blob: 77e265d7bb8f8a7a795e552aef4214570a40a7da [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * 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>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020089#include <linux/igmp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070090#include <linux/in.h>
91#include <linux/errno.h>
92#include <linux/timer.h>
93#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094#include <linux/inet.h>
95#include <linux/ipv6.h>
96#include <linux/netdevice.h>
97#include <net/snmp.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070098#include <net/ip.h>
99#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100#include <net/protocol.h>
101#include <linux/skbuff.h>
102#include <linux/proc_fs.h>
103#include <linux/seq_file.h>
104#include <net/sock.h>
105#include <net/udp.h>
106#include <net/icmp.h>
107#include <net/route.h>
108#include <net/inet_common.h>
109#include <net/checksum.h>
110#include <net/xfrm.h>
111
112/*
113 * Snmp MIB for the UDP layer
114 */
115
Eric Dumazetba899662005-08-26 12:05:31 -0700116DEFINE_SNMP_STAT(struct udp_mib, udp_statistics) __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
118struct hlist_head udp_hash[UDP_HTABLE_SIZE];
119DEFINE_RWLOCK(udp_hash_lock);
120
David S. Millerbed53ea2006-08-26 20:06:49 -0700121static int udp_port_rover;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122
Gerrit Renker25030a72006-08-26 20:06:05 -0700123static inline int udp_lport_inuse(u16 num)
124{
125 struct sock *sk;
126 struct hlist_node *node;
127
128 sk_for_each(sk, node, &udp_hash[num & (UDP_HTABLE_SIZE - 1)])
129 if (inet_sk(sk)->num == num)
130 return 1;
131 return 0;
132}
133
134/**
135 * udp_get_port - common port lookup for IPv4 and IPv6
136 *
137 * @sk: socket struct in question
138 * @snum: port number to look up
139 * @saddr_comp: AF-dependent comparison of bound local IP addresses
140 */
141int udp_get_port(struct sock *sk, unsigned short snum,
David S. Millere3b4ead2006-08-26 20:10:15 -0700142 int (*saddr_cmp)(const struct sock *sk1, const struct sock *sk2))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143{
144 struct hlist_node *node;
Gerrit Renker25030a72006-08-26 20:06:05 -0700145 struct hlist_head *head;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 struct sock *sk2;
Gerrit Renker25030a72006-08-26 20:06:05 -0700147 int error = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148
149 write_lock_bh(&udp_hash_lock);
150 if (snum == 0) {
151 int best_size_so_far, best, result, i;
152
153 if (udp_port_rover > sysctl_local_port_range[1] ||
154 udp_port_rover < sysctl_local_port_range[0])
155 udp_port_rover = sysctl_local_port_range[0];
156 best_size_so_far = 32767;
157 best = result = udp_port_rover;
158 for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 int size;
160
Gerrit Renker25030a72006-08-26 20:06:05 -0700161 head = &udp_hash[result & (UDP_HTABLE_SIZE - 1)];
162 if (hlist_empty(head)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 if (result > sysctl_local_port_range[1])
164 result = sysctl_local_port_range[0] +
165 ((result - sysctl_local_port_range[0]) &
166 (UDP_HTABLE_SIZE - 1));
167 goto gotit;
168 }
169 size = 0;
Gerrit Renker25030a72006-08-26 20:06:05 -0700170 sk_for_each(sk2, node, head)
171 if (++size < best_size_so_far) {
172 best_size_so_far = size;
173 best = result;
174 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 }
176 result = best;
177 for(i = 0; i < (1 << 16) / UDP_HTABLE_SIZE; i++, result += UDP_HTABLE_SIZE) {
178 if (result > sysctl_local_port_range[1])
179 result = sysctl_local_port_range[0]
180 + ((result - sysctl_local_port_range[0]) &
181 (UDP_HTABLE_SIZE - 1));
182 if (!udp_lport_inuse(result))
183 break;
184 }
185 if (i >= (1 << 16) / UDP_HTABLE_SIZE)
186 goto fail;
187gotit:
188 udp_port_rover = snum = result;
189 } else {
Gerrit Renker25030a72006-08-26 20:06:05 -0700190 head = &udp_hash[snum & (UDP_HTABLE_SIZE - 1)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191
Gerrit Renker25030a72006-08-26 20:06:05 -0700192 sk_for_each(sk2, node, head)
193 if (inet_sk(sk2)->num == snum &&
194 sk2 != sk &&
195 (!sk2->sk_reuse || !sk->sk_reuse) &&
196 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if
197 || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
198 (*saddr_cmp)(sk, sk2) )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700201 inet_sk(sk)->num = snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 if (sk_unhashed(sk)) {
Gerrit Renker25030a72006-08-26 20:06:05 -0700203 head = &udp_hash[snum & (UDP_HTABLE_SIZE - 1)];
204 sk_add_node(sk, head);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 sock_prot_inc_use(sk->sk_prot);
206 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700207 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208fail:
209 write_unlock_bh(&udp_hash_lock);
Gerrit Renker25030a72006-08-26 20:06:05 -0700210 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211}
212
David S. Millere3b4ead2006-08-26 20:10:15 -0700213static inline int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
Gerrit Renker25030a72006-08-26 20:06:05 -0700214{
215 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
216
217 return ( !ipv6_only_sock(sk2) &&
218 (!inet1->rcv_saddr || !inet2->rcv_saddr ||
219 inet1->rcv_saddr == inet2->rcv_saddr ));
220}
221
222static inline int udp_v4_get_port(struct sock *sk, unsigned short snum)
223{
224 return udp_get_port(sk, snum, ipv4_rcv_saddr_equal);
225}
226
227
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228static void udp_v4_hash(struct sock *sk)
229{
230 BUG();
231}
232
233static void udp_v4_unhash(struct sock *sk)
234{
235 write_lock_bh(&udp_hash_lock);
236 if (sk_del_node_init(sk)) {
237 inet_sk(sk)->num = 0;
238 sock_prot_dec_use(sk->sk_prot);
239 }
240 write_unlock_bh(&udp_hash_lock);
241}
242
243/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
244 * harder than this. -DaveM
245 */
246static struct sock *udp_v4_lookup_longway(u32 saddr, u16 sport,
247 u32 daddr, u16 dport, int dif)
248{
249 struct sock *sk, *result = NULL;
250 struct hlist_node *node;
251 unsigned short hnum = ntohs(dport);
252 int badness = -1;
253
254 sk_for_each(sk, node, &udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]) {
255 struct inet_sock *inet = inet_sk(sk);
256
257 if (inet->num == hnum && !ipv6_only_sock(sk)) {
258 int score = (sk->sk_family == PF_INET ? 1 : 0);
259 if (inet->rcv_saddr) {
260 if (inet->rcv_saddr != daddr)
261 continue;
262 score+=2;
263 }
264 if (inet->daddr) {
265 if (inet->daddr != saddr)
266 continue;
267 score+=2;
268 }
269 if (inet->dport) {
270 if (inet->dport != sport)
271 continue;
272 score+=2;
273 }
274 if (sk->sk_bound_dev_if) {
275 if (sk->sk_bound_dev_if != dif)
276 continue;
277 score+=2;
278 }
279 if(score == 9) {
280 result = sk;
281 break;
282 } else if(score > badness) {
283 result = sk;
284 badness = score;
285 }
286 }
287 }
288 return result;
289}
290
291static __inline__ struct sock *udp_v4_lookup(u32 saddr, u16 sport,
292 u32 daddr, u16 dport, int dif)
293{
294 struct sock *sk;
295
296 read_lock(&udp_hash_lock);
297 sk = udp_v4_lookup_longway(saddr, sport, daddr, dport, dif);
298 if (sk)
299 sock_hold(sk);
300 read_unlock(&udp_hash_lock);
301 return sk;
302}
303
304static inline struct sock *udp_v4_mcast_next(struct sock *sk,
305 u16 loc_port, u32 loc_addr,
306 u16 rmt_port, u32 rmt_addr,
307 int dif)
308{
309 struct hlist_node *node;
310 struct sock *s = sk;
311 unsigned short hnum = ntohs(loc_port);
312
313 sk_for_each_from(s, node) {
314 struct inet_sock *inet = inet_sk(s);
315
316 if (inet->num != hnum ||
317 (inet->daddr && inet->daddr != rmt_addr) ||
318 (inet->dport != rmt_port && inet->dport) ||
319 (inet->rcv_saddr && inet->rcv_saddr != loc_addr) ||
320 ipv6_only_sock(s) ||
321 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
322 continue;
323 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
324 continue;
325 goto found;
326 }
327 s = NULL;
328found:
329 return s;
330}
331
332/*
333 * This routine is called by the ICMP module when it gets some
334 * sort of error condition. If err < 0 then the socket should
335 * be closed and the error returned to the user. If err > 0
336 * it's just the icmp type << 8 | icmp code.
337 * Header points to the ip header of the error packet. We move
338 * on past this. Then (as it used to claim before adjustment)
339 * header points to the first 8 bytes of the udp header. We need
340 * to find the appropriate port.
341 */
342
343void udp_err(struct sk_buff *skb, u32 info)
344{
345 struct inet_sock *inet;
346 struct iphdr *iph = (struct iphdr*)skb->data;
347 struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2));
348 int type = skb->h.icmph->type;
349 int code = skb->h.icmph->code;
350 struct sock *sk;
351 int harderr;
352 int err;
353
354 sk = udp_v4_lookup(iph->daddr, uh->dest, iph->saddr, uh->source, skb->dev->ifindex);
355 if (sk == NULL) {
356 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
357 return; /* No socket for error */
358 }
359
360 err = 0;
361 harderr = 0;
362 inet = inet_sk(sk);
363
364 switch (type) {
365 default:
366 case ICMP_TIME_EXCEEDED:
367 err = EHOSTUNREACH;
368 break;
369 case ICMP_SOURCE_QUENCH:
370 goto out;
371 case ICMP_PARAMETERPROB:
372 err = EPROTO;
373 harderr = 1;
374 break;
375 case ICMP_DEST_UNREACH:
376 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
377 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
378 err = EMSGSIZE;
379 harderr = 1;
380 break;
381 }
382 goto out;
383 }
384 err = EHOSTUNREACH;
385 if (code <= NR_ICMP_UNREACH) {
386 harderr = icmp_err_convert[code].fatal;
387 err = icmp_err_convert[code].errno;
388 }
389 break;
390 }
391
392 /*
393 * RFC1122: OK. Passes ICMP errors back to application, as per
394 * 4.1.3.3.
395 */
396 if (!inet->recverr) {
397 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
398 goto out;
399 } else {
400 ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1));
401 }
402 sk->sk_err = err;
403 sk->sk_error_report(sk);
404out:
405 sock_put(sk);
406}
407
408/*
409 * Throw away all pending data and cancel the corking. Socket is locked.
410 */
411static void udp_flush_pending_frames(struct sock *sk)
412{
413 struct udp_sock *up = udp_sk(sk);
414
415 if (up->pending) {
416 up->len = 0;
417 up->pending = 0;
418 ip_flush_pending_frames(sk);
419 }
420}
421
422/*
423 * Push out all pending data as one UDP datagram. Socket is locked.
424 */
425static int udp_push_pending_frames(struct sock *sk, struct udp_sock *up)
426{
427 struct inet_sock *inet = inet_sk(sk);
428 struct flowi *fl = &inet->cork.fl;
429 struct sk_buff *skb;
430 struct udphdr *uh;
431 int err = 0;
432
433 /* Grab the skbuff where UDP header space exists. */
434 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
435 goto out;
436
437 /*
438 * Create a UDP header
439 */
440 uh = skb->h.uh;
441 uh->source = fl->fl_ip_sport;
442 uh->dest = fl->fl_ip_dport;
443 uh->len = htons(up->len);
444 uh->check = 0;
445
446 if (sk->sk_no_check == UDP_CSUM_NOXMIT) {
447 skb->ip_summed = CHECKSUM_NONE;
448 goto send;
449 }
450
451 if (skb_queue_len(&sk->sk_write_queue) == 1) {
452 /*
453 * Only one fragment on the socket.
454 */
Patrick McHardy84fa7932006-08-29 16:44:56 -0700455 if (skb->ip_summed == CHECKSUM_PARTIAL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 skb->csum = offsetof(struct udphdr, check);
457 uh->check = ~csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
458 up->len, IPPROTO_UDP, 0);
459 } else {
460 skb->csum = csum_partial((char *)uh,
461 sizeof(struct udphdr), skb->csum);
462 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
463 up->len, IPPROTO_UDP, skb->csum);
464 if (uh->check == 0)
465 uh->check = -1;
466 }
467 } else {
468 unsigned int csum = 0;
469 /*
470 * HW-checksum won't work as there are two or more
471 * fragments on the socket so that all csums of sk_buffs
472 * should be together.
473 */
Patrick McHardy84fa7932006-08-29 16:44:56 -0700474 if (skb->ip_summed == CHECKSUM_PARTIAL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 int offset = (unsigned char *)uh - skb->data;
476 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
477
478 skb->ip_summed = CHECKSUM_NONE;
479 } else {
480 skb->csum = csum_partial((char *)uh,
481 sizeof(struct udphdr), skb->csum);
482 }
483
484 skb_queue_walk(&sk->sk_write_queue, skb) {
485 csum = csum_add(csum, skb->csum);
486 }
487 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
488 up->len, IPPROTO_UDP, csum);
489 if (uh->check == 0)
490 uh->check = -1;
491 }
492send:
493 err = ip_push_pending_frames(sk);
494out:
495 up->len = 0;
496 up->pending = 0;
497 return err;
498}
499
500
501static unsigned short udp_check(struct udphdr *uh, int len, unsigned long saddr, unsigned long daddr, unsigned long base)
502{
503 return(csum_tcpudp_magic(saddr, daddr, len, IPPROTO_UDP, base));
504}
505
506int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
507 size_t len)
508{
509 struct inet_sock *inet = inet_sk(sk);
510 struct udp_sock *up = udp_sk(sk);
511 int ulen = len;
512 struct ipcm_cookie ipc;
513 struct rtable *rt = NULL;
514 int free = 0;
515 int connected = 0;
516 u32 daddr, faddr, saddr;
517 u16 dport;
518 u8 tos;
519 int err;
520 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
521
522 if (len > 0xFFFF)
523 return -EMSGSIZE;
524
525 /*
526 * Check the flags.
527 */
528
529 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */
530 return -EOPNOTSUPP;
531
532 ipc.opt = NULL;
533
534 if (up->pending) {
535 /*
536 * There are pending frames.
537 * The socket lock must be held while it's corked.
538 */
539 lock_sock(sk);
540 if (likely(up->pending)) {
541 if (unlikely(up->pending != AF_INET)) {
542 release_sock(sk);
543 return -EINVAL;
544 }
545 goto do_append_data;
546 }
547 release_sock(sk);
548 }
549 ulen += sizeof(struct udphdr);
550
551 /*
552 * Get and verify the address.
553 */
554 if (msg->msg_name) {
555 struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name;
556 if (msg->msg_namelen < sizeof(*usin))
557 return -EINVAL;
558 if (usin->sin_family != AF_INET) {
559 if (usin->sin_family != AF_UNSPEC)
560 return -EAFNOSUPPORT;
561 }
562
563 daddr = usin->sin_addr.s_addr;
564 dport = usin->sin_port;
565 if (dport == 0)
566 return -EINVAL;
567 } else {
568 if (sk->sk_state != TCP_ESTABLISHED)
569 return -EDESTADDRREQ;
570 daddr = inet->daddr;
571 dport = inet->dport;
572 /* Open fast path for connected socket.
573 Route will not be used, if at least one option is set.
574 */
575 connected = 1;
576 }
577 ipc.addr = inet->saddr;
578
579 ipc.oif = sk->sk_bound_dev_if;
580 if (msg->msg_controllen) {
581 err = ip_cmsg_send(msg, &ipc);
582 if (err)
583 return err;
584 if (ipc.opt)
585 free = 1;
586 connected = 0;
587 }
588 if (!ipc.opt)
589 ipc.opt = inet->opt;
590
591 saddr = ipc.addr;
592 ipc.addr = faddr = daddr;
593
594 if (ipc.opt && ipc.opt->srr) {
595 if (!daddr)
596 return -EINVAL;
597 faddr = ipc.opt->faddr;
598 connected = 0;
599 }
600 tos = RT_TOS(inet->tos);
601 if (sock_flag(sk, SOCK_LOCALROUTE) ||
602 (msg->msg_flags & MSG_DONTROUTE) ||
603 (ipc.opt && ipc.opt->is_strictroute)) {
604 tos |= RTO_ONLINK;
605 connected = 0;
606 }
607
608 if (MULTICAST(daddr)) {
609 if (!ipc.oif)
610 ipc.oif = inet->mc_index;
611 if (!saddr)
612 saddr = inet->mc_addr;
613 connected = 0;
614 }
615
616 if (connected)
617 rt = (struct rtable*)sk_dst_check(sk, 0);
618
619 if (rt == NULL) {
620 struct flowi fl = { .oif = ipc.oif,
621 .nl_u = { .ip4_u =
622 { .daddr = faddr,
623 .saddr = saddr,
624 .tos = tos } },
625 .proto = IPPROTO_UDP,
626 .uli_u = { .ports =
627 { .sport = inet->sport,
628 .dport = dport } } };
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -0700629 security_sk_classify_flow(sk, &fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
631 if (err)
632 goto out;
633
634 err = -EACCES;
635 if ((rt->rt_flags & RTCF_BROADCAST) &&
636 !sock_flag(sk, SOCK_BROADCAST))
637 goto out;
638 if (connected)
639 sk_dst_set(sk, dst_clone(&rt->u.dst));
640 }
641
642 if (msg->msg_flags&MSG_CONFIRM)
643 goto do_confirm;
644back_from_confirm:
645
646 saddr = rt->rt_src;
647 if (!ipc.addr)
648 daddr = ipc.addr = rt->rt_dst;
649
650 lock_sock(sk);
651 if (unlikely(up->pending)) {
652 /* The socket is already corked while preparing it. */
653 /* ... which is an evident application bug. --ANK */
654 release_sock(sk);
655
Patrick McHardy64ce2072005-08-09 20:50:53 -0700656 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 err = -EINVAL;
658 goto out;
659 }
660 /*
661 * Now cork the socket to pend data.
662 */
663 inet->cork.fl.fl4_dst = daddr;
664 inet->cork.fl.fl_ip_dport = dport;
665 inet->cork.fl.fl4_src = saddr;
666 inet->cork.fl.fl_ip_sport = inet->sport;
667 up->pending = AF_INET;
668
669do_append_data:
670 up->len += ulen;
671 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, ulen,
672 sizeof(struct udphdr), &ipc, rt,
673 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
674 if (err)
675 udp_flush_pending_frames(sk);
676 else if (!corkreq)
677 err = udp_push_pending_frames(sk, up);
678 release_sock(sk);
679
680out:
681 ip_rt_put(rt);
682 if (free)
683 kfree(ipc.opt);
684 if (!err) {
685 UDP_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS);
686 return len;
687 }
Martin Bligh81aa6462006-08-14 23:57:10 -0700688 /*
689 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
690 * ENOBUFS might not be good (it's not tunable per se), but otherwise
691 * we don't have a good statistic (IpOutDiscards but it can be too many
692 * things). We could add another new stat but at least for now that
693 * seems like overkill.
694 */
695 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
696 UDP_INC_STATS_USER(UDP_MIB_SNDBUFERRORS);
697 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 return err;
699
700do_confirm:
701 dst_confirm(&rt->u.dst);
702 if (!(msg->msg_flags&MSG_PROBE) || len)
703 goto back_from_confirm;
704 err = 0;
705 goto out;
706}
707
708static int udp_sendpage(struct sock *sk, struct page *page, int offset,
709 size_t size, int flags)
710{
711 struct udp_sock *up = udp_sk(sk);
712 int ret;
713
714 if (!up->pending) {
715 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
716
717 /* Call udp_sendmsg to specify destination address which
718 * sendpage interface can't pass.
719 * This will succeed only when the socket is connected.
720 */
721 ret = udp_sendmsg(NULL, sk, &msg, 0);
722 if (ret < 0)
723 return ret;
724 }
725
726 lock_sock(sk);
727
728 if (unlikely(!up->pending)) {
729 release_sock(sk);
730
Patrick McHardy64ce2072005-08-09 20:50:53 -0700731 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 return -EINVAL;
733 }
734
735 ret = ip_append_page(sk, page, offset, size, flags);
736 if (ret == -EOPNOTSUPP) {
737 release_sock(sk);
738 return sock_no_sendpage(sk->sk_socket, page, offset,
739 size, flags);
740 }
741 if (ret < 0) {
742 udp_flush_pending_frames(sk);
743 goto out;
744 }
745
746 up->len += size;
747 if (!(up->corkflag || (flags&MSG_MORE)))
748 ret = udp_push_pending_frames(sk, up);
749 if (!ret)
750 ret = size;
751out:
752 release_sock(sk);
753 return ret;
754}
755
756/*
757 * IOCTL requests applicable to the UDP protocol
758 */
759
760int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
761{
762 switch(cmd)
763 {
764 case SIOCOUTQ:
765 {
766 int amount = atomic_read(&sk->sk_wmem_alloc);
767 return put_user(amount, (int __user *)arg);
768 }
769
770 case SIOCINQ:
771 {
772 struct sk_buff *skb;
773 unsigned long amount;
774
775 amount = 0;
Herbert Xu208d8982005-05-30 15:50:15 -0700776 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 skb = skb_peek(&sk->sk_receive_queue);
778 if (skb != NULL) {
779 /*
780 * We will only return the amount
781 * of this packet since that is all
782 * that will be read.
783 */
784 amount = skb->len - sizeof(struct udphdr);
785 }
Herbert Xu208d8982005-05-30 15:50:15 -0700786 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 return put_user(amount, (int __user *)arg);
788 }
789
790 default:
791 return -ENOIOCTLCMD;
792 }
793 return(0);
794}
795
796static __inline__ int __udp_checksum_complete(struct sk_buff *skb)
797{
Herbert Xufb286bb2005-11-10 13:01:24 -0800798 return __skb_checksum_complete(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799}
800
801static __inline__ int udp_checksum_complete(struct sk_buff *skb)
802{
803 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
804 __udp_checksum_complete(skb);
805}
806
807/*
808 * This should be easy, if there is something there we
809 * return it, otherwise we block.
810 */
811
812static int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
813 size_t len, int noblock, int flags, int *addr_len)
814{
815 struct inet_sock *inet = inet_sk(sk);
816 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
817 struct sk_buff *skb;
818 int copied, err;
819
820 /*
821 * Check any passed addresses
822 */
823 if (addr_len)
824 *addr_len=sizeof(*sin);
825
826 if (flags & MSG_ERRQUEUE)
827 return ip_recv_error(sk, msg, len);
828
829try_again:
830 skb = skb_recv_datagram(sk, flags, noblock, &err);
831 if (!skb)
832 goto out;
833
834 copied = skb->len - sizeof(struct udphdr);
835 if (copied > len) {
836 copied = len;
837 msg->msg_flags |= MSG_TRUNC;
838 }
839
840 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
841 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
842 copied);
843 } else if (msg->msg_flags&MSG_TRUNC) {
844 if (__udp_checksum_complete(skb))
845 goto csum_copy_err;
846 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
847 copied);
848 } else {
849 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
850
851 if (err == -EINVAL)
852 goto csum_copy_err;
853 }
854
855 if (err)
856 goto out_free;
857
858 sock_recv_timestamp(msg, sk, skb);
859
860 /* Copy the address. */
861 if (sin)
862 {
863 sin->sin_family = AF_INET;
864 sin->sin_port = skb->h.uh->source;
865 sin->sin_addr.s_addr = skb->nh.iph->saddr;
866 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
867 }
868 if (inet->cmsg_flags)
869 ip_cmsg_recv(msg, skb);
870
871 err = copied;
872 if (flags & MSG_TRUNC)
873 err = skb->len - sizeof(struct udphdr);
874
875out_free:
876 skb_free_datagram(sk, skb);
877out:
878 return err;
879
880csum_copy_err:
881 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
882
Herbert Xu3305b802005-12-13 23:16:37 -0800883 skb_kill_datagram(sk, skb, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884
885 if (noblock)
886 return -EAGAIN;
887 goto try_again;
888}
889
890
891int udp_disconnect(struct sock *sk, int flags)
892{
893 struct inet_sock *inet = inet_sk(sk);
894 /*
895 * 1003.1g - break association.
896 */
897
898 sk->sk_state = TCP_CLOSE;
899 inet->daddr = 0;
900 inet->dport = 0;
901 sk->sk_bound_dev_if = 0;
902 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
903 inet_reset_saddr(sk);
904
905 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
906 sk->sk_prot->unhash(sk);
907 inet->sport = 0;
908 }
909 sk_dst_reset(sk);
910 return 0;
911}
912
913static void udp_close(struct sock *sk, long timeout)
914{
915 sk_common_release(sk);
916}
917
918/* return:
919 * 1 if the the UDP system should process it
920 * 0 if we should drop this packet
921 * -1 if it should get processed by xfrm4_rcv_encap
922 */
923static int udp_encap_rcv(struct sock * sk, struct sk_buff *skb)
924{
925#ifndef CONFIG_XFRM
926 return 1;
927#else
928 struct udp_sock *up = udp_sk(sk);
929 struct udphdr *uh = skb->h.uh;
930 struct iphdr *iph;
931 int iphlen, len;
932
933 __u8 *udpdata = (__u8 *)uh + sizeof(struct udphdr);
934 __u32 *udpdata32 = (__u32 *)udpdata;
935 __u16 encap_type = up->encap_type;
936
937 /* if we're overly short, let UDP handle it */
938 if (udpdata > skb->tail)
939 return 1;
940
941 /* if this is not encapsulated socket, then just return now */
942 if (!encap_type)
943 return 1;
944
945 len = skb->tail - udpdata;
946
947 switch (encap_type) {
948 default:
949 case UDP_ENCAP_ESPINUDP:
950 /* Check if this is a keepalive packet. If so, eat it. */
951 if (len == 1 && udpdata[0] == 0xff) {
952 return 0;
953 } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0 ) {
954 /* ESP Packet without Non-ESP header */
955 len = sizeof(struct udphdr);
956 } else
957 /* Must be an IKE packet.. pass it through */
958 return 1;
959 break;
960 case UDP_ENCAP_ESPINUDP_NON_IKE:
961 /* Check if this is a keepalive packet. If so, eat it. */
962 if (len == 1 && udpdata[0] == 0xff) {
963 return 0;
964 } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) &&
965 udpdata32[0] == 0 && udpdata32[1] == 0) {
966
967 /* ESP Packet with Non-IKE marker */
968 len = sizeof(struct udphdr) + 2 * sizeof(u32);
969 } else
970 /* Must be an IKE packet.. pass it through */
971 return 1;
972 break;
973 }
974
975 /* At this point we are sure that this is an ESPinUDP packet,
976 * so we need to remove 'len' bytes from the packet (the UDP
977 * header and optional ESP marker bytes) and then modify the
978 * protocol to ESP, and then call into the transform receiver.
979 */
Herbert Xu4d78b6c2005-04-19 22:48:59 -0700980 if (skb_cloned(skb) && pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
981 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
983 /* Now we can update and verify the packet length... */
984 iph = skb->nh.iph;
985 iphlen = iph->ihl << 2;
986 iph->tot_len = htons(ntohs(iph->tot_len) - len);
987 if (skb->len < iphlen + len) {
988 /* packet is too small!?! */
989 return 0;
990 }
991
992 /* pull the data buffer up to the ESP header and set the
993 * transport header to point to ESP. Keep UDP on the stack
994 * for later.
995 */
996 skb->h.raw = skb_pull(skb, len);
997
998 /* modify the protocol (it's ESP!) */
999 iph->protocol = IPPROTO_ESP;
1000
1001 /* and let the caller know to send this into the ESP processor... */
1002 return -1;
1003#endif
1004}
1005
1006/* returns:
1007 * -1: error
1008 * 0: success
1009 * >0: "udp encap" protocol resubmission
1010 *
1011 * Note that in the success and error cases, the skb is assumed to
1012 * have either been requeued or freed.
1013 */
1014static int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
1015{
1016 struct udp_sock *up = udp_sk(sk);
Martin Bligh81aa6462006-08-14 23:57:10 -07001017 int rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018
1019 /*
1020 * Charge it to the socket, dropping if the queue is full.
1021 */
1022 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
1023 kfree_skb(skb);
1024 return -1;
1025 }
Patrick McHardyb59c2702006-01-06 23:06:10 -08001026 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027
1028 if (up->encap_type) {
1029 /*
1030 * This is an encapsulation socket, so let's see if this is
1031 * an encapsulated packet.
1032 * If it's a keepalive packet, then just eat it.
1033 * If it's an encapsulateed packet, then pass it to the
1034 * IPsec xfrm input and return the response
1035 * appropriately. Otherwise, just fall through and
1036 * pass this up the UDP socket.
1037 */
1038 int ret;
1039
1040 ret = udp_encap_rcv(sk, skb);
1041 if (ret == 0) {
1042 /* Eat the packet .. */
1043 kfree_skb(skb);
1044 return 0;
1045 }
1046 if (ret < 0) {
1047 /* process the ESP packet */
1048 ret = xfrm4_rcv_encap(skb, up->encap_type);
1049 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS);
1050 return -ret;
1051 }
1052 /* FALLTHROUGH -- it's a UDP Packet */
1053 }
1054
1055 if (sk->sk_filter && skb->ip_summed != CHECKSUM_UNNECESSARY) {
1056 if (__udp_checksum_complete(skb)) {
1057 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1058 kfree_skb(skb);
1059 return -1;
1060 }
1061 skb->ip_summed = CHECKSUM_UNNECESSARY;
1062 }
1063
Martin Bligh81aa6462006-08-14 23:57:10 -07001064 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
1065 /* Note that an ENOMEM error is charged twice */
1066 if (rc == -ENOMEM)
1067 UDP_INC_STATS_BH(UDP_MIB_RCVBUFERRORS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1069 kfree_skb(skb);
1070 return -1;
1071 }
1072 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS);
1073 return 0;
1074}
1075
1076/*
1077 * Multicasts and broadcasts go to each listener.
1078 *
1079 * Note: called only from the BH handler context,
1080 * so we don't need to lock the hashes.
1081 */
1082static int udp_v4_mcast_deliver(struct sk_buff *skb, struct udphdr *uh,
1083 u32 saddr, u32 daddr)
1084{
1085 struct sock *sk;
1086 int dif;
1087
1088 read_lock(&udp_hash_lock);
1089 sk = sk_head(&udp_hash[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
1090 dif = skb->dev->ifindex;
1091 sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
1092 if (sk) {
1093 struct sock *sknext = NULL;
1094
1095 do {
1096 struct sk_buff *skb1 = skb;
1097
1098 sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr,
1099 uh->source, saddr, dif);
1100 if(sknext)
1101 skb1 = skb_clone(skb, GFP_ATOMIC);
1102
1103 if(skb1) {
1104 int ret = udp_queue_rcv_skb(sk, skb1);
1105 if (ret > 0)
1106 /* we should probably re-process instead
1107 * of dropping packets here. */
1108 kfree_skb(skb1);
1109 }
1110 sk = sknext;
1111 } while(sknext);
1112 } else
1113 kfree_skb(skb);
1114 read_unlock(&udp_hash_lock);
1115 return 0;
1116}
1117
1118/* Initialize UDP checksum. If exited with zero value (success),
1119 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1120 * Otherwise, csum completion requires chacksumming packet body,
1121 * including udp header and folding it to skb->csum.
1122 */
Stephen Hemminger65a45442005-12-13 23:17:02 -08001123static void udp_checksum_init(struct sk_buff *skb, struct udphdr *uh,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 unsigned short ulen, u32 saddr, u32 daddr)
1125{
1126 if (uh->check == 0) {
1127 skb->ip_summed = CHECKSUM_UNNECESSARY;
Patrick McHardy84fa7932006-08-29 16:44:56 -07001128 } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 if (!udp_check(uh, ulen, saddr, daddr, skb->csum))
Herbert Xufb286bb2005-11-10 13:01:24 -08001130 skb->ip_summed = CHECKSUM_UNNECESSARY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 }
1132 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
1133 skb->csum = csum_tcpudp_nofold(saddr, daddr, ulen, IPPROTO_UDP, 0);
1134 /* Probably, we should checksum udp header (it should be in cache
1135 * in any case) and data in tiny packets (< rx copybreak).
1136 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137}
1138
1139/*
1140 * All we need to do is get the socket, and then do a checksum.
1141 */
1142
1143int udp_rcv(struct sk_buff *skb)
1144{
1145 struct sock *sk;
1146 struct udphdr *uh;
1147 unsigned short ulen;
1148 struct rtable *rt = (struct rtable*)skb->dst;
1149 u32 saddr = skb->nh.iph->saddr;
1150 u32 daddr = skb->nh.iph->daddr;
1151 int len = skb->len;
1152
1153 /*
1154 * Validate the packet and the UDP length.
1155 */
1156 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1157 goto no_header;
1158
1159 uh = skb->h.uh;
1160
1161 ulen = ntohs(uh->len);
1162
1163 if (ulen > len || ulen < sizeof(*uh))
1164 goto short_packet;
1165
Stephen Hemmingere308e252005-09-08 12:32:21 -07001166 if (pskb_trim_rcsum(skb, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167 goto short_packet;
1168
Stephen Hemminger65a45442005-12-13 23:17:02 -08001169 udp_checksum_init(skb, uh, ulen, saddr, daddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170
1171 if(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1172 return udp_v4_mcast_deliver(skb, uh, saddr, daddr);
1173
1174 sk = udp_v4_lookup(saddr, uh->source, daddr, uh->dest, skb->dev->ifindex);
1175
1176 if (sk != NULL) {
1177 int ret = udp_queue_rcv_skb(sk, skb);
1178 sock_put(sk);
1179
1180 /* a return value > 0 means to resubmit the input, but
1181 * it it wants the return to be -protocol, or 0
1182 */
1183 if (ret > 0)
1184 return -ret;
1185 return 0;
1186 }
1187
1188 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1189 goto drop;
Patrick McHardyb59c2702006-01-06 23:06:10 -08001190 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191
1192 /* No socket. Drop packet silently, if checksum is wrong */
1193 if (udp_checksum_complete(skb))
1194 goto csum_error;
1195
1196 UDP_INC_STATS_BH(UDP_MIB_NOPORTS);
1197 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1198
1199 /*
1200 * Hmm. We got an UDP packet to a port to which we
1201 * don't wanna listen. Ignore it.
1202 */
1203 kfree_skb(skb);
1204 return(0);
1205
1206short_packet:
Patrick McHardy64ce2072005-08-09 20:50:53 -07001207 LIMIT_NETDEBUG(KERN_DEBUG "UDP: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
1208 NIPQUAD(saddr),
1209 ntohs(uh->source),
1210 ulen,
1211 len,
1212 NIPQUAD(daddr),
1213 ntohs(uh->dest));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214no_header:
1215 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1216 kfree_skb(skb);
1217 return(0);
1218
1219csum_error:
1220 /*
1221 * RFC1122: OK. Discards the bad packet silently (as far as
1222 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1223 */
Patrick McHardy64ce2072005-08-09 20:50:53 -07001224 LIMIT_NETDEBUG(KERN_DEBUG "UDP: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
1225 NIPQUAD(saddr),
1226 ntohs(uh->source),
1227 NIPQUAD(daddr),
1228 ntohs(uh->dest),
1229 ulen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230drop:
1231 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1232 kfree_skb(skb);
1233 return(0);
1234}
1235
1236static int udp_destroy_sock(struct sock *sk)
1237{
1238 lock_sock(sk);
1239 udp_flush_pending_frames(sk);
1240 release_sock(sk);
1241 return 0;
1242}
1243
1244/*
1245 * Socket option code for UDP
1246 */
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001247static int do_udp_setsockopt(struct sock *sk, int level, int optname,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 char __user *optval, int optlen)
1249{
1250 struct udp_sock *up = udp_sk(sk);
1251 int val;
1252 int err = 0;
1253
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 if(optlen<sizeof(int))
1255 return -EINVAL;
1256
1257 if (get_user(val, (int __user *)optval))
1258 return -EFAULT;
1259
1260 switch(optname) {
1261 case UDP_CORK:
1262 if (val != 0) {
1263 up->corkflag = 1;
1264 } else {
1265 up->corkflag = 0;
1266 lock_sock(sk);
1267 udp_push_pending_frames(sk, up);
1268 release_sock(sk);
1269 }
1270 break;
1271
1272 case UDP_ENCAP:
1273 switch (val) {
1274 case 0:
1275 case UDP_ENCAP_ESPINUDP:
1276 case UDP_ENCAP_ESPINUDP_NON_IKE:
1277 up->encap_type = val;
1278 break;
1279 default:
1280 err = -ENOPROTOOPT;
1281 break;
1282 }
1283 break;
1284
1285 default:
1286 err = -ENOPROTOOPT;
1287 break;
1288 };
1289
1290 return err;
1291}
1292
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001293static int udp_setsockopt(struct sock *sk, int level, int optname,
1294 char __user *optval, int optlen)
1295{
1296 if (level != SOL_UDP)
1297 return ip_setsockopt(sk, level, optname, optval, optlen);
1298 return do_udp_setsockopt(sk, level, optname, optval, optlen);
1299}
1300
1301#ifdef CONFIG_COMPAT
1302static int compat_udp_setsockopt(struct sock *sk, int level, int optname,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001303 char __user *optval, int optlen)
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001304{
1305 if (level != SOL_UDP)
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001306 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001307 return do_udp_setsockopt(sk, level, optname, optval, optlen);
1308}
1309#endif
1310
1311static int do_udp_getsockopt(struct sock *sk, int level, int optname,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 char __user *optval, int __user *optlen)
1313{
1314 struct udp_sock *up = udp_sk(sk);
1315 int val, len;
1316
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317 if(get_user(len,optlen))
1318 return -EFAULT;
1319
1320 len = min_t(unsigned int, len, sizeof(int));
1321
1322 if(len < 0)
1323 return -EINVAL;
1324
1325 switch(optname) {
1326 case UDP_CORK:
1327 val = up->corkflag;
1328 break;
1329
1330 case UDP_ENCAP:
1331 val = up->encap_type;
1332 break;
1333
1334 default:
1335 return -ENOPROTOOPT;
1336 };
1337
1338 if(put_user(len, optlen))
1339 return -EFAULT;
1340 if(copy_to_user(optval, &val,len))
1341 return -EFAULT;
1342 return 0;
1343}
1344
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001345static int udp_getsockopt(struct sock *sk, int level, int optname,
1346 char __user *optval, int __user *optlen)
1347{
1348 if (level != SOL_UDP)
1349 return ip_getsockopt(sk, level, optname, optval, optlen);
1350 return do_udp_getsockopt(sk, level, optname, optval, optlen);
1351}
1352
1353#ifdef CONFIG_COMPAT
1354static int compat_udp_getsockopt(struct sock *sk, int level, int optname,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001355 char __user *optval, int __user *optlen)
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001356{
1357 if (level != SOL_UDP)
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001358 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001359 return do_udp_getsockopt(sk, level, optname, optval, optlen);
1360}
1361#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362/**
1363 * udp_poll - wait for a UDP event.
1364 * @file - file struct
1365 * @sock - socket
1366 * @wait - poll table
1367 *
1368 * This is same as datagram poll, except for the special case of
1369 * blocking sockets. If application is using a blocking fd
1370 * and a packet with checksum error is in the queue;
1371 * then it could get return from select indicating data available
1372 * but then block when reading it. Add special case code
1373 * to work around these arguably broken applications.
1374 */
1375unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1376{
1377 unsigned int mask = datagram_poll(file, sock, wait);
1378 struct sock *sk = sock->sk;
1379
1380 /* Check for false positives due to checksum errors */
1381 if ( (mask & POLLRDNORM) &&
1382 !(file->f_flags & O_NONBLOCK) &&
1383 !(sk->sk_shutdown & RCV_SHUTDOWN)){
1384 struct sk_buff_head *rcvq = &sk->sk_receive_queue;
1385 struct sk_buff *skb;
1386
Herbert Xu208d8982005-05-30 15:50:15 -07001387 spin_lock_bh(&rcvq->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 while ((skb = skb_peek(rcvq)) != NULL) {
1389 if (udp_checksum_complete(skb)) {
1390 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1391 __skb_unlink(skb, rcvq);
1392 kfree_skb(skb);
1393 } else {
1394 skb->ip_summed = CHECKSUM_UNNECESSARY;
1395 break;
1396 }
1397 }
Herbert Xu208d8982005-05-30 15:50:15 -07001398 spin_unlock_bh(&rcvq->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399
1400 /* nothing to see, move along */
1401 if (skb == NULL)
1402 mask &= ~(POLLIN | POLLRDNORM);
1403 }
1404
1405 return mask;
1406
1407}
1408
1409struct proto udp_prot = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001410 .name = "UDP",
1411 .owner = THIS_MODULE,
1412 .close = udp_close,
1413 .connect = ip4_datagram_connect,
1414 .disconnect = udp_disconnect,
1415 .ioctl = udp_ioctl,
1416 .destroy = udp_destroy_sock,
1417 .setsockopt = udp_setsockopt,
1418 .getsockopt = udp_getsockopt,
1419 .sendmsg = udp_sendmsg,
1420 .recvmsg = udp_recvmsg,
1421 .sendpage = udp_sendpage,
1422 .backlog_rcv = udp_queue_rcv_skb,
1423 .hash = udp_v4_hash,
1424 .unhash = udp_v4_unhash,
1425 .get_port = udp_v4_get_port,
1426 .obj_size = sizeof(struct udp_sock),
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001427#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001428 .compat_setsockopt = compat_udp_setsockopt,
1429 .compat_getsockopt = compat_udp_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001430#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431};
1432
1433/* ------------------------------------------------------------------------ */
1434#ifdef CONFIG_PROC_FS
1435
1436static struct sock *udp_get_first(struct seq_file *seq)
1437{
1438 struct sock *sk;
1439 struct udp_iter_state *state = seq->private;
1440
1441 for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) {
1442 struct hlist_node *node;
1443 sk_for_each(sk, node, &udp_hash[state->bucket]) {
1444 if (sk->sk_family == state->family)
1445 goto found;
1446 }
1447 }
1448 sk = NULL;
1449found:
1450 return sk;
1451}
1452
1453static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1454{
1455 struct udp_iter_state *state = seq->private;
1456
1457 do {
1458 sk = sk_next(sk);
1459try_again:
1460 ;
1461 } while (sk && sk->sk_family != state->family);
1462
1463 if (!sk && ++state->bucket < UDP_HTABLE_SIZE) {
1464 sk = sk_head(&udp_hash[state->bucket]);
1465 goto try_again;
1466 }
1467 return sk;
1468}
1469
1470static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
1471{
1472 struct sock *sk = udp_get_first(seq);
1473
1474 if (sk)
1475 while(pos && (sk = udp_get_next(seq, sk)) != NULL)
1476 --pos;
1477 return pos ? NULL : sk;
1478}
1479
1480static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
1481{
1482 read_lock(&udp_hash_lock);
1483 return *pos ? udp_get_idx(seq, *pos-1) : (void *)1;
1484}
1485
1486static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1487{
1488 struct sock *sk;
1489
1490 if (v == (void *)1)
1491 sk = udp_get_idx(seq, 0);
1492 else
1493 sk = udp_get_next(seq, v);
1494
1495 ++*pos;
1496 return sk;
1497}
1498
1499static void udp_seq_stop(struct seq_file *seq, void *v)
1500{
1501 read_unlock(&udp_hash_lock);
1502}
1503
1504static int udp_seq_open(struct inode *inode, struct file *file)
1505{
1506 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
1507 struct seq_file *seq;
1508 int rc = -ENOMEM;
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07001509 struct udp_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510
1511 if (!s)
1512 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 s->family = afinfo->family;
1514 s->seq_ops.start = udp_seq_start;
1515 s->seq_ops.next = udp_seq_next;
1516 s->seq_ops.show = afinfo->seq_show;
1517 s->seq_ops.stop = udp_seq_stop;
1518
1519 rc = seq_open(file, &s->seq_ops);
1520 if (rc)
1521 goto out_kfree;
1522
1523 seq = file->private_data;
1524 seq->private = s;
1525out:
1526 return rc;
1527out_kfree:
1528 kfree(s);
1529 goto out;
1530}
1531
1532/* ------------------------------------------------------------------------ */
1533int udp_proc_register(struct udp_seq_afinfo *afinfo)
1534{
1535 struct proc_dir_entry *p;
1536 int rc = 0;
1537
1538 if (!afinfo)
1539 return -EINVAL;
1540 afinfo->seq_fops->owner = afinfo->owner;
1541 afinfo->seq_fops->open = udp_seq_open;
1542 afinfo->seq_fops->read = seq_read;
1543 afinfo->seq_fops->llseek = seq_lseek;
1544 afinfo->seq_fops->release = seq_release_private;
1545
1546 p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops);
1547 if (p)
1548 p->data = afinfo;
1549 else
1550 rc = -ENOMEM;
1551 return rc;
1552}
1553
1554void udp_proc_unregister(struct udp_seq_afinfo *afinfo)
1555{
1556 if (!afinfo)
1557 return;
1558 proc_net_remove(afinfo->name);
1559 memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops));
1560}
1561
1562/* ------------------------------------------------------------------------ */
1563static void udp4_format_sock(struct sock *sp, char *tmpbuf, int bucket)
1564{
1565 struct inet_sock *inet = inet_sk(sp);
1566 unsigned int dest = inet->daddr;
1567 unsigned int src = inet->rcv_saddr;
1568 __u16 destp = ntohs(inet->dport);
1569 __u16 srcp = ntohs(inet->sport);
1570
1571 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
1572 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
1573 bucket, src, srcp, dest, destp, sp->sk_state,
1574 atomic_read(&sp->sk_wmem_alloc),
1575 atomic_read(&sp->sk_rmem_alloc),
1576 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
1577 atomic_read(&sp->sk_refcnt), sp);
1578}
1579
1580static int udp4_seq_show(struct seq_file *seq, void *v)
1581{
1582 if (v == SEQ_START_TOKEN)
1583 seq_printf(seq, "%-127s\n",
1584 " sl local_address rem_address st tx_queue "
1585 "rx_queue tr tm->when retrnsmt uid timeout "
1586 "inode");
1587 else {
1588 char tmpbuf[129];
1589 struct udp_iter_state *state = seq->private;
1590
1591 udp4_format_sock(v, tmpbuf, state->bucket);
1592 seq_printf(seq, "%-127s\n", tmpbuf);
1593 }
1594 return 0;
1595}
1596
1597/* ------------------------------------------------------------------------ */
1598static struct file_operations udp4_seq_fops;
1599static struct udp_seq_afinfo udp4_seq_afinfo = {
1600 .owner = THIS_MODULE,
1601 .name = "udp",
1602 .family = AF_INET,
1603 .seq_show = udp4_seq_show,
1604 .seq_fops = &udp4_seq_fops,
1605};
1606
1607int __init udp4_proc_init(void)
1608{
1609 return udp_proc_register(&udp4_seq_afinfo);
1610}
1611
1612void udp4_proc_exit(void)
1613{
1614 udp_proc_unregister(&udp4_seq_afinfo);
1615}
1616#endif /* CONFIG_PROC_FS */
1617
1618EXPORT_SYMBOL(udp_disconnect);
1619EXPORT_SYMBOL(udp_hash);
1620EXPORT_SYMBOL(udp_hash_lock);
1621EXPORT_SYMBOL(udp_ioctl);
Gerrit Renker25030a72006-08-26 20:06:05 -07001622EXPORT_SYMBOL(udp_get_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623EXPORT_SYMBOL(udp_prot);
1624EXPORT_SYMBOL(udp_sendmsg);
1625EXPORT_SYMBOL(udp_poll);
1626
1627#ifdef CONFIG_PROC_FS
1628EXPORT_SYMBOL(udp_proc_register);
1629EXPORT_SYMBOL(udp_proc_unregister);
1630#endif