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 * $Id: udp.c,v 1.65 2002/02/01 22:01:04 davem Exp $ 11 * 12 * Fixes: 13 * Hideaki YOSHIFUJI : sin6_scope_id support 14 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which 15 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind 16 * a single port at the same time. 17 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data 18 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/udp6 to seq_file. 19 * 20 * This program is free software; you can redistribute it and/or 21 * modify it under the terms of the GNU General Public License 22 * as published by the Free Software Foundation; either version 23 * 2 of the License, or (at your option) any later version. 24 */ 25 26 #include <linux/errno.h> 27 #include <linux/types.h> 28 #include <linux/socket.h> 29 #include <linux/sockios.h> 30 #include <linux/net.h> 31 #include <linux/in6.h> 32 #include <linux/netdevice.h> 33 #include <linux/if_arp.h> 34 #include <linux/ipv6.h> 35 #include <linux/icmpv6.h> 36 #include <linux/init.h> 37 #include <linux/module.h> 38 #include <linux/skbuff.h> 39 #include <asm/uaccess.h> 40 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 50 #include <linux/proc_fs.h> 51 #include <linux/seq_file.h> 52 #include "udp_impl.h" 53 54 int udp_v6_get_port(struct sock *sk, unsigned short snum) 55 { 56 return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal); 57 } 58 59 static struct sock *__udp6_lib_lookup(struct net *net, 60 struct in6_addr *saddr, __be16 sport, 61 struct in6_addr *daddr, __be16 dport, 62 int dif, struct hlist_head udptable[]) 63 { 64 struct sock *sk, *result = NULL; 65 struct hlist_node *node; 66 unsigned short hnum = ntohs(dport); 67 int badness = -1; 68 69 read_lock(&udp_hash_lock); 70 sk_for_each(sk, node, &udptable[hnum & (UDP_HTABLE_SIZE - 1)]) { 71 struct inet_sock *inet = inet_sk(sk); 72 73 if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum && 74 sk->sk_family == PF_INET6) { 75 struct ipv6_pinfo *np = inet6_sk(sk); 76 int score = 0; 77 if (inet->dport) { 78 if (inet->dport != sport) 79 continue; 80 score++; 81 } 82 if (!ipv6_addr_any(&np->rcv_saddr)) { 83 if (!ipv6_addr_equal(&np->rcv_saddr, daddr)) 84 continue; 85 score++; 86 } 87 if (!ipv6_addr_any(&np->daddr)) { 88 if (!ipv6_addr_equal(&np->daddr, saddr)) 89 continue; 90 score++; 91 } 92 if (sk->sk_bound_dev_if) { 93 if (sk->sk_bound_dev_if != dif) 94 continue; 95 score++; 96 } 97 if (score == 4) { 98 result = sk; 99 break; 100 } else if (score > badness) { 101 result = sk; 102 badness = score; 103 } 104 } 105 } 106 if (result) 107 sock_hold(result); 108 read_unlock(&udp_hash_lock); 109 return result; 110 } 111 112 /* 113 * This should be easy, if there is something there we 114 * return it, otherwise we block. 115 */ 116 117 int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk, 118 struct msghdr *msg, size_t len, 119 int noblock, int flags, int *addr_len) 120 { 121 struct ipv6_pinfo *np = inet6_sk(sk); 122 struct inet_sock *inet = inet_sk(sk); 123 struct sk_buff *skb; 124 unsigned int ulen, copied; 125 int peeked; 126 int err; 127 int is_udplite = IS_UDPLITE(sk); 128 129 if (addr_len) 130 *addr_len=sizeof(struct sockaddr_in6); 131 132 if (flags & MSG_ERRQUEUE) 133 return ipv6_recv_error(sk, msg, len); 134 135 try_again: 136 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), 137 &peeked, &err); 138 if (!skb) 139 goto out; 140 141 ulen = skb->len - sizeof(struct udphdr); 142 copied = len; 143 if (copied > ulen) 144 copied = ulen; 145 else if (copied < ulen) 146 msg->msg_flags |= MSG_TRUNC; 147 148 /* 149 * If checksum is needed at all, try to do it while copying the 150 * data. If the data is truncated, or if we only want a partial 151 * coverage checksum (UDP-Lite), do it before the copy. 152 */ 153 154 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { 155 if (udp_lib_checksum_complete(skb)) 156 goto csum_copy_err; 157 } 158 159 if (skb_csum_unnecessary(skb)) 160 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), 161 msg->msg_iov, copied ); 162 else { 163 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov); 164 if (err == -EINVAL) 165 goto csum_copy_err; 166 } 167 if (err) 168 goto out_free; 169 170 if (!peeked) 171 UDP6_INC_STATS_USER(UDP_MIB_INDATAGRAMS, is_udplite); 172 173 sock_recv_timestamp(msg, sk, skb); 174 175 /* Copy the address. */ 176 if (msg->msg_name) { 177 struct sockaddr_in6 *sin6; 178 179 sin6 = (struct sockaddr_in6 *) msg->msg_name; 180 sin6->sin6_family = AF_INET6; 181 sin6->sin6_port = udp_hdr(skb)->source; 182 sin6->sin6_flowinfo = 0; 183 sin6->sin6_scope_id = 0; 184 185 if (skb->protocol == htons(ETH_P_IP)) 186 ipv6_addr_set(&sin6->sin6_addr, 0, 0, 187 htonl(0xffff), ip_hdr(skb)->saddr); 188 else { 189 ipv6_addr_copy(&sin6->sin6_addr, 190 &ipv6_hdr(skb)->saddr); 191 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL) 192 sin6->sin6_scope_id = IP6CB(skb)->iif; 193 } 194 195 } 196 if (skb->protocol == htons(ETH_P_IP)) { 197 if (inet->cmsg_flags) 198 ip_cmsg_recv(msg, skb); 199 } else { 200 if (np->rxopt.all) 201 datagram_recv_ctl(sk, msg, skb); 202 } 203 204 err = copied; 205 if (flags & MSG_TRUNC) 206 err = ulen; 207 208 out_free: 209 lock_sock(sk); 210 skb_free_datagram(sk, skb); 211 release_sock(sk); 212 out: 213 return err; 214 215 csum_copy_err: 216 lock_sock(sk); 217 if (!skb_kill_datagram(sk, skb, flags)) 218 UDP6_INC_STATS_USER(UDP_MIB_INERRORS, is_udplite); 219 release_sock(sk); 220 221 if (flags & MSG_DONTWAIT) 222 return -EAGAIN; 223 goto try_again; 224 } 225 226 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 227 int type, int code, int offset, __be32 info, 228 struct hlist_head udptable[] ) 229 { 230 struct ipv6_pinfo *np; 231 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data; 232 struct in6_addr *saddr = &hdr->saddr; 233 struct in6_addr *daddr = &hdr->daddr; 234 struct udphdr *uh = (struct udphdr*)(skb->data+offset); 235 struct sock *sk; 236 int err; 237 238 sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest, 239 saddr, uh->source, inet6_iif(skb), udptable); 240 if (sk == NULL) 241 return; 242 243 np = inet6_sk(sk); 244 245 if (!icmpv6_err_convert(type, code, &err) && !np->recverr) 246 goto out; 247 248 if (sk->sk_state != TCP_ESTABLISHED && !np->recverr) 249 goto out; 250 251 if (np->recverr) 252 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1)); 253 254 sk->sk_err = err; 255 sk->sk_error_report(sk); 256 out: 257 sock_put(sk); 258 } 259 260 static __inline__ void udpv6_err(struct sk_buff *skb, 261 struct inet6_skb_parm *opt, int type, 262 int code, int offset, __be32 info ) 263 { 264 __udp6_lib_err(skb, opt, type, code, offset, info, udp_hash); 265 } 266 267 int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb) 268 { 269 struct udp_sock *up = udp_sk(sk); 270 int rc; 271 int is_udplite = IS_UDPLITE(sk); 272 273 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) 274 goto drop; 275 276 /* 277 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c). 278 */ 279 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { 280 281 if (up->pcrlen == 0) { /* full coverage was set */ 282 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage" 283 " %d while full coverage %d requested\n", 284 UDP_SKB_CB(skb)->cscov, skb->len); 285 goto drop; 286 } 287 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { 288 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d " 289 "too small, need min %d\n", 290 UDP_SKB_CB(skb)->cscov, up->pcrlen); 291 goto drop; 292 } 293 } 294 295 if (sk->sk_filter) { 296 if (udp_lib_checksum_complete(skb)) 297 goto drop; 298 } 299 300 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) { 301 /* Note that an ENOMEM error is charged twice */ 302 if (rc == -ENOMEM) 303 UDP6_INC_STATS_BH(UDP_MIB_RCVBUFERRORS, is_udplite); 304 goto drop; 305 } 306 307 return 0; 308 drop: 309 UDP6_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite); 310 kfree_skb(skb); 311 return -1; 312 } 313 314 static struct sock *udp_v6_mcast_next(struct sock *sk, 315 __be16 loc_port, struct in6_addr *loc_addr, 316 __be16 rmt_port, struct in6_addr *rmt_addr, 317 int dif) 318 { 319 struct hlist_node *node; 320 struct sock *s = sk; 321 unsigned short num = ntohs(loc_port); 322 323 sk_for_each_from(s, node) { 324 struct inet_sock *inet = inet_sk(s); 325 326 if (sock_net(s) != sock_net(sk)) 327 continue; 328 329 if (s->sk_hash == num && s->sk_family == PF_INET6) { 330 struct ipv6_pinfo *np = inet6_sk(s); 331 if (inet->dport) { 332 if (inet->dport != rmt_port) 333 continue; 334 } 335 if (!ipv6_addr_any(&np->daddr) && 336 !ipv6_addr_equal(&np->daddr, rmt_addr)) 337 continue; 338 339 if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif) 340 continue; 341 342 if (!ipv6_addr_any(&np->rcv_saddr)) { 343 if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr)) 344 continue; 345 } 346 if (!inet6_mc_check(s, loc_addr, rmt_addr)) 347 continue; 348 return s; 349 } 350 } 351 return NULL; 352 } 353 354 /* 355 * Note: called only from the BH handler context, 356 * so we don't need to lock the hashes. 357 */ 358 static int __udp6_lib_mcast_deliver(struct sk_buff *skb, struct in6_addr *saddr, 359 struct in6_addr *daddr, struct hlist_head udptable[]) 360 { 361 struct sock *sk, *sk2; 362 const struct udphdr *uh = udp_hdr(skb); 363 int dif; 364 365 read_lock(&udp_hash_lock); 366 sk = sk_head(&udptable[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]); 367 dif = inet6_iif(skb); 368 sk = udp_v6_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif); 369 if (!sk) { 370 kfree_skb(skb); 371 goto out; 372 } 373 374 sk2 = sk; 375 while ((sk2 = udp_v6_mcast_next(sk_next(sk2), uh->dest, daddr, 376 uh->source, saddr, dif))) { 377 struct sk_buff *buff = skb_clone(skb, GFP_ATOMIC); 378 if (buff) { 379 bh_lock_sock_nested(sk2); 380 if (!sock_owned_by_user(sk2)) 381 udpv6_queue_rcv_skb(sk2, buff); 382 else 383 sk_add_backlog(sk2, buff); 384 bh_unlock_sock(sk2); 385 } 386 } 387 bh_lock_sock_nested(sk); 388 if (!sock_owned_by_user(sk)) 389 udpv6_queue_rcv_skb(sk, skb); 390 else 391 sk_add_backlog(sk, skb); 392 bh_unlock_sock(sk); 393 out: 394 read_unlock(&udp_hash_lock); 395 return 0; 396 } 397 398 static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh, 399 int proto) 400 { 401 int err; 402 403 UDP_SKB_CB(skb)->partial_cov = 0; 404 UDP_SKB_CB(skb)->cscov = skb->len; 405 406 if (proto == IPPROTO_UDPLITE) { 407 err = udplite_checksum_init(skb, uh); 408 if (err) 409 return err; 410 } 411 412 if (uh->check == 0) { 413 /* RFC 2460 section 8.1 says that we SHOULD log 414 this error. Well, it is reasonable. 415 */ 416 LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n"); 417 return 1; 418 } 419 if (skb->ip_summed == CHECKSUM_COMPLETE && 420 !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr, 421 skb->len, proto, skb->csum)) 422 skb->ip_summed = CHECKSUM_UNNECESSARY; 423 424 if (!skb_csum_unnecessary(skb)) 425 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, 426 &ipv6_hdr(skb)->daddr, 427 skb->len, proto, 0)); 428 429 return 0; 430 } 431 432 int __udp6_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[], 433 int proto) 434 { 435 struct sock *sk; 436 struct udphdr *uh; 437 struct net_device *dev = skb->dev; 438 struct in6_addr *saddr, *daddr; 439 u32 ulen = 0; 440 441 if (!pskb_may_pull(skb, sizeof(struct udphdr))) 442 goto short_packet; 443 444 saddr = &ipv6_hdr(skb)->saddr; 445 daddr = &ipv6_hdr(skb)->daddr; 446 uh = udp_hdr(skb); 447 448 ulen = ntohs(uh->len); 449 if (ulen > skb->len) 450 goto short_packet; 451 452 if (proto == IPPROTO_UDP) { 453 /* UDP validates ulen. */ 454 455 /* Check for jumbo payload */ 456 if (ulen == 0) 457 ulen = skb->len; 458 459 if (ulen < sizeof(*uh)) 460 goto short_packet; 461 462 if (ulen < skb->len) { 463 if (pskb_trim_rcsum(skb, ulen)) 464 goto short_packet; 465 saddr = &ipv6_hdr(skb)->saddr; 466 daddr = &ipv6_hdr(skb)->daddr; 467 uh = udp_hdr(skb); 468 } 469 } 470 471 if (udp6_csum_init(skb, uh, proto)) 472 goto discard; 473 474 /* 475 * Multicast receive code 476 */ 477 if (ipv6_addr_is_multicast(daddr)) 478 return __udp6_lib_mcast_deliver(skb, saddr, daddr, udptable); 479 480 /* Unicast */ 481 482 /* 483 * check socket cache ... must talk to Alan about his plans 484 * for sock caches... i'll skip this for now. 485 */ 486 sk = __udp6_lib_lookup(dev_net(skb->dev), saddr, uh->source, 487 daddr, uh->dest, inet6_iif(skb), udptable); 488 489 if (sk == NULL) { 490 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 491 goto discard; 492 493 if (udp_lib_checksum_complete(skb)) 494 goto discard; 495 UDP6_INC_STATS_BH(UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); 496 497 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0, dev); 498 499 kfree_skb(skb); 500 return 0; 501 } 502 503 /* deliver */ 504 505 bh_lock_sock_nested(sk); 506 if (!sock_owned_by_user(sk)) 507 udpv6_queue_rcv_skb(sk, skb); 508 else 509 sk_add_backlog(sk, skb); 510 bh_unlock_sock(sk); 511 sock_put(sk); 512 return 0; 513 514 short_packet: 515 LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: %d/%u\n", 516 proto == IPPROTO_UDPLITE ? "-Lite" : "", 517 ulen, skb->len); 518 519 discard: 520 UDP6_INC_STATS_BH(UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); 521 kfree_skb(skb); 522 return 0; 523 } 524 525 static __inline__ int udpv6_rcv(struct sk_buff *skb) 526 { 527 return __udp6_lib_rcv(skb, udp_hash, IPPROTO_UDP); 528 } 529 530 /* 531 * Throw away all pending data and cancel the corking. Socket is locked. 532 */ 533 static void udp_v6_flush_pending_frames(struct sock *sk) 534 { 535 struct udp_sock *up = udp_sk(sk); 536 537 if (up->pending == AF_INET) 538 udp_flush_pending_frames(sk); 539 else if (up->pending) { 540 up->len = 0; 541 up->pending = 0; 542 ip6_flush_pending_frames(sk); 543 } 544 } 545 546 /* 547 * Sending 548 */ 549 550 static int udp_v6_push_pending_frames(struct sock *sk) 551 { 552 struct sk_buff *skb; 553 struct udphdr *uh; 554 struct udp_sock *up = udp_sk(sk); 555 struct inet_sock *inet = inet_sk(sk); 556 struct flowi *fl = &inet->cork.fl; 557 int err = 0; 558 int is_udplite = IS_UDPLITE(sk); 559 __wsum csum = 0; 560 561 /* Grab the skbuff where UDP header space exists. */ 562 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 563 goto out; 564 565 /* 566 * Create a UDP header 567 */ 568 uh = udp_hdr(skb); 569 uh->source = fl->fl_ip_sport; 570 uh->dest = fl->fl_ip_dport; 571 uh->len = htons(up->len); 572 uh->check = 0; 573 574 if (is_udplite) 575 csum = udplite_csum_outgoing(sk, skb); 576 else 577 csum = udp_csum_outgoing(sk, skb); 578 579 /* add protocol-dependent pseudo-header */ 580 uh->check = csum_ipv6_magic(&fl->fl6_src, &fl->fl6_dst, 581 up->len, fl->proto, csum ); 582 if (uh->check == 0) 583 uh->check = CSUM_MANGLED_0; 584 585 err = ip6_push_pending_frames(sk); 586 out: 587 up->len = 0; 588 up->pending = 0; 589 if (!err) 590 UDP6_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS, is_udplite); 591 return err; 592 } 593 594 int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk, 595 struct msghdr *msg, size_t len) 596 { 597 struct ipv6_txoptions opt_space; 598 struct udp_sock *up = udp_sk(sk); 599 struct inet_sock *inet = inet_sk(sk); 600 struct ipv6_pinfo *np = inet6_sk(sk); 601 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name; 602 struct in6_addr *daddr, *final_p = NULL, final; 603 struct ipv6_txoptions *opt = NULL; 604 struct ip6_flowlabel *flowlabel = NULL; 605 struct flowi fl; 606 struct dst_entry *dst; 607 int addr_len = msg->msg_namelen; 608 int ulen = len; 609 int hlimit = -1; 610 int tclass = -1; 611 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; 612 int err; 613 int connected = 0; 614 int is_udplite = IS_UDPLITE(sk); 615 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); 616 617 /* destination address check */ 618 if (sin6) { 619 if (addr_len < offsetof(struct sockaddr, sa_data)) 620 return -EINVAL; 621 622 switch (sin6->sin6_family) { 623 case AF_INET6: 624 if (addr_len < SIN6_LEN_RFC2133) 625 return -EINVAL; 626 daddr = &sin6->sin6_addr; 627 break; 628 case AF_INET: 629 goto do_udp_sendmsg; 630 case AF_UNSPEC: 631 msg->msg_name = sin6 = NULL; 632 msg->msg_namelen = addr_len = 0; 633 daddr = NULL; 634 break; 635 default: 636 return -EINVAL; 637 } 638 } else if (!up->pending) { 639 if (sk->sk_state != TCP_ESTABLISHED) 640 return -EDESTADDRREQ; 641 daddr = &np->daddr; 642 } else 643 daddr = NULL; 644 645 if (daddr) { 646 if (ipv6_addr_v4mapped(daddr)) { 647 struct sockaddr_in sin; 648 sin.sin_family = AF_INET; 649 sin.sin_port = sin6 ? sin6->sin6_port : inet->dport; 650 sin.sin_addr.s_addr = daddr->s6_addr32[3]; 651 msg->msg_name = &sin; 652 msg->msg_namelen = sizeof(sin); 653 do_udp_sendmsg: 654 if (__ipv6_only_sock(sk)) 655 return -ENETUNREACH; 656 return udp_sendmsg(iocb, sk, msg, len); 657 } 658 } 659 660 if (up->pending == AF_INET) 661 return udp_sendmsg(iocb, sk, msg, len); 662 663 /* Rough check on arithmetic overflow, 664 better check is made in ip6_append_data(). 665 */ 666 if (len > INT_MAX - sizeof(struct udphdr)) 667 return -EMSGSIZE; 668 669 if (up->pending) { 670 /* 671 * There are pending frames. 672 * The socket lock must be held while it's corked. 673 */ 674 lock_sock(sk); 675 if (likely(up->pending)) { 676 if (unlikely(up->pending != AF_INET6)) { 677 release_sock(sk); 678 return -EAFNOSUPPORT; 679 } 680 dst = NULL; 681 goto do_append_data; 682 } 683 release_sock(sk); 684 } 685 ulen += sizeof(struct udphdr); 686 687 memset(&fl, 0, sizeof(fl)); 688 689 if (sin6) { 690 if (sin6->sin6_port == 0) 691 return -EINVAL; 692 693 fl.fl_ip_dport = sin6->sin6_port; 694 daddr = &sin6->sin6_addr; 695 696 if (np->sndflow) { 697 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; 698 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) { 699 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 700 if (flowlabel == NULL) 701 return -EINVAL; 702 daddr = &flowlabel->dst; 703 } 704 } 705 706 /* 707 * Otherwise it will be difficult to maintain 708 * sk->sk_dst_cache. 709 */ 710 if (sk->sk_state == TCP_ESTABLISHED && 711 ipv6_addr_equal(daddr, &np->daddr)) 712 daddr = &np->daddr; 713 714 if (addr_len >= sizeof(struct sockaddr_in6) && 715 sin6->sin6_scope_id && 716 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL) 717 fl.oif = sin6->sin6_scope_id; 718 } else { 719 if (sk->sk_state != TCP_ESTABLISHED) 720 return -EDESTADDRREQ; 721 722 fl.fl_ip_dport = inet->dport; 723 daddr = &np->daddr; 724 fl.fl6_flowlabel = np->flow_label; 725 connected = 1; 726 } 727 728 if (!fl.oif) 729 fl.oif = sk->sk_bound_dev_if; 730 731 if (msg->msg_controllen) { 732 opt = &opt_space; 733 memset(opt, 0, sizeof(struct ipv6_txoptions)); 734 opt->tot_len = sizeof(*opt); 735 736 err = datagram_send_ctl(sock_net(sk), msg, &fl, opt, &hlimit, &tclass); 737 if (err < 0) { 738 fl6_sock_release(flowlabel); 739 return err; 740 } 741 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { 742 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 743 if (flowlabel == NULL) 744 return -EINVAL; 745 } 746 if (!(opt->opt_nflen|opt->opt_flen)) 747 opt = NULL; 748 connected = 0; 749 } 750 if (opt == NULL) 751 opt = np->opt; 752 if (flowlabel) 753 opt = fl6_merge_options(&opt_space, flowlabel, opt); 754 opt = ipv6_fixup_options(&opt_space, opt); 755 756 fl.proto = sk->sk_protocol; 757 if (!ipv6_addr_any(daddr)) 758 ipv6_addr_copy(&fl.fl6_dst, daddr); 759 else 760 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ 761 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr)) 762 ipv6_addr_copy(&fl.fl6_src, &np->saddr); 763 fl.fl_ip_sport = inet->sport; 764 765 /* merge ip6_build_xmit from ip6_output */ 766 if (opt && opt->srcrt) { 767 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt; 768 ipv6_addr_copy(&final, &fl.fl6_dst); 769 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 770 final_p = &final; 771 connected = 0; 772 } 773 774 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst)) { 775 fl.oif = np->mcast_oif; 776 connected = 0; 777 } 778 779 security_sk_classify_flow(sk, &fl); 780 781 err = ip6_sk_dst_lookup(sk, &dst, &fl); 782 if (err) 783 goto out; 784 if (final_p) 785 ipv6_addr_copy(&fl.fl6_dst, final_p); 786 787 if ((err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT)) < 0) { 788 if (err == -EREMOTE) 789 err = ip6_dst_blackhole(sk, &dst, &fl); 790 if (err < 0) 791 goto out; 792 } 793 794 if (hlimit < 0) { 795 if (ipv6_addr_is_multicast(&fl.fl6_dst)) 796 hlimit = np->mcast_hops; 797 else 798 hlimit = np->hop_limit; 799 if (hlimit < 0) 800 hlimit = ip6_dst_hoplimit(dst); 801 } 802 803 if (tclass < 0) { 804 tclass = np->tclass; 805 if (tclass < 0) 806 tclass = 0; 807 } 808 809 if (msg->msg_flags&MSG_CONFIRM) 810 goto do_confirm; 811 back_from_confirm: 812 813 lock_sock(sk); 814 if (unlikely(up->pending)) { 815 /* The socket is already corked while preparing it. */ 816 /* ... which is an evident application bug. --ANK */ 817 release_sock(sk); 818 819 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n"); 820 err = -EINVAL; 821 goto out; 822 } 823 824 up->pending = AF_INET6; 825 826 do_append_data: 827 up->len += ulen; 828 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; 829 err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen, 830 sizeof(struct udphdr), hlimit, tclass, opt, &fl, 831 (struct rt6_info*)dst, 832 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); 833 if (err) 834 udp_v6_flush_pending_frames(sk); 835 else if (!corkreq) 836 err = udp_v6_push_pending_frames(sk); 837 else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) 838 up->pending = 0; 839 840 if (dst) { 841 if (connected) { 842 ip6_dst_store(sk, dst, 843 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ? 844 &np->daddr : NULL, 845 #ifdef CONFIG_IPV6_SUBTREES 846 ipv6_addr_equal(&fl.fl6_src, &np->saddr) ? 847 &np->saddr : 848 #endif 849 NULL); 850 } else { 851 dst_release(dst); 852 } 853 dst = NULL; 854 } 855 856 if (err > 0) 857 err = np->recverr ? net_xmit_errno(err) : 0; 858 release_sock(sk); 859 out: 860 dst_release(dst); 861 fl6_sock_release(flowlabel); 862 if (!err) 863 return len; 864 /* 865 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting 866 * ENOBUFS might not be good (it's not tunable per se), but otherwise 867 * we don't have a good statistic (IpOutDiscards but it can be too many 868 * things). We could add another new stat but at least for now that 869 * seems like overkill. 870 */ 871 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { 872 UDP6_INC_STATS_USER(UDP_MIB_SNDBUFERRORS, is_udplite); 873 } 874 return err; 875 876 do_confirm: 877 dst_confirm(dst); 878 if (!(msg->msg_flags&MSG_PROBE) || len) 879 goto back_from_confirm; 880 err = 0; 881 goto out; 882 } 883 884 int udpv6_destroy_sock(struct sock *sk) 885 { 886 lock_sock(sk); 887 udp_v6_flush_pending_frames(sk); 888 release_sock(sk); 889 890 inet6_destroy_sock(sk); 891 892 return 0; 893 } 894 895 /* 896 * Socket option code for UDP 897 */ 898 int udpv6_setsockopt(struct sock *sk, int level, int optname, 899 char __user *optval, int optlen) 900 { 901 if (level == SOL_UDP || level == SOL_UDPLITE) 902 return udp_lib_setsockopt(sk, level, optname, optval, optlen, 903 udp_v6_push_pending_frames); 904 return ipv6_setsockopt(sk, level, optname, optval, optlen); 905 } 906 907 #ifdef CONFIG_COMPAT 908 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname, 909 char __user *optval, int optlen) 910 { 911 if (level == SOL_UDP || level == SOL_UDPLITE) 912 return udp_lib_setsockopt(sk, level, optname, optval, optlen, 913 udp_v6_push_pending_frames); 914 return compat_ipv6_setsockopt(sk, level, optname, optval, optlen); 915 } 916 #endif 917 918 int udpv6_getsockopt(struct sock *sk, int level, int optname, 919 char __user *optval, int __user *optlen) 920 { 921 if (level == SOL_UDP || level == SOL_UDPLITE) 922 return udp_lib_getsockopt(sk, level, optname, optval, optlen); 923 return ipv6_getsockopt(sk, level, optname, optval, optlen); 924 } 925 926 #ifdef CONFIG_COMPAT 927 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname, 928 char __user *optval, int __user *optlen) 929 { 930 if (level == SOL_UDP || level == SOL_UDPLITE) 931 return udp_lib_getsockopt(sk, level, optname, optval, optlen); 932 return compat_ipv6_getsockopt(sk, level, optname, optval, optlen); 933 } 934 #endif 935 936 static struct inet6_protocol udpv6_protocol = { 937 .handler = udpv6_rcv, 938 .err_handler = udpv6_err, 939 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 940 }; 941 942 /* ------------------------------------------------------------------------ */ 943 #ifdef CONFIG_PROC_FS 944 945 static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket) 946 { 947 struct inet_sock *inet = inet_sk(sp); 948 struct ipv6_pinfo *np = inet6_sk(sp); 949 struct in6_addr *dest, *src; 950 __u16 destp, srcp; 951 952 dest = &np->daddr; 953 src = &np->rcv_saddr; 954 destp = ntohs(inet->dport); 955 srcp = ntohs(inet->sport); 956 seq_printf(seq, 957 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " 958 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n", 959 bucket, 960 src->s6_addr32[0], src->s6_addr32[1], 961 src->s6_addr32[2], src->s6_addr32[3], srcp, 962 dest->s6_addr32[0], dest->s6_addr32[1], 963 dest->s6_addr32[2], dest->s6_addr32[3], destp, 964 sp->sk_state, 965 atomic_read(&sp->sk_wmem_alloc), 966 atomic_read(&sp->sk_rmem_alloc), 967 0, 0L, 0, 968 sock_i_uid(sp), 0, 969 sock_i_ino(sp), 970 atomic_read(&sp->sk_refcnt), sp); 971 } 972 973 int udp6_seq_show(struct seq_file *seq, void *v) 974 { 975 if (v == SEQ_START_TOKEN) 976 seq_printf(seq, 977 " sl " 978 "local_address " 979 "remote_address " 980 "st tx_queue rx_queue tr tm->when retrnsmt" 981 " uid timeout inode\n"); 982 else 983 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket); 984 return 0; 985 } 986 987 static struct udp_seq_afinfo udp6_seq_afinfo = { 988 .name = "udp6", 989 .family = AF_INET6, 990 .hashtable = udp_hash, 991 .seq_fops = { 992 .owner = THIS_MODULE, 993 }, 994 .seq_ops = { 995 .show = udp6_seq_show, 996 }, 997 }; 998 999 int udp6_proc_init(struct net *net) 1000 { 1001 return udp_proc_register(net, &udp6_seq_afinfo); 1002 } 1003 1004 void udp6_proc_exit(struct net *net) { 1005 udp_proc_unregister(net, &udp6_seq_afinfo); 1006 } 1007 #endif /* CONFIG_PROC_FS */ 1008 1009 /* ------------------------------------------------------------------------ */ 1010 1011 struct proto udpv6_prot = { 1012 .name = "UDPv6", 1013 .owner = THIS_MODULE, 1014 .close = udp_lib_close, 1015 .connect = ip6_datagram_connect, 1016 .disconnect = udp_disconnect, 1017 .ioctl = udp_ioctl, 1018 .destroy = udpv6_destroy_sock, 1019 .setsockopt = udpv6_setsockopt, 1020 .getsockopt = udpv6_getsockopt, 1021 .sendmsg = udpv6_sendmsg, 1022 .recvmsg = udpv6_recvmsg, 1023 .backlog_rcv = udpv6_queue_rcv_skb, 1024 .hash = udp_lib_hash, 1025 .unhash = udp_lib_unhash, 1026 .get_port = udp_v6_get_port, 1027 .memory_allocated = &udp_memory_allocated, 1028 .sysctl_mem = sysctl_udp_mem, 1029 .sysctl_wmem = &sysctl_udp_wmem_min, 1030 .sysctl_rmem = &sysctl_udp_rmem_min, 1031 .obj_size = sizeof(struct udp6_sock), 1032 .h.udp_hash = udp_hash, 1033 #ifdef CONFIG_COMPAT 1034 .compat_setsockopt = compat_udpv6_setsockopt, 1035 .compat_getsockopt = compat_udpv6_getsockopt, 1036 #endif 1037 }; 1038 1039 static struct inet_protosw udpv6_protosw = { 1040 .type = SOCK_DGRAM, 1041 .protocol = IPPROTO_UDP, 1042 .prot = &udpv6_prot, 1043 .ops = &inet6_dgram_ops, 1044 .capability =-1, 1045 .no_check = UDP_CSUM_DEFAULT, 1046 .flags = INET_PROTOSW_PERMANENT, 1047 }; 1048 1049 1050 int __init udpv6_init(void) 1051 { 1052 int ret; 1053 1054 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP); 1055 if (ret) 1056 goto out; 1057 1058 ret = inet6_register_protosw(&udpv6_protosw); 1059 if (ret) 1060 goto out_udpv6_protocol; 1061 out: 1062 return ret; 1063 1064 out_udpv6_protocol: 1065 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); 1066 goto out; 1067 } 1068 1069 void udpv6_exit(void) 1070 { 1071 inet6_unregister_protosw(&udpv6_protosw); 1072 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); 1073 } 1074