blob: 8e42fbbd57616316df2692a39379398f9486ac22 [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 *
Jesper Juhl02c30a82005-05-05 16:16:16 -07008 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
11 * Alan Cox, <Alan.Cox@linux.org>
12 * Hirokazu Takahashi, <taka@valinux.co.jp>
13 *
14 * Fixes:
15 * Alan Cox : verify_area() calls
16 * Alan Cox : stopped close while in use off icmp
17 * messages. Not a fix but a botch that
18 * for udp at least is 'valid'.
19 * Alan Cox : Fixed icmp handling properly
20 * Alan Cox : Correct error for oversized datagrams
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090021 * Alan Cox : Tidied select() semantics.
22 * Alan Cox : udp_err() fixed properly, also now
Linus Torvalds1da177e2005-04-16 15:20:36 -070023 * select and read wake correctly on errors
24 * Alan Cox : udp_send verify_area moved to avoid mem leak
25 * Alan Cox : UDP can count its memory
26 * Alan Cox : send to an unknown connection causes
27 * an ECONNREFUSED off the icmp, but
28 * does NOT close.
29 * Alan Cox : Switched to new sk_buff handlers. No more backlog!
30 * Alan Cox : Using generic datagram code. Even smaller and the PEEK
31 * bug no longer crashes it.
32 * Fred Van Kempen : Net2e support for sk->broadcast.
33 * Alan Cox : Uses skb_free_datagram
34 * Alan Cox : Added get/set sockopt support.
35 * Alan Cox : Broadcasting without option set returns EACCES.
36 * Alan Cox : No wakeup calls. Instead we now use the callbacks.
37 * Alan Cox : Use ip_tos and ip_ttl
38 * Alan Cox : SNMP Mibs
39 * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support.
40 * Matt Dillon : UDP length checks.
41 * Alan Cox : Smarter af_inet used properly.
42 * Alan Cox : Use new kernel side addressing.
43 * Alan Cox : Incorrect return on truncated datagram receive.
44 * Arnt Gulbrandsen : New udp_send and stuff
45 * Alan Cox : Cache last socket
46 * Alan Cox : Route cache
47 * Jon Peatfield : Minor efficiency fix to sendto().
48 * Mike Shaver : RFC1122 checks.
49 * Alan Cox : Nonblocking error fix.
50 * Willy Konynenberg : Transparent proxying support.
51 * Mike McLagan : Routing by source
52 * David S. Miller : New socket lookup architecture.
53 * Last socket cache retained as it
54 * does have a high hit rate.
55 * Olaf Kirch : Don't linearise iovec on sendmsg.
56 * Andi Kleen : Some cleanups, cache destination entry
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090057 * for connect.
Linus Torvalds1da177e2005-04-16 15:20:36 -070058 * Vitaly E. Lavrov : Transparent proxy revived after year coma.
59 * Melvin Smith : Check msg_name not msg_namelen in sendto(),
60 * return ENOTCONN for unconnected sockets (POSIX)
61 * Janos Farkas : don't deliver multi/broadcasts to a different
62 * bound-to-device socket
63 * Hirokazu Takahashi : HW checksumming for outgoing UDP
64 * datagrams.
65 * Hirokazu Takahashi : sendfile() on UDP works now.
66 * Arnaldo C. Melo : convert /proc/net/udp to seq_file
67 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
68 * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind
69 * a single port at the same time.
70 * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support
James Chapman342f0232007-06-27 15:37:46 -070071 * James Chapman : Add L2TP encapsulation type.
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 *
73 *
74 * This program is free software; you can redistribute it and/or
75 * modify it under the terms of the GNU General Public License
76 * as published by the Free Software Foundation; either version
77 * 2 of the License, or (at your option) any later version.
78 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090079
Linus Torvalds1da177e2005-04-16 15:20:36 -070080#include <asm/system.h>
81#include <asm/uaccess.h>
82#include <asm/ioctls.h>
Hideo Aoki95766ff2007-12-31 00:29:24 -080083#include <linux/bootmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070084#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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070095#include <linux/netdevice.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070096#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070097#include <linux/skbuff.h>
98#include <linux/proc_fs.h>
99#include <linux/seq_file.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200100#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101#include <net/icmp.h>
102#include <net/route.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103#include <net/checksum.h>
104#include <net/xfrm.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800105#include "udp_impl.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106
107/*
108 * Snmp MIB for the UDP layer
109 */
110
Herbert Xu9055e052007-12-14 11:25:26 -0800111DEFINE_SNMP_STAT(struct udp_mib, udp_stats_in6) __read_mostly;
112EXPORT_SYMBOL(udp_stats_in6);
113
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114struct hlist_head udp_hash[UDP_HTABLE_SIZE];
115DEFINE_RWLOCK(udp_hash_lock);
116
Hideo Aoki95766ff2007-12-31 00:29:24 -0800117int sysctl_udp_mem[3] __read_mostly;
118int sysctl_udp_rmem_min __read_mostly;
119int sysctl_udp_wmem_min __read_mostly;
120
121EXPORT_SYMBOL(sysctl_udp_mem);
122EXPORT_SYMBOL(sysctl_udp_rmem_min);
123EXPORT_SYMBOL(sysctl_udp_wmem_min);
124
125atomic_t udp_memory_allocated;
126EXPORT_SYMBOL(udp_memory_allocated);
127
Pavel Emelyanovfa4d3c62008-01-31 05:07:57 -0800128static inline int __udp_lib_lport_inuse(struct net *net, __u16 num,
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700129 const struct hlist_head udptable[])
Gerrit Renker25030a72006-08-26 20:06:05 -0700130{
131 struct sock *sk;
132 struct hlist_node *node;
133
Pavel Emelyanov19c75782008-06-16 17:12:29 -0700134 sk_for_each(sk, node, &udptable[udp_hashfn(net, num)])
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900135 if (net_eq(sock_net(sk), net) && sk->sk_hash == num)
Gerrit Renker25030a72006-08-26 20:06:05 -0700136 return 1;
137 return 0;
138}
139
140/**
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700141 * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
Gerrit Renker25030a72006-08-26 20:06:05 -0700142 *
143 * @sk: socket struct in question
144 * @snum: port number to look up
David S. Millerdf2bc452007-06-05 15:18:43 -0700145 * @saddr_comp: AF-dependent comparison of bound local IP addresses
Gerrit Renker25030a72006-08-26 20:06:05 -0700146 */
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700147int udp_lib_get_port(struct sock *sk, unsigned short snum,
David S. Millerdf2bc452007-06-05 15:18:43 -0700148 int (*saddr_comp)(const struct sock *sk1,
149 const struct sock *sk2 ) )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150{
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700151 struct hlist_head *udptable = sk->sk_prot->h.udp_hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 struct hlist_node *node;
Gerrit Renker25030a72006-08-26 20:06:05 -0700153 struct hlist_head *head;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154 struct sock *sk2;
Gerrit Renker25030a72006-08-26 20:06:05 -0700155 int error = 1;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900156 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157
158 write_lock_bh(&udp_hash_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700160 if (!snum) {
Anton Arapova25de532007-10-18 22:00:17 -0700161 int i, low, high, remaining;
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700162 unsigned rover, best, best_size_so_far;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700164 inet_get_local_port_range(&low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700165 remaining = (high - low) + 1;
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700166
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700167 best_size_so_far = UINT_MAX;
Anton Arapova25de532007-10-18 22:00:17 -0700168 best = rover = net_random() % remaining + low;
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700169
170 /* 1st pass: look for empty (or shortest) hash chain */
171 for (i = 0; i < UDP_HTABLE_SIZE; i++) {
172 int size = 0;
173
Pavel Emelyanov19c75782008-06-16 17:12:29 -0700174 head = &udptable[udp_hashfn(net, rover)];
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700175 if (hlist_empty(head))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 goto gotit;
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700177
David S. Miller5c668702006-12-22 11:42:26 -0800178 sk_for_each(sk2, node, head) {
179 if (++size >= best_size_so_far)
180 goto next;
181 }
182 best_size_so_far = size;
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700183 best = rover;
David S. Miller5c668702006-12-22 11:42:26 -0800184 next:
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700185 /* fold back if end of range */
186 if (++rover > high)
187 rover = low + ((rover - low)
188 & (UDP_HTABLE_SIZE - 1));
189
190
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 }
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700192
193 /* 2nd pass: find hole in shortest hash chain */
194 rover = best;
195 for (i = 0; i < (1 << 16) / UDP_HTABLE_SIZE; i++) {
Pavel Emelyanovfa4d3c62008-01-31 05:07:57 -0800196 if (! __udp_lib_lport_inuse(net, rover, udptable))
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700197 goto gotit;
198 rover += UDP_HTABLE_SIZE;
199 if (rover > high)
200 rover = low + ((rover - low)
201 & (UDP_HTABLE_SIZE - 1));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 }
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700203
204
205 /* All ports in use! */
206 goto fail;
207
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208gotit:
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700209 snum = rover;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 } else {
Pavel Emelyanov19c75782008-06-16 17:12:29 -0700211 head = &udptable[udp_hashfn(net, snum)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212
Gerrit Renker25030a72006-08-26 20:06:05 -0700213 sk_for_each(sk2, node, head)
David S. Millerdf2bc452007-06-05 15:18:43 -0700214 if (sk2->sk_hash == snum &&
215 sk2 != sk &&
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900216 net_eq(sock_net(sk2), net) &&
David S. Millerdf2bc452007-06-05 15:18:43 -0700217 (!sk2->sk_reuse || !sk->sk_reuse) &&
218 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if
219 || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
220 (*saddr_comp)(sk, sk2) )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 }
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700223
Gerrit Renker25030a72006-08-26 20:06:05 -0700224 inet_sk(sk)->num = snum;
David S. Millerdf2bc452007-06-05 15:18:43 -0700225 sk->sk_hash = snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 if (sk_unhashed(sk)) {
Pavel Emelyanov19c75782008-06-16 17:12:29 -0700227 head = &udptable[udp_hashfn(net, snum)];
Gerrit Renker25030a72006-08-26 20:06:05 -0700228 sk_add_node(sk, head);
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700229 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700231 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232fail:
233 write_unlock_bh(&udp_hash_lock);
Gerrit Renker25030a72006-08-26 20:06:05 -0700234 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235}
236
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700237static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
David S. Millerdb8dac22008-03-06 16:22:02 -0800238{
239 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
240
241 return ( !ipv6_only_sock(sk2) &&
242 (!inet1->rcv_saddr || !inet2->rcv_saddr ||
243 inet1->rcv_saddr == inet2->rcv_saddr ));
244}
245
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700246int udp_v4_get_port(struct sock *sk, unsigned short snum)
David S. Millerdb8dac22008-03-06 16:22:02 -0800247{
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700248 return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal);
David S. Millerdb8dac22008-03-06 16:22:02 -0800249}
250
251/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
252 * harder than this. -DaveM
253 */
254static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
255 __be16 sport, __be32 daddr, __be16 dport,
256 int dif, struct hlist_head udptable[])
257{
258 struct sock *sk, *result = NULL;
259 struct hlist_node *node;
260 unsigned short hnum = ntohs(dport);
261 int badness = -1;
262
263 read_lock(&udp_hash_lock);
Pavel Emelyanov19c75782008-06-16 17:12:29 -0700264 sk_for_each(sk, node, &udptable[udp_hashfn(net, hnum)]) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800265 struct inet_sock *inet = inet_sk(sk);
266
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900267 if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum &&
David S. Millerdb8dac22008-03-06 16:22:02 -0800268 !ipv6_only_sock(sk)) {
269 int score = (sk->sk_family == PF_INET ? 1 : 0);
270 if (inet->rcv_saddr) {
271 if (inet->rcv_saddr != daddr)
272 continue;
273 score+=2;
274 }
275 if (inet->daddr) {
276 if (inet->daddr != saddr)
277 continue;
278 score+=2;
279 }
280 if (inet->dport) {
281 if (inet->dport != sport)
282 continue;
283 score+=2;
284 }
285 if (sk->sk_bound_dev_if) {
286 if (sk->sk_bound_dev_if != dif)
287 continue;
288 score+=2;
289 }
290 if (score == 9) {
291 result = sk;
292 break;
293 } else if (score > badness) {
294 result = sk;
295 badness = score;
296 }
297 }
298 }
299 if (result)
300 sock_hold(result);
301 read_unlock(&udp_hash_lock);
302 return result;
303}
304
305static inline struct sock *udp_v4_mcast_next(struct sock *sk,
306 __be16 loc_port, __be32 loc_addr,
307 __be16 rmt_port, __be32 rmt_addr,
308 int dif)
309{
310 struct hlist_node *node;
311 struct sock *s = sk;
312 unsigned short hnum = ntohs(loc_port);
313
314 sk_for_each_from(s, node) {
315 struct inet_sock *inet = inet_sk(s);
316
317 if (s->sk_hash != hnum ||
318 (inet->daddr && inet->daddr != rmt_addr) ||
319 (inet->dport != rmt_port && inet->dport) ||
320 (inet->rcv_saddr && inet->rcv_saddr != loc_addr) ||
321 ipv6_only_sock(s) ||
322 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
323 continue;
324 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
325 continue;
326 goto found;
327 }
328 s = NULL;
329found:
330 return s;
331}
332
333/*
334 * This routine is called by the ICMP module when it gets some
335 * sort of error condition. If err < 0 then the socket should
336 * be closed and the error returned to the user. If err > 0
337 * it's just the icmp type << 8 | icmp code.
338 * Header points to the ip header of the error packet. We move
339 * on past this. Then (as it used to claim before adjustment)
340 * header points to the first 8 bytes of the udp header. We need
341 * to find the appropriate port.
342 */
343
344void __udp4_lib_err(struct sk_buff *skb, u32 info, struct hlist_head udptable[])
345{
346 struct inet_sock *inet;
347 struct iphdr *iph = (struct iphdr*)skb->data;
348 struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2));
349 const int type = icmp_hdr(skb)->type;
350 const int code = icmp_hdr(skb)->code;
351 struct sock *sk;
352 int harderr;
353 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700354 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800355
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700356 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
David S. Millerdb8dac22008-03-06 16:22:02 -0800357 iph->saddr, uh->source, skb->dev->ifindex, udptable);
358 if (sk == NULL) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700359 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800360 return; /* No socket for error */
361 }
362
363 err = 0;
364 harderr = 0;
365 inet = inet_sk(sk);
366
367 switch (type) {
368 default:
369 case ICMP_TIME_EXCEEDED:
370 err = EHOSTUNREACH;
371 break;
372 case ICMP_SOURCE_QUENCH:
373 goto out;
374 case ICMP_PARAMETERPROB:
375 err = EPROTO;
376 harderr = 1;
377 break;
378 case ICMP_DEST_UNREACH:
379 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
380 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
381 err = EMSGSIZE;
382 harderr = 1;
383 break;
384 }
385 goto out;
386 }
387 err = EHOSTUNREACH;
388 if (code <= NR_ICMP_UNREACH) {
389 harderr = icmp_err_convert[code].fatal;
390 err = icmp_err_convert[code].errno;
391 }
392 break;
393 }
394
395 /*
396 * RFC1122: OK. Passes ICMP errors back to application, as per
397 * 4.1.3.3.
398 */
399 if (!inet->recverr) {
400 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
401 goto out;
402 } else {
403 ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1));
404 }
405 sk->sk_err = err;
406 sk->sk_error_report(sk);
407out:
408 sock_put(sk);
409}
410
411void udp_err(struct sk_buff *skb, u32 info)
412{
413 __udp4_lib_err(skb, info, udp_hash);
414}
415
416/*
417 * Throw away all pending data and cancel the corking. Socket is locked.
418 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400419void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800420{
421 struct udp_sock *up = udp_sk(sk);
422
423 if (up->pending) {
424 up->len = 0;
425 up->pending = 0;
426 ip_flush_pending_frames(sk);
427 }
428}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400429EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800430
431/**
432 * udp4_hwcsum_outgoing - handle outgoing HW checksumming
433 * @sk: socket we are sending on
434 * @skb: sk_buff containing the filled-in UDP header
435 * (checksum field must be zeroed out)
436 */
437static void udp4_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
438 __be32 src, __be32 dst, int len )
439{
440 unsigned int offset;
441 struct udphdr *uh = udp_hdr(skb);
442 __wsum csum = 0;
443
444 if (skb_queue_len(&sk->sk_write_queue) == 1) {
445 /*
446 * Only one fragment on the socket.
447 */
448 skb->csum_start = skb_transport_header(skb) - skb->head;
449 skb->csum_offset = offsetof(struct udphdr, check);
450 uh->check = ~csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, 0);
451 } else {
452 /*
453 * HW-checksum won't work as there are two or more
454 * fragments on the socket so that all csums of sk_buffs
455 * should be together
456 */
457 offset = skb_transport_offset(skb);
458 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
459
460 skb->ip_summed = CHECKSUM_NONE;
461
462 skb_queue_walk(&sk->sk_write_queue, skb) {
463 csum = csum_add(csum, skb->csum);
464 }
465
466 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
467 if (uh->check == 0)
468 uh->check = CSUM_MANGLED_0;
469 }
470}
471
472/*
473 * Push out all pending data as one UDP datagram. Socket is locked.
474 */
475static int udp_push_pending_frames(struct sock *sk)
476{
477 struct udp_sock *up = udp_sk(sk);
478 struct inet_sock *inet = inet_sk(sk);
479 struct flowi *fl = &inet->cork.fl;
480 struct sk_buff *skb;
481 struct udphdr *uh;
482 int err = 0;
483 int is_udplite = IS_UDPLITE(sk);
484 __wsum csum = 0;
485
486 /* Grab the skbuff where UDP header space exists. */
487 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
488 goto out;
489
490 /*
491 * Create a UDP header
492 */
493 uh = udp_hdr(skb);
494 uh->source = fl->fl_ip_sport;
495 uh->dest = fl->fl_ip_dport;
496 uh->len = htons(up->len);
497 uh->check = 0;
498
499 if (is_udplite) /* UDP-Lite */
500 csum = udplite_csum_outgoing(sk, skb);
501
502 else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */
503
504 skb->ip_summed = CHECKSUM_NONE;
505 goto send;
506
507 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
508
509 udp4_hwcsum_outgoing(sk, skb, fl->fl4_src,fl->fl4_dst, up->len);
510 goto send;
511
512 } else /* `normal' UDP */
513 csum = udp_csum_outgoing(sk, skb);
514
515 /* add protocol-dependent pseudo-header */
516 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, up->len,
517 sk->sk_protocol, csum );
518 if (uh->check == 0)
519 uh->check = CSUM_MANGLED_0;
520
521send:
522 err = ip_push_pending_frames(sk);
523out:
524 up->len = 0;
525 up->pending = 0;
526 if (!err)
Pavel Emelyanov629ca232008-07-05 21:18:07 -0700527 UDP_INC_STATS_USER(sock_net(sk),
528 UDP_MIB_OUTDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -0800529 return err;
530}
531
532int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
533 size_t len)
534{
535 struct inet_sock *inet = inet_sk(sk);
536 struct udp_sock *up = udp_sk(sk);
537 int ulen = len;
538 struct ipcm_cookie ipc;
539 struct rtable *rt = NULL;
540 int free = 0;
541 int connected = 0;
542 __be32 daddr, faddr, saddr;
543 __be16 dport;
544 u8 tos;
545 int err, is_udplite = IS_UDPLITE(sk);
546 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
547 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
548
549 if (len > 0xFFFF)
550 return -EMSGSIZE;
551
552 /*
553 * Check the flags.
554 */
555
556 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */
557 return -EOPNOTSUPP;
558
559 ipc.opt = NULL;
560
561 if (up->pending) {
562 /*
563 * There are pending frames.
564 * The socket lock must be held while it's corked.
565 */
566 lock_sock(sk);
567 if (likely(up->pending)) {
568 if (unlikely(up->pending != AF_INET)) {
569 release_sock(sk);
570 return -EINVAL;
571 }
572 goto do_append_data;
573 }
574 release_sock(sk);
575 }
576 ulen += sizeof(struct udphdr);
577
578 /*
579 * Get and verify the address.
580 */
581 if (msg->msg_name) {
582 struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name;
583 if (msg->msg_namelen < sizeof(*usin))
584 return -EINVAL;
585 if (usin->sin_family != AF_INET) {
586 if (usin->sin_family != AF_UNSPEC)
587 return -EAFNOSUPPORT;
588 }
589
590 daddr = usin->sin_addr.s_addr;
591 dport = usin->sin_port;
592 if (dport == 0)
593 return -EINVAL;
594 } else {
595 if (sk->sk_state != TCP_ESTABLISHED)
596 return -EDESTADDRREQ;
597 daddr = inet->daddr;
598 dport = inet->dport;
599 /* Open fast path for connected socket.
600 Route will not be used, if at least one option is set.
601 */
602 connected = 1;
603 }
604 ipc.addr = inet->saddr;
605
606 ipc.oif = sk->sk_bound_dev_if;
607 if (msg->msg_controllen) {
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900608 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
David S. Millerdb8dac22008-03-06 16:22:02 -0800609 if (err)
610 return err;
611 if (ipc.opt)
612 free = 1;
613 connected = 0;
614 }
615 if (!ipc.opt)
616 ipc.opt = inet->opt;
617
618 saddr = ipc.addr;
619 ipc.addr = faddr = daddr;
620
621 if (ipc.opt && ipc.opt->srr) {
622 if (!daddr)
623 return -EINVAL;
624 faddr = ipc.opt->faddr;
625 connected = 0;
626 }
627 tos = RT_TOS(inet->tos);
628 if (sock_flag(sk, SOCK_LOCALROUTE) ||
629 (msg->msg_flags & MSG_DONTROUTE) ||
630 (ipc.opt && ipc.opt->is_strictroute)) {
631 tos |= RTO_ONLINK;
632 connected = 0;
633 }
634
635 if (ipv4_is_multicast(daddr)) {
636 if (!ipc.oif)
637 ipc.oif = inet->mc_index;
638 if (!saddr)
639 saddr = inet->mc_addr;
640 connected = 0;
641 }
642
643 if (connected)
644 rt = (struct rtable*)sk_dst_check(sk, 0);
645
646 if (rt == NULL) {
647 struct flowi fl = { .oif = ipc.oif,
648 .nl_u = { .ip4_u =
649 { .daddr = faddr,
650 .saddr = saddr,
651 .tos = tos } },
652 .proto = sk->sk_protocol,
653 .uli_u = { .ports =
654 { .sport = inet->sport,
655 .dport = dport } } };
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -0700656 struct net *net = sock_net(sk);
657
David S. Millerdb8dac22008-03-06 16:22:02 -0800658 security_sk_classify_flow(sk, &fl);
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -0700659 err = ip_route_output_flow(net, &rt, &fl, sk, 1);
David S. Millerdb8dac22008-03-06 16:22:02 -0800660 if (err) {
661 if (err == -ENETUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700662 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -0800663 goto out;
664 }
665
666 err = -EACCES;
667 if ((rt->rt_flags & RTCF_BROADCAST) &&
668 !sock_flag(sk, SOCK_BROADCAST))
669 goto out;
670 if (connected)
671 sk_dst_set(sk, dst_clone(&rt->u.dst));
672 }
673
674 if (msg->msg_flags&MSG_CONFIRM)
675 goto do_confirm;
676back_from_confirm:
677
678 saddr = rt->rt_src;
679 if (!ipc.addr)
680 daddr = ipc.addr = rt->rt_dst;
681
682 lock_sock(sk);
683 if (unlikely(up->pending)) {
684 /* The socket is already corked while preparing it. */
685 /* ... which is an evident application bug. --ANK */
686 release_sock(sk);
687
688 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
689 err = -EINVAL;
690 goto out;
691 }
692 /*
693 * Now cork the socket to pend data.
694 */
695 inet->cork.fl.fl4_dst = daddr;
696 inet->cork.fl.fl_ip_dport = dport;
697 inet->cork.fl.fl4_src = saddr;
698 inet->cork.fl.fl_ip_sport = inet->sport;
699 up->pending = AF_INET;
700
701do_append_data:
702 up->len += ulen;
703 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
704 err = ip_append_data(sk, getfrag, msg->msg_iov, ulen,
705 sizeof(struct udphdr), &ipc, rt,
706 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
707 if (err)
708 udp_flush_pending_frames(sk);
709 else if (!corkreq)
710 err = udp_push_pending_frames(sk);
711 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
712 up->pending = 0;
713 release_sock(sk);
714
715out:
716 ip_rt_put(rt);
717 if (free)
718 kfree(ipc.opt);
719 if (!err)
720 return len;
721 /*
722 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
723 * ENOBUFS might not be good (it's not tunable per se), but otherwise
724 * we don't have a good statistic (IpOutDiscards but it can be too many
725 * things). We could add another new stat but at least for now that
726 * seems like overkill.
727 */
728 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Pavel Emelyanov629ca232008-07-05 21:18:07 -0700729 UDP_INC_STATS_USER(sock_net(sk),
730 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -0800731 }
732 return err;
733
734do_confirm:
735 dst_confirm(&rt->u.dst);
736 if (!(msg->msg_flags&MSG_PROBE) || len)
737 goto back_from_confirm;
738 err = 0;
739 goto out;
740}
741
742int udp_sendpage(struct sock *sk, struct page *page, int offset,
743 size_t size, int flags)
744{
745 struct udp_sock *up = udp_sk(sk);
746 int ret;
747
748 if (!up->pending) {
749 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
750
751 /* Call udp_sendmsg to specify destination address which
752 * sendpage interface can't pass.
753 * This will succeed only when the socket is connected.
754 */
755 ret = udp_sendmsg(NULL, sk, &msg, 0);
756 if (ret < 0)
757 return ret;
758 }
759
760 lock_sock(sk);
761
762 if (unlikely(!up->pending)) {
763 release_sock(sk);
764
765 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n");
766 return -EINVAL;
767 }
768
769 ret = ip_append_page(sk, page, offset, size, flags);
770 if (ret == -EOPNOTSUPP) {
771 release_sock(sk);
772 return sock_no_sendpage(sk->sk_socket, page, offset,
773 size, flags);
774 }
775 if (ret < 0) {
776 udp_flush_pending_frames(sk);
777 goto out;
778 }
779
780 up->len += size;
781 if (!(up->corkflag || (flags&MSG_MORE)))
782 ret = udp_push_pending_frames(sk);
783 if (!ret)
784 ret = size;
785out:
786 release_sock(sk);
787 return ret;
788}
789
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790/*
791 * IOCTL requests applicable to the UDP protocol
792 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900793
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
795{
Stephen Hemminger6516c652007-03-08 20:41:55 -0800796 switch (cmd) {
797 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 {
Stephen Hemminger6516c652007-03-08 20:41:55 -0800799 int amount = atomic_read(&sk->sk_wmem_alloc);
800 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 }
Stephen Hemminger6516c652007-03-08 20:41:55 -0800802
803 case SIOCINQ:
804 {
805 struct sk_buff *skb;
806 unsigned long amount;
807
808 amount = 0;
809 spin_lock_bh(&sk->sk_receive_queue.lock);
810 skb = skb_peek(&sk->sk_receive_queue);
811 if (skb != NULL) {
812 /*
813 * We will only return the amount
814 * of this packet since that is all
815 * that will be read.
816 */
817 amount = skb->len - sizeof(struct udphdr);
818 }
819 spin_unlock_bh(&sk->sk_receive_queue.lock);
820 return put_user(amount, (int __user *)arg);
821 }
822
823 default:
824 return -ENOIOCTLCMD;
825 }
826
827 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828}
829
David S. Millerdb8dac22008-03-06 16:22:02 -0800830/*
831 * This should be easy, if there is something there we
832 * return it, otherwise we block.
833 */
834
835int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
836 size_t len, int noblock, int flags, int *addr_len)
837{
838 struct inet_sock *inet = inet_sk(sk);
839 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
840 struct sk_buff *skb;
841 unsigned int ulen, copied;
842 int peeked;
843 int err;
844 int is_udplite = IS_UDPLITE(sk);
845
846 /*
847 * Check any passed addresses
848 */
849 if (addr_len)
850 *addr_len=sizeof(*sin);
851
852 if (flags & MSG_ERRQUEUE)
853 return ip_recv_error(sk, msg, len);
854
855try_again:
856 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
857 &peeked, &err);
858 if (!skb)
859 goto out;
860
861 ulen = skb->len - sizeof(struct udphdr);
862 copied = len;
863 if (copied > ulen)
864 copied = ulen;
865 else if (copied < ulen)
866 msg->msg_flags |= MSG_TRUNC;
867
868 /*
869 * If checksum is needed at all, try to do it while copying the
870 * data. If the data is truncated, or if we only want a partial
871 * coverage checksum (UDP-Lite), do it before the copy.
872 */
873
874 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
875 if (udp_lib_checksum_complete(skb))
876 goto csum_copy_err;
877 }
878
879 if (skb_csum_unnecessary(skb))
880 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
881 msg->msg_iov, copied );
882 else {
883 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
884
885 if (err == -EINVAL)
886 goto csum_copy_err;
887 }
888
889 if (err)
890 goto out_free;
891
892 if (!peeked)
Pavel Emelyanov629ca232008-07-05 21:18:07 -0700893 UDP_INC_STATS_USER(sock_net(sk),
894 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -0800895
896 sock_recv_timestamp(msg, sk, skb);
897
898 /* Copy the address. */
899 if (sin)
900 {
901 sin->sin_family = AF_INET;
902 sin->sin_port = udp_hdr(skb)->source;
903 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
904 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
905 }
906 if (inet->cmsg_flags)
907 ip_cmsg_recv(msg, skb);
908
909 err = copied;
910 if (flags & MSG_TRUNC)
911 err = ulen;
912
913out_free:
914 lock_sock(sk);
915 skb_free_datagram(sk, skb);
916 release_sock(sk);
917out:
918 return err;
919
920csum_copy_err:
921 lock_sock(sk);
922 if (!skb_kill_datagram(sk, skb, flags))
Pavel Emelyanov629ca232008-07-05 21:18:07 -0700923 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -0800924 release_sock(sk);
925
926 if (noblock)
927 return -EAGAIN;
928 goto try_again;
929}
930
931
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932int udp_disconnect(struct sock *sk, int flags)
933{
934 struct inet_sock *inet = inet_sk(sk);
935 /*
936 * 1003.1g - break association.
937 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900938
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 sk->sk_state = TCP_CLOSE;
940 inet->daddr = 0;
941 inet->dport = 0;
942 sk->sk_bound_dev_if = 0;
943 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
944 inet_reset_saddr(sk);
945
946 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
947 sk->sk_prot->unhash(sk);
948 inet->sport = 0;
949 }
950 sk_dst_reset(sk);
951 return 0;
952}
953
David S. Millerdb8dac22008-03-06 16:22:02 -0800954/* returns:
955 * -1: error
956 * 0: success
957 * >0: "udp encap" protocol resubmission
958 *
959 * Note that in the success and error cases, the skb is assumed to
960 * have either been requeued or freed.
961 */
962int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
963{
964 struct udp_sock *up = udp_sk(sk);
965 int rc;
966 int is_udplite = IS_UDPLITE(sk);
967
968 /*
969 * Charge it to the socket, dropping if the queue is full.
970 */
971 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
972 goto drop;
973 nf_reset(skb);
974
975 if (up->encap_type) {
976 /*
977 * This is an encapsulation socket so pass the skb to
978 * the socket's udp_encap_rcv() hook. Otherwise, just
979 * fall through and pass this up the UDP socket.
980 * up->encap_rcv() returns the following value:
981 * =0 if skb was successfully passed to the encap
982 * handler or was discarded by it.
983 * >0 if skb should be passed on to UDP.
984 * <0 if skb should be resubmitted as proto -N
985 */
986
987 /* if we're overly short, let UDP handle it */
988 if (skb->len > sizeof(struct udphdr) &&
989 up->encap_rcv != NULL) {
990 int ret;
991
Herbert Xud97106e2008-08-09 00:35:05 -0700992 bh_unlock_sock(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -0800993 ret = (*up->encap_rcv)(sk, skb);
Herbert Xud97106e2008-08-09 00:35:05 -0700994 bh_lock_sock(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -0800995 if (ret <= 0) {
Pavel Emelyanov02833282008-07-05 21:18:48 -0700996 UDP_INC_STATS_BH(sock_net(sk),
997 UDP_MIB_INDATAGRAMS,
David S. Millerdb8dac22008-03-06 16:22:02 -0800998 is_udplite);
999 return -ret;
1000 }
1001 }
1002
1003 /* FALLTHROUGH -- it's a UDP Packet */
1004 }
1005
1006 /*
1007 * UDP-Lite specific tests, ignored on UDP sockets
1008 */
1009 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1010
1011 /*
1012 * MIB statistics other than incrementing the error count are
1013 * disabled for the following two types of errors: these depend
1014 * on the application settings, not on the functioning of the
1015 * protocol stack as such.
1016 *
1017 * RFC 3828 here recommends (sec 3.3): "There should also be a
1018 * way ... to ... at least let the receiving application block
1019 * delivery of packets with coverage values less than a value
1020 * provided by the application."
1021 */
1022 if (up->pcrlen == 0) { /* full coverage was set */
1023 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage "
1024 "%d while full coverage %d requested\n",
1025 UDP_SKB_CB(skb)->cscov, skb->len);
1026 goto drop;
1027 }
1028 /* The next case involves violating the min. coverage requested
1029 * by the receiver. This is subtle: if receiver wants x and x is
1030 * greater than the buffersize/MTU then receiver will complain
1031 * that it wants x while sender emits packets of smaller size y.
1032 * Therefore the above ...()->partial_cov statement is essential.
1033 */
1034 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
1035 LIMIT_NETDEBUG(KERN_WARNING
1036 "UDPLITE: coverage %d too small, need min %d\n",
1037 UDP_SKB_CB(skb)->cscov, up->pcrlen);
1038 goto drop;
1039 }
1040 }
1041
1042 if (sk->sk_filter) {
1043 if (udp_lib_checksum_complete(skb))
1044 goto drop;
1045 }
1046
1047 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
1048 /* Note that an ENOMEM error is charged twice */
Eric Dumazetcb61cb92008-06-17 21:04:56 -07001049 if (rc == -ENOMEM) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001050 UDP_INC_STATS_BH(sock_net(sk),
1051 UDP_MIB_RCVBUFERRORS, is_udplite);
Eric Dumazetcb61cb92008-06-17 21:04:56 -07001052 atomic_inc(&sk->sk_drops);
1053 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001054 goto drop;
1055 }
1056
1057 return 0;
1058
1059drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001060 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001061 kfree_skb(skb);
1062 return -1;
1063}
1064
1065/*
1066 * Multicasts and broadcasts go to each listener.
1067 *
1068 * Note: called only from the BH handler context,
1069 * so we don't need to lock the hashes.
1070 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001071static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001072 struct udphdr *uh,
1073 __be32 saddr, __be32 daddr,
1074 struct hlist_head udptable[])
1075{
1076 struct sock *sk;
1077 int dif;
1078
1079 read_lock(&udp_hash_lock);
Pavel Emelyanov19c75782008-06-16 17:12:29 -07001080 sk = sk_head(&udptable[udp_hashfn(net, ntohs(uh->dest))]);
David S. Millerdb8dac22008-03-06 16:22:02 -08001081 dif = skb->dev->ifindex;
1082 sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
1083 if (sk) {
1084 struct sock *sknext = NULL;
1085
1086 do {
1087 struct sk_buff *skb1 = skb;
1088
1089 sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr,
1090 uh->source, saddr, dif);
1091 if (sknext)
1092 skb1 = skb_clone(skb, GFP_ATOMIC);
1093
1094 if (skb1) {
1095 int ret = 0;
1096
Herbert Xud97106e2008-08-09 00:35:05 -07001097 bh_lock_sock(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001098 if (!sock_owned_by_user(sk))
1099 ret = udp_queue_rcv_skb(sk, skb1);
1100 else
1101 sk_add_backlog(sk, skb1);
1102 bh_unlock_sock(sk);
1103
1104 if (ret > 0)
1105 /* we should probably re-process instead
1106 * of dropping packets here. */
1107 kfree_skb(skb1);
1108 }
1109 sk = sknext;
1110 } while (sknext);
1111 } else
1112 kfree_skb(skb);
1113 read_unlock(&udp_hash_lock);
1114 return 0;
1115}
1116
1117/* Initialize UDP checksum. If exited with zero value (success),
1118 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1119 * Otherwise, csum completion requires chacksumming packet body,
1120 * including udp header and folding it to skb->csum.
1121 */
1122static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1123 int proto)
1124{
1125 const struct iphdr *iph;
1126 int err;
1127
1128 UDP_SKB_CB(skb)->partial_cov = 0;
1129 UDP_SKB_CB(skb)->cscov = skb->len;
1130
1131 if (proto == IPPROTO_UDPLITE) {
1132 err = udplite_checksum_init(skb, uh);
1133 if (err)
1134 return err;
1135 }
1136
1137 iph = ip_hdr(skb);
1138 if (uh->check == 0) {
1139 skb->ip_summed = CHECKSUM_UNNECESSARY;
1140 } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
1141 if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len,
1142 proto, skb->csum))
1143 skb->ip_summed = CHECKSUM_UNNECESSARY;
1144 }
1145 if (!skb_csum_unnecessary(skb))
1146 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
1147 skb->len, proto, 0);
1148 /* Probably, we should checksum udp header (it should be in cache
1149 * in any case) and data in tiny packets (< rx copybreak).
1150 */
1151
1152 return 0;
1153}
1154
1155/*
1156 * All we need to do is get the socket, and then do a checksum.
1157 */
1158
1159int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[],
1160 int proto)
1161{
1162 struct sock *sk;
1163 struct udphdr *uh = udp_hdr(skb);
1164 unsigned short ulen;
1165 struct rtable *rt = (struct rtable*)skb->dst;
1166 __be32 saddr = ip_hdr(skb)->saddr;
1167 __be32 daddr = ip_hdr(skb)->daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001168 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001169
1170 /*
1171 * Validate the packet.
1172 */
1173 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1174 goto drop; /* No space for header. */
1175
1176 ulen = ntohs(uh->len);
1177 if (ulen > skb->len)
1178 goto short_packet;
1179
1180 if (proto == IPPROTO_UDP) {
1181 /* UDP validates ulen. */
1182 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1183 goto short_packet;
1184 uh = udp_hdr(skb);
1185 }
1186
1187 if (udp4_csum_init(skb, uh, proto))
1188 goto csum_error;
1189
1190 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
Pavel Emelyanove3163492008-06-16 17:12:11 -07001191 return __udp4_lib_mcast_deliver(net, skb, uh,
1192 saddr, daddr, udptable);
David S. Millerdb8dac22008-03-06 16:22:02 -08001193
Pavel Emelyanove3163492008-06-16 17:12:11 -07001194 sk = __udp4_lib_lookup(net, saddr, uh->source, daddr,
David S. Millerdb8dac22008-03-06 16:22:02 -08001195 uh->dest, inet_iif(skb), udptable);
1196
1197 if (sk != NULL) {
1198 int ret = 0;
Herbert Xud97106e2008-08-09 00:35:05 -07001199 bh_lock_sock(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001200 if (!sock_owned_by_user(sk))
1201 ret = udp_queue_rcv_skb(sk, skb);
1202 else
1203 sk_add_backlog(sk, skb);
1204 bh_unlock_sock(sk);
1205 sock_put(sk);
1206
1207 /* a return value > 0 means to resubmit the input, but
1208 * it wants the return to be -protocol, or 0
1209 */
1210 if (ret > 0)
1211 return -ret;
1212 return 0;
1213 }
1214
1215 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1216 goto drop;
1217 nf_reset(skb);
1218
1219 /* No socket. Drop packet silently, if checksum is wrong */
1220 if (udp_lib_checksum_complete(skb))
1221 goto csum_error;
1222
Pavel Emelyanov02833282008-07-05 21:18:48 -07001223 UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001224 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1225
1226 /*
1227 * Hmm. We got an UDP packet to a port to which we
1228 * don't wanna listen. Ignore it.
1229 */
1230 kfree_skb(skb);
1231 return 0;
1232
1233short_packet:
YOSHIFUJI Hideakia7d632b2008-04-14 04:09:00 -07001234 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From " NIPQUAD_FMT ":%u %d/%d to " NIPQUAD_FMT ":%u\n",
David S. Millerdb8dac22008-03-06 16:22:02 -08001235 proto == IPPROTO_UDPLITE ? "-Lite" : "",
1236 NIPQUAD(saddr),
1237 ntohs(uh->source),
1238 ulen,
1239 skb->len,
1240 NIPQUAD(daddr),
1241 ntohs(uh->dest));
1242 goto drop;
1243
1244csum_error:
1245 /*
1246 * RFC1122: OK. Discards the bad packet silently (as far as
1247 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1248 */
YOSHIFUJI Hideakia7d632b2008-04-14 04:09:00 -07001249 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From " NIPQUAD_FMT ":%u to " NIPQUAD_FMT ":%u ulen %d\n",
David S. Millerdb8dac22008-03-06 16:22:02 -08001250 proto == IPPROTO_UDPLITE ? "-Lite" : "",
1251 NIPQUAD(saddr),
1252 ntohs(uh->source),
1253 NIPQUAD(daddr),
1254 ntohs(uh->dest),
1255 ulen);
1256drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001257 UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001258 kfree_skb(skb);
1259 return 0;
1260}
1261
1262int udp_rcv(struct sk_buff *skb)
1263{
1264 return __udp4_lib_rcv(skb, udp_hash, IPPROTO_UDP);
1265}
1266
Brian Haley7d06b2e2008-06-14 17:04:49 -07001267void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08001268{
1269 lock_sock(sk);
1270 udp_flush_pending_frames(sk);
1271 release_sock(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001272}
1273
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274/*
1275 * Socket option code for UDP
1276 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001277int udp_lib_setsockopt(struct sock *sk, int level, int optname,
1278 char __user *optval, int optlen,
1279 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280{
1281 struct udp_sock *up = udp_sk(sk);
1282 int val;
1283 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11001284 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285
Stephen Hemminger6516c652007-03-08 20:41:55 -08001286 if (optlen<sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 return -EINVAL;
1288
1289 if (get_user(val, (int __user *)optval))
1290 return -EFAULT;
1291
Stephen Hemminger6516c652007-03-08 20:41:55 -08001292 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 case UDP_CORK:
1294 if (val != 0) {
1295 up->corkflag = 1;
1296 } else {
1297 up->corkflag = 0;
1298 lock_sock(sk);
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001299 (*push_pending_frames)(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 release_sock(sk);
1301 }
1302 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001303
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 case UDP_ENCAP:
1305 switch (val) {
1306 case 0:
1307 case UDP_ENCAP_ESPINUDP:
1308 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07001309 up->encap_rcv = xfrm4_udp_encap_rcv;
1310 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07001311 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 up->encap_type = val;
1313 break;
1314 default:
1315 err = -ENOPROTOOPT;
1316 break;
1317 }
1318 break;
1319
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001320 /*
1321 * UDP-Lite's partial checksum coverage (RFC 3828).
1322 */
1323 /* The sender sets actual checksum coverage length via this option.
1324 * The case coverage > packet length is handled by send module. */
1325 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11001326 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001327 return -ENOPROTOOPT;
1328 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
1329 val = 8;
Gerrit Renker47112e22008-07-21 13:35:08 -07001330 else if (val > USHORT_MAX)
1331 val = USHORT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001332 up->pcslen = val;
1333 up->pcflag |= UDPLITE_SEND_CC;
1334 break;
1335
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001336 /* The receiver specifies a minimum checksum coverage value. To make
1337 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001338 * used, this again means full checksum coverage. */
1339 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11001340 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001341 return -ENOPROTOOPT;
1342 if (val != 0 && val < 8) /* Avoid silly minimal values. */
1343 val = 8;
Gerrit Renker47112e22008-07-21 13:35:08 -07001344 else if (val > USHORT_MAX)
1345 val = USHORT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001346 up->pcrlen = val;
1347 up->pcflag |= UDPLITE_RECV_CC;
1348 break;
1349
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 default:
1351 err = -ENOPROTOOPT;
1352 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001353 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354
1355 return err;
1356}
1357
David S. Millerdb8dac22008-03-06 16:22:02 -08001358int udp_setsockopt(struct sock *sk, int level, int optname,
1359 char __user *optval, int optlen)
1360{
1361 if (level == SOL_UDP || level == SOL_UDPLITE)
1362 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1363 udp_push_pending_frames);
1364 return ip_setsockopt(sk, level, optname, optval, optlen);
1365}
1366
1367#ifdef CONFIG_COMPAT
1368int compat_udp_setsockopt(struct sock *sk, int level, int optname,
1369 char __user *optval, int optlen)
1370{
1371 if (level == SOL_UDP || level == SOL_UDPLITE)
1372 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1373 udp_push_pending_frames);
1374 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
1375}
1376#endif
1377
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001378int udp_lib_getsockopt(struct sock *sk, int level, int optname,
1379 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380{
1381 struct udp_sock *up = udp_sk(sk);
1382 int val, len;
1383
Stephen Hemminger6516c652007-03-08 20:41:55 -08001384 if (get_user(len,optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 return -EFAULT;
1386
1387 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001388
Stephen Hemminger6516c652007-03-08 20:41:55 -08001389 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 return -EINVAL;
1391
Stephen Hemminger6516c652007-03-08 20:41:55 -08001392 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393 case UDP_CORK:
1394 val = up->corkflag;
1395 break;
1396
1397 case UDP_ENCAP:
1398 val = up->encap_type;
1399 break;
1400
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001401 /* The following two cannot be changed on UDP sockets, the return is
1402 * always 0 (which corresponds to the full checksum coverage of UDP). */
1403 case UDPLITE_SEND_CSCOV:
1404 val = up->pcslen;
1405 break;
1406
1407 case UDPLITE_RECV_CSCOV:
1408 val = up->pcrlen;
1409 break;
1410
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 default:
1412 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001413 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414
Stephen Hemminger6516c652007-03-08 20:41:55 -08001415 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001416 return -EFAULT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001417 if (copy_to_user(optval, &val,len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001419 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420}
1421
David S. Millerdb8dac22008-03-06 16:22:02 -08001422int udp_getsockopt(struct sock *sk, int level, int optname,
1423 char __user *optval, int __user *optlen)
1424{
1425 if (level == SOL_UDP || level == SOL_UDPLITE)
1426 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1427 return ip_getsockopt(sk, level, optname, optval, optlen);
1428}
1429
1430#ifdef CONFIG_COMPAT
1431int compat_udp_getsockopt(struct sock *sk, int level, int optname,
1432 char __user *optval, int __user *optlen)
1433{
1434 if (level == SOL_UDP || level == SOL_UDPLITE)
1435 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1436 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
1437}
1438#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439/**
1440 * udp_poll - wait for a UDP event.
1441 * @file - file struct
1442 * @sock - socket
1443 * @wait - poll table
1444 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001445 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 * blocking sockets. If application is using a blocking fd
1447 * and a packet with checksum error is in the queue;
1448 * then it could get return from select indicating data available
1449 * but then block when reading it. Add special case code
1450 * to work around these arguably broken applications.
1451 */
1452unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1453{
1454 unsigned int mask = datagram_poll(file, sock, wait);
1455 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001456 int is_lite = IS_UDPLITE(sk);
1457
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458 /* Check for false positives due to checksum errors */
1459 if ( (mask & POLLRDNORM) &&
1460 !(file->f_flags & O_NONBLOCK) &&
1461 !(sk->sk_shutdown & RCV_SHUTDOWN)){
1462 struct sk_buff_head *rcvq = &sk->sk_receive_queue;
1463 struct sk_buff *skb;
1464
Herbert Xu208d8982005-05-30 15:50:15 -07001465 spin_lock_bh(&rcvq->lock);
Herbert Xu759e5d02007-03-25 20:10:56 -07001466 while ((skb = skb_peek(rcvq)) != NULL &&
1467 udp_lib_checksum_complete(skb)) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001468 UDP_INC_STATS_BH(sock_net(sk),
1469 UDP_MIB_INERRORS, is_lite);
Herbert Xu759e5d02007-03-25 20:10:56 -07001470 __skb_unlink(skb, rcvq);
1471 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 }
Herbert Xu208d8982005-05-30 15:50:15 -07001473 spin_unlock_bh(&rcvq->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474
1475 /* nothing to see, move along */
1476 if (skb == NULL)
1477 mask &= ~(POLLIN | POLLRDNORM);
1478 }
1479
1480 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001481
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482}
1483
David S. Millerdb8dac22008-03-06 16:22:02 -08001484struct proto udp_prot = {
1485 .name = "UDP",
1486 .owner = THIS_MODULE,
1487 .close = udp_lib_close,
1488 .connect = ip4_datagram_connect,
1489 .disconnect = udp_disconnect,
1490 .ioctl = udp_ioctl,
1491 .destroy = udp_destroy_sock,
1492 .setsockopt = udp_setsockopt,
1493 .getsockopt = udp_getsockopt,
1494 .sendmsg = udp_sendmsg,
1495 .recvmsg = udp_recvmsg,
1496 .sendpage = udp_sendpage,
1497 .backlog_rcv = udp_queue_rcv_skb,
1498 .hash = udp_lib_hash,
1499 .unhash = udp_lib_unhash,
1500 .get_port = udp_v4_get_port,
1501 .memory_allocated = &udp_memory_allocated,
1502 .sysctl_mem = sysctl_udp_mem,
1503 .sysctl_wmem = &sysctl_udp_wmem_min,
1504 .sysctl_rmem = &sysctl_udp_rmem_min,
1505 .obj_size = sizeof(struct udp_sock),
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -07001506 .h.udp_hash = udp_hash,
David S. Millerdb8dac22008-03-06 16:22:02 -08001507#ifdef CONFIG_COMPAT
1508 .compat_setsockopt = compat_udp_setsockopt,
1509 .compat_getsockopt = compat_udp_getsockopt,
1510#endif
David S. Millerdb8dac22008-03-06 16:22:02 -08001511};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512
1513/* ------------------------------------------------------------------------ */
1514#ifdef CONFIG_PROC_FS
1515
1516static struct sock *udp_get_first(struct seq_file *seq)
1517{
1518 struct sock *sk;
1519 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07001520 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521
1522 for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) {
1523 struct hlist_node *node;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001524 sk_for_each(sk, node, state->hashtable + state->bucket) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09001525 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07001526 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 if (sk->sk_family == state->family)
1528 goto found;
1529 }
1530 }
1531 sk = NULL;
1532found:
1533 return sk;
1534}
1535
1536static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1537{
1538 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07001539 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540
1541 do {
1542 sk = sk_next(sk);
1543try_again:
1544 ;
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09001545 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546
1547 if (!sk && ++state->bucket < UDP_HTABLE_SIZE) {
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001548 sk = sk_head(state->hashtable + state->bucket);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 goto try_again;
1550 }
1551 return sk;
1552}
1553
1554static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
1555{
1556 struct sock *sk = udp_get_first(seq);
1557
1558 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08001559 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 --pos;
1561 return pos ? NULL : sk;
1562}
1563
1564static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08001565 __acquires(udp_hash_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566{
1567 read_lock(&udp_hash_lock);
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07001568 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569}
1570
1571static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1572{
1573 struct sock *sk;
1574
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07001575 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 sk = udp_get_idx(seq, 0);
1577 else
1578 sk = udp_get_next(seq, v);
1579
1580 ++*pos;
1581 return sk;
1582}
1583
1584static void udp_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08001585 __releases(udp_hash_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586{
1587 read_unlock(&udp_hash_lock);
1588}
1589
1590static int udp_seq_open(struct inode *inode, struct file *file)
1591{
1592 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07001593 struct udp_iter_state *s;
1594 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595
Denis V. Luneva2be75c2008-03-28 18:25:06 -07001596 err = seq_open_net(inode, file, &afinfo->seq_ops,
1597 sizeof(struct udp_iter_state));
1598 if (err < 0)
1599 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07001600
Denis V. Luneva2be75c2008-03-28 18:25:06 -07001601 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 s->family = afinfo->family;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001603 s->hashtable = afinfo->hashtable;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07001604 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07001605}
1606
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07001608int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609{
1610 struct proc_dir_entry *p;
1611 int rc = 0;
1612
Denis V. Lunev3ba94412008-03-28 18:25:32 -07001613 afinfo->seq_fops.open = udp_seq_open;
1614 afinfo->seq_fops.read = seq_read;
1615 afinfo->seq_fops.llseek = seq_lseek;
1616 afinfo->seq_fops.release = seq_release_net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617
Denis V. Lunevdda61922008-03-28 18:24:26 -07001618 afinfo->seq_ops.start = udp_seq_start;
1619 afinfo->seq_ops.next = udp_seq_next;
1620 afinfo->seq_ops.stop = udp_seq_stop;
1621
Denis V. Lunev84841c32008-05-02 04:10:08 -07001622 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1623 &afinfo->seq_fops, afinfo);
1624 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 rc = -ENOMEM;
1626 return rc;
1627}
1628
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07001629void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630{
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07001631 proc_net_remove(net, afinfo->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632}
David S. Millerdb8dac22008-03-06 16:22:02 -08001633
1634/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07001635static void udp4_format_sock(struct sock *sp, struct seq_file *f,
1636 int bucket, int *len)
David S. Millerdb8dac22008-03-06 16:22:02 -08001637{
1638 struct inet_sock *inet = inet_sk(sp);
1639 __be32 dest = inet->daddr;
1640 __be32 src = inet->rcv_saddr;
1641 __u16 destp = ntohs(inet->dport);
1642 __u16 srcp = ntohs(inet->sport);
1643
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07001644 seq_printf(f, "%4d: %08X:%04X %08X:%04X"
Eric Dumazetcb61cb92008-06-17 21:04:56 -07001645 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d%n",
David S. Millerdb8dac22008-03-06 16:22:02 -08001646 bucket, src, srcp, dest, destp, sp->sk_state,
1647 atomic_read(&sp->sk_wmem_alloc),
1648 atomic_read(&sp->sk_rmem_alloc),
1649 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07001650 atomic_read(&sp->sk_refcnt), sp,
1651 atomic_read(&sp->sk_drops), len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001652}
1653
1654int udp4_seq_show(struct seq_file *seq, void *v)
1655{
1656 if (v == SEQ_START_TOKEN)
1657 seq_printf(seq, "%-127s\n",
1658 " sl local_address rem_address st tx_queue "
1659 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07001660 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08001661 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08001662 struct udp_iter_state *state = seq->private;
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07001663 int len;
David S. Millerdb8dac22008-03-06 16:22:02 -08001664
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07001665 udp4_format_sock(v, seq, state->bucket, &len);
1666 seq_printf(seq, "%*s\n", 127 - len ,"");
David S. Millerdb8dac22008-03-06 16:22:02 -08001667 }
1668 return 0;
1669}
1670
1671/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08001672static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08001673 .name = "udp",
1674 .family = AF_INET,
1675 .hashtable = udp_hash,
Denis V. Lunev4ad96d32008-03-28 18:25:53 -07001676 .seq_fops = {
1677 .owner = THIS_MODULE,
1678 },
Denis V. Lunevdda61922008-03-28 18:24:26 -07001679 .seq_ops = {
1680 .show = udp4_seq_show,
1681 },
David S. Millerdb8dac22008-03-06 16:22:02 -08001682};
1683
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07001684static int udp4_proc_init_net(struct net *net)
1685{
1686 return udp_proc_register(net, &udp4_seq_afinfo);
1687}
1688
1689static void udp4_proc_exit_net(struct net *net)
1690{
1691 udp_proc_unregister(net, &udp4_seq_afinfo);
1692}
1693
1694static struct pernet_operations udp4_net_ops = {
1695 .init = udp4_proc_init_net,
1696 .exit = udp4_proc_exit_net,
1697};
1698
David S. Millerdb8dac22008-03-06 16:22:02 -08001699int __init udp4_proc_init(void)
1700{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07001701 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001702}
1703
1704void udp4_proc_exit(void)
1705{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07001706 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001707}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708#endif /* CONFIG_PROC_FS */
1709
Hideo Aoki95766ff2007-12-31 00:29:24 -08001710void __init udp_init(void)
1711{
1712 unsigned long limit;
1713
1714 /* Set the pressure threshold up by the same strategy of TCP. It is a
1715 * fraction of global memory that is up to 1/2 at 256 MB, decreasing
1716 * toward zero with the amount of memory, with a floor of 128 pages.
1717 */
1718 limit = min(nr_all_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
1719 limit = (limit * (nr_all_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
1720 limit = max(limit, 128UL);
1721 sysctl_udp_mem[0] = limit / 4 * 3;
1722 sysctl_udp_mem[1] = limit;
1723 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
1724
1725 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
1726 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
1727}
1728
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729EXPORT_SYMBOL(udp_disconnect);
1730EXPORT_SYMBOL(udp_hash);
1731EXPORT_SYMBOL(udp_hash_lock);
1732EXPORT_SYMBOL(udp_ioctl);
David S. Millerdb8dac22008-03-06 16:22:02 -08001733EXPORT_SYMBOL(udp_prot);
1734EXPORT_SYMBOL(udp_sendmsg);
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001735EXPORT_SYMBOL(udp_lib_getsockopt);
1736EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737EXPORT_SYMBOL(udp_poll);
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -07001738EXPORT_SYMBOL(udp_lib_get_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739
1740#ifdef CONFIG_PROC_FS
1741EXPORT_SYMBOL(udp_proc_register);
1742EXPORT_SYMBOL(udp_proc_unregister);
1743#endif