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