1 /* 2 * UDP over IPv6 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Pedro Roque <roque@di.fc.ul.pt> 7 * 8 * Based on linux/ipv4/udp.c 9 * 10 * Fixes: 11 * Hideaki YOSHIFUJI : sin6_scope_id support 12 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which 13 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind 14 * a single port at the same time. 15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data 16 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/udp6 to seq_file. 17 * 18 * This program is free software; you can redistribute it and/or 19 * modify it under the terms of the GNU General Public License 20 * as published by the Free Software Foundation; either version 21 * 2 of the License, or (at your option) any later version. 22 */ 23 24 #include <linux/errno.h> 25 #include <linux/types.h> 26 #include <linux/socket.h> 27 #include <linux/sockios.h> 28 #include <linux/net.h> 29 #include <linux/in6.h> 30 #include <linux/netdevice.h> 31 #include <linux/if_arp.h> 32 #include <linux/ipv6.h> 33 #include <linux/icmpv6.h> 34 #include <linux/init.h> 35 #include <linux/module.h> 36 #include <linux/skbuff.h> 37 #include <linux/slab.h> 38 #include <asm/uaccess.h> 39 40 #include <net/ndisc.h> 41 #include <net/protocol.h> 42 #include <net/transp_v6.h> 43 #include <net/ip6_route.h> 44 #include <net/raw.h> 45 #include <net/tcp_states.h> 46 #include <net/ip6_checksum.h> 47 #include <net/xfrm.h> 48 49 #include <linux/proc_fs.h> 50 #include <linux/seq_file.h> 51 #include "udp_impl.h" 52 53 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2) 54 { 55 const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr; 56 const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2); 57 __be32 sk1_rcv_saddr = inet_sk(sk)->inet_rcv_saddr; 58 __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2); 59 int sk_ipv6only = ipv6_only_sock(sk); 60 int sk2_ipv6only = inet_v6_ipv6only(sk2); 61 int addr_type = ipv6_addr_type(sk_rcv_saddr6); 62 int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED; 63 64 /* if both are mapped, treat as IPv4 */ 65 if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED) 66 return (!sk2_ipv6only && 67 (!sk1_rcv_saddr || !sk2_rcv_saddr || 68 sk1_rcv_saddr == sk2_rcv_saddr)); 69 70 if (addr_type2 == IPV6_ADDR_ANY && 71 !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED)) 72 return 1; 73 74 if (addr_type == IPV6_ADDR_ANY && 75 !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED)) 76 return 1; 77 78 if (sk2_rcv_saddr6 && 79 ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6)) 80 return 1; 81 82 return 0; 83 } 84 85 static unsigned int udp6_portaddr_hash(struct net *net, 86 const struct in6_addr *addr6, 87 unsigned int port) 88 { 89 unsigned int hash, mix = net_hash_mix(net); 90 91 if (ipv6_addr_any(addr6)) 92 hash = jhash_1word(0, mix); 93 else if (ipv6_addr_v4mapped(addr6)) 94 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix); 95 else 96 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix); 97 98 return hash ^ port; 99 } 100 101 102 int udp_v6_get_port(struct sock *sk, unsigned short snum) 103 { 104 unsigned int hash2_nulladdr = 105 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum); 106 unsigned int hash2_partial = 107 udp6_portaddr_hash(sock_net(sk), &inet6_sk(sk)->rcv_saddr, 0); 108 109 /* precompute partial secondary hash */ 110 udp_sk(sk)->udp_portaddr_hash = hash2_partial; 111 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr); 112 } 113 114 static inline int compute_score(struct sock *sk, struct net *net, 115 unsigned short hnum, 116 struct in6_addr *saddr, __be16 sport, 117 struct in6_addr *daddr, __be16 dport, 118 int dif) 119 { 120 int score = -1; 121 122 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum && 123 sk->sk_family == PF_INET6) { 124 struct ipv6_pinfo *np = inet6_sk(sk); 125 struct inet_sock *inet = inet_sk(sk); 126 127 score = 0; 128 if (inet->inet_dport) { 129 if (inet->inet_dport != sport) 130 return -1; 131 score++; 132 } 133 if (!ipv6_addr_any(&np->rcv_saddr)) { 134 if (!ipv6_addr_equal(&np->rcv_saddr, daddr)) 135 return -1; 136 score++; 137 } 138 if (!ipv6_addr_any(&np->daddr)) { 139 if (!ipv6_addr_equal(&np->daddr, saddr)) 140 return -1; 141 score++; 142 } 143 if (sk->sk_bound_dev_if) { 144 if (sk->sk_bound_dev_if != dif) 145 return -1; 146 score++; 147 } 148 } 149 return score; 150 } 151 152 #define SCORE2_MAX (1 + 1 + 1) 153 static inline int compute_score2(struct sock *sk, struct net *net, 154 const struct in6_addr *saddr, __be16 sport, 155 const struct in6_addr *daddr, unsigned short hnum, 156 int dif) 157 { 158 int score = -1; 159 160 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum && 161 sk->sk_family == PF_INET6) { 162 struct ipv6_pinfo *np = inet6_sk(sk); 163 struct inet_sock *inet = inet_sk(sk); 164 165 if (!ipv6_addr_equal(&np->rcv_saddr, daddr)) 166 return -1; 167 score = 0; 168 if (inet->inet_dport) { 169 if (inet->inet_dport != sport) 170 return -1; 171 score++; 172 } 173 if (!ipv6_addr_any(&np->daddr)) { 174 if (!ipv6_addr_equal(&np->daddr, saddr)) 175 return -1; 176 score++; 177 } 178 if (sk->sk_bound_dev_if) { 179 if (sk->sk_bound_dev_if != dif) 180 return -1; 181 score++; 182 } 183 } 184 return score; 185 } 186 187 188 /* called with read_rcu_lock() */ 189 static struct sock *udp6_lib_lookup2(struct net *net, 190 const struct in6_addr *saddr, __be16 sport, 191 const struct in6_addr *daddr, unsigned int hnum, int dif, 192 struct udp_hslot *hslot2, unsigned int slot2) 193 { 194 struct sock *sk, *result; 195 struct hlist_nulls_node *node; 196 int score, badness; 197 198 begin: 199 result = NULL; 200 badness = -1; 201 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { 202 score = compute_score2(sk, net, saddr, sport, 203 daddr, hnum, dif); 204 if (score > badness) { 205 result = sk; 206 badness = score; 207 if (score == SCORE2_MAX) 208 goto exact_match; 209 } 210 } 211 /* 212 * if the nulls value we got at the end of this lookup is 213 * not the expected one, we must restart lookup. 214 * We probably met an item that was moved to another chain. 215 */ 216 if (get_nulls_value(node) != slot2) 217 goto begin; 218 219 if (result) { 220 exact_match: 221 if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt))) 222 result = NULL; 223 else if (unlikely(compute_score2(result, net, saddr, sport, 224 daddr, hnum, dif) < badness)) { 225 sock_put(result); 226 goto begin; 227 } 228 } 229 return result; 230 } 231 232 static struct sock *__udp6_lib_lookup(struct net *net, 233 struct in6_addr *saddr, __be16 sport, 234 struct in6_addr *daddr, __be16 dport, 235 int dif, struct udp_table *udptable) 236 { 237 struct sock *sk, *result; 238 struct hlist_nulls_node *node; 239 unsigned short hnum = ntohs(dport); 240 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); 241 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; 242 int score, badness; 243 244 rcu_read_lock(); 245 if (hslot->count > 10) { 246 hash2 = udp6_portaddr_hash(net, daddr, hnum); 247 slot2 = hash2 & udptable->mask; 248 hslot2 = &udptable->hash2[slot2]; 249 if (hslot->count < hslot2->count) 250 goto begin; 251 252 result = udp6_lib_lookup2(net, saddr, sport, 253 daddr, hnum, dif, 254 hslot2, slot2); 255 if (!result) { 256 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum); 257 slot2 = hash2 & udptable->mask; 258 hslot2 = &udptable->hash2[slot2]; 259 if (hslot->count < hslot2->count) 260 goto begin; 261 262 result = udp6_lib_lookup2(net, saddr, sport, 263 &in6addr_any, hnum, dif, 264 hslot2, slot2); 265 } 266 rcu_read_unlock(); 267 return result; 268 } 269 begin: 270 result = NULL; 271 badness = -1; 272 sk_nulls_for_each_rcu(sk, node, &hslot->head) { 273 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif); 274 if (score > badness) { 275 result = sk; 276 badness = score; 277 } 278 } 279 /* 280 * if the nulls value we got at the end of this lookup is 281 * not the expected one, we must restart lookup. 282 * We probably met an item that was moved to another chain. 283 */ 284 if (get_nulls_value(node) != slot) 285 goto begin; 286 287 if (result) { 288 if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt))) 289 result = NULL; 290 else if (unlikely(compute_score(result, net, hnum, saddr, sport, 291 daddr, dport, dif) < badness)) { 292 sock_put(result); 293 goto begin; 294 } 295 } 296 rcu_read_unlock(); 297 return result; 298 } 299 300 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb, 301 __be16 sport, __be16 dport, 302 struct udp_table *udptable) 303 { 304 struct sock *sk; 305 struct ipv6hdr *iph = ipv6_hdr(skb); 306 307 if (unlikely(sk = skb_steal_sock(skb))) 308 return sk; 309 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport, 310 &iph->daddr, dport, inet6_iif(skb), 311 udptable); 312 } 313 314 /* 315 * This should be easy, if there is something there we 316 * return it, otherwise we block. 317 */ 318 319 int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk, 320 struct msghdr *msg, size_t len, 321 int noblock, int flags, int *addr_len) 322 { 323 struct ipv6_pinfo *np = inet6_sk(sk); 324 struct inet_sock *inet = inet_sk(sk); 325 struct sk_buff *skb; 326 unsigned int ulen; 327 int peeked; 328 int err; 329 int is_udplite = IS_UDPLITE(sk); 330 int is_udp4; 331 bool slow; 332 333 if (addr_len) 334 *addr_len=sizeof(struct sockaddr_in6); 335 336 if (flags & MSG_ERRQUEUE) 337 return ipv6_recv_error(sk, msg, len); 338 339 if (np->rxpmtu && np->rxopt.bits.rxpmtu) 340 return ipv6_recv_rxpmtu(sk, msg, len); 341 342 try_again: 343 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), 344 &peeked, &err); 345 if (!skb) 346 goto out; 347 348 ulen = skb->len - sizeof(struct udphdr); 349 if (len > ulen) 350 len = ulen; 351 else if (len < ulen) 352 msg->msg_flags |= MSG_TRUNC; 353 354 is_udp4 = (skb->protocol == htons(ETH_P_IP)); 355 356 /* 357 * If checksum is needed at all, try to do it while copying the 358 * data. If the data is truncated, or if we only want a partial 359 * coverage checksum (UDP-Lite), do it before the copy. 360 */ 361 362 if (len < ulen || UDP_SKB_CB(skb)->partial_cov) { 363 if (udp_lib_checksum_complete(skb)) 364 goto csum_copy_err; 365 } 366 367 if (skb_csum_unnecessary(skb)) 368 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), 369 msg->msg_iov,len); 370 else { 371 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov); 372 if (err == -EINVAL) 373 goto csum_copy_err; 374 } 375 if (err) 376 goto out_free; 377 378 if (!peeked) { 379 if (is_udp4) 380 UDP_INC_STATS_USER(sock_net(sk), 381 UDP_MIB_INDATAGRAMS, is_udplite); 382 else 383 UDP6_INC_STATS_USER(sock_net(sk), 384 UDP_MIB_INDATAGRAMS, is_udplite); 385 } 386 387 sock_recv_ts_and_drops(msg, sk, skb); 388 389 /* Copy the address. */ 390 if (msg->msg_name) { 391 struct sockaddr_in6 *sin6; 392 393 sin6 = (struct sockaddr_in6 *) msg->msg_name; 394 sin6->sin6_family = AF_INET6; 395 sin6->sin6_port = udp_hdr(skb)->source; 396 sin6->sin6_flowinfo = 0; 397 sin6->sin6_scope_id = 0; 398 399 if (is_udp4) 400 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, 401 &sin6->sin6_addr); 402 else { 403 ipv6_addr_copy(&sin6->sin6_addr, 404 &ipv6_hdr(skb)->saddr); 405 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL) 406 sin6->sin6_scope_id = IP6CB(skb)->iif; 407 } 408 409 } 410 if (is_udp4) { 411 if (inet->cmsg_flags) 412 ip_cmsg_recv(msg, skb); 413 } else { 414 if (np->rxopt.all) 415 datagram_recv_ctl(sk, msg, skb); 416 } 417 418 err = len; 419 if (flags & MSG_TRUNC) 420 err = ulen; 421 422 out_free: 423 skb_free_datagram_locked(sk, skb); 424 out: 425 return err; 426 427 csum_copy_err: 428 slow = lock_sock_fast(sk); 429 if (!skb_kill_datagram(sk, skb, flags)) { 430 if (is_udp4) 431 UDP_INC_STATS_USER(sock_net(sk), 432 UDP_MIB_INERRORS, is_udplite); 433 else 434 UDP6_INC_STATS_USER(sock_net(sk), 435 UDP_MIB_INERRORS, is_udplite); 436 } 437 unlock_sock_fast(sk, slow); 438 439 if (flags & MSG_DONTWAIT) 440 return -EAGAIN; 441 goto try_again; 442 } 443 444 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 445 u8 type, u8 code, int offset, __be32 info, 446 struct udp_table *udptable) 447 { 448 struct ipv6_pinfo *np; 449 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data; 450 struct in6_addr *saddr = &hdr->saddr; 451 struct in6_addr *daddr = &hdr->daddr; 452 struct udphdr *uh = (struct udphdr*)(skb->data+offset); 453 struct sock *sk; 454 int err; 455 456 sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest, 457 saddr, uh->source, inet6_iif(skb), udptable); 458 if (sk == NULL) 459 return; 460 461 np = inet6_sk(sk); 462 463 if (!icmpv6_err_convert(type, code, &err) && !np->recverr) 464 goto out; 465 466 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr) 467 goto out; 468 469 if (np->recverr) 470 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1)); 471 472 sk->sk_err = err; 473 sk->sk_error_report(sk); 474 out: 475 sock_put(sk); 476 } 477 478 static __inline__ void udpv6_err(struct sk_buff *skb, 479 struct inet6_skb_parm *opt, u8 type, 480 u8 code, int offset, __be32 info ) 481 { 482 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table); 483 } 484 485 int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb) 486 { 487 struct udp_sock *up = udp_sk(sk); 488 int rc; 489 int is_udplite = IS_UDPLITE(sk); 490 491 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) 492 goto drop; 493 494 /* 495 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c). 496 */ 497 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { 498 499 if (up->pcrlen == 0) { /* full coverage was set */ 500 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage" 501 " %d while full coverage %d requested\n", 502 UDP_SKB_CB(skb)->cscov, skb->len); 503 goto drop; 504 } 505 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { 506 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d " 507 "too small, need min %d\n", 508 UDP_SKB_CB(skb)->cscov, up->pcrlen); 509 goto drop; 510 } 511 } 512 513 if (sk->sk_filter) { 514 if (udp_lib_checksum_complete(skb)) 515 goto drop; 516 } 517 518 if ((rc = ip_queue_rcv_skb(sk, skb)) < 0) { 519 /* Note that an ENOMEM error is charged twice */ 520 if (rc == -ENOMEM) 521 UDP6_INC_STATS_BH(sock_net(sk), 522 UDP_MIB_RCVBUFERRORS, is_udplite); 523 goto drop_no_sk_drops_inc; 524 } 525 526 return 0; 527 drop: 528 atomic_inc(&sk->sk_drops); 529 drop_no_sk_drops_inc: 530 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); 531 kfree_skb(skb); 532 return -1; 533 } 534 535 static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk, 536 __be16 loc_port, struct in6_addr *loc_addr, 537 __be16 rmt_port, struct in6_addr *rmt_addr, 538 int dif) 539 { 540 struct hlist_nulls_node *node; 541 struct sock *s = sk; 542 unsigned short num = ntohs(loc_port); 543 544 sk_nulls_for_each_from(s, node) { 545 struct inet_sock *inet = inet_sk(s); 546 547 if (!net_eq(sock_net(s), net)) 548 continue; 549 550 if (udp_sk(s)->udp_port_hash == num && 551 s->sk_family == PF_INET6) { 552 struct ipv6_pinfo *np = inet6_sk(s); 553 if (inet->inet_dport) { 554 if (inet->inet_dport != rmt_port) 555 continue; 556 } 557 if (!ipv6_addr_any(&np->daddr) && 558 !ipv6_addr_equal(&np->daddr, rmt_addr)) 559 continue; 560 561 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif) 562 continue; 563 564 if (!ipv6_addr_any(&np->rcv_saddr)) { 565 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr)) 566 continue; 567 } 568 if (!inet6_mc_check(s, loc_addr, rmt_addr)) 569 continue; 570 return s; 571 } 572 } 573 return NULL; 574 } 575 576 static void flush_stack(struct sock **stack, unsigned int count, 577 struct sk_buff *skb, unsigned int final) 578 { 579 unsigned int i; 580 struct sock *sk; 581 struct sk_buff *skb1; 582 583 for (i = 0; i < count; i++) { 584 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC); 585 586 sk = stack[i]; 587 if (skb1) { 588 if (sk_rcvqueues_full(sk, skb)) { 589 kfree_skb(skb1); 590 goto drop; 591 } 592 bh_lock_sock(sk); 593 if (!sock_owned_by_user(sk)) 594 udpv6_queue_rcv_skb(sk, skb1); 595 else if (sk_add_backlog(sk, skb1)) { 596 kfree_skb(skb1); 597 bh_unlock_sock(sk); 598 goto drop; 599 } 600 bh_unlock_sock(sk); 601 continue; 602 } 603 drop: 604 atomic_inc(&sk->sk_drops); 605 UDP6_INC_STATS_BH(sock_net(sk), 606 UDP_MIB_RCVBUFERRORS, IS_UDPLITE(sk)); 607 UDP6_INC_STATS_BH(sock_net(sk), 608 UDP_MIB_INERRORS, IS_UDPLITE(sk)); 609 } 610 } 611 /* 612 * Note: called only from the BH handler context, 613 * so we don't need to lock the hashes. 614 */ 615 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, 616 struct in6_addr *saddr, struct in6_addr *daddr, 617 struct udp_table *udptable) 618 { 619 struct sock *sk, *stack[256 / sizeof(struct sock *)]; 620 const struct udphdr *uh = udp_hdr(skb); 621 struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest)); 622 int dif; 623 unsigned int i, count = 0; 624 625 spin_lock(&hslot->lock); 626 sk = sk_nulls_head(&hslot->head); 627 dif = inet6_iif(skb); 628 sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif); 629 while (sk) { 630 stack[count++] = sk; 631 sk = udp_v6_mcast_next(net, sk_nulls_next(sk), uh->dest, daddr, 632 uh->source, saddr, dif); 633 if (unlikely(count == ARRAY_SIZE(stack))) { 634 if (!sk) 635 break; 636 flush_stack(stack, count, skb, ~0); 637 count = 0; 638 } 639 } 640 /* 641 * before releasing the lock, we must take reference on sockets 642 */ 643 for (i = 0; i < count; i++) 644 sock_hold(stack[i]); 645 646 spin_unlock(&hslot->lock); 647 648 if (count) { 649 flush_stack(stack, count, skb, count - 1); 650 651 for (i = 0; i < count; i++) 652 sock_put(stack[i]); 653 } else { 654 kfree_skb(skb); 655 } 656 return 0; 657 } 658 659 static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh, 660 int proto) 661 { 662 int err; 663 664 UDP_SKB_CB(skb)->partial_cov = 0; 665 UDP_SKB_CB(skb)->cscov = skb->len; 666 667 if (proto == IPPROTO_UDPLITE) { 668 err = udplite_checksum_init(skb, uh); 669 if (err) 670 return err; 671 } 672 673 if (uh->check == 0) { 674 /* RFC 2460 section 8.1 says that we SHOULD log 675 this error. Well, it is reasonable. 676 */ 677 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n"); 678 return 1; 679 } 680 if (skb->ip_summed == CHECKSUM_COMPLETE && 681 !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, 682 skb->len, proto, skb->csum)) 683 skb->ip_summed = CHECKSUM_UNNECESSARY; 684 685 if (!skb_csum_unnecessary(skb)) 686 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 687 &ipv6_hdr(skb)->daddr, 688 skb->len, proto, 0)); 689 690 return 0; 691 } 692 693 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, 694 int proto) 695 { 696 struct net *net = dev_net(skb->dev); 697 struct sock *sk; 698 struct udphdr *uh; 699 struct in6_addr *saddr, *daddr; 700 u32 ulen = 0; 701 702 if (!pskb_may_pull(skb, sizeof(struct udphdr))) 703 goto discard; 704 705 saddr = &ipv6_hdr(skb)->saddr; 706 daddr = &ipv6_hdr(skb)->daddr; 707 uh = udp_hdr(skb); 708 709 ulen = ntohs(uh->len); 710 if (ulen > skb->len) 711 goto short_packet; 712 713 if (proto == IPPROTO_UDP) { 714 /* UDP validates ulen. */ 715 716 /* Check for jumbo payload */ 717 if (ulen == 0) 718 ulen = skb->len; 719 720 if (ulen < sizeof(*uh)) 721 goto short_packet; 722 723 if (ulen < skb->len) { 724 if (pskb_trim_rcsum(skb, ulen)) 725 goto short_packet; 726 saddr = &ipv6_hdr(skb)->saddr; 727 daddr = &ipv6_hdr(skb)->daddr; 728 uh = udp_hdr(skb); 729 } 730 } 731 732 if (udp6_csum_init(skb, uh, proto)) 733 goto discard; 734 735 /* 736 * Multicast receive code 737 */ 738 if (ipv6_addr_is_multicast(daddr)) 739 return __udp6_lib_mcast_deliver(net, skb, 740 saddr, daddr, udptable); 741 742 /* Unicast */ 743 744 /* 745 * check socket cache ... must talk to Alan about his plans 746 * for sock caches... i'll skip this for now. 747 */ 748 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable); 749 750 if (sk == NULL) { 751 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 752 goto discard; 753 754 if (udp_lib_checksum_complete(skb)) 755 goto discard; 756 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, 757 proto == IPPROTO_UDPLITE); 758 759 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 760 761 kfree_skb(skb); 762 return 0; 763 } 764 765 /* deliver */ 766 767 if (sk_rcvqueues_full(sk, skb)) { 768 sock_put(sk); 769 goto discard; 770 } 771 bh_lock_sock(sk); 772 if (!sock_owned_by_user(sk)) 773 udpv6_queue_rcv_skb(sk, skb); 774 else if (sk_add_backlog(sk, skb)) { 775 atomic_inc(&sk->sk_drops); 776 bh_unlock_sock(sk); 777 sock_put(sk); 778 goto discard; 779 } 780 bh_unlock_sock(sk); 781 sock_put(sk); 782 return 0; 783 784 short_packet: 785 LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n", 786 proto == IPPROTO_UDPLITE ? "-Lite" : "", 787 saddr, 788 ntohs(uh->source), 789 ulen, 790 skb->len, 791 daddr, 792 ntohs(uh->dest)); 793 794 discard: 795 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); 796 kfree_skb(skb); 797 return 0; 798 } 799 800 static __inline__ int udpv6_rcv(struct sk_buff *skb) 801 { 802 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP); 803 } 804 805 /* 806 * Throw away all pending data and cancel the corking. Socket is locked. 807 */ 808 static void udp_v6_flush_pending_frames(struct sock *sk) 809 { 810 struct udp_sock *up = udp_sk(sk); 811 812 if (up->pending == AF_INET) 813 udp_flush_pending_frames(sk); 814 else if (up->pending) { 815 up->len = 0; 816 up->pending = 0; 817 ip6_flush_pending_frames(sk); 818 } 819 } 820 821 /** 822 * udp6_hwcsum_outgoing - handle outgoing HW checksumming 823 * @sk: socket we are sending on 824 * @skb: sk_buff containing the filled-in UDP header 825 * (checksum field must be zeroed out) 826 */ 827 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, 828 const struct in6_addr *saddr, 829 const struct in6_addr *daddr, int len) 830 { 831 unsigned int offset; 832 struct udphdr *uh = udp_hdr(skb); 833 __wsum csum = 0; 834 835 if (skb_queue_len(&sk->sk_write_queue) == 1) { 836 /* Only one fragment on the socket. */ 837 skb->csum_start = skb_transport_header(skb) - skb->head; 838 skb->csum_offset = offsetof(struct udphdr, check); 839 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0); 840 } else { 841 /* 842 * HW-checksum won't work as there are two or more 843 * fragments on the socket so that all csums of sk_buffs 844 * should be together 845 */ 846 offset = skb_transport_offset(skb); 847 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); 848 849 skb->ip_summed = CHECKSUM_NONE; 850 851 skb_queue_walk(&sk->sk_write_queue, skb) { 852 csum = csum_add(csum, skb->csum); 853 } 854 855 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 856 csum); 857 if (uh->check == 0) 858 uh->check = CSUM_MANGLED_0; 859 } 860 } 861 862 /* 863 * Sending 864 */ 865 866 static int udp_v6_push_pending_frames(struct sock *sk) 867 { 868 struct sk_buff *skb; 869 struct udphdr *uh; 870 struct udp_sock *up = udp_sk(sk); 871 struct inet_sock *inet = inet_sk(sk); 872 struct flowi *fl = &inet->cork.fl; 873 int err = 0; 874 int is_udplite = IS_UDPLITE(sk); 875 __wsum csum = 0; 876 877 /* Grab the skbuff where UDP header space exists. */ 878 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 879 goto out; 880 881 /* 882 * Create a UDP header 883 */ 884 uh = udp_hdr(skb); 885 uh->source = fl->fl_ip_sport; 886 uh->dest = fl->fl_ip_dport; 887 uh->len = htons(up->len); 888 uh->check = 0; 889 890 if (is_udplite) 891 csum = udplite_csum_outgoing(sk, skb); 892 else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ 893 udp6_hwcsum_outgoing(sk, skb, &fl->fl6_src, &fl->fl6_dst, 894 up->len); 895 goto send; 896 } else 897 csum = udp_csum_outgoing(sk, skb); 898 899 /* add protocol-dependent pseudo-header */ 900 uh->check = csum_ipv6_magic(&fl->fl6_src, &fl->fl6_dst, 901 up->len, fl->proto, csum ); 902 if (uh->check == 0) 903 uh->check = CSUM_MANGLED_0; 904 905 send: 906 err = ip6_push_pending_frames(sk); 907 if (err) { 908 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) { 909 UDP6_INC_STATS_USER(sock_net(sk), 910 UDP_MIB_SNDBUFERRORS, is_udplite); 911 err = 0; 912 } 913 } else 914 UDP6_INC_STATS_USER(sock_net(sk), 915 UDP_MIB_OUTDATAGRAMS, is_udplite); 916 out: 917 up->len = 0; 918 up->pending = 0; 919 return err; 920 } 921 922 int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk, 923 struct msghdr *msg, size_t len) 924 { 925 struct ipv6_txoptions opt_space; 926 struct udp_sock *up = udp_sk(sk); 927 struct inet_sock *inet = inet_sk(sk); 928 struct ipv6_pinfo *np = inet6_sk(sk); 929 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name; 930 struct in6_addr *daddr, *final_p = NULL, final; 931 struct ipv6_txoptions *opt = NULL; 932 struct ip6_flowlabel *flowlabel = NULL; 933 struct flowi fl; 934 struct dst_entry *dst; 935 int addr_len = msg->msg_namelen; 936 int ulen = len; 937 int hlimit = -1; 938 int tclass = -1; 939 int dontfrag = -1; 940 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; 941 int err; 942 int connected = 0; 943 int is_udplite = IS_UDPLITE(sk); 944 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); 945 946 /* destination address check */ 947 if (sin6) { 948 if (addr_len < offsetof(struct sockaddr, sa_data)) 949 return -EINVAL; 950 951 switch (sin6->sin6_family) { 952 case AF_INET6: 953 if (addr_len < SIN6_LEN_RFC2133) 954 return -EINVAL; 955 daddr = &sin6->sin6_addr; 956 break; 957 case AF_INET: 958 goto do_udp_sendmsg; 959 case AF_UNSPEC: 960 msg->msg_name = sin6 = NULL; 961 msg->msg_namelen = addr_len = 0; 962 daddr = NULL; 963 break; 964 default: 965 return -EINVAL; 966 } 967 } else if (!up->pending) { 968 if (sk->sk_state != TCP_ESTABLISHED) 969 return -EDESTADDRREQ; 970 daddr = &np->daddr; 971 } else 972 daddr = NULL; 973 974 if (daddr) { 975 if (ipv6_addr_v4mapped(daddr)) { 976 struct sockaddr_in sin; 977 sin.sin_family = AF_INET; 978 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport; 979 sin.sin_addr.s_addr = daddr->s6_addr32[3]; 980 msg->msg_name = &sin; 981 msg->msg_namelen = sizeof(sin); 982 do_udp_sendmsg: 983 if (__ipv6_only_sock(sk)) 984 return -ENETUNREACH; 985 return udp_sendmsg(iocb, sk, msg, len); 986 } 987 } 988 989 if (up->pending == AF_INET) 990 return udp_sendmsg(iocb, sk, msg, len); 991 992 /* Rough check on arithmetic overflow, 993 better check is made in ip6_append_data(). 994 */ 995 if (len > INT_MAX - sizeof(struct udphdr)) 996 return -EMSGSIZE; 997 998 if (up->pending) { 999 /* 1000 * There are pending frames. 1001 * The socket lock must be held while it's corked. 1002 */ 1003 lock_sock(sk); 1004 if (likely(up->pending)) { 1005 if (unlikely(up->pending != AF_INET6)) { 1006 release_sock(sk); 1007 return -EAFNOSUPPORT; 1008 } 1009 dst = NULL; 1010 goto do_append_data; 1011 } 1012 release_sock(sk); 1013 } 1014 ulen += sizeof(struct udphdr); 1015 1016 memset(&fl, 0, sizeof(fl)); 1017 1018 if (sin6) { 1019 if (sin6->sin6_port == 0) 1020 return -EINVAL; 1021 1022 fl.fl_ip_dport = sin6->sin6_port; 1023 daddr = &sin6->sin6_addr; 1024 1025 if (np->sndflow) { 1026 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 1027 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) { 1028 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 1029 if (flowlabel == NULL) 1030 return -EINVAL; 1031 daddr = &flowlabel->dst; 1032 } 1033 } 1034 1035 /* 1036 * Otherwise it will be difficult to maintain 1037 * sk->sk_dst_cache. 1038 */ 1039 if (sk->sk_state == TCP_ESTABLISHED && 1040 ipv6_addr_equal(daddr, &np->daddr)) 1041 daddr = &np->daddr; 1042 1043 if (addr_len >= sizeof(struct sockaddr_in6) && 1044 sin6->sin6_scope_id && 1045 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL) 1046 fl.oif = sin6->sin6_scope_id; 1047 } else { 1048 if (sk->sk_state != TCP_ESTABLISHED) 1049 return -EDESTADDRREQ; 1050 1051 fl.fl_ip_dport = inet->inet_dport; 1052 daddr = &np->daddr; 1053 fl.fl6_flowlabel = np->flow_label; 1054 connected = 1; 1055 } 1056 1057 if (!fl.oif) 1058 fl.oif = sk->sk_bound_dev_if; 1059 1060 if (!fl.oif) 1061 fl.oif = np->sticky_pktinfo.ipi6_ifindex; 1062 1063 fl.mark = sk->sk_mark; 1064 1065 if (msg->msg_controllen) { 1066 opt = &opt_space; 1067 memset(opt, 0, sizeof(struct ipv6_txoptions)); 1068 opt->tot_len = sizeof(*opt); 1069 1070 err = datagram_send_ctl(sock_net(sk), msg, &fl, opt, &hlimit, 1071 &tclass, &dontfrag); 1072 if (err < 0) { 1073 fl6_sock_release(flowlabel); 1074 return err; 1075 } 1076 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 1077 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 1078 if (flowlabel == NULL) 1079 return -EINVAL; 1080 } 1081 if (!(opt->opt_nflen|opt->opt_flen)) 1082 opt = NULL; 1083 connected = 0; 1084 } 1085 if (opt == NULL) 1086 opt = np->opt; 1087 if (flowlabel) 1088 opt = fl6_merge_options(&opt_space, flowlabel, opt); 1089 opt = ipv6_fixup_options(&opt_space, opt); 1090 1091 fl.proto = sk->sk_protocol; 1092 if (!ipv6_addr_any(daddr)) 1093 ipv6_addr_copy(&fl.fl6_dst, daddr); 1094 else 1095 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 1096 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr)) 1097 ipv6_addr_copy(&fl.fl6_src, &np->saddr); 1098 fl.fl_ip_sport = inet->inet_sport; 1099 1100 /* merge ip6_build_xmit from ip6_output */ 1101 if (opt && opt->srcrt) { 1102 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt; 1103 ipv6_addr_copy(&final, &fl.fl6_dst); 1104 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 1105 final_p = &final; 1106 connected = 0; 1107 } 1108 1109 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) { 1110 fl.oif = np->mcast_oif; 1111 connected = 0; 1112 } 1113 1114 security_sk_classify_flow(sk, &fl); 1115 1116 err = ip6_sk_dst_lookup(sk, &dst, &fl); 1117 if (err) 1118 goto out; 1119 if (final_p) 1120 ipv6_addr_copy(&fl.fl6_dst, final_p); 1121 1122 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT); 1123 if (err < 0) { 1124 if (err == -EREMOTE) 1125 err = ip6_dst_blackhole(sk, &dst, &fl); 1126 if (err < 0) 1127 goto out; 1128 } 1129 1130 if (hlimit < 0) { 1131 if (ipv6_addr_is_multicast(&fl.fl6_dst)) 1132 hlimit = np->mcast_hops; 1133 else 1134 hlimit = np->hop_limit; 1135 if (hlimit < 0) 1136 hlimit = ip6_dst_hoplimit(dst); 1137 } 1138 1139 if (tclass < 0) 1140 tclass = np->tclass; 1141 1142 if (dontfrag < 0) 1143 dontfrag = np->dontfrag; 1144 1145 if (msg->msg_flags&MSG_CONFIRM) 1146 goto do_confirm; 1147 back_from_confirm: 1148 1149 lock_sock(sk); 1150 if (unlikely(up->pending)) { 1151 /* The socket is already corked while preparing it. */ 1152 /* ... which is an evident application bug. --ANK */ 1153 release_sock(sk); 1154 1155 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n"); 1156 err = -EINVAL; 1157 goto out; 1158 } 1159 1160 up->pending = AF_INET6; 1161 1162 do_append_data: 1163 up->len += ulen; 1164 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; 1165 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen, 1166 sizeof(struct udphdr), hlimit, tclass, opt, &fl, 1167 (struct rt6_info*)dst, 1168 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag); 1169 if (err) 1170 udp_v6_flush_pending_frames(sk); 1171 else if (!corkreq) 1172 err = udp_v6_push_pending_frames(sk); 1173 else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) 1174 up->pending = 0; 1175 1176 if (dst) { 1177 if (connected) { 1178 ip6_dst_store(sk, dst, 1179 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ? 1180 &np->daddr : NULL, 1181 #ifdef CONFIG_IPV6_SUBTREES 1182 ipv6_addr_equal(&fl.fl6_src, &np->saddr) ? 1183 &np->saddr : 1184 #endif 1185 NULL); 1186 } else { 1187 dst_release(dst); 1188 } 1189 dst = NULL; 1190 } 1191 1192 if (err > 0) 1193 err = np->recverr ? net_xmit_errno(err) : 0; 1194 release_sock(sk); 1195 out: 1196 dst_release(dst); 1197 fl6_sock_release(flowlabel); 1198 if (!err) 1199 return len; 1200 /* 1201 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting 1202 * ENOBUFS might not be good (it's not tunable per se), but otherwise 1203 * we don't have a good statistic (IpOutDiscards but it can be too many 1204 * things). We could add another new stat but at least for now that 1205 * seems like overkill. 1206 */ 1207 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { 1208 UDP6_INC_STATS_USER(sock_net(sk), 1209 UDP_MIB_SNDBUFERRORS, is_udplite); 1210 } 1211 return err; 1212 1213 do_confirm: 1214 dst_confirm(dst); 1215 if (!(msg->msg_flags&MSG_PROBE) || len) 1216 goto back_from_confirm; 1217 err = 0; 1218 goto out; 1219 } 1220 1221 void udpv6_destroy_sock(struct sock *sk) 1222 { 1223 lock_sock(sk); 1224 udp_v6_flush_pending_frames(sk); 1225 release_sock(sk); 1226 1227 inet6_destroy_sock(sk); 1228 } 1229 1230 /* 1231 * Socket option code for UDP 1232 */ 1233 int udpv6_setsockopt(struct sock *sk, int level, int optname, 1234 char __user *optval, unsigned int optlen) 1235 { 1236 if (level == SOL_UDP || level == SOL_UDPLITE) 1237 return udp_lib_setsockopt(sk, level, optname, optval, optlen, 1238 udp_v6_push_pending_frames); 1239 return ipv6_setsockopt(sk, level, optname, optval, optlen); 1240 } 1241 1242 #ifdef CONFIG_COMPAT 1243 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname, 1244 char __user *optval, unsigned int optlen) 1245 { 1246 if (level == SOL_UDP || level == SOL_UDPLITE) 1247 return udp_lib_setsockopt(sk, level, optname, optval, optlen, 1248 udp_v6_push_pending_frames); 1249 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen); 1250 } 1251 #endif 1252 1253 int udpv6_getsockopt(struct sock *sk, int level, int optname, 1254 char __user *optval, int __user *optlen) 1255 { 1256 if (level == SOL_UDP || level == SOL_UDPLITE) 1257 return udp_lib_getsockopt(sk, level, optname, optval, optlen); 1258 return ipv6_getsockopt(sk, level, optname, optval, optlen); 1259 } 1260 1261 #ifdef CONFIG_COMPAT 1262 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname, 1263 char __user *optval, int __user *optlen) 1264 { 1265 if (level == SOL_UDP || level == SOL_UDPLITE) 1266 return udp_lib_getsockopt(sk, level, optname, optval, optlen); 1267 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen); 1268 } 1269 #endif 1270 1271 static int udp6_ufo_send_check(struct sk_buff *skb) 1272 { 1273 struct ipv6hdr *ipv6h; 1274 struct udphdr *uh; 1275 1276 if (!pskb_may_pull(skb, sizeof(*uh))) 1277 return -EINVAL; 1278 1279 ipv6h = ipv6_hdr(skb); 1280 uh = udp_hdr(skb); 1281 1282 uh->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len, 1283 IPPROTO_UDP, 0); 1284 skb->csum_start = skb_transport_header(skb) - skb->head; 1285 skb->csum_offset = offsetof(struct udphdr, check); 1286 skb->ip_summed = CHECKSUM_PARTIAL; 1287 return 0; 1288 } 1289 1290 static struct sk_buff *udp6_ufo_fragment(struct sk_buff *skb, int features) 1291 { 1292 struct sk_buff *segs = ERR_PTR(-EINVAL); 1293 unsigned int mss; 1294 unsigned int unfrag_ip6hlen, unfrag_len; 1295 struct frag_hdr *fptr; 1296 u8 *mac_start, *prevhdr; 1297 u8 nexthdr; 1298 u8 frag_hdr_sz = sizeof(struct frag_hdr); 1299 int offset; 1300 __wsum csum; 1301 1302 mss = skb_shinfo(skb)->gso_size; 1303 if (unlikely(skb->len <= mss)) 1304 goto out; 1305 1306 if (skb_gso_ok(skb, features | NETIF_F_GSO_ROBUST)) { 1307 /* Packet is from an untrusted source, reset gso_segs. */ 1308 int type = skb_shinfo(skb)->gso_type; 1309 1310 if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY) || 1311 !(type & (SKB_GSO_UDP)))) 1312 goto out; 1313 1314 skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss); 1315 1316 segs = NULL; 1317 goto out; 1318 } 1319 1320 /* Do software UFO. Complete and fill in the UDP checksum as HW cannot 1321 * do checksum of UDP packets sent as multiple IP fragments. 1322 */ 1323 offset = skb->csum_start - skb_headroom(skb); 1324 csum = skb_checksum(skb, offset, skb->len- offset, 0); 1325 offset += skb->csum_offset; 1326 *(__sum16 *)(skb->data + offset) = csum_fold(csum); 1327 skb->ip_summed = CHECKSUM_NONE; 1328 1329 /* Check if there is enough headroom to insert fragment header. */ 1330 if ((skb_headroom(skb) < frag_hdr_sz) && 1331 pskb_expand_head(skb, frag_hdr_sz, 0, GFP_ATOMIC)) 1332 goto out; 1333 1334 /* Find the unfragmentable header and shift it left by frag_hdr_sz 1335 * bytes to insert fragment header. 1336 */ 1337 unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr); 1338 nexthdr = *prevhdr; 1339 *prevhdr = NEXTHDR_FRAGMENT; 1340 unfrag_len = skb_network_header(skb) - skb_mac_header(skb) + 1341 unfrag_ip6hlen; 1342 mac_start = skb_mac_header(skb); 1343 memmove(mac_start-frag_hdr_sz, mac_start, unfrag_len); 1344 1345 skb->mac_header -= frag_hdr_sz; 1346 skb->network_header -= frag_hdr_sz; 1347 1348 fptr = (struct frag_hdr *)(skb_network_header(skb) + unfrag_ip6hlen); 1349 fptr->nexthdr = nexthdr; 1350 fptr->reserved = 0; 1351 ipv6_select_ident(fptr); 1352 1353 /* Fragment the skb. ipv6 header and the remaining fields of the 1354 * fragment header are updated in ipv6_gso_segment() 1355 */ 1356 segs = skb_segment(skb, features); 1357 1358 out: 1359 return segs; 1360 } 1361 1362 static const struct inet6_protocol udpv6_protocol = { 1363 .handler = udpv6_rcv, 1364 .err_handler = udpv6_err, 1365 .gso_send_check = udp6_ufo_send_check, 1366 .gso_segment = udp6_ufo_fragment, 1367 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1368 }; 1369 1370 /* ------------------------------------------------------------------------ */ 1371 #ifdef CONFIG_PROC_FS 1372 1373 static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket) 1374 { 1375 struct inet_sock *inet = inet_sk(sp); 1376 struct ipv6_pinfo *np = inet6_sk(sp); 1377 struct in6_addr *dest, *src; 1378 __u16 destp, srcp; 1379 1380 dest = &np->daddr; 1381 src = &np->rcv_saddr; 1382 destp = ntohs(inet->inet_dport); 1383 srcp = ntohs(inet->inet_sport); 1384 seq_printf(seq, 1385 "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " 1386 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n", 1387 bucket, 1388 src->s6_addr32[0], src->s6_addr32[1], 1389 src->s6_addr32[2], src->s6_addr32[3], srcp, 1390 dest->s6_addr32[0], dest->s6_addr32[1], 1391 dest->s6_addr32[2], dest->s6_addr32[3], destp, 1392 sp->sk_state, 1393 sk_wmem_alloc_get(sp), 1394 sk_rmem_alloc_get(sp), 1395 0, 0L, 0, 1396 sock_i_uid(sp), 0, 1397 sock_i_ino(sp), 1398 atomic_read(&sp->sk_refcnt), sp, 1399 atomic_read(&sp->sk_drops)); 1400 } 1401 1402 int udp6_seq_show(struct seq_file *seq, void *v) 1403 { 1404 if (v == SEQ_START_TOKEN) 1405 seq_printf(seq, 1406 " sl " 1407 "local_address " 1408 "remote_address " 1409 "st tx_queue rx_queue tr tm->when retrnsmt" 1410 " uid timeout inode ref pointer drops\n"); 1411 else 1412 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket); 1413 return 0; 1414 } 1415 1416 static struct udp_seq_afinfo udp6_seq_afinfo = { 1417 .name = "udp6", 1418 .family = AF_INET6, 1419 .udp_table = &udp_table, 1420 .seq_fops = { 1421 .owner = THIS_MODULE, 1422 }, 1423 .seq_ops = { 1424 .show = udp6_seq_show, 1425 }, 1426 }; 1427 1428 int __net_init udp6_proc_init(struct net *net) 1429 { 1430 return udp_proc_register(net, &udp6_seq_afinfo); 1431 } 1432 1433 void udp6_proc_exit(struct net *net) { 1434 udp_proc_unregister(net, &udp6_seq_afinfo); 1435 } 1436 #endif /* CONFIG_PROC_FS */ 1437 1438 /* ------------------------------------------------------------------------ */ 1439 1440 struct proto udpv6_prot = { 1441 .name = "UDPv6", 1442 .owner = THIS_MODULE, 1443 .close = udp_lib_close, 1444 .connect = ip6_datagram_connect, 1445 .disconnect = udp_disconnect, 1446 .ioctl = udp_ioctl, 1447 .destroy = udpv6_destroy_sock, 1448 .setsockopt = udpv6_setsockopt, 1449 .getsockopt = udpv6_getsockopt, 1450 .sendmsg = udpv6_sendmsg, 1451 .recvmsg = udpv6_recvmsg, 1452 .backlog_rcv = udpv6_queue_rcv_skb, 1453 .hash = udp_lib_hash, 1454 .unhash = udp_lib_unhash, 1455 .get_port = udp_v6_get_port, 1456 .memory_allocated = &udp_memory_allocated, 1457 .sysctl_mem = sysctl_udp_mem, 1458 .sysctl_wmem = &sysctl_udp_wmem_min, 1459 .sysctl_rmem = &sysctl_udp_rmem_min, 1460 .obj_size = sizeof(struct udp6_sock), 1461 .slab_flags = SLAB_DESTROY_BY_RCU, 1462 .h.udp_table = &udp_table, 1463 #ifdef CONFIG_COMPAT 1464 .compat_setsockopt = compat_udpv6_setsockopt, 1465 .compat_getsockopt = compat_udpv6_getsockopt, 1466 #endif 1467 }; 1468 1469 static struct inet_protosw udpv6_protosw = { 1470 .type = SOCK_DGRAM, 1471 .protocol = IPPROTO_UDP, 1472 .prot = &udpv6_prot, 1473 .ops = &inet6_dgram_ops, 1474 .no_check = UDP_CSUM_DEFAULT, 1475 .flags = INET_PROTOSW_PERMANENT, 1476 }; 1477 1478 1479 int __init udpv6_init(void) 1480 { 1481 int ret; 1482 1483 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP); 1484 if (ret) 1485 goto out; 1486 1487 ret = inet6_register_protosw(&udpv6_protosw); 1488 if (ret) 1489 goto out_udpv6_protocol; 1490 out: 1491 return ret; 1492 1493 out_udpv6_protocol: 1494 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); 1495 goto out; 1496 } 1497 1498 void udpv6_exit(void) 1499 { 1500 inet6_unregister_protosw(&udpv6_protosw); 1501 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); 1502 } 1503