1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * The User Datagram Protocol (UDP). 7 * 8 * Version: $Id: udp.c,v 1.102 2002/02/01 22:01:04 davem Exp $ 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * Arnt Gulbrandsen, <agulbra@nvg.unit.no> 13 * Alan Cox, <Alan.Cox@linux.org> 14 * Hirokazu Takahashi, <taka@valinux.co.jp> 15 * 16 * Fixes: 17 * Alan Cox : verify_area() calls 18 * Alan Cox : stopped close while in use off icmp 19 * messages. Not a fix but a botch that 20 * for udp at least is 'valid'. 21 * Alan Cox : Fixed icmp handling properly 22 * Alan Cox : Correct error for oversized datagrams 23 * Alan Cox : Tidied select() semantics. 24 * Alan Cox : udp_err() fixed properly, also now 25 * select and read wake correctly on errors 26 * Alan Cox : udp_send verify_area moved to avoid mem leak 27 * Alan Cox : UDP can count its memory 28 * Alan Cox : send to an unknown connection causes 29 * an ECONNREFUSED off the icmp, but 30 * does NOT close. 31 * Alan Cox : Switched to new sk_buff handlers. No more backlog! 32 * Alan Cox : Using generic datagram code. Even smaller and the PEEK 33 * bug no longer crashes it. 34 * Fred Van Kempen : Net2e support for sk->broadcast. 35 * Alan Cox : Uses skb_free_datagram 36 * Alan Cox : Added get/set sockopt support. 37 * Alan Cox : Broadcasting without option set returns EACCES. 38 * Alan Cox : No wakeup calls. Instead we now use the callbacks. 39 * Alan Cox : Use ip_tos and ip_ttl 40 * Alan Cox : SNMP Mibs 41 * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support. 42 * Matt Dillon : UDP length checks. 43 * Alan Cox : Smarter af_inet used properly. 44 * Alan Cox : Use new kernel side addressing. 45 * Alan Cox : Incorrect return on truncated datagram receive. 46 * Arnt Gulbrandsen : New udp_send and stuff 47 * Alan Cox : Cache last socket 48 * Alan Cox : Route cache 49 * Jon Peatfield : Minor efficiency fix to sendto(). 50 * Mike Shaver : RFC1122 checks. 51 * Alan Cox : Nonblocking error fix. 52 * Willy Konynenberg : Transparent proxying support. 53 * Mike McLagan : Routing by source 54 * David S. Miller : New socket lookup architecture. 55 * Last socket cache retained as it 56 * does have a high hit rate. 57 * Olaf Kirch : Don't linearise iovec on sendmsg. 58 * Andi Kleen : Some cleanups, cache destination entry 59 * for connect. 60 * Vitaly E. Lavrov : Transparent proxy revived after year coma. 61 * Melvin Smith : Check msg_name not msg_namelen in sendto(), 62 * return ENOTCONN for unconnected sockets (POSIX) 63 * Janos Farkas : don't deliver multi/broadcasts to a different 64 * bound-to-device socket 65 * Hirokazu Takahashi : HW checksumming for outgoing UDP 66 * datagrams. 67 * Hirokazu Takahashi : sendfile() on UDP works now. 68 * Arnaldo C. Melo : convert /proc/net/udp to seq_file 69 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which 70 * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind 71 * a single port at the same time. 72 * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support 73 * 74 * 75 * This program is free software; you can redistribute it and/or 76 * modify it under the terms of the GNU General Public License 77 * as published by the Free Software Foundation; either version 78 * 2 of the License, or (at your option) any later version. 79 */ 80 81 #include <asm/system.h> 82 #include <asm/uaccess.h> 83 #include <asm/ioctls.h> 84 #include <linux/types.h> 85 #include <linux/fcntl.h> 86 #include <linux/module.h> 87 #include <linux/socket.h> 88 #include <linux/sockios.h> 89 #include <linux/igmp.h> 90 #include <linux/in.h> 91 #include <linux/errno.h> 92 #include <linux/timer.h> 93 #include <linux/mm.h> 94 #include <linux/config.h> 95 #include <linux/inet.h> 96 #include <linux/ipv6.h> 97 #include <linux/netdevice.h> 98 #include <net/snmp.h> 99 #include <net/ip.h> 100 #include <net/tcp_states.h> 101 #include <net/protocol.h> 102 #include <linux/skbuff.h> 103 #include <linux/proc_fs.h> 104 #include <linux/seq_file.h> 105 #include <net/sock.h> 106 #include <net/udp.h> 107 #include <net/icmp.h> 108 #include <net/route.h> 109 #include <net/inet_common.h> 110 #include <net/checksum.h> 111 #include <net/xfrm.h> 112 113 /* 114 * Snmp MIB for the UDP layer 115 */ 116 117 DEFINE_SNMP_STAT(struct udp_mib, udp_statistics) __read_mostly; 118 119 struct hlist_head udp_hash[UDP_HTABLE_SIZE]; 120 DEFINE_RWLOCK(udp_hash_lock); 121 122 /* Shared by v4/v6 udp. */ 123 int udp_port_rover; 124 125 static int udp_v4_get_port(struct sock *sk, unsigned short snum) 126 { 127 struct hlist_node *node; 128 struct sock *sk2; 129 struct inet_sock *inet = inet_sk(sk); 130 131 write_lock_bh(&udp_hash_lock); 132 if (snum == 0) { 133 int best_size_so_far, best, result, i; 134 135 if (udp_port_rover > sysctl_local_port_range[1] || 136 udp_port_rover < sysctl_local_port_range[0]) 137 udp_port_rover = sysctl_local_port_range[0]; 138 best_size_so_far = 32767; 139 best = result = udp_port_rover; 140 for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) { 141 struct hlist_head *list; 142 int size; 143 144 list = &udp_hash[result & (UDP_HTABLE_SIZE - 1)]; 145 if (hlist_empty(list)) { 146 if (result > sysctl_local_port_range[1]) 147 result = sysctl_local_port_range[0] + 148 ((result - sysctl_local_port_range[0]) & 149 (UDP_HTABLE_SIZE - 1)); 150 goto gotit; 151 } 152 size = 0; 153 sk_for_each(sk2, node, list) 154 if (++size >= best_size_so_far) 155 goto next; 156 best_size_so_far = size; 157 best = result; 158 next:; 159 } 160 result = best; 161 for(i = 0; i < (1 << 16) / UDP_HTABLE_SIZE; i++, result += UDP_HTABLE_SIZE) { 162 if (result > sysctl_local_port_range[1]) 163 result = sysctl_local_port_range[0] 164 + ((result - sysctl_local_port_range[0]) & 165 (UDP_HTABLE_SIZE - 1)); 166 if (!udp_lport_inuse(result)) 167 break; 168 } 169 if (i >= (1 << 16) / UDP_HTABLE_SIZE) 170 goto fail; 171 gotit: 172 udp_port_rover = snum = result; 173 } else { 174 sk_for_each(sk2, node, 175 &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]) { 176 struct inet_sock *inet2 = inet_sk(sk2); 177 178 if (inet2->num == snum && 179 sk2 != sk && 180 !ipv6_only_sock(sk2) && 181 (!sk2->sk_bound_dev_if || 182 !sk->sk_bound_dev_if || 183 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && 184 (!inet2->rcv_saddr || 185 !inet->rcv_saddr || 186 inet2->rcv_saddr == inet->rcv_saddr) && 187 (!sk2->sk_reuse || !sk->sk_reuse)) 188 goto fail; 189 } 190 } 191 inet->num = snum; 192 if (sk_unhashed(sk)) { 193 struct hlist_head *h = &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]; 194 195 sk_add_node(sk, h); 196 sock_prot_inc_use(sk->sk_prot); 197 } 198 write_unlock_bh(&udp_hash_lock); 199 return 0; 200 201 fail: 202 write_unlock_bh(&udp_hash_lock); 203 return 1; 204 } 205 206 static void udp_v4_hash(struct sock *sk) 207 { 208 BUG(); 209 } 210 211 static void udp_v4_unhash(struct sock *sk) 212 { 213 write_lock_bh(&udp_hash_lock); 214 if (sk_del_node_init(sk)) { 215 inet_sk(sk)->num = 0; 216 sock_prot_dec_use(sk->sk_prot); 217 } 218 write_unlock_bh(&udp_hash_lock); 219 } 220 221 /* UDP is nearly always wildcards out the wazoo, it makes no sense to try 222 * harder than this. -DaveM 223 */ 224 static struct sock *udp_v4_lookup_longway(u32 saddr, u16 sport, 225 u32 daddr, u16 dport, int dif) 226 { 227 struct sock *sk, *result = NULL; 228 struct hlist_node *node; 229 unsigned short hnum = ntohs(dport); 230 int badness = -1; 231 232 sk_for_each(sk, node, &udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]) { 233 struct inet_sock *inet = inet_sk(sk); 234 235 if (inet->num == hnum && !ipv6_only_sock(sk)) { 236 int score = (sk->sk_family == PF_INET ? 1 : 0); 237 if (inet->rcv_saddr) { 238 if (inet->rcv_saddr != daddr) 239 continue; 240 score+=2; 241 } 242 if (inet->daddr) { 243 if (inet->daddr != saddr) 244 continue; 245 score+=2; 246 } 247 if (inet->dport) { 248 if (inet->dport != sport) 249 continue; 250 score+=2; 251 } 252 if (sk->sk_bound_dev_if) { 253 if (sk->sk_bound_dev_if != dif) 254 continue; 255 score+=2; 256 } 257 if(score == 9) { 258 result = sk; 259 break; 260 } else if(score > badness) { 261 result = sk; 262 badness = score; 263 } 264 } 265 } 266 return result; 267 } 268 269 static __inline__ struct sock *udp_v4_lookup(u32 saddr, u16 sport, 270 u32 daddr, u16 dport, int dif) 271 { 272 struct sock *sk; 273 274 read_lock(&udp_hash_lock); 275 sk = udp_v4_lookup_longway(saddr, sport, daddr, dport, dif); 276 if (sk) 277 sock_hold(sk); 278 read_unlock(&udp_hash_lock); 279 return sk; 280 } 281 282 static inline struct sock *udp_v4_mcast_next(struct sock *sk, 283 u16 loc_port, u32 loc_addr, 284 u16 rmt_port, u32 rmt_addr, 285 int dif) 286 { 287 struct hlist_node *node; 288 struct sock *s = sk; 289 unsigned short hnum = ntohs(loc_port); 290 291 sk_for_each_from(s, node) { 292 struct inet_sock *inet = inet_sk(s); 293 294 if (inet->num != hnum || 295 (inet->daddr && inet->daddr != rmt_addr) || 296 (inet->dport != rmt_port && inet->dport) || 297 (inet->rcv_saddr && inet->rcv_saddr != loc_addr) || 298 ipv6_only_sock(s) || 299 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)) 300 continue; 301 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif)) 302 continue; 303 goto found; 304 } 305 s = NULL; 306 found: 307 return s; 308 } 309 310 /* 311 * This routine is called by the ICMP module when it gets some 312 * sort of error condition. If err < 0 then the socket should 313 * be closed and the error returned to the user. If err > 0 314 * it's just the icmp type << 8 | icmp code. 315 * Header points to the ip header of the error packet. We move 316 * on past this. Then (as it used to claim before adjustment) 317 * header points to the first 8 bytes of the udp header. We need 318 * to find the appropriate port. 319 */ 320 321 void udp_err(struct sk_buff *skb, u32 info) 322 { 323 struct inet_sock *inet; 324 struct iphdr *iph = (struct iphdr*)skb->data; 325 struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2)); 326 int type = skb->h.icmph->type; 327 int code = skb->h.icmph->code; 328 struct sock *sk; 329 int harderr; 330 int err; 331 332 sk = udp_v4_lookup(iph->daddr, uh->dest, iph->saddr, uh->source, skb->dev->ifindex); 333 if (sk == NULL) { 334 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS); 335 return; /* No socket for error */ 336 } 337 338 err = 0; 339 harderr = 0; 340 inet = inet_sk(sk); 341 342 switch (type) { 343 default: 344 case ICMP_TIME_EXCEEDED: 345 err = EHOSTUNREACH; 346 break; 347 case ICMP_SOURCE_QUENCH: 348 goto out; 349 case ICMP_PARAMETERPROB: 350 err = EPROTO; 351 harderr = 1; 352 break; 353 case ICMP_DEST_UNREACH: 354 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ 355 if (inet->pmtudisc != IP_PMTUDISC_DONT) { 356 err = EMSGSIZE; 357 harderr = 1; 358 break; 359 } 360 goto out; 361 } 362 err = EHOSTUNREACH; 363 if (code <= NR_ICMP_UNREACH) { 364 harderr = icmp_err_convert[code].fatal; 365 err = icmp_err_convert[code].errno; 366 } 367 break; 368 } 369 370 /* 371 * RFC1122: OK. Passes ICMP errors back to application, as per 372 * 4.1.3.3. 373 */ 374 if (!inet->recverr) { 375 if (!harderr || sk->sk_state != TCP_ESTABLISHED) 376 goto out; 377 } else { 378 ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1)); 379 } 380 sk->sk_err = err; 381 sk->sk_error_report(sk); 382 out: 383 sock_put(sk); 384 } 385 386 /* 387 * Throw away all pending data and cancel the corking. Socket is locked. 388 */ 389 static void udp_flush_pending_frames(struct sock *sk) 390 { 391 struct udp_sock *up = udp_sk(sk); 392 393 if (up->pending) { 394 up->len = 0; 395 up->pending = 0; 396 ip_flush_pending_frames(sk); 397 } 398 } 399 400 /* 401 * Push out all pending data as one UDP datagram. Socket is locked. 402 */ 403 static int udp_push_pending_frames(struct sock *sk, struct udp_sock *up) 404 { 405 struct inet_sock *inet = inet_sk(sk); 406 struct flowi *fl = &inet->cork.fl; 407 struct sk_buff *skb; 408 struct udphdr *uh; 409 int err = 0; 410 411 /* Grab the skbuff where UDP header space exists. */ 412 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) 413 goto out; 414 415 /* 416 * Create a UDP header 417 */ 418 uh = skb->h.uh; 419 uh->source = fl->fl_ip_sport; 420 uh->dest = fl->fl_ip_dport; 421 uh->len = htons(up->len); 422 uh->check = 0; 423 424 if (sk->sk_no_check == UDP_CSUM_NOXMIT) { 425 skb->ip_summed = CHECKSUM_NONE; 426 goto send; 427 } 428 429 if (skb_queue_len(&sk->sk_write_queue) == 1) { 430 /* 431 * Only one fragment on the socket. 432 */ 433 if (skb->ip_summed == CHECKSUM_HW) { 434 skb->csum = offsetof(struct udphdr, check); 435 uh->check = ~csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, 436 up->len, IPPROTO_UDP, 0); 437 } else { 438 skb->csum = csum_partial((char *)uh, 439 sizeof(struct udphdr), skb->csum); 440 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, 441 up->len, IPPROTO_UDP, skb->csum); 442 if (uh->check == 0) 443 uh->check = -1; 444 } 445 } else { 446 unsigned int csum = 0; 447 /* 448 * HW-checksum won't work as there are two or more 449 * fragments on the socket so that all csums of sk_buffs 450 * should be together. 451 */ 452 if (skb->ip_summed == CHECKSUM_HW) { 453 int offset = (unsigned char *)uh - skb->data; 454 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); 455 456 skb->ip_summed = CHECKSUM_NONE; 457 } else { 458 skb->csum = csum_partial((char *)uh, 459 sizeof(struct udphdr), skb->csum); 460 } 461 462 skb_queue_walk(&sk->sk_write_queue, skb) { 463 csum = csum_add(csum, skb->csum); 464 } 465 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, 466 up->len, IPPROTO_UDP, csum); 467 if (uh->check == 0) 468 uh->check = -1; 469 } 470 send: 471 err = ip_push_pending_frames(sk); 472 out: 473 up->len = 0; 474 up->pending = 0; 475 return err; 476 } 477 478 479 static unsigned short udp_check(struct udphdr *uh, int len, unsigned long saddr, unsigned long daddr, unsigned long base) 480 { 481 return(csum_tcpudp_magic(saddr, daddr, len, IPPROTO_UDP, base)); 482 } 483 484 int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, 485 size_t len) 486 { 487 struct inet_sock *inet = inet_sk(sk); 488 struct udp_sock *up = udp_sk(sk); 489 int ulen = len; 490 struct ipcm_cookie ipc; 491 struct rtable *rt = NULL; 492 int free = 0; 493 int connected = 0; 494 u32 daddr, faddr, saddr; 495 u16 dport; 496 u8 tos; 497 int err; 498 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; 499 500 if (len > 0xFFFF) 501 return -EMSGSIZE; 502 503 /* 504 * Check the flags. 505 */ 506 507 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */ 508 return -EOPNOTSUPP; 509 510 ipc.opt = NULL; 511 512 if (up->pending) { 513 /* 514 * There are pending frames. 515 * The socket lock must be held while it's corked. 516 */ 517 lock_sock(sk); 518 if (likely(up->pending)) { 519 if (unlikely(up->pending != AF_INET)) { 520 release_sock(sk); 521 return -EINVAL; 522 } 523 goto do_append_data; 524 } 525 release_sock(sk); 526 } 527 ulen += sizeof(struct udphdr); 528 529 /* 530 * Get and verify the address. 531 */ 532 if (msg->msg_name) { 533 struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name; 534 if (msg->msg_namelen < sizeof(*usin)) 535 return -EINVAL; 536 if (usin->sin_family != AF_INET) { 537 if (usin->sin_family != AF_UNSPEC) 538 return -EAFNOSUPPORT; 539 } 540 541 daddr = usin->sin_addr.s_addr; 542 dport = usin->sin_port; 543 if (dport == 0) 544 return -EINVAL; 545 } else { 546 if (sk->sk_state != TCP_ESTABLISHED) 547 return -EDESTADDRREQ; 548 daddr = inet->daddr; 549 dport = inet->dport; 550 /* Open fast path for connected socket. 551 Route will not be used, if at least one option is set. 552 */ 553 connected = 1; 554 } 555 ipc.addr = inet->saddr; 556 557 ipc.oif = sk->sk_bound_dev_if; 558 if (msg->msg_controllen) { 559 err = ip_cmsg_send(msg, &ipc); 560 if (err) 561 return err; 562 if (ipc.opt) 563 free = 1; 564 connected = 0; 565 } 566 if (!ipc.opt) 567 ipc.opt = inet->opt; 568 569 saddr = ipc.addr; 570 ipc.addr = faddr = daddr; 571 572 if (ipc.opt && ipc.opt->srr) { 573 if (!daddr) 574 return -EINVAL; 575 faddr = ipc.opt->faddr; 576 connected = 0; 577 } 578 tos = RT_TOS(inet->tos); 579 if (sock_flag(sk, SOCK_LOCALROUTE) || 580 (msg->msg_flags & MSG_DONTROUTE) || 581 (ipc.opt && ipc.opt->is_strictroute)) { 582 tos |= RTO_ONLINK; 583 connected = 0; 584 } 585 586 if (MULTICAST(daddr)) { 587 if (!ipc.oif) 588 ipc.oif = inet->mc_index; 589 if (!saddr) 590 saddr = inet->mc_addr; 591 connected = 0; 592 } 593 594 if (connected) 595 rt = (struct rtable*)sk_dst_check(sk, 0); 596 597 if (rt == NULL) { 598 struct flowi fl = { .oif = ipc.oif, 599 .nl_u = { .ip4_u = 600 { .daddr = faddr, 601 .saddr = saddr, 602 .tos = tos } }, 603 .proto = IPPROTO_UDP, 604 .uli_u = { .ports = 605 { .sport = inet->sport, 606 .dport = dport } } }; 607 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT)); 608 if (err) 609 goto out; 610 611 err = -EACCES; 612 if ((rt->rt_flags & RTCF_BROADCAST) && 613 !sock_flag(sk, SOCK_BROADCAST)) 614 goto out; 615 if (connected) 616 sk_dst_set(sk, dst_clone(&rt->u.dst)); 617 } 618 619 if (msg->msg_flags&MSG_CONFIRM) 620 goto do_confirm; 621 back_from_confirm: 622 623 saddr = rt->rt_src; 624 if (!ipc.addr) 625 daddr = ipc.addr = rt->rt_dst; 626 627 lock_sock(sk); 628 if (unlikely(up->pending)) { 629 /* The socket is already corked while preparing it. */ 630 /* ... which is an evident application bug. --ANK */ 631 release_sock(sk); 632 633 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n"); 634 err = -EINVAL; 635 goto out; 636 } 637 /* 638 * Now cork the socket to pend data. 639 */ 640 inet->cork.fl.fl4_dst = daddr; 641 inet->cork.fl.fl_ip_dport = dport; 642 inet->cork.fl.fl4_src = saddr; 643 inet->cork.fl.fl_ip_sport = inet->sport; 644 up->pending = AF_INET; 645 646 do_append_data: 647 up->len += ulen; 648 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, ulen, 649 sizeof(struct udphdr), &ipc, rt, 650 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); 651 if (err) 652 udp_flush_pending_frames(sk); 653 else if (!corkreq) 654 err = udp_push_pending_frames(sk, up); 655 release_sock(sk); 656 657 out: 658 ip_rt_put(rt); 659 if (free) 660 kfree(ipc.opt); 661 if (!err) { 662 UDP_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS); 663 return len; 664 } 665 return err; 666 667 do_confirm: 668 dst_confirm(&rt->u.dst); 669 if (!(msg->msg_flags&MSG_PROBE) || len) 670 goto back_from_confirm; 671 err = 0; 672 goto out; 673 } 674 675 static int udp_sendpage(struct sock *sk, struct page *page, int offset, 676 size_t size, int flags) 677 { 678 struct udp_sock *up = udp_sk(sk); 679 int ret; 680 681 if (!up->pending) { 682 struct msghdr msg = { .msg_flags = flags|MSG_MORE }; 683 684 /* Call udp_sendmsg to specify destination address which 685 * sendpage interface can't pass. 686 * This will succeed only when the socket is connected. 687 */ 688 ret = udp_sendmsg(NULL, sk, &msg, 0); 689 if (ret < 0) 690 return ret; 691 } 692 693 lock_sock(sk); 694 695 if (unlikely(!up->pending)) { 696 release_sock(sk); 697 698 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n"); 699 return -EINVAL; 700 } 701 702 ret = ip_append_page(sk, page, offset, size, flags); 703 if (ret == -EOPNOTSUPP) { 704 release_sock(sk); 705 return sock_no_sendpage(sk->sk_socket, page, offset, 706 size, flags); 707 } 708 if (ret < 0) { 709 udp_flush_pending_frames(sk); 710 goto out; 711 } 712 713 up->len += size; 714 if (!(up->corkflag || (flags&MSG_MORE))) 715 ret = udp_push_pending_frames(sk, up); 716 if (!ret) 717 ret = size; 718 out: 719 release_sock(sk); 720 return ret; 721 } 722 723 /* 724 * IOCTL requests applicable to the UDP protocol 725 */ 726 727 int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) 728 { 729 switch(cmd) 730 { 731 case SIOCOUTQ: 732 { 733 int amount = atomic_read(&sk->sk_wmem_alloc); 734 return put_user(amount, (int __user *)arg); 735 } 736 737 case SIOCINQ: 738 { 739 struct sk_buff *skb; 740 unsigned long amount; 741 742 amount = 0; 743 spin_lock_bh(&sk->sk_receive_queue.lock); 744 skb = skb_peek(&sk->sk_receive_queue); 745 if (skb != NULL) { 746 /* 747 * We will only return the amount 748 * of this packet since that is all 749 * that will be read. 750 */ 751 amount = skb->len - sizeof(struct udphdr); 752 } 753 spin_unlock_bh(&sk->sk_receive_queue.lock); 754 return put_user(amount, (int __user *)arg); 755 } 756 757 default: 758 return -ENOIOCTLCMD; 759 } 760 return(0); 761 } 762 763 static __inline__ int __udp_checksum_complete(struct sk_buff *skb) 764 { 765 return __skb_checksum_complete(skb); 766 } 767 768 static __inline__ int udp_checksum_complete(struct sk_buff *skb) 769 { 770 return skb->ip_summed != CHECKSUM_UNNECESSARY && 771 __udp_checksum_complete(skb); 772 } 773 774 /* 775 * This should be easy, if there is something there we 776 * return it, otherwise we block. 777 */ 778 779 static int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, 780 size_t len, int noblock, int flags, int *addr_len) 781 { 782 struct inet_sock *inet = inet_sk(sk); 783 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; 784 struct sk_buff *skb; 785 int copied, err; 786 787 /* 788 * Check any passed addresses 789 */ 790 if (addr_len) 791 *addr_len=sizeof(*sin); 792 793 if (flags & MSG_ERRQUEUE) 794 return ip_recv_error(sk, msg, len); 795 796 try_again: 797 skb = skb_recv_datagram(sk, flags, noblock, &err); 798 if (!skb) 799 goto out; 800 801 copied = skb->len - sizeof(struct udphdr); 802 if (copied > len) { 803 copied = len; 804 msg->msg_flags |= MSG_TRUNC; 805 } 806 807 if (skb->ip_summed==CHECKSUM_UNNECESSARY) { 808 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov, 809 copied); 810 } else if (msg->msg_flags&MSG_TRUNC) { 811 if (__udp_checksum_complete(skb)) 812 goto csum_copy_err; 813 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov, 814 copied); 815 } else { 816 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov); 817 818 if (err == -EINVAL) 819 goto csum_copy_err; 820 } 821 822 if (err) 823 goto out_free; 824 825 sock_recv_timestamp(msg, sk, skb); 826 827 /* Copy the address. */ 828 if (sin) 829 { 830 sin->sin_family = AF_INET; 831 sin->sin_port = skb->h.uh->source; 832 sin->sin_addr.s_addr = skb->nh.iph->saddr; 833 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 834 } 835 if (inet->cmsg_flags) 836 ip_cmsg_recv(msg, skb); 837 838 err = copied; 839 if (flags & MSG_TRUNC) 840 err = skb->len - sizeof(struct udphdr); 841 842 out_free: 843 skb_free_datagram(sk, skb); 844 out: 845 return err; 846 847 csum_copy_err: 848 UDP_INC_STATS_BH(UDP_MIB_INERRORS); 849 850 skb_kill_datagram(sk, skb, flags); 851 852 if (noblock) 853 return -EAGAIN; 854 goto try_again; 855 } 856 857 858 int udp_disconnect(struct sock *sk, int flags) 859 { 860 struct inet_sock *inet = inet_sk(sk); 861 /* 862 * 1003.1g - break association. 863 */ 864 865 sk->sk_state = TCP_CLOSE; 866 inet->daddr = 0; 867 inet->dport = 0; 868 sk->sk_bound_dev_if = 0; 869 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) 870 inet_reset_saddr(sk); 871 872 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { 873 sk->sk_prot->unhash(sk); 874 inet->sport = 0; 875 } 876 sk_dst_reset(sk); 877 return 0; 878 } 879 880 static void udp_close(struct sock *sk, long timeout) 881 { 882 sk_common_release(sk); 883 } 884 885 /* return: 886 * 1 if the the UDP system should process it 887 * 0 if we should drop this packet 888 * -1 if it should get processed by xfrm4_rcv_encap 889 */ 890 static int udp_encap_rcv(struct sock * sk, struct sk_buff *skb) 891 { 892 #ifndef CONFIG_XFRM 893 return 1; 894 #else 895 struct udp_sock *up = udp_sk(sk); 896 struct udphdr *uh = skb->h.uh; 897 struct iphdr *iph; 898 int iphlen, len; 899 900 __u8 *udpdata = (__u8 *)uh + sizeof(struct udphdr); 901 __u32 *udpdata32 = (__u32 *)udpdata; 902 __u16 encap_type = up->encap_type; 903 904 /* if we're overly short, let UDP handle it */ 905 if (udpdata > skb->tail) 906 return 1; 907 908 /* if this is not encapsulated socket, then just return now */ 909 if (!encap_type) 910 return 1; 911 912 len = skb->tail - udpdata; 913 914 switch (encap_type) { 915 default: 916 case UDP_ENCAP_ESPINUDP: 917 /* Check if this is a keepalive packet. If so, eat it. */ 918 if (len == 1 && udpdata[0] == 0xff) { 919 return 0; 920 } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0 ) { 921 /* ESP Packet without Non-ESP header */ 922 len = sizeof(struct udphdr); 923 } else 924 /* Must be an IKE packet.. pass it through */ 925 return 1; 926 break; 927 case UDP_ENCAP_ESPINUDP_NON_IKE: 928 /* Check if this is a keepalive packet. If so, eat it. */ 929 if (len == 1 && udpdata[0] == 0xff) { 930 return 0; 931 } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) && 932 udpdata32[0] == 0 && udpdata32[1] == 0) { 933 934 /* ESP Packet with Non-IKE marker */ 935 len = sizeof(struct udphdr) + 2 * sizeof(u32); 936 } else 937 /* Must be an IKE packet.. pass it through */ 938 return 1; 939 break; 940 } 941 942 /* At this point we are sure that this is an ESPinUDP packet, 943 * so we need to remove 'len' bytes from the packet (the UDP 944 * header and optional ESP marker bytes) and then modify the 945 * protocol to ESP, and then call into the transform receiver. 946 */ 947 if (skb_cloned(skb) && pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) 948 return 0; 949 950 /* Now we can update and verify the packet length... */ 951 iph = skb->nh.iph; 952 iphlen = iph->ihl << 2; 953 iph->tot_len = htons(ntohs(iph->tot_len) - len); 954 if (skb->len < iphlen + len) { 955 /* packet is too small!?! */ 956 return 0; 957 } 958 959 /* pull the data buffer up to the ESP header and set the 960 * transport header to point to ESP. Keep UDP on the stack 961 * for later. 962 */ 963 skb->h.raw = skb_pull(skb, len); 964 965 /* modify the protocol (it's ESP!) */ 966 iph->protocol = IPPROTO_ESP; 967 968 /* and let the caller know to send this into the ESP processor... */ 969 return -1; 970 #endif 971 } 972 973 /* returns: 974 * -1: error 975 * 0: success 976 * >0: "udp encap" protocol resubmission 977 * 978 * Note that in the success and error cases, the skb is assumed to 979 * have either been requeued or freed. 980 */ 981 static int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb) 982 { 983 struct udp_sock *up = udp_sk(sk); 984 985 /* 986 * Charge it to the socket, dropping if the queue is full. 987 */ 988 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) { 989 kfree_skb(skb); 990 return -1; 991 } 992 nf_reset(skb); 993 994 if (up->encap_type) { 995 /* 996 * This is an encapsulation socket, so let's see if this is 997 * an encapsulated packet. 998 * If it's a keepalive packet, then just eat it. 999 * If it's an encapsulateed packet, then pass it to the 1000 * IPsec xfrm input and return the response 1001 * appropriately. Otherwise, just fall through and 1002 * pass this up the UDP socket. 1003 */ 1004 int ret; 1005 1006 ret = udp_encap_rcv(sk, skb); 1007 if (ret == 0) { 1008 /* Eat the packet .. */ 1009 kfree_skb(skb); 1010 return 0; 1011 } 1012 if (ret < 0) { 1013 /* process the ESP packet */ 1014 ret = xfrm4_rcv_encap(skb, up->encap_type); 1015 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS); 1016 return -ret; 1017 } 1018 /* FALLTHROUGH -- it's a UDP Packet */ 1019 } 1020 1021 if (sk->sk_filter && skb->ip_summed != CHECKSUM_UNNECESSARY) { 1022 if (__udp_checksum_complete(skb)) { 1023 UDP_INC_STATS_BH(UDP_MIB_INERRORS); 1024 kfree_skb(skb); 1025 return -1; 1026 } 1027 skb->ip_summed = CHECKSUM_UNNECESSARY; 1028 } 1029 1030 if (sock_queue_rcv_skb(sk,skb)<0) { 1031 UDP_INC_STATS_BH(UDP_MIB_INERRORS); 1032 kfree_skb(skb); 1033 return -1; 1034 } 1035 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS); 1036 return 0; 1037 } 1038 1039 /* 1040 * Multicasts and broadcasts go to each listener. 1041 * 1042 * Note: called only from the BH handler context, 1043 * so we don't need to lock the hashes. 1044 */ 1045 static int udp_v4_mcast_deliver(struct sk_buff *skb, struct udphdr *uh, 1046 u32 saddr, u32 daddr) 1047 { 1048 struct sock *sk; 1049 int dif; 1050 1051 read_lock(&udp_hash_lock); 1052 sk = sk_head(&udp_hash[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]); 1053 dif = skb->dev->ifindex; 1054 sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif); 1055 if (sk) { 1056 struct sock *sknext = NULL; 1057 1058 do { 1059 struct sk_buff *skb1 = skb; 1060 1061 sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr, 1062 uh->source, saddr, dif); 1063 if(sknext) 1064 skb1 = skb_clone(skb, GFP_ATOMIC); 1065 1066 if(skb1) { 1067 int ret = udp_queue_rcv_skb(sk, skb1); 1068 if (ret > 0) 1069 /* we should probably re-process instead 1070 * of dropping packets here. */ 1071 kfree_skb(skb1); 1072 } 1073 sk = sknext; 1074 } while(sknext); 1075 } else 1076 kfree_skb(skb); 1077 read_unlock(&udp_hash_lock); 1078 return 0; 1079 } 1080 1081 /* Initialize UDP checksum. If exited with zero value (success), 1082 * CHECKSUM_UNNECESSARY means, that no more checks are required. 1083 * Otherwise, csum completion requires chacksumming packet body, 1084 * including udp header and folding it to skb->csum. 1085 */ 1086 static void udp_checksum_init(struct sk_buff *skb, struct udphdr *uh, 1087 unsigned short ulen, u32 saddr, u32 daddr) 1088 { 1089 if (uh->check == 0) { 1090 skb->ip_summed = CHECKSUM_UNNECESSARY; 1091 } else if (skb->ip_summed == CHECKSUM_HW) { 1092 if (!udp_check(uh, ulen, saddr, daddr, skb->csum)) 1093 skb->ip_summed = CHECKSUM_UNNECESSARY; 1094 } 1095 if (skb->ip_summed != CHECKSUM_UNNECESSARY) 1096 skb->csum = csum_tcpudp_nofold(saddr, daddr, ulen, IPPROTO_UDP, 0); 1097 /* Probably, we should checksum udp header (it should be in cache 1098 * in any case) and data in tiny packets (< rx copybreak). 1099 */ 1100 } 1101 1102 /* 1103 * All we need to do is get the socket, and then do a checksum. 1104 */ 1105 1106 int udp_rcv(struct sk_buff *skb) 1107 { 1108 struct sock *sk; 1109 struct udphdr *uh; 1110 unsigned short ulen; 1111 struct rtable *rt = (struct rtable*)skb->dst; 1112 u32 saddr = skb->nh.iph->saddr; 1113 u32 daddr = skb->nh.iph->daddr; 1114 int len = skb->len; 1115 1116 /* 1117 * Validate the packet and the UDP length. 1118 */ 1119 if (!pskb_may_pull(skb, sizeof(struct udphdr))) 1120 goto no_header; 1121 1122 uh = skb->h.uh; 1123 1124 ulen = ntohs(uh->len); 1125 1126 if (ulen > len || ulen < sizeof(*uh)) 1127 goto short_packet; 1128 1129 if (pskb_trim_rcsum(skb, ulen)) 1130 goto short_packet; 1131 1132 udp_checksum_init(skb, uh, ulen, saddr, daddr); 1133 1134 if(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) 1135 return udp_v4_mcast_deliver(skb, uh, saddr, daddr); 1136 1137 sk = udp_v4_lookup(saddr, uh->source, daddr, uh->dest, skb->dev->ifindex); 1138 1139 if (sk != NULL) { 1140 int ret = udp_queue_rcv_skb(sk, skb); 1141 sock_put(sk); 1142 1143 /* a return value > 0 means to resubmit the input, but 1144 * it it wants the return to be -protocol, or 0 1145 */ 1146 if (ret > 0) 1147 return -ret; 1148 return 0; 1149 } 1150 1151 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) 1152 goto drop; 1153 nf_reset(skb); 1154 1155 /* No socket. Drop packet silently, if checksum is wrong */ 1156 if (udp_checksum_complete(skb)) 1157 goto csum_error; 1158 1159 UDP_INC_STATS_BH(UDP_MIB_NOPORTS); 1160 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); 1161 1162 /* 1163 * Hmm. We got an UDP packet to a port to which we 1164 * don't wanna listen. Ignore it. 1165 */ 1166 kfree_skb(skb); 1167 return(0); 1168 1169 short_packet: 1170 LIMIT_NETDEBUG(KERN_DEBUG "UDP: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n", 1171 NIPQUAD(saddr), 1172 ntohs(uh->source), 1173 ulen, 1174 len, 1175 NIPQUAD(daddr), 1176 ntohs(uh->dest)); 1177 no_header: 1178 UDP_INC_STATS_BH(UDP_MIB_INERRORS); 1179 kfree_skb(skb); 1180 return(0); 1181 1182 csum_error: 1183 /* 1184 * RFC1122: OK. Discards the bad packet silently (as far as 1185 * the network is concerned, anyway) as per 4.1.3.4 (MUST). 1186 */ 1187 LIMIT_NETDEBUG(KERN_DEBUG "UDP: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n", 1188 NIPQUAD(saddr), 1189 ntohs(uh->source), 1190 NIPQUAD(daddr), 1191 ntohs(uh->dest), 1192 ulen); 1193 drop: 1194 UDP_INC_STATS_BH(UDP_MIB_INERRORS); 1195 kfree_skb(skb); 1196 return(0); 1197 } 1198 1199 static int udp_destroy_sock(struct sock *sk) 1200 { 1201 lock_sock(sk); 1202 udp_flush_pending_frames(sk); 1203 release_sock(sk); 1204 return 0; 1205 } 1206 1207 /* 1208 * Socket option code for UDP 1209 */ 1210 static int do_udp_setsockopt(struct sock *sk, int level, int optname, 1211 char __user *optval, int optlen) 1212 { 1213 struct udp_sock *up = udp_sk(sk); 1214 int val; 1215 int err = 0; 1216 1217 if(optlen<sizeof(int)) 1218 return -EINVAL; 1219 1220 if (get_user(val, (int __user *)optval)) 1221 return -EFAULT; 1222 1223 switch(optname) { 1224 case UDP_CORK: 1225 if (val != 0) { 1226 up->corkflag = 1; 1227 } else { 1228 up->corkflag = 0; 1229 lock_sock(sk); 1230 udp_push_pending_frames(sk, up); 1231 release_sock(sk); 1232 } 1233 break; 1234 1235 case UDP_ENCAP: 1236 switch (val) { 1237 case 0: 1238 case UDP_ENCAP_ESPINUDP: 1239 case UDP_ENCAP_ESPINUDP_NON_IKE: 1240 up->encap_type = val; 1241 break; 1242 default: 1243 err = -ENOPROTOOPT; 1244 break; 1245 } 1246 break; 1247 1248 default: 1249 err = -ENOPROTOOPT; 1250 break; 1251 }; 1252 1253 return err; 1254 } 1255 1256 static int udp_setsockopt(struct sock *sk, int level, int optname, 1257 char __user *optval, int optlen) 1258 { 1259 if (level != SOL_UDP) 1260 return ip_setsockopt(sk, level, optname, optval, optlen); 1261 return do_udp_setsockopt(sk, level, optname, optval, optlen); 1262 } 1263 1264 #ifdef CONFIG_COMPAT 1265 static int compat_udp_setsockopt(struct sock *sk, int level, int optname, 1266 char __user *optval, int optlen) 1267 { 1268 if (level != SOL_UDP) 1269 return compat_ip_setsockopt(sk, level, optname, optval, optlen); 1270 return do_udp_setsockopt(sk, level, optname, optval, optlen); 1271 } 1272 #endif 1273 1274 static int do_udp_getsockopt(struct sock *sk, int level, int optname, 1275 char __user *optval, int __user *optlen) 1276 { 1277 struct udp_sock *up = udp_sk(sk); 1278 int val, len; 1279 1280 if(get_user(len,optlen)) 1281 return -EFAULT; 1282 1283 len = min_t(unsigned int, len, sizeof(int)); 1284 1285 if(len < 0) 1286 return -EINVAL; 1287 1288 switch(optname) { 1289 case UDP_CORK: 1290 val = up->corkflag; 1291 break; 1292 1293 case UDP_ENCAP: 1294 val = up->encap_type; 1295 break; 1296 1297 default: 1298 return -ENOPROTOOPT; 1299 }; 1300 1301 if(put_user(len, optlen)) 1302 return -EFAULT; 1303 if(copy_to_user(optval, &val,len)) 1304 return -EFAULT; 1305 return 0; 1306 } 1307 1308 static int udp_getsockopt(struct sock *sk, int level, int optname, 1309 char __user *optval, int __user *optlen) 1310 { 1311 if (level != SOL_UDP) 1312 return ip_getsockopt(sk, level, optname, optval, optlen); 1313 return do_udp_getsockopt(sk, level, optname, optval, optlen); 1314 } 1315 1316 #ifdef CONFIG_COMPAT 1317 static int compat_udp_getsockopt(struct sock *sk, int level, int optname, 1318 char __user *optval, int __user *optlen) 1319 { 1320 if (level != SOL_UDP) 1321 return compat_ip_getsockopt(sk, level, optname, optval, optlen); 1322 return do_udp_getsockopt(sk, level, optname, optval, optlen); 1323 } 1324 #endif 1325 /** 1326 * udp_poll - wait for a UDP event. 1327 * @file - file struct 1328 * @sock - socket 1329 * @wait - poll table 1330 * 1331 * This is same as datagram poll, except for the special case of 1332 * blocking sockets. If application is using a blocking fd 1333 * and a packet with checksum error is in the queue; 1334 * then it could get return from select indicating data available 1335 * but then block when reading it. Add special case code 1336 * to work around these arguably broken applications. 1337 */ 1338 unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) 1339 { 1340 unsigned int mask = datagram_poll(file, sock, wait); 1341 struct sock *sk = sock->sk; 1342 1343 /* Check for false positives due to checksum errors */ 1344 if ( (mask & POLLRDNORM) && 1345 !(file->f_flags & O_NONBLOCK) && 1346 !(sk->sk_shutdown & RCV_SHUTDOWN)){ 1347 struct sk_buff_head *rcvq = &sk->sk_receive_queue; 1348 struct sk_buff *skb; 1349 1350 spin_lock_bh(&rcvq->lock); 1351 while ((skb = skb_peek(rcvq)) != NULL) { 1352 if (udp_checksum_complete(skb)) { 1353 UDP_INC_STATS_BH(UDP_MIB_INERRORS); 1354 __skb_unlink(skb, rcvq); 1355 kfree_skb(skb); 1356 } else { 1357 skb->ip_summed = CHECKSUM_UNNECESSARY; 1358 break; 1359 } 1360 } 1361 spin_unlock_bh(&rcvq->lock); 1362 1363 /* nothing to see, move along */ 1364 if (skb == NULL) 1365 mask &= ~(POLLIN | POLLRDNORM); 1366 } 1367 1368 return mask; 1369 1370 } 1371 1372 struct proto udp_prot = { 1373 .name = "UDP", 1374 .owner = THIS_MODULE, 1375 .close = udp_close, 1376 .connect = ip4_datagram_connect, 1377 .disconnect = udp_disconnect, 1378 .ioctl = udp_ioctl, 1379 .destroy = udp_destroy_sock, 1380 .setsockopt = udp_setsockopt, 1381 .getsockopt = udp_getsockopt, 1382 .sendmsg = udp_sendmsg, 1383 .recvmsg = udp_recvmsg, 1384 .sendpage = udp_sendpage, 1385 .backlog_rcv = udp_queue_rcv_skb, 1386 .hash = udp_v4_hash, 1387 .unhash = udp_v4_unhash, 1388 .get_port = udp_v4_get_port, 1389 .obj_size = sizeof(struct udp_sock), 1390 #ifdef CONFIG_COMPAT 1391 .compat_setsockopt = compat_udp_setsockopt, 1392 .compat_getsockopt = compat_udp_getsockopt, 1393 #endif 1394 }; 1395 1396 /* ------------------------------------------------------------------------ */ 1397 #ifdef CONFIG_PROC_FS 1398 1399 static struct sock *udp_get_first(struct seq_file *seq) 1400 { 1401 struct sock *sk; 1402 struct udp_iter_state *state = seq->private; 1403 1404 for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) { 1405 struct hlist_node *node; 1406 sk_for_each(sk, node, &udp_hash[state->bucket]) { 1407 if (sk->sk_family == state->family) 1408 goto found; 1409 } 1410 } 1411 sk = NULL; 1412 found: 1413 return sk; 1414 } 1415 1416 static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) 1417 { 1418 struct udp_iter_state *state = seq->private; 1419 1420 do { 1421 sk = sk_next(sk); 1422 try_again: 1423 ; 1424 } while (sk && sk->sk_family != state->family); 1425 1426 if (!sk && ++state->bucket < UDP_HTABLE_SIZE) { 1427 sk = sk_head(&udp_hash[state->bucket]); 1428 goto try_again; 1429 } 1430 return sk; 1431 } 1432 1433 static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) 1434 { 1435 struct sock *sk = udp_get_first(seq); 1436 1437 if (sk) 1438 while(pos && (sk = udp_get_next(seq, sk)) != NULL) 1439 --pos; 1440 return pos ? NULL : sk; 1441 } 1442 1443 static void *udp_seq_start(struct seq_file *seq, loff_t *pos) 1444 { 1445 read_lock(&udp_hash_lock); 1446 return *pos ? udp_get_idx(seq, *pos-1) : (void *)1; 1447 } 1448 1449 static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1450 { 1451 struct sock *sk; 1452 1453 if (v == (void *)1) 1454 sk = udp_get_idx(seq, 0); 1455 else 1456 sk = udp_get_next(seq, v); 1457 1458 ++*pos; 1459 return sk; 1460 } 1461 1462 static void udp_seq_stop(struct seq_file *seq, void *v) 1463 { 1464 read_unlock(&udp_hash_lock); 1465 } 1466 1467 static int udp_seq_open(struct inode *inode, struct file *file) 1468 { 1469 struct udp_seq_afinfo *afinfo = PDE(inode)->data; 1470 struct seq_file *seq; 1471 int rc = -ENOMEM; 1472 struct udp_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL); 1473 1474 if (!s) 1475 goto out; 1476 memset(s, 0, sizeof(*s)); 1477 s->family = afinfo->family; 1478 s->seq_ops.start = udp_seq_start; 1479 s->seq_ops.next = udp_seq_next; 1480 s->seq_ops.show = afinfo->seq_show; 1481 s->seq_ops.stop = udp_seq_stop; 1482 1483 rc = seq_open(file, &s->seq_ops); 1484 if (rc) 1485 goto out_kfree; 1486 1487 seq = file->private_data; 1488 seq->private = s; 1489 out: 1490 return rc; 1491 out_kfree: 1492 kfree(s); 1493 goto out; 1494 } 1495 1496 /* ------------------------------------------------------------------------ */ 1497 int udp_proc_register(struct udp_seq_afinfo *afinfo) 1498 { 1499 struct proc_dir_entry *p; 1500 int rc = 0; 1501 1502 if (!afinfo) 1503 return -EINVAL; 1504 afinfo->seq_fops->owner = afinfo->owner; 1505 afinfo->seq_fops->open = udp_seq_open; 1506 afinfo->seq_fops->read = seq_read; 1507 afinfo->seq_fops->llseek = seq_lseek; 1508 afinfo->seq_fops->release = seq_release_private; 1509 1510 p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops); 1511 if (p) 1512 p->data = afinfo; 1513 else 1514 rc = -ENOMEM; 1515 return rc; 1516 } 1517 1518 void udp_proc_unregister(struct udp_seq_afinfo *afinfo) 1519 { 1520 if (!afinfo) 1521 return; 1522 proc_net_remove(afinfo->name); 1523 memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops)); 1524 } 1525 1526 /* ------------------------------------------------------------------------ */ 1527 static void udp4_format_sock(struct sock *sp, char *tmpbuf, int bucket) 1528 { 1529 struct inet_sock *inet = inet_sk(sp); 1530 unsigned int dest = inet->daddr; 1531 unsigned int src = inet->rcv_saddr; 1532 __u16 destp = ntohs(inet->dport); 1533 __u16 srcp = ntohs(inet->sport); 1534 1535 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X" 1536 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p", 1537 bucket, src, srcp, dest, destp, sp->sk_state, 1538 atomic_read(&sp->sk_wmem_alloc), 1539 atomic_read(&sp->sk_rmem_alloc), 1540 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), 1541 atomic_read(&sp->sk_refcnt), sp); 1542 } 1543 1544 static int udp4_seq_show(struct seq_file *seq, void *v) 1545 { 1546 if (v == SEQ_START_TOKEN) 1547 seq_printf(seq, "%-127s\n", 1548 " sl local_address rem_address st tx_queue " 1549 "rx_queue tr tm->when retrnsmt uid timeout " 1550 "inode"); 1551 else { 1552 char tmpbuf[129]; 1553 struct udp_iter_state *state = seq->private; 1554 1555 udp4_format_sock(v, tmpbuf, state->bucket); 1556 seq_printf(seq, "%-127s\n", tmpbuf); 1557 } 1558 return 0; 1559 } 1560 1561 /* ------------------------------------------------------------------------ */ 1562 static struct file_operations udp4_seq_fops; 1563 static struct udp_seq_afinfo udp4_seq_afinfo = { 1564 .owner = THIS_MODULE, 1565 .name = "udp", 1566 .family = AF_INET, 1567 .seq_show = udp4_seq_show, 1568 .seq_fops = &udp4_seq_fops, 1569 }; 1570 1571 int __init udp4_proc_init(void) 1572 { 1573 return udp_proc_register(&udp4_seq_afinfo); 1574 } 1575 1576 void udp4_proc_exit(void) 1577 { 1578 udp_proc_unregister(&udp4_seq_afinfo); 1579 } 1580 #endif /* CONFIG_PROC_FS */ 1581 1582 EXPORT_SYMBOL(udp_disconnect); 1583 EXPORT_SYMBOL(udp_hash); 1584 EXPORT_SYMBOL(udp_hash_lock); 1585 EXPORT_SYMBOL(udp_ioctl); 1586 EXPORT_SYMBOL(udp_port_rover); 1587 EXPORT_SYMBOL(udp_prot); 1588 EXPORT_SYMBOL(udp_sendmsg); 1589 EXPORT_SYMBOL(udp_poll); 1590 1591 #ifdef CONFIG_PROC_FS 1592 EXPORT_SYMBOL(udp_proc_register); 1593 EXPORT_SYMBOL(udp_proc_unregister); 1594 #endif 1595