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/addrconf.h> 41 #include <net/ndisc.h> 42 #include <net/protocol.h> 43 #include <net/transp_v6.h> 44 #include <net/ip6_route.h> 45 #include <net/raw.h> 46 #include <net/tcp_states.h> 47 #include <net/ip6_checksum.h> 48 #include <net/xfrm.h> 49 #include <net/inet6_hashtables.h> 50 #include <net/busy_poll.h> 51 #include <net/sock_reuseport.h> 52 53 #include <linux/proc_fs.h> 54 #include <linux/seq_file.h> 55 #include <trace/events/skb.h> 56 #include "udp_impl.h" 57 58 static u32 udp6_ehashfn(const struct net *net, 59 const struct in6_addr *laddr, 60 const u16 lport, 61 const struct in6_addr *faddr, 62 const __be16 fport) 63 { 64 static u32 udp6_ehash_secret __read_mostly; 65 static u32 udp_ipv6_hash_secret __read_mostly; 66 67 u32 lhash, fhash; 68 69 net_get_random_once(&udp6_ehash_secret, 70 sizeof(udp6_ehash_secret)); 71 net_get_random_once(&udp_ipv6_hash_secret, 72 sizeof(udp_ipv6_hash_secret)); 73 74 lhash = (__force u32)laddr->s6_addr32[3]; 75 fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret); 76 77 return __inet6_ehashfn(lhash, lport, fhash, fport, 78 udp_ipv6_hash_secret + net_hash_mix(net)); 79 } 80 81 static u32 udp6_portaddr_hash(const struct net *net, 82 const struct in6_addr *addr6, 83 unsigned int port) 84 { 85 unsigned int hash, mix = net_hash_mix(net); 86 87 if (ipv6_addr_any(addr6)) 88 hash = jhash_1word(0, mix); 89 else if (ipv6_addr_v4mapped(addr6)) 90 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix); 91 else 92 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix); 93 94 return hash ^ port; 95 } 96 97 int udp_v6_get_port(struct sock *sk, unsigned short snum) 98 { 99 unsigned int hash2_nulladdr = 100 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum); 101 unsigned int hash2_partial = 102 udp6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0); 103 104 /* precompute partial secondary hash */ 105 udp_sk(sk)->udp_portaddr_hash = hash2_partial; 106 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr); 107 } 108 109 static void udp_v6_rehash(struct sock *sk) 110 { 111 u16 new_hash = udp6_portaddr_hash(sock_net(sk), 112 &sk->sk_v6_rcv_saddr, 113 inet_sk(sk)->inet_num); 114 115 udp_lib_rehash(sk, new_hash); 116 } 117 118 static inline int compute_score(struct sock *sk, struct net *net, 119 unsigned short hnum, 120 const struct in6_addr *saddr, __be16 sport, 121 const struct in6_addr *daddr, __be16 dport, 122 int dif) 123 { 124 int score; 125 struct inet_sock *inet; 126 127 if (!net_eq(sock_net(sk), net) || 128 udp_sk(sk)->udp_port_hash != hnum || 129 sk->sk_family != PF_INET6) 130 return -1; 131 132 score = 0; 133 inet = inet_sk(sk); 134 135 if (inet->inet_dport) { 136 if (inet->inet_dport != sport) 137 return -1; 138 score++; 139 } 140 141 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) { 142 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr)) 143 return -1; 144 score++; 145 } 146 147 if (!ipv6_addr_any(&sk->sk_v6_daddr)) { 148 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr)) 149 return -1; 150 score++; 151 } 152 153 if (sk->sk_bound_dev_if) { 154 if (sk->sk_bound_dev_if != dif) 155 return -1; 156 score++; 157 } 158 159 if (sk->sk_incoming_cpu == raw_smp_processor_id()) 160 score++; 161 162 return score; 163 } 164 165 static inline int compute_score2(struct sock *sk, struct net *net, 166 const struct in6_addr *saddr, __be16 sport, 167 const struct in6_addr *daddr, 168 unsigned short hnum, int dif) 169 { 170 int score; 171 struct inet_sock *inet; 172 173 if (!net_eq(sock_net(sk), net) || 174 udp_sk(sk)->udp_port_hash != hnum || 175 sk->sk_family != PF_INET6) 176 return -1; 177 178 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr)) 179 return -1; 180 181 score = 0; 182 inet = inet_sk(sk); 183 184 if (inet->inet_dport) { 185 if (inet->inet_dport != sport) 186 return -1; 187 score++; 188 } 189 190 if (!ipv6_addr_any(&sk->sk_v6_daddr)) { 191 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr)) 192 return -1; 193 score++; 194 } 195 196 if (sk->sk_bound_dev_if) { 197 if (sk->sk_bound_dev_if != dif) 198 return -1; 199 score++; 200 } 201 202 if (sk->sk_incoming_cpu == raw_smp_processor_id()) 203 score++; 204 205 return score; 206 } 207 208 /* called with read_rcu_lock() */ 209 static struct sock *udp6_lib_lookup2(struct net *net, 210 const struct in6_addr *saddr, __be16 sport, 211 const struct in6_addr *daddr, unsigned int hnum, int dif, 212 struct udp_hslot *hslot2, unsigned int slot2, 213 struct sk_buff *skb) 214 { 215 struct sock *sk, *result; 216 int score, badness, matches = 0, reuseport = 0; 217 u32 hash = 0; 218 219 result = NULL; 220 badness = -1; 221 udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { 222 score = compute_score2(sk, net, saddr, sport, 223 daddr, hnum, dif); 224 if (score > badness) { 225 reuseport = sk->sk_reuseport; 226 if (reuseport) { 227 hash = udp6_ehashfn(net, daddr, hnum, 228 saddr, sport); 229 230 result = reuseport_select_sock(sk, hash, skb, 231 sizeof(struct udphdr)); 232 if (result) 233 return result; 234 matches = 1; 235 } 236 result = sk; 237 badness = score; 238 } else if (score == badness && reuseport) { 239 matches++; 240 if (reciprocal_scale(hash, matches) == 0) 241 result = sk; 242 hash = next_pseudo_random32(hash); 243 } 244 } 245 return result; 246 } 247 248 /* rcu_read_lock() must be held */ 249 struct sock *__udp6_lib_lookup(struct net *net, 250 const struct in6_addr *saddr, __be16 sport, 251 const struct in6_addr *daddr, __be16 dport, 252 int dif, struct udp_table *udptable, 253 struct sk_buff *skb) 254 { 255 struct sock *sk, *result; 256 unsigned short hnum = ntohs(dport); 257 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); 258 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; 259 int score, badness, matches = 0, reuseport = 0; 260 u32 hash = 0; 261 262 if (hslot->count > 10) { 263 hash2 = udp6_portaddr_hash(net, daddr, hnum); 264 slot2 = hash2 & udptable->mask; 265 hslot2 = &udptable->hash2[slot2]; 266 if (hslot->count < hslot2->count) 267 goto begin; 268 269 result = udp6_lib_lookup2(net, saddr, sport, 270 daddr, hnum, dif, 271 hslot2, slot2, skb); 272 if (!result) { 273 hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum); 274 slot2 = hash2 & udptable->mask; 275 hslot2 = &udptable->hash2[slot2]; 276 if (hslot->count < hslot2->count) 277 goto begin; 278 279 result = udp6_lib_lookup2(net, saddr, sport, 280 &in6addr_any, hnum, dif, 281 hslot2, slot2, skb); 282 } 283 return result; 284 } 285 begin: 286 result = NULL; 287 badness = -1; 288 sk_for_each_rcu(sk, &hslot->head) { 289 score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif); 290 if (score > badness) { 291 reuseport = sk->sk_reuseport; 292 if (reuseport) { 293 hash = udp6_ehashfn(net, daddr, hnum, 294 saddr, sport); 295 result = reuseport_select_sock(sk, hash, skb, 296 sizeof(struct udphdr)); 297 if (result) 298 return result; 299 matches = 1; 300 } 301 result = sk; 302 badness = score; 303 } else if (score == badness && reuseport) { 304 matches++; 305 if (reciprocal_scale(hash, matches) == 0) 306 result = sk; 307 hash = next_pseudo_random32(hash); 308 } 309 } 310 return result; 311 } 312 EXPORT_SYMBOL_GPL(__udp6_lib_lookup); 313 314 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb, 315 __be16 sport, __be16 dport, 316 struct udp_table *udptable) 317 { 318 struct sock *sk; 319 const struct ipv6hdr *iph = ipv6_hdr(skb); 320 321 sk = skb_steal_sock(skb); 322 if (unlikely(sk)) 323 return sk; 324 return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport, 325 &iph->daddr, dport, inet6_iif(skb), 326 udptable, skb); 327 } 328 329 struct sock *udp6_lib_lookup_skb(struct sk_buff *skb, 330 __be16 sport, __be16 dport) 331 { 332 const struct ipv6hdr *iph = ipv6_hdr(skb); 333 const struct net_device *dev = 334 skb_dst(skb) ? skb_dst(skb)->dev : skb->dev; 335 336 return __udp6_lib_lookup(dev_net(dev), &iph->saddr, sport, 337 &iph->daddr, dport, inet6_iif(skb), 338 &udp_table, skb); 339 } 340 EXPORT_SYMBOL_GPL(udp6_lib_lookup_skb); 341 342 /* Must be called under rcu_read_lock(). 343 * Does increment socket refcount. 344 */ 345 #if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \ 346 IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY) 347 struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport, 348 const struct in6_addr *daddr, __be16 dport, int dif) 349 { 350 struct sock *sk; 351 352 sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport, 353 dif, &udp_table, NULL); 354 if (sk && !atomic_inc_not_zero(&sk->sk_refcnt)) 355 sk = NULL; 356 return sk; 357 } 358 EXPORT_SYMBOL_GPL(udp6_lib_lookup); 359 #endif 360 361 /* 362 * This should be easy, if there is something there we 363 * return it, otherwise we block. 364 */ 365 366 int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, 367 int noblock, int flags, int *addr_len) 368 { 369 struct ipv6_pinfo *np = inet6_sk(sk); 370 struct inet_sock *inet = inet_sk(sk); 371 struct sk_buff *skb; 372 unsigned int ulen, copied; 373 int peeked, peeking, off; 374 int err; 375 int is_udplite = IS_UDPLITE(sk); 376 bool checksum_valid = false; 377 int is_udp4; 378 bool slow; 379 380 if (flags & MSG_ERRQUEUE) 381 return ipv6_recv_error(sk, msg, len, addr_len); 382 383 if (np->rxpmtu && np->rxopt.bits.rxpmtu) 384 return ipv6_recv_rxpmtu(sk, msg, len, addr_len); 385 386 try_again: 387 peeking = off = sk_peek_offset(sk, flags); 388 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), 389 &peeked, &off, &err); 390 if (!skb) 391 return err; 392 393 ulen = skb->len; 394 copied = len; 395 if (copied > ulen - off) 396 copied = ulen - off; 397 else if (copied < ulen) 398 msg->msg_flags |= MSG_TRUNC; 399 400 is_udp4 = (skb->protocol == htons(ETH_P_IP)); 401 402 /* 403 * If checksum is needed at all, try to do it while copying the 404 * data. If the data is truncated, or if we only want a partial 405 * coverage checksum (UDP-Lite), do it before the copy. 406 */ 407 408 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov || peeking) { 409 checksum_valid = !udp_lib_checksum_complete(skb); 410 if (!checksum_valid) 411 goto csum_copy_err; 412 } 413 414 if (checksum_valid || skb_csum_unnecessary(skb)) 415 err = skb_copy_datagram_msg(skb, off, msg, copied); 416 else { 417 err = skb_copy_and_csum_datagram_msg(skb, off, msg); 418 if (err == -EINVAL) 419 goto csum_copy_err; 420 } 421 if (unlikely(err)) { 422 trace_kfree_skb(skb, udpv6_recvmsg); 423 if (!peeked) { 424 atomic_inc(&sk->sk_drops); 425 if (is_udp4) 426 UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, 427 is_udplite); 428 else 429 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, 430 is_udplite); 431 } 432 skb_free_datagram_locked(sk, skb); 433 return err; 434 } 435 if (!peeked) { 436 if (is_udp4) 437 UDP_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS, 438 is_udplite); 439 else 440 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS, 441 is_udplite); 442 } 443 444 sock_recv_ts_and_drops(msg, sk, skb); 445 446 /* Copy the address. */ 447 if (msg->msg_name) { 448 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 449 sin6->sin6_family = AF_INET6; 450 sin6->sin6_port = udp_hdr(skb)->source; 451 sin6->sin6_flowinfo = 0; 452 453 if (is_udp4) { 454 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, 455 &sin6->sin6_addr); 456 sin6->sin6_scope_id = 0; 457 } else { 458 sin6->sin6_addr = ipv6_hdr(skb)->saddr; 459 sin6->sin6_scope_id = 460 ipv6_iface_scope_id(&sin6->sin6_addr, 461 inet6_iif(skb)); 462 } 463 *addr_len = sizeof(*sin6); 464 } 465 466 if (np->rxopt.all) 467 ip6_datagram_recv_common_ctl(sk, msg, skb); 468 469 if (is_udp4) { 470 if (inet->cmsg_flags) 471 ip_cmsg_recv(msg, skb); 472 } else { 473 if (np->rxopt.all) 474 ip6_datagram_recv_specific_ctl(sk, msg, skb); 475 } 476 477 err = copied; 478 if (flags & MSG_TRUNC) 479 err = ulen; 480 481 __skb_free_datagram_locked(sk, skb, peeking ? -err : err); 482 return err; 483 484 csum_copy_err: 485 slow = lock_sock_fast(sk); 486 if (!skb_kill_datagram(sk, skb, flags)) { 487 if (is_udp4) { 488 UDP_INC_STATS(sock_net(sk), 489 UDP_MIB_CSUMERRORS, is_udplite); 490 UDP_INC_STATS(sock_net(sk), 491 UDP_MIB_INERRORS, is_udplite); 492 } else { 493 UDP6_INC_STATS(sock_net(sk), 494 UDP_MIB_CSUMERRORS, is_udplite); 495 UDP6_INC_STATS(sock_net(sk), 496 UDP_MIB_INERRORS, is_udplite); 497 } 498 } 499 unlock_sock_fast(sk, slow); 500 501 /* starting over for a new packet, but check if we need to yield */ 502 cond_resched(); 503 msg->msg_flags &= ~MSG_TRUNC; 504 goto try_again; 505 } 506 507 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 508 u8 type, u8 code, int offset, __be32 info, 509 struct udp_table *udptable) 510 { 511 struct ipv6_pinfo *np; 512 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data; 513 const struct in6_addr *saddr = &hdr->saddr; 514 const struct in6_addr *daddr = &hdr->daddr; 515 struct udphdr *uh = (struct udphdr *)(skb->data+offset); 516 struct sock *sk; 517 int harderr; 518 int err; 519 struct net *net = dev_net(skb->dev); 520 521 sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source, 522 inet6_iif(skb), udptable, skb); 523 if (!sk) { 524 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), 525 ICMP6_MIB_INERRORS); 526 return; 527 } 528 529 harderr = icmpv6_err_convert(type, code, &err); 530 np = inet6_sk(sk); 531 532 if (type == ICMPV6_PKT_TOOBIG) { 533 if (!ip6_sk_accept_pmtu(sk)) 534 goto out; 535 ip6_sk_update_pmtu(skb, sk, info); 536 if (np->pmtudisc != IPV6_PMTUDISC_DONT) 537 harderr = 1; 538 } 539 if (type == NDISC_REDIRECT) { 540 ip6_sk_redirect(skb, sk); 541 goto out; 542 } 543 544 if (!np->recverr) { 545 if (!harderr || sk->sk_state != TCP_ESTABLISHED) 546 goto out; 547 } else { 548 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1)); 549 } 550 551 sk->sk_err = err; 552 sk->sk_error_report(sk); 553 out: 554 return; 555 } 556 557 static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 558 { 559 int rc; 560 561 if (!ipv6_addr_any(&sk->sk_v6_daddr)) { 562 sock_rps_save_rxhash(sk, skb); 563 sk_mark_napi_id(sk, skb); 564 sk_incoming_cpu_update(sk); 565 } 566 567 rc = __sock_queue_rcv_skb(sk, skb); 568 if (rc < 0) { 569 int is_udplite = IS_UDPLITE(sk); 570 571 /* Note that an ENOMEM error is charged twice */ 572 if (rc == -ENOMEM) 573 UDP6_INC_STATS(sock_net(sk), 574 UDP_MIB_RCVBUFERRORS, is_udplite); 575 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); 576 kfree_skb(skb); 577 return -1; 578 } 579 return 0; 580 } 581 582 static __inline__ void udpv6_err(struct sk_buff *skb, 583 struct inet6_skb_parm *opt, u8 type, 584 u8 code, int offset, __be32 info) 585 { 586 __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table); 587 } 588 589 static struct static_key udpv6_encap_needed __read_mostly; 590 void udpv6_encap_enable(void) 591 { 592 if (!static_key_enabled(&udpv6_encap_needed)) 593 static_key_slow_inc(&udpv6_encap_needed); 594 } 595 EXPORT_SYMBOL(udpv6_encap_enable); 596 597 int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 598 { 599 struct udp_sock *up = udp_sk(sk); 600 int rc; 601 int is_udplite = IS_UDPLITE(sk); 602 603 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) 604 goto drop; 605 606 if (static_key_false(&udpv6_encap_needed) && up->encap_type) { 607 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); 608 609 /* 610 * This is an encapsulation socket so pass the skb to 611 * the socket's udp_encap_rcv() hook. Otherwise, just 612 * fall through and pass this up the UDP socket. 613 * up->encap_rcv() returns the following value: 614 * =0 if skb was successfully passed to the encap 615 * handler or was discarded by it. 616 * >0 if skb should be passed on to UDP. 617 * <0 if skb should be resubmitted as proto -N 618 */ 619 620 /* if we're overly short, let UDP handle it */ 621 encap_rcv = ACCESS_ONCE(up->encap_rcv); 622 if (skb->len > sizeof(struct udphdr) && encap_rcv) { 623 int ret; 624 625 /* Verify checksum before giving to encap */ 626 if (udp_lib_checksum_complete(skb)) 627 goto csum_error; 628 629 ret = encap_rcv(sk, skb); 630 if (ret <= 0) { 631 __UDP_INC_STATS(sock_net(sk), 632 UDP_MIB_INDATAGRAMS, 633 is_udplite); 634 return -ret; 635 } 636 } 637 638 /* FALLTHROUGH -- it's a UDP Packet */ 639 } 640 641 /* 642 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c). 643 */ 644 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { 645 646 if (up->pcrlen == 0) { /* full coverage was set */ 647 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n", 648 UDP_SKB_CB(skb)->cscov, skb->len); 649 goto drop; 650 } 651 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { 652 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n", 653 UDP_SKB_CB(skb)->cscov, up->pcrlen); 654 goto drop; 655 } 656 } 657 658 if (rcu_access_pointer(sk->sk_filter)) { 659 if (udp_lib_checksum_complete(skb)) 660 goto csum_error; 661 if (sk_filter(sk, skb)) 662 goto drop; 663 } 664 665 udp_csum_pull_header(skb); 666 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) { 667 __UDP6_INC_STATS(sock_net(sk), 668 UDP_MIB_RCVBUFERRORS, is_udplite); 669 goto drop; 670 } 671 672 skb_dst_drop(skb); 673 674 bh_lock_sock(sk); 675 rc = 0; 676 if (!sock_owned_by_user(sk)) 677 rc = __udpv6_queue_rcv_skb(sk, skb); 678 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) { 679 bh_unlock_sock(sk); 680 goto drop; 681 } 682 bh_unlock_sock(sk); 683 684 return rc; 685 686 csum_error: 687 __UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); 688 drop: 689 __UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); 690 atomic_inc(&sk->sk_drops); 691 kfree_skb(skb); 692 return -1; 693 } 694 695 static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk, 696 __be16 loc_port, const struct in6_addr *loc_addr, 697 __be16 rmt_port, const struct in6_addr *rmt_addr, 698 int dif, unsigned short hnum) 699 { 700 struct inet_sock *inet = inet_sk(sk); 701 702 if (!net_eq(sock_net(sk), net)) 703 return false; 704 705 if (udp_sk(sk)->udp_port_hash != hnum || 706 sk->sk_family != PF_INET6 || 707 (inet->inet_dport && inet->inet_dport != rmt_port) || 708 (!ipv6_addr_any(&sk->sk_v6_daddr) && 709 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) || 710 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) || 711 (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) && 712 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))) 713 return false; 714 if (!inet6_mc_check(sk, loc_addr, rmt_addr)) 715 return false; 716 return true; 717 } 718 719 static void udp6_csum_zero_error(struct sk_buff *skb) 720 { 721 /* RFC 2460 section 8.1 says that we SHOULD log 722 * this error. Well, it is reasonable. 723 */ 724 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n", 725 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source), 726 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest)); 727 } 728 729 /* 730 * Note: called only from the BH handler context, 731 * so we don't need to lock the hashes. 732 */ 733 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, 734 const struct in6_addr *saddr, const struct in6_addr *daddr, 735 struct udp_table *udptable, int proto) 736 { 737 struct sock *sk, *first = NULL; 738 const struct udphdr *uh = udp_hdr(skb); 739 unsigned short hnum = ntohs(uh->dest); 740 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum); 741 unsigned int offset = offsetof(typeof(*sk), sk_node); 742 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10); 743 int dif = inet6_iif(skb); 744 struct hlist_node *node; 745 struct sk_buff *nskb; 746 747 if (use_hash2) { 748 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) & 749 udp_table.mask; 750 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask; 751 start_lookup: 752 hslot = &udp_table.hash2[hash2]; 753 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node); 754 } 755 756 sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) { 757 if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr, 758 uh->source, saddr, dif, hnum)) 759 continue; 760 /* If zero checksum and no_check is not on for 761 * the socket then skip it. 762 */ 763 if (!uh->check && !udp_sk(sk)->no_check6_rx) 764 continue; 765 if (!first) { 766 first = sk; 767 continue; 768 } 769 nskb = skb_clone(skb, GFP_ATOMIC); 770 if (unlikely(!nskb)) { 771 atomic_inc(&sk->sk_drops); 772 __UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS, 773 IS_UDPLITE(sk)); 774 __UDP6_INC_STATS(net, UDP_MIB_INERRORS, 775 IS_UDPLITE(sk)); 776 continue; 777 } 778 779 if (udpv6_queue_rcv_skb(sk, nskb) > 0) 780 consume_skb(nskb); 781 } 782 783 /* Also lookup *:port if we are using hash2 and haven't done so yet. */ 784 if (use_hash2 && hash2 != hash2_any) { 785 hash2 = hash2_any; 786 goto start_lookup; 787 } 788 789 if (first) { 790 if (udpv6_queue_rcv_skb(first, skb) > 0) 791 consume_skb(skb); 792 } else { 793 kfree_skb(skb); 794 __UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI, 795 proto == IPPROTO_UDPLITE); 796 } 797 return 0; 798 } 799 800 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, 801 int proto) 802 { 803 const struct in6_addr *saddr, *daddr; 804 struct net *net = dev_net(skb->dev); 805 struct udphdr *uh; 806 struct sock *sk; 807 u32 ulen = 0; 808 809 if (!pskb_may_pull(skb, sizeof(struct udphdr))) 810 goto discard; 811 812 saddr = &ipv6_hdr(skb)->saddr; 813 daddr = &ipv6_hdr(skb)->daddr; 814 uh = udp_hdr(skb); 815 816 ulen = ntohs(uh->len); 817 if (ulen > skb->len) 818 goto short_packet; 819 820 if (proto == IPPROTO_UDP) { 821 /* UDP validates ulen. */ 822 823 /* Check for jumbo payload */ 824 if (ulen == 0) 825 ulen = skb->len; 826 827 if (ulen < sizeof(*uh)) 828 goto short_packet; 829 830 if (ulen < skb->len) { 831 if (pskb_trim_rcsum(skb, ulen)) 832 goto short_packet; 833 saddr = &ipv6_hdr(skb)->saddr; 834 daddr = &ipv6_hdr(skb)->daddr; 835 uh = udp_hdr(skb); 836 } 837 } 838 839 if (udp6_csum_init(skb, uh, proto)) 840 goto csum_error; 841 842 /* 843 * Multicast receive code 844 */ 845 if (ipv6_addr_is_multicast(daddr)) 846 return __udp6_lib_mcast_deliver(net, skb, 847 saddr, daddr, udptable, proto); 848 849 /* Unicast */ 850 851 /* 852 * check socket cache ... must talk to Alan about his plans 853 * for sock caches... i'll skip this for now. 854 */ 855 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable); 856 if (sk) { 857 int ret; 858 859 if (!uh->check && !udp_sk(sk)->no_check6_rx) { 860 udp6_csum_zero_error(skb); 861 goto csum_error; 862 } 863 864 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk)) 865 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check, 866 ip6_compute_pseudo); 867 868 ret = udpv6_queue_rcv_skb(sk, skb); 869 870 /* a return value > 0 means to resubmit the input */ 871 if (ret > 0) 872 return ret; 873 874 return 0; 875 } 876 877 if (!uh->check) { 878 udp6_csum_zero_error(skb); 879 goto csum_error; 880 } 881 882 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 883 goto discard; 884 885 if (udp_lib_checksum_complete(skb)) 886 goto csum_error; 887 888 __UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); 889 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 890 891 kfree_skb(skb); 892 return 0; 893 894 short_packet: 895 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n", 896 proto == IPPROTO_UDPLITE ? "-Lite" : "", 897 saddr, ntohs(uh->source), 898 ulen, skb->len, 899 daddr, ntohs(uh->dest)); 900 goto discard; 901 csum_error: 902 __UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); 903 discard: 904 __UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); 905 kfree_skb(skb); 906 return 0; 907 } 908 909 static __inline__ int udpv6_rcv(struct sk_buff *skb) 910 { 911 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP); 912 } 913 914 /* 915 * Throw away all pending data and cancel the corking. Socket is locked. 916 */ 917 static void udp_v6_flush_pending_frames(struct sock *sk) 918 { 919 struct udp_sock *up = udp_sk(sk); 920 921 if (up->pending == AF_INET) 922 udp_flush_pending_frames(sk); 923 else if (up->pending) { 924 up->len = 0; 925 up->pending = 0; 926 ip6_flush_pending_frames(sk); 927 } 928 } 929 930 /** 931 * udp6_hwcsum_outgoing - handle outgoing HW checksumming 932 * @sk: socket we are sending on 933 * @skb: sk_buff containing the filled-in UDP header 934 * (checksum field must be zeroed out) 935 */ 936 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, 937 const struct in6_addr *saddr, 938 const struct in6_addr *daddr, int len) 939 { 940 unsigned int offset; 941 struct udphdr *uh = udp_hdr(skb); 942 struct sk_buff *frags = skb_shinfo(skb)->frag_list; 943 __wsum csum = 0; 944 945 if (!frags) { 946 /* Only one fragment on the socket. */ 947 skb->csum_start = skb_transport_header(skb) - skb->head; 948 skb->csum_offset = offsetof(struct udphdr, check); 949 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0); 950 } else { 951 /* 952 * HW-checksum won't work as there are two or more 953 * fragments on the socket so that all csums of sk_buffs 954 * should be together 955 */ 956 offset = skb_transport_offset(skb); 957 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); 958 959 skb->ip_summed = CHECKSUM_NONE; 960 961 do { 962 csum = csum_add(csum, frags->csum); 963 } while ((frags = frags->next)); 964 965 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 966 csum); 967 if (uh->check == 0) 968 uh->check = CSUM_MANGLED_0; 969 } 970 } 971 972 /* 973 * Sending 974 */ 975 976 static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6) 977 { 978 struct sock *sk = skb->sk; 979 struct udphdr *uh; 980 int err = 0; 981 int is_udplite = IS_UDPLITE(sk); 982 __wsum csum = 0; 983 int offset = skb_transport_offset(skb); 984 int len = skb->len - offset; 985 986 /* 987 * Create a UDP header 988 */ 989 uh = udp_hdr(skb); 990 uh->source = fl6->fl6_sport; 991 uh->dest = fl6->fl6_dport; 992 uh->len = htons(len); 993 uh->check = 0; 994 995 if (is_udplite) 996 csum = udplite_csum(skb); 997 else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */ 998 skb->ip_summed = CHECKSUM_NONE; 999 goto send; 1000 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ 1001 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len); 1002 goto send; 1003 } else 1004 csum = udp_csum(skb); 1005 1006 /* add protocol-dependent pseudo-header */ 1007 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, 1008 len, fl6->flowi6_proto, csum); 1009 if (uh->check == 0) 1010 uh->check = CSUM_MANGLED_0; 1011 1012 send: 1013 err = ip6_send_skb(skb); 1014 if (err) { 1015 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) { 1016 UDP6_INC_STATS(sock_net(sk), 1017 UDP_MIB_SNDBUFERRORS, is_udplite); 1018 err = 0; 1019 } 1020 } else { 1021 UDP6_INC_STATS(sock_net(sk), 1022 UDP_MIB_OUTDATAGRAMS, is_udplite); 1023 } 1024 return err; 1025 } 1026 1027 static int udp_v6_push_pending_frames(struct sock *sk) 1028 { 1029 struct sk_buff *skb; 1030 struct udp_sock *up = udp_sk(sk); 1031 struct flowi6 fl6; 1032 int err = 0; 1033 1034 if (up->pending == AF_INET) 1035 return udp_push_pending_frames(sk); 1036 1037 /* ip6_finish_skb will release the cork, so make a copy of 1038 * fl6 here. 1039 */ 1040 fl6 = inet_sk(sk)->cork.fl.u.ip6; 1041 1042 skb = ip6_finish_skb(sk); 1043 if (!skb) 1044 goto out; 1045 1046 err = udp_v6_send_skb(skb, &fl6); 1047 1048 out: 1049 up->len = 0; 1050 up->pending = 0; 1051 return err; 1052 } 1053 1054 int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 1055 { 1056 struct ipv6_txoptions opt_space; 1057 struct udp_sock *up = udp_sk(sk); 1058 struct inet_sock *inet = inet_sk(sk); 1059 struct ipv6_pinfo *np = inet6_sk(sk); 1060 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 1061 struct in6_addr *daddr, *final_p, final; 1062 struct ipv6_txoptions *opt = NULL; 1063 struct ipv6_txoptions *opt_to_free = NULL; 1064 struct ip6_flowlabel *flowlabel = NULL; 1065 struct flowi6 fl6; 1066 struct dst_entry *dst; 1067 struct ipcm6_cookie ipc6; 1068 int addr_len = msg->msg_namelen; 1069 int ulen = len; 1070 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; 1071 int err; 1072 int connected = 0; 1073 int is_udplite = IS_UDPLITE(sk); 1074 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); 1075 struct sockcm_cookie sockc; 1076 1077 ipc6.hlimit = -1; 1078 ipc6.tclass = -1; 1079 ipc6.dontfrag = -1; 1080 1081 /* destination address check */ 1082 if (sin6) { 1083 if (addr_len < offsetof(struct sockaddr, sa_data)) 1084 return -EINVAL; 1085 1086 switch (sin6->sin6_family) { 1087 case AF_INET6: 1088 if (addr_len < SIN6_LEN_RFC2133) 1089 return -EINVAL; 1090 daddr = &sin6->sin6_addr; 1091 break; 1092 case AF_INET: 1093 goto do_udp_sendmsg; 1094 case AF_UNSPEC: 1095 msg->msg_name = sin6 = NULL; 1096 msg->msg_namelen = addr_len = 0; 1097 daddr = NULL; 1098 break; 1099 default: 1100 return -EINVAL; 1101 } 1102 } else if (!up->pending) { 1103 if (sk->sk_state != TCP_ESTABLISHED) 1104 return -EDESTADDRREQ; 1105 daddr = &sk->sk_v6_daddr; 1106 } else 1107 daddr = NULL; 1108 1109 if (daddr) { 1110 if (ipv6_addr_v4mapped(daddr)) { 1111 struct sockaddr_in sin; 1112 sin.sin_family = AF_INET; 1113 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport; 1114 sin.sin_addr.s_addr = daddr->s6_addr32[3]; 1115 msg->msg_name = &sin; 1116 msg->msg_namelen = sizeof(sin); 1117 do_udp_sendmsg: 1118 if (__ipv6_only_sock(sk)) 1119 return -ENETUNREACH; 1120 return udp_sendmsg(sk, msg, len); 1121 } 1122 } 1123 1124 if (up->pending == AF_INET) 1125 return udp_sendmsg(sk, msg, len); 1126 1127 /* Rough check on arithmetic overflow, 1128 better check is made in ip6_append_data(). 1129 */ 1130 if (len > INT_MAX - sizeof(struct udphdr)) 1131 return -EMSGSIZE; 1132 1133 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; 1134 if (up->pending) { 1135 /* 1136 * There are pending frames. 1137 * The socket lock must be held while it's corked. 1138 */ 1139 lock_sock(sk); 1140 if (likely(up->pending)) { 1141 if (unlikely(up->pending != AF_INET6)) { 1142 release_sock(sk); 1143 return -EAFNOSUPPORT; 1144 } 1145 dst = NULL; 1146 goto do_append_data; 1147 } 1148 release_sock(sk); 1149 } 1150 ulen += sizeof(struct udphdr); 1151 1152 memset(&fl6, 0, sizeof(fl6)); 1153 1154 if (sin6) { 1155 if (sin6->sin6_port == 0) 1156 return -EINVAL; 1157 1158 fl6.fl6_dport = sin6->sin6_port; 1159 daddr = &sin6->sin6_addr; 1160 1161 if (np->sndflow) { 1162 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 1163 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { 1164 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 1165 if (!flowlabel) 1166 return -EINVAL; 1167 } 1168 } 1169 1170 /* 1171 * Otherwise it will be difficult to maintain 1172 * sk->sk_dst_cache. 1173 */ 1174 if (sk->sk_state == TCP_ESTABLISHED && 1175 ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) 1176 daddr = &sk->sk_v6_daddr; 1177 1178 if (addr_len >= sizeof(struct sockaddr_in6) && 1179 sin6->sin6_scope_id && 1180 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) 1181 fl6.flowi6_oif = sin6->sin6_scope_id; 1182 } else { 1183 if (sk->sk_state != TCP_ESTABLISHED) 1184 return -EDESTADDRREQ; 1185 1186 fl6.fl6_dport = inet->inet_dport; 1187 daddr = &sk->sk_v6_daddr; 1188 fl6.flowlabel = np->flow_label; 1189 connected = 1; 1190 } 1191 1192 if (!fl6.flowi6_oif) 1193 fl6.flowi6_oif = sk->sk_bound_dev_if; 1194 1195 if (!fl6.flowi6_oif) 1196 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex; 1197 1198 fl6.flowi6_mark = sk->sk_mark; 1199 sockc.tsflags = sk->sk_tsflags; 1200 1201 if (msg->msg_controllen) { 1202 opt = &opt_space; 1203 memset(opt, 0, sizeof(struct ipv6_txoptions)); 1204 opt->tot_len = sizeof(*opt); 1205 ipc6.opt = opt; 1206 1207 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6, &sockc); 1208 if (err < 0) { 1209 fl6_sock_release(flowlabel); 1210 return err; 1211 } 1212 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 1213 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 1214 if (!flowlabel) 1215 return -EINVAL; 1216 } 1217 if (!(opt->opt_nflen|opt->opt_flen)) 1218 opt = NULL; 1219 connected = 0; 1220 } 1221 if (!opt) { 1222 opt = txopt_get(np); 1223 opt_to_free = opt; 1224 } 1225 if (flowlabel) 1226 opt = fl6_merge_options(&opt_space, flowlabel, opt); 1227 opt = ipv6_fixup_options(&opt_space, opt); 1228 ipc6.opt = opt; 1229 1230 fl6.flowi6_proto = sk->sk_protocol; 1231 if (!ipv6_addr_any(daddr)) 1232 fl6.daddr = *daddr; 1233 else 1234 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 1235 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) 1236 fl6.saddr = np->saddr; 1237 fl6.fl6_sport = inet->inet_sport; 1238 1239 final_p = fl6_update_dst(&fl6, opt, &final); 1240 if (final_p) 1241 connected = 0; 1242 1243 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) { 1244 fl6.flowi6_oif = np->mcast_oif; 1245 connected = 0; 1246 } else if (!fl6.flowi6_oif) 1247 fl6.flowi6_oif = np->ucast_oif; 1248 1249 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 1250 1251 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p); 1252 if (IS_ERR(dst)) { 1253 err = PTR_ERR(dst); 1254 dst = NULL; 1255 goto out; 1256 } 1257 1258 if (ipc6.hlimit < 0) 1259 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); 1260 1261 if (ipc6.tclass < 0) 1262 ipc6.tclass = np->tclass; 1263 1264 if (msg->msg_flags&MSG_CONFIRM) 1265 goto do_confirm; 1266 back_from_confirm: 1267 1268 /* Lockless fast path for the non-corking case */ 1269 if (!corkreq) { 1270 struct sk_buff *skb; 1271 1272 skb = ip6_make_skb(sk, getfrag, msg, ulen, 1273 sizeof(struct udphdr), &ipc6, 1274 &fl6, (struct rt6_info *)dst, 1275 msg->msg_flags, &sockc); 1276 err = PTR_ERR(skb); 1277 if (!IS_ERR_OR_NULL(skb)) 1278 err = udp_v6_send_skb(skb, &fl6); 1279 goto release_dst; 1280 } 1281 1282 lock_sock(sk); 1283 if (unlikely(up->pending)) { 1284 /* The socket is already corked while preparing it. */ 1285 /* ... which is an evident application bug. --ANK */ 1286 release_sock(sk); 1287 1288 net_dbg_ratelimited("udp cork app bug 2\n"); 1289 err = -EINVAL; 1290 goto out; 1291 } 1292 1293 up->pending = AF_INET6; 1294 1295 do_append_data: 1296 if (ipc6.dontfrag < 0) 1297 ipc6.dontfrag = np->dontfrag; 1298 up->len += ulen; 1299 err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr), 1300 &ipc6, &fl6, (struct rt6_info *)dst, 1301 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, &sockc); 1302 if (err) 1303 udp_v6_flush_pending_frames(sk); 1304 else if (!corkreq) 1305 err = udp_v6_push_pending_frames(sk); 1306 else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) 1307 up->pending = 0; 1308 1309 if (err > 0) 1310 err = np->recverr ? net_xmit_errno(err) : 0; 1311 release_sock(sk); 1312 1313 release_dst: 1314 if (dst) { 1315 if (connected) { 1316 ip6_dst_store(sk, dst, 1317 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ? 1318 &sk->sk_v6_daddr : NULL, 1319 #ifdef CONFIG_IPV6_SUBTREES 1320 ipv6_addr_equal(&fl6.saddr, &np->saddr) ? 1321 &np->saddr : 1322 #endif 1323 NULL); 1324 } else { 1325 dst_release(dst); 1326 } 1327 dst = NULL; 1328 } 1329 1330 out: 1331 dst_release(dst); 1332 fl6_sock_release(flowlabel); 1333 txopt_put(opt_to_free); 1334 if (!err) 1335 return len; 1336 /* 1337 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting 1338 * ENOBUFS might not be good (it's not tunable per se), but otherwise 1339 * we don't have a good statistic (IpOutDiscards but it can be too many 1340 * things). We could add another new stat but at least for now that 1341 * seems like overkill. 1342 */ 1343 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { 1344 UDP6_INC_STATS(sock_net(sk), 1345 UDP_MIB_SNDBUFERRORS, is_udplite); 1346 } 1347 return err; 1348 1349 do_confirm: 1350 dst_confirm(dst); 1351 if (!(msg->msg_flags&MSG_PROBE) || len) 1352 goto back_from_confirm; 1353 err = 0; 1354 goto out; 1355 } 1356 1357 void udpv6_destroy_sock(struct sock *sk) 1358 { 1359 struct udp_sock *up = udp_sk(sk); 1360 lock_sock(sk); 1361 udp_v6_flush_pending_frames(sk); 1362 release_sock(sk); 1363 1364 if (static_key_false(&udpv6_encap_needed) && up->encap_type) { 1365 void (*encap_destroy)(struct sock *sk); 1366 encap_destroy = ACCESS_ONCE(up->encap_destroy); 1367 if (encap_destroy) 1368 encap_destroy(sk); 1369 } 1370 1371 inet6_destroy_sock(sk); 1372 } 1373 1374 /* 1375 * Socket option code for UDP 1376 */ 1377 int udpv6_setsockopt(struct sock *sk, int level, int optname, 1378 char __user *optval, unsigned int optlen) 1379 { 1380 if (level == SOL_UDP || level == SOL_UDPLITE) 1381 return udp_lib_setsockopt(sk, level, optname, optval, optlen, 1382 udp_v6_push_pending_frames); 1383 return ipv6_setsockopt(sk, level, optname, optval, optlen); 1384 } 1385 1386 #ifdef CONFIG_COMPAT 1387 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname, 1388 char __user *optval, unsigned int optlen) 1389 { 1390 if (level == SOL_UDP || level == SOL_UDPLITE) 1391 return udp_lib_setsockopt(sk, level, optname, optval, optlen, 1392 udp_v6_push_pending_frames); 1393 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen); 1394 } 1395 #endif 1396 1397 int udpv6_getsockopt(struct sock *sk, int level, int optname, 1398 char __user *optval, int __user *optlen) 1399 { 1400 if (level == SOL_UDP || level == SOL_UDPLITE) 1401 return udp_lib_getsockopt(sk, level, optname, optval, optlen); 1402 return ipv6_getsockopt(sk, level, optname, optval, optlen); 1403 } 1404 1405 #ifdef CONFIG_COMPAT 1406 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname, 1407 char __user *optval, int __user *optlen) 1408 { 1409 if (level == SOL_UDP || level == SOL_UDPLITE) 1410 return udp_lib_getsockopt(sk, level, optname, optval, optlen); 1411 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen); 1412 } 1413 #endif 1414 1415 static const struct inet6_protocol udpv6_protocol = { 1416 .handler = udpv6_rcv, 1417 .err_handler = udpv6_err, 1418 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1419 }; 1420 1421 /* ------------------------------------------------------------------------ */ 1422 #ifdef CONFIG_PROC_FS 1423 int udp6_seq_show(struct seq_file *seq, void *v) 1424 { 1425 if (v == SEQ_START_TOKEN) { 1426 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); 1427 } else { 1428 int bucket = ((struct udp_iter_state *)seq->private)->bucket; 1429 struct inet_sock *inet = inet_sk(v); 1430 __u16 srcp = ntohs(inet->inet_sport); 1431 __u16 destp = ntohs(inet->inet_dport); 1432 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket); 1433 } 1434 return 0; 1435 } 1436 1437 static const struct file_operations udp6_afinfo_seq_fops = { 1438 .owner = THIS_MODULE, 1439 .open = udp_seq_open, 1440 .read = seq_read, 1441 .llseek = seq_lseek, 1442 .release = seq_release_net 1443 }; 1444 1445 static struct udp_seq_afinfo udp6_seq_afinfo = { 1446 .name = "udp6", 1447 .family = AF_INET6, 1448 .udp_table = &udp_table, 1449 .seq_fops = &udp6_afinfo_seq_fops, 1450 .seq_ops = { 1451 .show = udp6_seq_show, 1452 }, 1453 }; 1454 1455 int __net_init udp6_proc_init(struct net *net) 1456 { 1457 return udp_proc_register(net, &udp6_seq_afinfo); 1458 } 1459 1460 void udp6_proc_exit(struct net *net) 1461 { 1462 udp_proc_unregister(net, &udp6_seq_afinfo); 1463 } 1464 #endif /* CONFIG_PROC_FS */ 1465 1466 void udp_v6_clear_sk(struct sock *sk, int size) 1467 { 1468 struct inet_sock *inet = inet_sk(sk); 1469 1470 /* we do not want to clear pinet6 field, because of RCU lookups */ 1471 sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6)); 1472 1473 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6); 1474 memset(&inet->pinet6 + 1, 0, size); 1475 } 1476 1477 /* ------------------------------------------------------------------------ */ 1478 1479 struct proto udpv6_prot = { 1480 .name = "UDPv6", 1481 .owner = THIS_MODULE, 1482 .close = udp_lib_close, 1483 .connect = ip6_datagram_connect, 1484 .disconnect = udp_disconnect, 1485 .ioctl = udp_ioctl, 1486 .destroy = udpv6_destroy_sock, 1487 .setsockopt = udpv6_setsockopt, 1488 .getsockopt = udpv6_getsockopt, 1489 .sendmsg = udpv6_sendmsg, 1490 .recvmsg = udpv6_recvmsg, 1491 .backlog_rcv = __udpv6_queue_rcv_skb, 1492 .release_cb = ip6_datagram_release_cb, 1493 .hash = udp_lib_hash, 1494 .unhash = udp_lib_unhash, 1495 .rehash = udp_v6_rehash, 1496 .get_port = udp_v6_get_port, 1497 .memory_allocated = &udp_memory_allocated, 1498 .sysctl_mem = sysctl_udp_mem, 1499 .sysctl_wmem = &sysctl_udp_wmem_min, 1500 .sysctl_rmem = &sysctl_udp_rmem_min, 1501 .obj_size = sizeof(struct udp6_sock), 1502 .slab_flags = SLAB_DESTROY_BY_RCU, 1503 .h.udp_table = &udp_table, 1504 #ifdef CONFIG_COMPAT 1505 .compat_setsockopt = compat_udpv6_setsockopt, 1506 .compat_getsockopt = compat_udpv6_getsockopt, 1507 #endif 1508 .clear_sk = udp_v6_clear_sk, 1509 }; 1510 1511 static struct inet_protosw udpv6_protosw = { 1512 .type = SOCK_DGRAM, 1513 .protocol = IPPROTO_UDP, 1514 .prot = &udpv6_prot, 1515 .ops = &inet6_dgram_ops, 1516 .flags = INET_PROTOSW_PERMANENT, 1517 }; 1518 1519 int __init udpv6_init(void) 1520 { 1521 int ret; 1522 1523 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP); 1524 if (ret) 1525 goto out; 1526 1527 ret = inet6_register_protosw(&udpv6_protosw); 1528 if (ret) 1529 goto out_udpv6_protocol; 1530 out: 1531 return ret; 1532 1533 out_udpv6_protocol: 1534 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); 1535 goto out; 1536 } 1537 1538 void udpv6_exit(void) 1539 { 1540 inet6_unregister_protosw(&udpv6_protosw); 1541 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); 1542 } 1543