG: 1000 { 1001 int df; 1002 1003 err = -EINVAL; 1004 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) 1005 goto exit_f; 1006 1007 df = *(int *)CMSG_DATA(cmsg); 1008 if (df < 0 || df > 1) 1009 goto exit_f; 1010 1011 err = 0; 1012 ipc6->dontfrag = df; 1013 1014 break; 1015 } 1016 default: 1017 net_dbg_ratelimited("invalid cmsg type: %d\n", 1018 cmsg->cmsg_type); 1019 err = -EINVAL; 1020 goto exit_f; 1021 } 1022 } 1023 1024 exit_f: 1025 return err; 1026 } 1027 EXPORT_SYMBOL_GPL(ip6_datagram_send_ctl); 1028 1029 void ip6_dgram_sock_seq_show(struct seq_file *seq, struct sock *sp, 1030 __u16 srcp, __u16 destp, int bucket) 1031 { 1032 const struct in6_addr *dest, *src; 1033 1034 dest = &sp->sk_v6_daddr; 1035 src = &sp->sk_v6_rcv_saddr; 1036 seq_printf(seq, 1037 "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " 1038 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n", 1039 bucket, 1040 src->s6_addr32[0], src->s6_addr32[1], 1041 src->s6_addr32[2], src->s6_addr32[3], srcp, 1042 dest->s6_addr32[0], dest->s6_addr32[1], 1043 dest->s6_addr32[2], dest->s6_addr32[3], destp, 1044 sp->sk_state, 1045 sk_wmem_alloc_get(sp), 1046 sk_rmem_alloc_get(sp), 1047 0, 0L, 0, 1048 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), 1049 0, 1050 sock_i_ino(sp), 1051 refcount_read(&sp->sk_refcnt), sp, 1052 atomic_read(&sp->sk_drops)); 1053 } 1054