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 = sk_rcv_saddr(sk); 58 __be32 sk2_rcv_saddr = sk_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 void udp_v6_rehash(struct sock *sk) 115 { 116 u16 new_hash = udp6_portaddr_hash(sock_net(sk), 117 &inet6_sk(sk)->rcv_saddr, 118 inet_sk(sk)->inet_num); 119 120 udp_lib_rehash(sk, new_hash); 121 } 122 123 static inline int compute_score(struct sock *sk, struct net *net, 124 unsigned short hnum, 125 const struct in6_addr *saddr, __be16 sport, 126 const struct in6_addr *daddr, __be16 dport, 127 int dif) 128 { 129 int score = -1; 130 131 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum && 132 sk->sk_family == PF_INET6) { 133 struct ipv6_pinfo *np = inet6_sk(sk); 134 struct inet_sock *inet = inet_sk(sk); 135 136 score = 0; 137 if (inet->inet_dport) { 138 if (inet->inet_dport != sport) 139 return -1; 140 score++; 141 } 142 if (!ipv6_addr_any(&np->rcv_saddr)) { 143 if (!ipv6_addr_equal(&np->rcv_saddr, daddr)) 144 return -1; 145 score++; 146 } 147 if (!ipv6_addr_any(&np->daddr)) { 148 if (!ipv6_addr_equal(&np->daddr, saddr)) 149 return -1; 150 score++; 151 } 152 if (sk->sk_bound_dev_if) { 153 if (sk->sk_bound_dev_if != dif) 154 return -1; 155 score++; 156 } 157 } 158 return score; 159 } 160 161 #define SCORE2_MAX (1 + 1 + 1) 162 static inline int compute_score2(struct sock *sk, struct net *net, 163 const struct in6_addr *saddr, __be16 sport, 164 const struct in6_addr *daddr, unsigned short hnum, 165 int dif) 166 { 167 int score = -1; 168 169 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum && 170 sk->sk_family == PF_INET6) { 171 struct ipv6_pinfo *np = inet6_sk(sk); 172 struct inet_sock *inet = inet_sk(sk); 173 174 if (!ipv6_addr_equal(&np->rcv_saddr, daddr)) 175 return -1; 176 score = 0; 177 if (inet->inet_dport) { 178 if (inet->inet_dport != sport) 179 return -1; 180 score++; 181 } 182 if (!ipv6_addr_any(&np->daddr)) { 183 if (!ipv6_addr_equal(&np->daddr, saddr)) 184 return -1; 185 score++; 186 } 187 if (sk->sk_bound_dev_if) { 188 if (sk->sk_bound_dev_if != dif) 189 return -1; 190 score++; 191 } 192 } 193 return score; 194 } 195 196 197 /* called with read_rcu_lock() */ 198 static struct sock *udp6_lib_lookup2(struct net *net, 199 const struct in6_addr *saddr, __be16 sport, 200 const struct in6_addr *daddr, unsigned int hnum, int dif, 201 struct udp_hslot *hslot2, unsigned int slot2) 202 { 203 struct sock *sk, *result; 204 struct hlist_nulls_node *node; 205 int score, badness; 206 207 begin: 208 result = NULL; 209 badness = -1; 210 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { 211 score = compute_score2(sk, net, saddr, sport, 212 daddr, hnum, dif); 213 if (score > badness) { 214 result = sk; 215 badness = score; 216 if (score == SCORE2_MAX) 217 goto exact_match; 218 } 219 } 220 /* 221 * if the nulls value we got at the end of this lookup is 222 * not the expected one, we must restart lookup. 223 * We probably met an item that was moved to another chain. 224 */ 225 if (get_nulls_value(node) != slot2) 226 goto begin; 227 228 if (result) { 229 exact_match: 230 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) 231 result = NULL; 232 else if (unlikely(compute_score2(result, net, saddr, sport, 233 daddr, hnum, dif) < badness)) { 234 sock_put(result); 235 goto begin; 236 } 237 } 238 return result; 239 } 240 241 static struct sock *__udp6_lib_lookup(struct net *net, 242 const struct in6_addr *saddr, __be16 sport, 243 const struct in6_addr *daddr, __be16 dport, 244 int dif, struct udp_table *udptable) 245 { 246 struct sock *sk, *result; 247 struct hlist_nulls_node *node; 248 unsigned short hnum = ntohs(dport); 249 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); 250 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; 251 int score, badness; 252 253 rcu_read_lock(); 254 if (hslot->count > 10) { 255 hash2 = udp6_portaddr_hash(net, daddr, hnum); 256 slot2 = hash2 & udptable->mask; 257 hslot2 = &udptable->hash2[slot2]; 258 if (hslot->count < hslot2->count) 259 goto begin; 260 261 result = udp6_lib_lookup2(net, saddr, sport, 262 daddr, hnum, dif, 263 hslot2, slot2); 264 if (!result) { 265 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum); 266 slot2 = hash2 & udptable->mask; 267 hslot2 = &udptable->hash2[slot2]; 268 if (hslot->count < hslot2->count) 269 goto begin; 270 271 result = udp6_lib_lookup2(net, saddr, sport, 272 &in6addr_any, hnum, dif, 273 hslot2, slot2); 274 } 275 rcu_read_unlock(); 276 return result; 277 } 278 begin: 279 result = NULL; 280 badness = -1; 281 sk_nulls_for_each_rcu(sk, node, &hslot->head) { 282 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif); 283 if (score > badness) { 284 result = sk; 285 badness = score; 286 } 287 } 288 /* 289 * if the nulls value we got at the end of this lookup is 290 * not the expected one, we must restart lookup. 291 * We probably met an item that was moved to another chain. 292 */ 293 if (get_nulls_value(node) != slot) 294 goto begin; 295 296 if (result) { 297 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) 298 result = NULL; 299 else if (unlikely(compute_score(result, net, hnum, saddr, sport, 300 daddr, dport, dif) < badness)) { 301 sock_put(result); 302 goto begin; 303 } 304 } 305 rcu_read_unlock(); 306 return result; 307 } 308 309 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb, 310 __be16 sport, __be16 dport, 311 struct udp_table *udptable) 312 { 313 struct sock *sk; 314 const struct ipv6hdr *iph = ipv6_hdr(skb); 315 316 if (unlikely(sk = skb_steal_sock(skb))) 317 return sk; 318 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport, 319 &iph->daddr, dport, inet6_iif(skb), 320 udptable); 321 } 322 323 struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport, 324 const struct in6_addr *daddr, __be16 dport, int dif) 325 { 326 return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table); 327 } 328 EXPORT_SYMBOL_GPL(udp6_lib_lookup); 329 330 331 /* 332 * This should be easy, if there is something there we 333 * return it, otherwise we block. 334 */ 335 336 int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk, 337 struct msghdr *msg, size_t len, 338 int noblock, int flags, int *addr_len) 339 { 340 struct ipv6_pinfo *np = inet6_sk(sk); 341 struct inet_sock *inet = inet_sk(sk); 342 struct sk_buff *skb; 343 unsigned int ulen, copied; 344 int peeked; 345 int err; 346 int is_udplite = IS_UDPLITE(sk); 347 int is_udp4; 348 bool slow; 349 350 if (addr_len) 351 *addr_len=sizeof(struct sockaddr_in6); 352 353 if (flags & MSG_ERRQUEUE) 354 return ipv6_recv_error(sk, msg, len); 355 356 if (np->rxpmtu && np->rxopt.bits.rxpmtu) 357 return ipv6_recv_rxpmtu(sk, msg, len); 358 359 try_again: 360 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), 361 &peeked, &err); 362 if (!skb) 363 goto out; 364 365 ulen = skb->len - sizeof(struct udphdr); 366 copied = len; 367 if (copied > ulen) 368 copied = ulen; 369 else if (copied < ulen) 370 msg->msg_flags |= MSG_TRUNC; 371 372 is_udp4 = (skb->protocol == htons(ETH_P_IP)); 373 374 /* 375 * If checksum is needed at all, try to do it while copying the 376 * data. If the data is truncated, or if we only want a partial 377 * coverage checksum (UDP-Lite), do it before the copy. 378 */ 379 380 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { 381 if (udp_lib_checksum_complete(skb)) 382 goto csum_copy_err; 383 } 384 385 if (skb_csum_unnecessary(skb)) 386 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), 387 msg->msg_iov, copied ); 388 else { 389 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov); 390 if (err == -EINVAL) 391 goto csum_copy_err; 392 } 393 if (err) 394 goto out_free; 395 396 if (!peeked) { 397 if (is_udp4) 398 UDP_INC_STATS_USER(sock_net(sk), 399 UDP_MIB_INDATAGRAMS, is_udplite); 400 else 401 UDP6_INC_STATS_USER(sock_net(sk), 402 UDP_MIB_INDATAGRAMS, is_udplite); 403 } 404 405 sock_recv_ts_and_drops(msg, sk, skb); 406 407 /* Copy the address. */ 408 if (msg->msg_name) { 409 struct sockaddr_in6 *sin6; 410 411 sin6 = (struct sockaddr_in6 *) msg->msg_name; 412 sin6->sin6_family = AF_INET6; 413 sin6->sin6_port = udp_hdr(skb)->source; 414 sin6->sin6_flowinfo = 0; 415 sin6->sin6_scope_id = 0; 416 417 if (is_udp4) 418 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, 419 &sin6->sin6_addr); 420 else { 421 ipv6_addr_copy(&sin6->sin6_addr, 422 &ipv6_hdr(skb)->saddr); 423 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL) 424 sin6->sin6_scope_id = IP6CB(skb)->iif; 425 } 426 427 } 428 if (is_udp4) { 429 if (inet->cmsg_flags) 430 ip_cmsg_recv(msg, skb); 431 } else { 432 if (np->rxopt.all) 433 datagram_recv_ctl(sk, msg, skb); 434 } 435 436 err = copied; 437 if (flags & MSG_TRUNC) 438 err = ulen; 439 440 out_free: 441 skb_free_datagram_locked(sk, skb); 442 out: 443 return err; 444 445 csum_copy_err: 446 slow = lock_sock_fast(sk); 447 if (!skb_kill_datagram(sk, skb, flags)) { 448 if (is_udp4) 449 UDP_INC_STATS_USER(sock_net(sk), 450 UDP_MIB_INERRORS, is_udplite); 451 else 452 UDP6_INC_STATS_USER(sock_net(sk), 453 UDP_MIB_INERRORS, is_udplite); 454 } 455 unlock_sock_fast(sk, slow); 456 457 if (noblock) 458 return -EAGAIN; 459 460 /* starting over for a new packet */ 461 msg->msg_flags &= ~MSG_TRUNC; 462 goto try_again; 463 } 464 465 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 466 u8 type, u8 code, int offset, __be32 info, 467 struct udp_table *udptable) 468 { 469 struct ipv6_pinfo *np; 470 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data; 471 const struct in6_addr *saddr = &hdr->saddr; 472 const struct in6_addr *daddr = &hdr->daddr; 473 struct udphdr *uh = (struct udphdr*)(skb->data+offset); 474 struct sock *sk; 475 int err; 476 477 sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest, 478 saddr, uh->source, inet6_iif(skb), udptable); 479 if (sk == NULL) 480 return; 481 482 np = inet6_sk(sk); 483 484 if (!icmpv6_err_convert(type, code, &err) && !np->recverr) 485 goto out; 486 487 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr) 488 goto out; 489 490 if (np->recverr) 491 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1)); 492 493 sk->sk_err = err; 494 sk->sk_error_report(sk); 495 out: 496 sock_put(sk); 497 } 498 499 static __inline__ void udpv6_err(struct sk_buff *skb, 500 struct inet6_skb_parm *opt, u8 type, 501 u8 code, int offset, __be32 info ) 502 { 503 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table); 504 } 505 506 int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb) 507 { 508 struct udp_sock *up = udp_sk(sk); 509 int rc; 510 int is_udplite = IS_UDPLITE(sk); 511 512 if (!ipv6_addr_any(&inet6_sk(sk)->daddr)) 513 sock_rps_save_rxhash(sk, skb); 514 515 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) 516 goto drop; 517 518 /* 519 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c). 520 */ 521 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { 522 523 if (up->pcrlen == 0) { /* full coverage was set */ 524 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage" 525 " %d while full coverage %d requested\n", 526 UDP_SKB_CB(skb)->cscov, skb->len); 527 goto drop; 528 } 529 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { 530 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d " 531 "too small, need min %d\n", 532 UDP_SKB_CB(skb)->cscov, up->pcrlen); 533 goto drop; 534 } 535 } 536 537 if (rcu_access_pointer(sk->sk_filter)) { 538 if (udp_lib_checksum_complete(skb)) 539 goto drop; 540 } 541 542 if ((rc = ip_queue_rcv_skb(sk, skb)) < 0) { 543 /* Note that an ENOMEM error is charged twice */ 544 if (rc == -ENOMEM) 545 UDP6_INC_STATS_BH(sock_net(sk), 546 UDP_MIB_RCVBUFERRORS, is_udplite); 547 goto drop_no_sk_drops_inc; 548 } 549 550 return 0; 551 drop: 552 atomic_inc(&sk->sk_drops); 553 drop_no_sk_drops_inc: 554 UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); 555 kfree_skb(skb); 556 return -1; 557 } 558 559 static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk, 560 __be16 loc_port, const struct in6_addr *loc_addr, 561 __be16 rmt_port, const struct in6_addr *rmt_addr, 562 int dif) 563 { 564 struct hlist_nulls_node *node; 565 struct sock *s = sk; 566 unsigned short num = ntohs(loc_port); 567 568 sk_nulls_for_each_from(s, node) { 569 struct inet_sock *inet = inet_sk(s); 570 571 if (!net_eq(sock_net(s), net)) 572 continue; 573 574 if (udp_sk(s)->udp_port_hash == num && 575 s->sk_family == PF_INET6) { 576 struct ipv6_pinfo *np = inet6_sk(s); 577 if (inet->inet_dport) { 578 if (inet->inet_dport != rmt_port) 579 continue; 580 } 581 if (!ipv6_addr_any(&np->daddr) && 582 !ipv6_addr_equal(&np->daddr, rmt_addr)) 583 continue; 584 585 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif) 586 continue; 587 588 if (!ipv6_addr_any(&np->rcv_saddr)) { 589 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr)) 590 continue; 591 } 592 if (!inet6_mc_check(s, loc_addr, rmt_addr)) 593 continue; 594 return s; 595 } 596 } 597 return NULL; 598 } 599 600 static void flush_stack(struct sock **stack, unsigned int count, 601 struct sk_buff *skb, unsigned int final) 602 { 603 unsigned int i; 604 struct sock *sk; 605 struct sk_buff *skb1; 606 607 for (i = 0; i < count; i++) { 608 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC); 609 610 sk = stack[i]; 611 if (skb1) { 612 if (sk_rcvqueues_full(sk, skb1)) { 613 kfree_skb(skb1); 614 goto drop; 615 } 616 bh_lock_sock(sk); 617 if (!sock_owned_by_user(sk)) 618 udpv6_queue_rcv_skb(sk, skb1); 619 else if (sk_add_backlog(sk, skb1)) { 620 kfree_skb(skb1); 621 bh_unlock_sock(sk); 622 goto drop; 623 } 624 bh_unlock_sock(sk); 625 continue; 626 } 627 drop: 628 atomic_inc(&sk->sk_drops); 629 UDP6_INC_STATS_BH(sock_net(sk), 630 UDP_MIB_RCVBUFERRORS, IS_UDPLITE(sk)); 631 UDP6_INC_STATS_BH(sock_net(sk), 632 UDP_MIB_INERRORS, IS_UDPLITE(sk)); 633 } 634 } 635 /* 636 * Note: called only from the BH handler context, 637 * so we don't need to lock the hashes. 638 */ 639 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, 640 const struct in6_addr *saddr, const struct in6_addr *daddr, 641 struct udp_table *udptable) 642 { 643 struct sock *sk, *stack[256 / sizeof(struct sock *)]; 644 const struct udphdr *uh = udp_hdr(skb); 645 struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest)); 646 int dif; 647 unsigned int i, count = 0; 648 649 spin_lock(&hslot->lock); 650 sk = sk_nulls_head(&hslot->head); 651 dif = inet6_iif(skb); 652 sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif); 653 while (sk) { 654 stack[count++] = sk; 655 sk = udp_v6_mcast_next(net, sk_nulls_next(sk), uh->dest, daddr, 656 uh->source, saddr, dif); 657 if (unlikely(count == ARRAY_SIZE(stack))) { 658 if (!sk) 659 break; 660 flush_stack(stack, count, skb, ~0); 661 count = 0; 662 } 663 } 664 /* 665 * before releasing the lock, we must take reference on sockets 666 */ 667 for (i = 0; i < count; i++) 668 sock_hold(stack[i]); 669 670 spin_unlock(&hslot->lock); 671 672 if (count) { 673 flush_stack(stack, count, skb, count - 1); 674 675 for (i = 0; i < count; i++) 676 sock_put(stack[i]); 677 } else { 678 kfree_skb(skb); 679 } 680 return 0; 681 } 682 683 static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh, 684 int proto) 685 { 686 int err; 687 688 UDP_SKB_CB(skb)->partial_cov = 0; 689 UDP_SKB_CB(skb)->cscov = skb->len; 690 691 if (proto == IPPROTO_UDPLITE) { 692 err = udplite_checksum_init(skb, uh); 693 if (err) 694 return err; 695 } 696 697 if (uh->check == 0) { 698 /* RFC 2460 section 8.1 says that we SHOULD log 699 this error. Well, it is reasonable. 700 */ 701 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n"); 702 return 1; 703 } 704 if (skb->ip_summed == CHECKSUM_COMPLETE && 705 !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, 706 skb->len, proto, skb->csum)) 707 skb->ip_summed = CHECKSUM_UNNECESSARY; 708 709 if (!skb_csum_unnecessary(skb)) 710 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 711 &ipv6_hdr(skb)->daddr, 712 skb->len, proto, 0)); 713 714 return 0; 715 } 716 717 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, 718 int proto) 719 { 720 struct net *net = dev_net(skb->dev); 721 struct sock *sk; 722 struct udphdr *uh; 723 const struct in6_addr *saddr, *daddr; 724 u32 ulen = 0; 725 726 if (!pskb_may_pull(skb, sizeof(struct udphdr))) 727 goto discard; 728 729 saddr = &ipv6_hdr(skb)->saddr; 730 daddr = &ipv6_hdr(skb)->daddr; 731 uh = udp_hdr(skb); 732 733 ulen = ntohs(uh->len); 734 if (ulen > skb->len) 735 goto short_packet; 736 737 if (proto == IPPROTO_UDP) { 738 /* UDP validates ulen. */ 739 740 /* Check for jumbo payload */ 741 if (ulen == 0) 742 ulen = skb->len; 743 744 if (ulen < sizeof(*uh)) 745 goto short_packet; 746 747 if (ulen < skb->len) { 748 if (pskb_trim_rcsum(skb, ulen)) 749 goto short_packet; 750 saddr = &ipv6_hdr(skb)->saddr; 751 daddr = &ipv6_hdr(skb)->daddr; 752 uh = udp_hdr(skb); 753 } 754 } 755 756 if (udp6_csum_init(skb, uh, proto)) 757 goto discard; 758 759 /* 760 * Multicast receive code 761 */ 762 if (ipv6_addr_is_multicast(daddr)) 763 return __udp6_lib_mcast_deliver(net, skb, 764 saddr, daddr, udptable); 765 766 /* Unicast */ 767 768 /* 769 * check socket cache ... must talk to Alan about his plans 770 * for sock caches... i'll skip this for now. 771 */ 772 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable); 773 774 if (sk == NULL) { 775 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 776 goto discard; 777 778 if (udp_lib_checksum_complete(skb)) 779 goto discard; 780 UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS, 781 proto == IPPROTO_UDPLITE); 782 783 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 784 785 kfree_skb(skb); 786 return 0; 787 } 788 789 /* deliver */ 790 791 if (sk_rcvqueues_full(sk, skb)) { 792 sock_put(sk); 793 goto discard; 794 } 795 bh_lock_sock(sk); 796 if (!sock_owned_by_user(sk)) 797 udpv6_queue_rcv_skb(sk, skb); 798 else if (sk_add_backlog(sk, skb)) { 799 atomic_inc(&sk->sk_drops); 800 bh_unlock_sock(sk); 801 sock_put(sk); 802 goto discard; 803 } 804 bh_unlock_sock(sk); 805 sock_put(sk); 806 return 0; 807 808 short_packet: 809 LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n", 810 proto == IPPROTO_UDPLITE ? "-Lite" : "", 811 saddr, 812 ntohs(uh->source), 813 ulen, 814 skb->len, 815 daddr, 816 ntohs(uh->dest)); 817 818 discard: 819 UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); 820 kfree_skb(skb); 821 return 0; 822 } 823 824 static __inline__ int udpv6_rcv(struct sk_buff *skb) 825 { 826 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP); 827 } 828 829 /* 830 * Throw away all pending data and cancel the corking. Socket is locked. 831 */ 832 static void udp_v6_flush_pending_frames(struct sock *sk) 833 { 834 struct udp_sock *up = udp_sk(sk); 835 836 if (up->pending == AF_INET) 837 udp_flush_pending_frames(sk); 838 else if (up->pending) { 839 up->len = 0; 840 up->pending = 0; 841 ip6_flush_pending_frames(sk); 842 } 843 } 844 845 /** 846 * udp6_hwcsum_outgoing - handle outgoing HW checksumming 847 * @sk: socket we are sending on 848 * @skb: sk_buff containing the filled-in UDP header 849 * (checksum field must be zeroed out) 850 */ 851 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, 852 const struct in6_addr *saddr, 853 const struct in6_addr *daddr, int len) 854 { 855 unsigned int offset; 856 struct udphdr *uh = udp_hdr(skb); 857 __wsum csum = 0; 858 859 if (skb_queue_len(&sk->sk_write_queue) == 1) { 860 /* Only one fragment on the socket. */ 861 skb->csum_start = skb_transport_header(skb) - skb->head; 862 skb->csum_offset = offsetof(struct udphdr, check); 863 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0); 864 } else { 865 /* 866 * HW-checksum won't work as there are two or more 867 * fragments on the socket so that all csums of sk_buffs 868 * should be together 869 */ 870 offset = skb_transport_offset(skb); 871 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); 872 873 skb->ip_summed = CHECKSUM_NONE; 874 875 skb_queue_walk(&sk->sk_write_queue, skb) { 876 csum = csum_add(csum, skb->csum); 877 } 878 879 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 880 csum); 881 if (uh->check == 0) 882 uh->check = CSUM_MANGLED_0; 883 } 884 } 885 886 /* 887 * Sending 888 */ 889 890 static int udp_v6_push_pending_frames(struct sock *sk) 891 { 892 struct sk_buff *skb; 893 struct udphdr *uh; 894 struct udp_sock *up = udp_sk(sk); 895 struct inet_sock *inet = inet_sk(sk); 896 struct flowi6 *fl6 = &inet->cork.fl.u.ip6; 897 int err = 0; 898 int is_udplite = IS_UDPLITE(sk); 899 __wsum csum = 0; 900 901 /* Grab the skbuff where UDP header space exists. */ 902 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 903 goto out; 904 905 /* 906 * Create a UDP header 907 */ 908 uh = udp_hdr(skb); 909 uh->source = fl6->fl6_sport; 910 uh->dest = fl6->fl6_dport; 911 uh->len = htons(up->len); 912 uh->check = 0; 913 914 if (is_udplite) 915 csum = udplite_csum_outgoing(sk, skb); 916 else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ 917 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, 918 up->len); 919 goto send; 920 } else 921 csum = udp_csum_outgoing(sk, skb); 922 923 /* add protocol-dependent pseudo-header */ 924 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, 925 up->len, fl6->flowi6_proto, csum); 926 if (uh->check == 0) 927 uh->check = CSUM_MANGLED_0; 928 929 send: 930 err = ip6_push_pending_frames(sk); 931 if (err) { 932 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) { 933 UDP6_INC_STATS_USER(sock_net(sk), 934 UDP_MIB_SNDBUFERRORS, is_udplite); 935 err = 0; 936 } 937 } else 938 UDP6_INC_STATS_USER(sock_net(sk), 939 UDP_MIB_OUTDATAGRAMS, is_udplite); 940 out: 941 up->len = 0; 942 up->pending = 0; 943 return err; 944 } 945 946 int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk, 947 struct msghdr *msg, size_t len) 948 { 949 struct ipv6_txoptions opt_space; 950 struct udp_sock *up = udp_sk(sk); 951 struct inet_sock *inet = inet_sk(sk); 952 struct ipv6_pinfo *np = inet6_sk(sk); 953 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name; 954 struct in6_addr *daddr, *final_p, final; 955 struct ipv6_txoptions *opt = NULL; 956 struct ip6_flowlabel *flowlabel = NULL; 957 struct flowi6 fl6; 958 struct dst_entry *dst; 959 int addr_len = msg->msg_namelen; 960 int ulen = len; 961 int hlimit = -1; 962 int tclass = -1; 963 int dontfrag = -1; 964 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; 965 int err; 966 int connected = 0; 967 int is_udplite = IS_UDPLITE(sk); 968 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); 969 970 /* destination address check */ 971 if (sin6) { 972 if (addr_len < offsetof(struct sockaddr, sa_data)) 973 return -EINVAL; 974 975 switch (sin6->sin6_family) { 976 case AF_INET6: 977 if (addr_len < SIN6_LEN_RFC2133) 978 return -EINVAL; 979 daddr = &sin6->sin6_addr; 980 break; 981 case AF_INET: 982 goto do_udp_sendmsg; 983 case AF_UNSPEC: 984 msg->msg_name = sin6 = NULL; 985 msg->msg_namelen = addr_len = 0; 986 daddr = NULL; 987 break; 988 default: 989 return -EINVAL; 990 } 991 } else if (!up->pending) { 992 if (sk->sk_state != TCP_ESTABLISHED) 993 return -EDESTADDRREQ; 994 daddr = &np->daddr; 995 } else 996 daddr = NULL; 997 998 if (daddr) { 999 if (ipv6_addr_v4mapped(daddr)) { 1000 struct sockaddr_in sin; 1001 sin.sin_family = AF_INET; 1002 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport; 1003 sin.sin_addr.s_addr = daddr->s6_addr32[3]; 1004 msg->msg_name = &sin; 1005 msg->msg_namelen = sizeof(sin); 1006 do_udp_sendmsg: 1007 if (__ipv6_only_sock(sk)) 1008 return -ENETUNREACH; 1009 return udp_sendmsg(iocb, sk, msg, len); 1010 } 1011 } 1012 1013 if (up->pending == AF_INET) 1014 return udp_sendmsg(iocb, sk, msg, len); 1015 1016 /* Rough check on arithmetic overflow, 1017 better check is made in ip6_append_data(). 1018 */ 1019 if (len > INT_MAX - sizeof(struct udphdr)) 1020 return -EMSGSIZE; 1021 1022 if (up->pending) { 1023 /* 1024 * There are pending frames. 1025 * The socket lock must be held while it's corked. 1026 */ 1027 lock_sock(sk); 1028 if (likely(up->pending)) { 1029 if (unlikely(up->pending != AF_INET6)) { 1030 release_sock(sk); 1031 return -EAFNOSUPPORT; 1032 } 1033 dst = NULL; 1034 goto do_append_data; 1035 } 1036 release_sock(sk); 1037 } 1038 ulen += sizeof(struct udphdr); 1039 1040 memset(&fl6, 0, sizeof(fl6)); 1041 1042 if (sin6) { 1043 if (sin6->sin6_port == 0) 1044 return -EINVAL; 1045 1046 fl6.fl6_dport = sin6->sin6_port; 1047 daddr = &sin6->sin6_addr; 1048 1049 if (np->sndflow) { 1050 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 1051 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { 1052 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 1053 if (flowlabel == NULL) 1054 return -EINVAL; 1055 daddr = &flowlabel->dst; 1056 } 1057 } 1058 1059 /* 1060 * Otherwise it will be difficult to maintain 1061 * sk->sk_dst_cache. 1062 */ 1063 if (sk->sk_state == TCP_ESTABLISHED && 1064 ipv6_addr_equal(daddr, &np->daddr)) 1065 daddr = &np->daddr; 1066 1067 if (addr_len >= sizeof(struct sockaddr_in6) && 1068 sin6->sin6_scope_id && 1069 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL) 1070 fl6.flowi6_oif = sin6->sin6_scope_id; 1071 } else { 1072 if (sk->sk_state != TCP_ESTABLISHED) 1073 return -EDESTADDRREQ; 1074 1075 fl6.fl6_dport = inet->inet_dport; 1076 daddr = &np->daddr; 1077 fl6.flowlabel = np->flow_label; 1078 connected = 1; 1079 } 1080 1081 if (!fl6.flowi6_oif) 1082 fl6.flowi6_oif = sk->sk_bound_dev_if; 1083 1084 if (!fl6.flowi6_oif) 1085 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex; 1086 1087 fl6.flowi6_mark = sk->sk_mark; 1088 1089 if (msg->msg_controllen) { 1090 opt = &opt_space; 1091 memset(opt, 0, sizeof(struct ipv6_txoptions)); 1092 opt->tot_len = sizeof(*opt); 1093 1094 err = datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt, 1095 &hlimit, &tclass, &dontfrag); 1096 if (err < 0) { 1097 fl6_sock_release(flowlabel); 1098 return err; 1099 } 1100 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 1101 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 1102 if (flowlabel == NULL) 1103 return -EINVAL; 1104 } 1105 if (!(opt->opt_nflen|opt->opt_flen)) 1106 opt = NULL; 1107 connected = 0; 1108 } 1109 if (opt == NULL) 1110 opt = np->opt; 1111 if (flowlabel) 1112 opt = fl6_merge_options(&opt_space, flowlabel, opt); 1113 opt = ipv6_fixup_options(&opt_space, opt); 1114 1115 fl6.flowi6_proto = sk->sk_protocol; 1116 if (!ipv6_addr_any(daddr)) 1117 ipv6_addr_copy(&fl6.daddr, daddr); 1118 else 1119 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 1120 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) 1121 ipv6_addr_copy(&fl6.saddr, &np->saddr); 1122 fl6.fl6_sport = inet->inet_sport; 1123 1124 final_p = fl6_update_dst(&fl6, opt, &final); 1125 if (final_p) 1126 connected = 0; 1127 1128 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) { 1129 fl6.flowi6_oif = np->mcast_oif; 1130 connected = 0; 1131 } 1132 1133 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 1134 1135 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, true); 1136 if (IS_ERR(dst)) { 1137 err = PTR_ERR(dst); 1138 dst = NULL; 1139 goto out; 1140 } 1141 1142 if (hlimit < 0) { 1143 if (ipv6_addr_is_multicast(&fl6.daddr)) 1144 hlimit = np->mcast_hops; 1145 else 1146 hlimit = np->hop_limit; 1147 if (hlimit < 0) 1148 hlimit = ip6_dst_hoplimit(dst); 1149 } 1150 1151 if (tclass < 0) 1152 tclass = np->tclass; 1153 1154 if (dontfrag < 0) 1155 dontfrag = np->dontfrag; 1156 1157 if (msg->msg_flags&MSG_CONFIRM) 1158 goto do_confirm; 1159 back_from_confirm: 1160 1161 lock_sock(sk); 1162 if (unlikely(up->pending)) { 1163 /* The socket is already corked while preparing it. */ 1164 /* ... which is an evident application bug. --ANK */ 1165 release_sock(sk); 1166 1167 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n"); 1168 err = -EINVAL; 1169 goto out; 1170 } 1171 1172 up->pending = AF_INET6; 1173 1174 do_append_data: 1175 up->len += ulen; 1176 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; 1177 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen, 1178 sizeof(struct udphdr), hlimit, tclass, opt, &fl6, 1179 (struct rt6_info*)dst, 1180 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag); 1181 if (err) 1182 udp_v6_flush_pending_frames(sk); 1183 else if (!corkreq) 1184 err = udp_v6_push_pending_frames(sk); 1185 else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) 1186 up->pending = 0; 1187 1188 if (dst) { 1189 if (connected) { 1190 ip6_dst_store(sk, dst, 1191 ipv6_addr_equal(&fl6.daddr, &np->daddr) ? 1192 &np->daddr : NULL, 1193 #ifdef CONFIG_IPV6_SUBTREES 1194 ipv6_addr_equal(&fl6.saddr, &np->saddr) ? 1195 &np->saddr : 1196 #endif 1197 NULL); 1198 } else { 1199 dst_release(dst); 1200 } 1201 dst = NULL; 1202 } 1203 1204 if (err > 0) 1205 err = np->recverr ? net_xmit_errno(err) : 0; 1206 release_sock(sk); 1207 out: 1208 dst_release(dst); 1209 fl6_sock_release(flowlabel); 1210 if (!err) 1211 return len; 1212 /* 1213 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting 1214 * ENOBUFS might not be good (it's not tunable per se), but otherwise 1215 * we don't have a good statistic (IpOutDiscards but it can be too many 1216 * things). We could add another new stat but at least for now that 1217 * seems like overkill. 1218 */ 1219 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { 1220 UDP6_INC_STATS_USER(sock_net(sk), 1221 UDP_MIB_SNDBUFERRORS, is_udplite); 1222 } 1223 return err; 1224 1225 do_confirm: 1226 dst_confirm(dst); 1227 if (!(msg->msg_flags&MSG_PROBE) || len) 1228 goto back_from_confirm; 1229 err = 0; 1230 goto out; 1231 } 1232 1233 void udpv6_destroy_sock(struct sock *sk) 1234 { 1235 lock_sock(sk); 1236 udp_v6_flush_pending_frames(sk); 1237 release_sock(sk); 1238 1239 inet6_destroy_sock(sk); 1240 } 1241 1242 /* 1243 * Socket option code for UDP 1244 */ 1245 int udpv6_setsockopt(struct sock *sk, int level, int optname, 1246 char __user *optval, unsigned int optlen) 1247 { 1248 if (level == SOL_UDP || level == SOL_UDPLITE) 1249 return udp_lib_setsockopt(sk, level, optname, optval, optlen, 1250 udp_v6_push_pending_frames); 1251 return ipv6_setsockopt(sk, level, optname, optval, optlen); 1252 } 1253 1254 #ifdef CONFIG_COMPAT 1255 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname, 1256 char __user *optval, unsigned int optlen) 1257 { 1258 if (level == SOL_UDP || level == SOL_UDPLITE) 1259 return udp_lib_setsockopt(sk, level, optname, optval, optlen, 1260 udp_v6_push_pending_frames); 1261 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen); 1262 } 1263 #endif 1264 1265 int udpv6_getsockopt(struct sock *sk, int level, int optname, 1266 char __user *optval, int __user *optlen) 1267 { 1268 if (level == SOL_UDP || level == SOL_UDPLITE) 1269 return udp_lib_getsockopt(sk, level, optname, optval, optlen); 1270 return ipv6_getsockopt(sk, level, optname, optval, optlen); 1271 } 1272 1273 #ifdef CONFIG_COMPAT 1274 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname, 1275 char __user *optval, int __user *optlen) 1276 { 1277 if (level == SOL_UDP || level == SOL_UDPLITE) 1278 return udp_lib_getsockopt(sk, level, optname, optval, optlen); 1279 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen); 1280 } 1281 #endif 1282 1283 static int udp6_ufo_send_check(struct sk_buff *skb) 1284 { 1285 const struct ipv6hdr *ipv6h; 1286 struct udphdr *uh; 1287 1288 if (!pskb_may_pull(skb, sizeof(*uh))) 1289 return -EINVAL; 1290 1291 ipv6h = ipv6_hdr(skb); 1292 uh = udp_hdr(skb); 1293 1294 uh->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len, 1295 IPPROTO_UDP, 0); 1296 skb->csum_start = skb_transport_header(skb) - skb->head; 1297 skb->csum_offset = offsetof(struct udphdr, check); 1298 skb->ip_summed = CHECKSUM_PARTIAL; 1299 return 0; 1300 } 1301 1302 static struct sk_buff *udp6_ufo_fragment(struct sk_buff *skb, u32 features) 1303 { 1304 struct sk_buff *segs = ERR_PTR(-EINVAL); 1305 unsigned int mss; 1306 unsigned int unfrag_ip6hlen, unfrag_len; 1307 struct frag_hdr *fptr; 1308 u8 *mac_start, *prevhdr; 1309 u8 nexthdr; 1310 u8 frag_hdr_sz = sizeof(struct frag_hdr); 1311 int offset; 1312 __wsum csum; 1313 1314 mss = skb_shinfo(skb)->gso_size; 1315 if (unlikely(skb->len <= mss)) 1316 goto out; 1317 1318 if (skb_gso_ok(skb, features | NETIF_F_GSO_ROBUST)) { 1319 /* Packet is from an untrusted source, reset gso_segs. */ 1320 int type = skb_shinfo(skb)->gso_type; 1321 1322 if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY) || 1323 !(type & (SKB_GSO_UDP)))) 1324 goto out; 1325 1326 skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss); 1327 1328 segs = NULL; 1329 goto out; 1330 } 1331 1332 /* Do software UFO. Complete and fill in the UDP checksum as HW cannot 1333 * do checksum of UDP packets sent as multiple IP fragments. 1334 */ 1335 offset = skb_checksum_start_offset(skb); 1336 csum = skb_checksum(skb, offset, skb->len- offset, 0); 1337 offset += skb->csum_offset; 1338 *(__sum16 *)(skb->data + offset) = csum_fold(csum); 1339 skb->ip_summed = CHECKSUM_NONE; 1340 1341 /* Check if there is enough headroom to insert fragment header. */ 1342 if ((skb_mac_header(skb) < skb->head + frag_hdr_sz) && 1343 pskb_expand_head(skb, frag_hdr_sz, 0, GFP_ATOMIC)) 1344 goto out; 1345 1346 /* Find the unfragmentable header and shift it left by frag_hdr_sz 1347 * bytes to insert fragment header. 1348 */ 1349 unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr); 1350 nexthdr = *prevhdr; 1351 *prevhdr = NEXTHDR_FRAGMENT; 1352 unfrag_len = skb_network_header(skb) - skb_mac_header(skb) + 1353 unfrag_ip6hlen; 1354 mac_start = skb_mac_header(skb); 1355 memmove(mac_start-frag_hdr_sz, mac_start, unfrag_len); 1356 1357 skb->mac_header -= frag_hdr_sz; 1358 skb->network_header -= frag_hdr_sz; 1359 1360 fptr = (struct frag_hdr *)(skb_network_header(skb) + unfrag_ip6hlen); 1361 fptr->nexthdr = nexthdr; 1362 fptr->reserved = 0; 1363 ipv6_select_ident(fptr, (struct rt6_info *)skb_dst(skb)); 1364 1365 /* Fragment the skb. ipv6 header and the remaining fields of the 1366 * fragment header are updated in ipv6_gso_segment() 1367 */ 1368 segs = skb_segment(skb, features); 1369 1370 out: 1371 return segs; 1372 } 1373 1374 static const struct inet6_protocol udpv6_protocol = { 1375 .handler = udpv6_rcv, 1376 .err_handler = udpv6_err, 1377 .gso_send_check = udp6_ufo_send_check, 1378 .gso_segment = udp6_ufo_fragment, 1379 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1380 }; 1381 1382 /* ------------------------------------------------------------------------ */ 1383 #ifdef CONFIG_PROC_FS 1384 1385 static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket) 1386 { 1387 struct inet_sock *inet = inet_sk(sp); 1388 struct ipv6_pinfo *np = inet6_sk(sp); 1389 const struct in6_addr *dest, *src; 1390 __u16 destp, srcp; 1391 1392 dest = &np->daddr; 1393 src = &np->rcv_saddr; 1394 destp = ntohs(inet->inet_dport); 1395 srcp = ntohs(inet->inet_sport); 1396 seq_printf(seq, 1397 "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " 1398 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n", 1399 bucket, 1400 src->s6_addr32[0], src->s6_addr32[1], 1401 src->s6_addr32[2], src->s6_addr32[3], srcp, 1402 dest->s6_addr32[0], dest->s6_addr32[1], 1403 dest->s6_addr32[2], dest->s6_addr32[3], destp, 1404 sp->sk_state, 1405 sk_wmem_alloc_get(sp), 1406 sk_rmem_alloc_get(sp), 1407 0, 0L, 0, 1408 sock_i_uid(sp), 0, 1409 sock_i_ino(sp), 1410 atomic_read(&sp->sk_refcnt), sp, 1411 atomic_read(&sp->sk_drops)); 1412 } 1413 1414 int udp6_seq_show(struct seq_file *seq, void *v) 1415 { 1416 if (v == SEQ_START_TOKEN) 1417 seq_printf(seq, 1418 " sl " 1419 "local_address " 1420 "remote_address " 1421 "st tx_queue rx_queue tr tm->when retrnsmt" 1422 " uid timeout inode ref pointer drops\n"); 1423 else 1424 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket); 1425 return 0; 1426 } 1427 1428 static const struct file_operations udp6_afinfo_seq_fops = { 1429 .owner = THIS_MODULE, 1430 .open = udp_seq_open, 1431 .read = seq_read, 1432 .llseek = seq_lseek, 1433 .release = seq_release_net 1434 }; 1435 1436 static struct udp_seq_afinfo udp6_seq_afinfo = { 1437 .name = "udp6", 1438 .family = AF_INET6, 1439 .udp_table = &udp_table, 1440 .seq_fops = &udp6_afinfo_seq_fops, 1441 .seq_ops = { 1442 .show = udp6_seq_show, 1443 }, 1444 }; 1445 1446 int __net_init udp6_proc_init(struct net *net) 1447 { 1448 return udp_proc_register(net, &udp6_seq_afinfo); 1449 } 1450 1451 void udp6_proc_exit(struct net *net) { 1452 udp_proc_unregister(net, &udp6_seq_afinfo); 1453 } 1454 #endif /* CONFIG_PROC_FS */ 1455 1456 /* ------------------------------------------------------------------------ */ 1457 1458 struct proto udpv6_prot = { 1459 .name = "UDPv6", 1460 .owner = THIS_MODULE, 1461 .close = udp_lib_close, 1462 .connect = ip6_datagram_connect, 1463 .disconnect = udp_disconnect, 1464 .ioctl = udp_ioctl, 1465 .destroy = udpv6_destroy_sock, 1466 .setsockopt = udpv6_setsockopt, 1467 .getsockopt = udpv6_getsockopt, 1468 .sendmsg = udpv6_sendmsg, 1469 .recvmsg = udpv6_recvmsg, 1470 .backlog_rcv = udpv6_queue_rcv_skb, 1471 .hash = udp_lib_hash, 1472 .unhash = udp_lib_unhash, 1473 .rehash = udp_v6_rehash, 1474 .get_port = udp_v6_get_port, 1475 .memory_allocated = &udp_memory_allocated, 1476 .sysctl_mem = sysctl_udp_mem, 1477 .sysctl_wmem = &sysctl_udp_wmem_min, 1478 .sysctl_rmem = &sysctl_udp_rmem_min, 1479 .obj_size = sizeof(struct udp6_sock), 1480 .slab_flags = SLAB_DESTROY_BY_RCU, 1481 .h.udp_table = &udp_table, 1482 #ifdef CONFIG_COMPAT 1483 .compat_setsockopt = compat_udpv6_setsockopt, 1484 .compat_getsockopt = compat_udpv6_getsockopt, 1485 #endif 1486 .clear_sk = sk_prot_clear_portaddr_nulls, 1487 }; 1488 1489 static struct inet_protosw udpv6_protosw = { 1490 .type = SOCK_DGRAM, 1491 .protocol = IPPROTO_UDP, 1492 .prot = &udpv6_prot, 1493 .ops = &inet6_dgram_ops, 1494 .no_check = UDP_CSUM_DEFAULT, 1495 .flags = INET_PROTOSW_PERMANENT, 1496 }; 1497 1498 1499 int __init udpv6_init(void) 1500 { 1501 int ret; 1502 1503 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP); 1504 if (ret) 1505 goto out; 1506 1507 ret = inet6_register_protosw(&udpv6_protosw); 1508 if (ret) 1509 goto out_udpv6_protocol; 1510 out: 1511 return ret; 1512 1513 out_udpv6_protocol: 1514 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); 1515 goto out; 1516 } 1517 1518 void udpv6_exit(void) 1519 { 1520 inet6_unregister_protosw(&udpv6_protosw); 1521 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); 1522 } 1523