blob: c47c989cb1fb162b9fb71f8f27a580f4ba8088d0 [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>
Alan Cox113aa832008-10-13 19:01:08 -070011 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012 * 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>
Eric Dumazet8203efb2008-10-29 02:32:32 -070084#include <linux/highmem.h>
85#include <linux/swap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070086#include <linux/types.h>
87#include <linux/fcntl.h>
88#include <linux/module.h>
89#include <linux/socket.h>
90#include <linux/sockios.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020091#include <linux/igmp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070092#include <linux/in.h>
93#include <linux/errno.h>
94#include <linux/timer.h>
95#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070096#include <linux/inet.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070097#include <linux/netdevice.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070098#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099#include <linux/skbuff.h>
100#include <linux/proc_fs.h>
101#include <linux/seq_file.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200102#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103#include <net/icmp.h>
104#include <net/route.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#include <net/checksum.h>
106#include <net/xfrm.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800107#include "udp_impl.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108
Eric Dumazet645ca702008-10-29 01:41:45 -0700109struct udp_table udp_table;
110EXPORT_SYMBOL(udp_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111
Hideo Aoki95766ff2007-12-31 00:29:24 -0800112int sysctl_udp_mem[3] __read_mostly;
113int sysctl_udp_rmem_min __read_mostly;
114int sysctl_udp_wmem_min __read_mostly;
115
116EXPORT_SYMBOL(sysctl_udp_mem);
117EXPORT_SYMBOL(sysctl_udp_rmem_min);
118EXPORT_SYMBOL(sysctl_udp_wmem_min);
119
120atomic_t udp_memory_allocated;
121EXPORT_SYMBOL(udp_memory_allocated);
122
Eric Dumazet98322f22009-01-26 21:35:35 -0800123#define PORTS_PER_CHAIN (65536 / UDP_HTABLE_SIZE)
124
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700125static int udp_lib_lport_inuse(struct net *net, __u16 num,
Eric Dumazet645ca702008-10-29 01:41:45 -0700126 const struct udp_hslot *hslot,
Eric Dumazet98322f22009-01-26 21:35:35 -0800127 unsigned long *bitmap,
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700128 struct sock *sk,
129 int (*saddr_comp)(const struct sock *sk1,
130 const struct sock *sk2))
Gerrit Renker25030a72006-08-26 20:06:05 -0700131{
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700132 struct sock *sk2;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800133 struct hlist_nulls_node *node;
Gerrit Renker25030a72006-08-26 20:06:05 -0700134
Eric Dumazet88ab1932008-11-16 19:39:21 -0800135 sk_nulls_for_each(sk2, node, &hslot->head)
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700136 if (net_eq(sock_net(sk2), net) &&
137 sk2 != sk &&
Eric Dumazet98322f22009-01-26 21:35:35 -0800138 (bitmap || sk2->sk_hash == num) &&
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700139 (!sk2->sk_reuse || !sk->sk_reuse) &&
140 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if
141 || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Eric Dumazet98322f22009-01-26 21:35:35 -0800142 (*saddr_comp)(sk, sk2)) {
143 if (bitmap)
144 __set_bit(sk2->sk_hash / UDP_HTABLE_SIZE,
145 bitmap);
146 else
147 return 1;
148 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700149 return 0;
150}
151
152/**
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700153 * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
Gerrit Renker25030a72006-08-26 20:06:05 -0700154 *
155 * @sk: socket struct in question
156 * @snum: port number to look up
David S. Millerdf2bc452007-06-05 15:18:43 -0700157 * @saddr_comp: AF-dependent comparison of bound local IP addresses
Gerrit Renker25030a72006-08-26 20:06:05 -0700158 */
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700159int udp_lib_get_port(struct sock *sk, unsigned short snum,
David S. Millerdf2bc452007-06-05 15:18:43 -0700160 int (*saddr_comp)(const struct sock *sk1,
161 const struct sock *sk2 ) )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162{
Eric Dumazet645ca702008-10-29 01:41:45 -0700163 struct udp_hslot *hslot;
164 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Gerrit Renker25030a72006-08-26 20:06:05 -0700165 int error = 1;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900166 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700168 if (!snum) {
Eric Dumazet9088c562008-10-08 11:44:17 -0700169 int low, high, remaining;
170 unsigned rand;
Eric Dumazet98322f22009-01-26 21:35:35 -0800171 unsigned short first, last;
172 DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700174 inet_get_local_port_range(&low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700175 remaining = (high - low) + 1;
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700176
Eric Dumazet9088c562008-10-08 11:44:17 -0700177 rand = net_random();
Eric Dumazet98322f22009-01-26 21:35:35 -0800178 first = (((u64)rand * remaining) >> 32) + low;
179 /*
180 * force rand to be an odd multiple of UDP_HTABLE_SIZE
181 */
182 rand = (rand | 1) * UDP_HTABLE_SIZE;
183 for (last = first + UDP_HTABLE_SIZE; first != last; first++) {
184 hslot = &udptable->hash[udp_hashfn(net, first)];
185 bitmap_zero(bitmap, PORTS_PER_CHAIN);
Eric Dumazet645ca702008-10-29 01:41:45 -0700186 spin_lock_bh(&hslot->lock);
Eric Dumazet98322f22009-01-26 21:35:35 -0800187 udp_lib_lport_inuse(net, snum, hslot, bitmap, sk,
188 saddr_comp);
189
190 snum = first;
191 /*
192 * Iterate on all possible values of snum for this hash.
193 * Using steps of an odd multiple of UDP_HTABLE_SIZE
194 * give us randomization and full range coverage.
195 */
Eric Dumazet9088c562008-10-08 11:44:17 -0700196 do {
Eric Dumazet98322f22009-01-26 21:35:35 -0800197 if (low <= snum && snum <= high &&
198 !test_bit(snum / UDP_HTABLE_SIZE, bitmap))
199 goto found;
200 snum += rand;
201 } while (snum != first);
202 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 }
Eric Dumazet98322f22009-01-26 21:35:35 -0800204 goto fail;
Eric Dumazet645ca702008-10-29 01:41:45 -0700205 } else {
206 hslot = &udptable->hash[udp_hashfn(net, snum)];
207 spin_lock_bh(&hslot->lock);
Eric Dumazet98322f22009-01-26 21:35:35 -0800208 if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, saddr_comp))
Eric Dumazet645ca702008-10-29 01:41:45 -0700209 goto fail_unlock;
210 }
Eric Dumazet98322f22009-01-26 21:35:35 -0800211found:
Gerrit Renker25030a72006-08-26 20:06:05 -0700212 inet_sk(sk)->num = snum;
David S. Millerdf2bc452007-06-05 15:18:43 -0700213 sk->sk_hash = snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 if (sk_unhashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800215 sk_nulls_add_node_rcu(sk, &hslot->head);
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700216 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700218 error = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -0700219fail_unlock:
220 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221fail:
Gerrit Renker25030a72006-08-26 20:06:05 -0700222 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223}
224
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700225static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
David S. Millerdb8dac22008-03-06 16:22:02 -0800226{
227 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
228
229 return ( !ipv6_only_sock(sk2) &&
230 (!inet1->rcv_saddr || !inet2->rcv_saddr ||
231 inet1->rcv_saddr == inet2->rcv_saddr ));
232}
233
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700234int udp_v4_get_port(struct sock *sk, unsigned short snum)
David S. Millerdb8dac22008-03-06 16:22:02 -0800235{
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700236 return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal);
David S. Millerdb8dac22008-03-06 16:22:02 -0800237}
238
Eric Dumazet645ca702008-10-29 01:41:45 -0700239static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr,
240 unsigned short hnum,
241 __be16 sport, __be32 daddr, __be16 dport, int dif)
242{
243 int score = -1;
244
245 if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum &&
246 !ipv6_only_sock(sk)) {
247 struct inet_sock *inet = inet_sk(sk);
248
249 score = (sk->sk_family == PF_INET ? 1 : 0);
250 if (inet->rcv_saddr) {
251 if (inet->rcv_saddr != daddr)
252 return -1;
253 score += 2;
254 }
255 if (inet->daddr) {
256 if (inet->daddr != saddr)
257 return -1;
258 score += 2;
259 }
260 if (inet->dport) {
261 if (inet->dport != sport)
262 return -1;
263 score += 2;
264 }
265 if (sk->sk_bound_dev_if) {
266 if (sk->sk_bound_dev_if != dif)
267 return -1;
268 score += 2;
269 }
270 }
271 return score;
272}
273
David S. Millerdb8dac22008-03-06 16:22:02 -0800274/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
275 * harder than this. -DaveM
276 */
277static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
278 __be16 sport, __be32 daddr, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700279 int dif, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800280{
Eric Dumazet271b72c2008-10-29 02:11:14 -0700281 struct sock *sk, *result;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800282 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800283 unsigned short hnum = ntohs(dport);
Eric Dumazet645ca702008-10-29 01:41:45 -0700284 unsigned int hash = udp_hashfn(net, hnum);
285 struct udp_hslot *hslot = &udptable->hash[hash];
Eric Dumazet271b72c2008-10-29 02:11:14 -0700286 int score, badness;
David S. Millerdb8dac22008-03-06 16:22:02 -0800287
Eric Dumazet271b72c2008-10-29 02:11:14 -0700288 rcu_read_lock();
289begin:
290 result = NULL;
291 badness = -1;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800292 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
Eric Dumazet645ca702008-10-29 01:41:45 -0700293 score = compute_score(sk, net, saddr, hnum, sport,
294 daddr, dport, dif);
295 if (score > badness) {
296 result = sk;
297 badness = score;
David S. Millerdb8dac22008-03-06 16:22:02 -0800298 }
299 }
Eric Dumazet88ab1932008-11-16 19:39:21 -0800300 /*
301 * if the nulls value we got at the end of this lookup is
302 * not the expected one, we must restart lookup.
303 * We probably met an item that was moved to another chain.
304 */
305 if (get_nulls_value(node) != hash)
306 goto begin;
307
Eric Dumazet271b72c2008-10-29 02:11:14 -0700308 if (result) {
309 if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt)))
310 result = NULL;
311 else if (unlikely(compute_score(result, net, saddr, hnum, sport,
312 daddr, dport, dif) < badness)) {
313 sock_put(result);
314 goto begin;
315 }
316 }
317 rcu_read_unlock();
David S. Millerdb8dac22008-03-06 16:22:02 -0800318 return result;
319}
320
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700321static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
322 __be16 sport, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700323 struct udp_table *udptable)
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700324{
KOVACS Krisztian23542612008-10-07 12:41:01 -0700325 struct sock *sk;
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700326 const struct iphdr *iph = ip_hdr(skb);
327
KOVACS Krisztian23542612008-10-07 12:41:01 -0700328 if (unlikely(sk = skb_steal_sock(skb)))
329 return sk;
330 else
331 return __udp4_lib_lookup(dev_net(skb->dst->dev), iph->saddr, sport,
332 iph->daddr, dport, inet_iif(skb),
333 udptable);
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700334}
335
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700336struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
337 __be32 daddr, __be16 dport, int dif)
338{
Eric Dumazet645ca702008-10-29 01:41:45 -0700339 return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700340}
341EXPORT_SYMBOL_GPL(udp4_lib_lookup);
342
Eric Dumazet920a4612008-11-01 21:22:23 -0700343static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk,
David S. Millerdb8dac22008-03-06 16:22:02 -0800344 __be16 loc_port, __be32 loc_addr,
345 __be16 rmt_port, __be32 rmt_addr,
346 int dif)
347{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800348 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800349 struct sock *s = sk;
350 unsigned short hnum = ntohs(loc_port);
351
Eric Dumazet88ab1932008-11-16 19:39:21 -0800352 sk_nulls_for_each_from(s, node) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800353 struct inet_sock *inet = inet_sk(s);
354
Eric Dumazet920a4612008-11-01 21:22:23 -0700355 if (!net_eq(sock_net(s), net) ||
356 s->sk_hash != hnum ||
David S. Millerdb8dac22008-03-06 16:22:02 -0800357 (inet->daddr && inet->daddr != rmt_addr) ||
358 (inet->dport != rmt_port && inet->dport) ||
359 (inet->rcv_saddr && inet->rcv_saddr != loc_addr) ||
360 ipv6_only_sock(s) ||
361 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
362 continue;
363 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
364 continue;
365 goto found;
366 }
367 s = NULL;
368found:
369 return s;
370}
371
372/*
373 * This routine is called by the ICMP module when it gets some
374 * sort of error condition. If err < 0 then the socket should
375 * be closed and the error returned to the user. If err > 0
376 * it's just the icmp type << 8 | icmp code.
377 * Header points to the ip header of the error packet. We move
378 * on past this. Then (as it used to claim before adjustment)
379 * header points to the first 8 bytes of the udp header. We need
380 * to find the appropriate port.
381 */
382
Eric Dumazet645ca702008-10-29 01:41:45 -0700383void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800384{
385 struct inet_sock *inet;
386 struct iphdr *iph = (struct iphdr*)skb->data;
387 struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2));
388 const int type = icmp_hdr(skb)->type;
389 const int code = icmp_hdr(skb)->code;
390 struct sock *sk;
391 int harderr;
392 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700393 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800394
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700395 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
David S. Millerdb8dac22008-03-06 16:22:02 -0800396 iph->saddr, uh->source, skb->dev->ifindex, udptable);
397 if (sk == NULL) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700398 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800399 return; /* No socket for error */
400 }
401
402 err = 0;
403 harderr = 0;
404 inet = inet_sk(sk);
405
406 switch (type) {
407 default:
408 case ICMP_TIME_EXCEEDED:
409 err = EHOSTUNREACH;
410 break;
411 case ICMP_SOURCE_QUENCH:
412 goto out;
413 case ICMP_PARAMETERPROB:
414 err = EPROTO;
415 harderr = 1;
416 break;
417 case ICMP_DEST_UNREACH:
418 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
419 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
420 err = EMSGSIZE;
421 harderr = 1;
422 break;
423 }
424 goto out;
425 }
426 err = EHOSTUNREACH;
427 if (code <= NR_ICMP_UNREACH) {
428 harderr = icmp_err_convert[code].fatal;
429 err = icmp_err_convert[code].errno;
430 }
431 break;
432 }
433
434 /*
435 * RFC1122: OK. Passes ICMP errors back to application, as per
436 * 4.1.3.3.
437 */
438 if (!inet->recverr) {
439 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
440 goto out;
441 } else {
442 ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1));
443 }
444 sk->sk_err = err;
445 sk->sk_error_report(sk);
446out:
447 sock_put(sk);
448}
449
450void udp_err(struct sk_buff *skb, u32 info)
451{
Eric Dumazet645ca702008-10-29 01:41:45 -0700452 __udp4_lib_err(skb, info, &udp_table);
David S. Millerdb8dac22008-03-06 16:22:02 -0800453}
454
455/*
456 * Throw away all pending data and cancel the corking. Socket is locked.
457 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400458void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800459{
460 struct udp_sock *up = udp_sk(sk);
461
462 if (up->pending) {
463 up->len = 0;
464 up->pending = 0;
465 ip_flush_pending_frames(sk);
466 }
467}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400468EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800469
470/**
471 * udp4_hwcsum_outgoing - handle outgoing HW checksumming
472 * @sk: socket we are sending on
473 * @skb: sk_buff containing the filled-in UDP header
474 * (checksum field must be zeroed out)
475 */
476static void udp4_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
477 __be32 src, __be32 dst, int len )
478{
479 unsigned int offset;
480 struct udphdr *uh = udp_hdr(skb);
481 __wsum csum = 0;
482
483 if (skb_queue_len(&sk->sk_write_queue) == 1) {
484 /*
485 * Only one fragment on the socket.
486 */
487 skb->csum_start = skb_transport_header(skb) - skb->head;
488 skb->csum_offset = offsetof(struct udphdr, check);
489 uh->check = ~csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, 0);
490 } else {
491 /*
492 * HW-checksum won't work as there are two or more
493 * fragments on the socket so that all csums of sk_buffs
494 * should be together
495 */
496 offset = skb_transport_offset(skb);
497 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
498
499 skb->ip_summed = CHECKSUM_NONE;
500
501 skb_queue_walk(&sk->sk_write_queue, skb) {
502 csum = csum_add(csum, skb->csum);
503 }
504
505 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
506 if (uh->check == 0)
507 uh->check = CSUM_MANGLED_0;
508 }
509}
510
511/*
512 * Push out all pending data as one UDP datagram. Socket is locked.
513 */
514static int udp_push_pending_frames(struct sock *sk)
515{
516 struct udp_sock *up = udp_sk(sk);
517 struct inet_sock *inet = inet_sk(sk);
518 struct flowi *fl = &inet->cork.fl;
519 struct sk_buff *skb;
520 struct udphdr *uh;
521 int err = 0;
522 int is_udplite = IS_UDPLITE(sk);
523 __wsum csum = 0;
524
525 /* Grab the skbuff where UDP header space exists. */
526 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
527 goto out;
528
529 /*
530 * Create a UDP header
531 */
532 uh = udp_hdr(skb);
533 uh->source = fl->fl_ip_sport;
534 uh->dest = fl->fl_ip_dport;
535 uh->len = htons(up->len);
536 uh->check = 0;
537
538 if (is_udplite) /* UDP-Lite */
539 csum = udplite_csum_outgoing(sk, skb);
540
541 else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */
542
543 skb->ip_summed = CHECKSUM_NONE;
544 goto send;
545
546 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
547
548 udp4_hwcsum_outgoing(sk, skb, fl->fl4_src,fl->fl4_dst, up->len);
549 goto send;
550
551 } else /* `normal' UDP */
552 csum = udp_csum_outgoing(sk, skb);
553
554 /* add protocol-dependent pseudo-header */
555 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, up->len,
556 sk->sk_protocol, csum );
557 if (uh->check == 0)
558 uh->check = CSUM_MANGLED_0;
559
560send:
561 err = ip_push_pending_frames(sk);
562out:
563 up->len = 0;
564 up->pending = 0;
565 if (!err)
Pavel Emelyanov629ca232008-07-05 21:18:07 -0700566 UDP_INC_STATS_USER(sock_net(sk),
567 UDP_MIB_OUTDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -0800568 return err;
569}
570
571int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
572 size_t len)
573{
574 struct inet_sock *inet = inet_sk(sk);
575 struct udp_sock *up = udp_sk(sk);
576 int ulen = len;
577 struct ipcm_cookie ipc;
578 struct rtable *rt = NULL;
579 int free = 0;
580 int connected = 0;
581 __be32 daddr, faddr, saddr;
582 __be16 dport;
583 u8 tos;
584 int err, is_udplite = IS_UDPLITE(sk);
585 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
586 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
587
588 if (len > 0xFFFF)
589 return -EMSGSIZE;
590
591 /*
592 * Check the flags.
593 */
594
595 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */
596 return -EOPNOTSUPP;
597
598 ipc.opt = NULL;
599
600 if (up->pending) {
601 /*
602 * There are pending frames.
603 * The socket lock must be held while it's corked.
604 */
605 lock_sock(sk);
606 if (likely(up->pending)) {
607 if (unlikely(up->pending != AF_INET)) {
608 release_sock(sk);
609 return -EINVAL;
610 }
611 goto do_append_data;
612 }
613 release_sock(sk);
614 }
615 ulen += sizeof(struct udphdr);
616
617 /*
618 * Get and verify the address.
619 */
620 if (msg->msg_name) {
621 struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name;
622 if (msg->msg_namelen < sizeof(*usin))
623 return -EINVAL;
624 if (usin->sin_family != AF_INET) {
625 if (usin->sin_family != AF_UNSPEC)
626 return -EAFNOSUPPORT;
627 }
628
629 daddr = usin->sin_addr.s_addr;
630 dport = usin->sin_port;
631 if (dport == 0)
632 return -EINVAL;
633 } else {
634 if (sk->sk_state != TCP_ESTABLISHED)
635 return -EDESTADDRREQ;
636 daddr = inet->daddr;
637 dport = inet->dport;
638 /* Open fast path for connected socket.
639 Route will not be used, if at least one option is set.
640 */
641 connected = 1;
642 }
643 ipc.addr = inet->saddr;
644
645 ipc.oif = sk->sk_bound_dev_if;
646 if (msg->msg_controllen) {
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900647 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
David S. Millerdb8dac22008-03-06 16:22:02 -0800648 if (err)
649 return err;
650 if (ipc.opt)
651 free = 1;
652 connected = 0;
653 }
654 if (!ipc.opt)
655 ipc.opt = inet->opt;
656
657 saddr = ipc.addr;
658 ipc.addr = faddr = daddr;
659
660 if (ipc.opt && ipc.opt->srr) {
661 if (!daddr)
662 return -EINVAL;
663 faddr = ipc.opt->faddr;
664 connected = 0;
665 }
666 tos = RT_TOS(inet->tos);
667 if (sock_flag(sk, SOCK_LOCALROUTE) ||
668 (msg->msg_flags & MSG_DONTROUTE) ||
669 (ipc.opt && ipc.opt->is_strictroute)) {
670 tos |= RTO_ONLINK;
671 connected = 0;
672 }
673
674 if (ipv4_is_multicast(daddr)) {
675 if (!ipc.oif)
676 ipc.oif = inet->mc_index;
677 if (!saddr)
678 saddr = inet->mc_addr;
679 connected = 0;
680 }
681
682 if (connected)
683 rt = (struct rtable*)sk_dst_check(sk, 0);
684
685 if (rt == NULL) {
686 struct flowi fl = { .oif = ipc.oif,
687 .nl_u = { .ip4_u =
688 { .daddr = faddr,
689 .saddr = saddr,
690 .tos = tos } },
691 .proto = sk->sk_protocol,
Balazs Scheidlera134f852008-11-20 01:07:24 -0800692 .flags = inet_sk_flowi_flags(sk),
David S. Millerdb8dac22008-03-06 16:22:02 -0800693 .uli_u = { .ports =
694 { .sport = inet->sport,
695 .dport = dport } } };
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -0700696 struct net *net = sock_net(sk);
697
David S. Millerdb8dac22008-03-06 16:22:02 -0800698 security_sk_classify_flow(sk, &fl);
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -0700699 err = ip_route_output_flow(net, &rt, &fl, sk, 1);
David S. Millerdb8dac22008-03-06 16:22:02 -0800700 if (err) {
701 if (err == -ENETUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700702 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -0800703 goto out;
704 }
705
706 err = -EACCES;
707 if ((rt->rt_flags & RTCF_BROADCAST) &&
708 !sock_flag(sk, SOCK_BROADCAST))
709 goto out;
710 if (connected)
711 sk_dst_set(sk, dst_clone(&rt->u.dst));
712 }
713
714 if (msg->msg_flags&MSG_CONFIRM)
715 goto do_confirm;
716back_from_confirm:
717
718 saddr = rt->rt_src;
719 if (!ipc.addr)
720 daddr = ipc.addr = rt->rt_dst;
721
722 lock_sock(sk);
723 if (unlikely(up->pending)) {
724 /* The socket is already corked while preparing it. */
725 /* ... which is an evident application bug. --ANK */
726 release_sock(sk);
727
728 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
729 err = -EINVAL;
730 goto out;
731 }
732 /*
733 * Now cork the socket to pend data.
734 */
735 inet->cork.fl.fl4_dst = daddr;
736 inet->cork.fl.fl_ip_dport = dport;
737 inet->cork.fl.fl4_src = saddr;
738 inet->cork.fl.fl_ip_sport = inet->sport;
739 up->pending = AF_INET;
740
741do_append_data:
742 up->len += ulen;
743 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
744 err = ip_append_data(sk, getfrag, msg->msg_iov, ulen,
Eric Dumazet2e77d892008-11-24 15:52:46 -0800745 sizeof(struct udphdr), &ipc, &rt,
David S. Millerdb8dac22008-03-06 16:22:02 -0800746 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
747 if (err)
748 udp_flush_pending_frames(sk);
749 else if (!corkreq)
750 err = udp_push_pending_frames(sk);
751 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
752 up->pending = 0;
753 release_sock(sk);
754
755out:
756 ip_rt_put(rt);
757 if (free)
758 kfree(ipc.opt);
759 if (!err)
760 return len;
761 /*
762 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
763 * ENOBUFS might not be good (it's not tunable per se), but otherwise
764 * we don't have a good statistic (IpOutDiscards but it can be too many
765 * things). We could add another new stat but at least for now that
766 * seems like overkill.
767 */
768 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Pavel Emelyanov629ca232008-07-05 21:18:07 -0700769 UDP_INC_STATS_USER(sock_net(sk),
770 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -0800771 }
772 return err;
773
774do_confirm:
775 dst_confirm(&rt->u.dst);
776 if (!(msg->msg_flags&MSG_PROBE) || len)
777 goto back_from_confirm;
778 err = 0;
779 goto out;
780}
781
782int udp_sendpage(struct sock *sk, struct page *page, int offset,
783 size_t size, int flags)
784{
785 struct udp_sock *up = udp_sk(sk);
786 int ret;
787
788 if (!up->pending) {
789 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
790
791 /* Call udp_sendmsg to specify destination address which
792 * sendpage interface can't pass.
793 * This will succeed only when the socket is connected.
794 */
795 ret = udp_sendmsg(NULL, sk, &msg, 0);
796 if (ret < 0)
797 return ret;
798 }
799
800 lock_sock(sk);
801
802 if (unlikely(!up->pending)) {
803 release_sock(sk);
804
805 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n");
806 return -EINVAL;
807 }
808
809 ret = ip_append_page(sk, page, offset, size, flags);
810 if (ret == -EOPNOTSUPP) {
811 release_sock(sk);
812 return sock_no_sendpage(sk->sk_socket, page, offset,
813 size, flags);
814 }
815 if (ret < 0) {
816 udp_flush_pending_frames(sk);
817 goto out;
818 }
819
820 up->len += size;
821 if (!(up->corkflag || (flags&MSG_MORE)))
822 ret = udp_push_pending_frames(sk);
823 if (!ret)
824 ret = size;
825out:
826 release_sock(sk);
827 return ret;
828}
829
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830/*
831 * IOCTL requests applicable to the UDP protocol
832 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900833
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
835{
Stephen Hemminger6516c652007-03-08 20:41:55 -0800836 switch (cmd) {
837 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 {
Stephen Hemminger6516c652007-03-08 20:41:55 -0800839 int amount = atomic_read(&sk->sk_wmem_alloc);
840 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 }
Stephen Hemminger6516c652007-03-08 20:41:55 -0800842
843 case SIOCINQ:
844 {
845 struct sk_buff *skb;
846 unsigned long amount;
847
848 amount = 0;
849 spin_lock_bh(&sk->sk_receive_queue.lock);
850 skb = skb_peek(&sk->sk_receive_queue);
851 if (skb != NULL) {
852 /*
853 * We will only return the amount
854 * of this packet since that is all
855 * that will be read.
856 */
857 amount = skb->len - sizeof(struct udphdr);
858 }
859 spin_unlock_bh(&sk->sk_receive_queue.lock);
860 return put_user(amount, (int __user *)arg);
861 }
862
863 default:
864 return -ENOIOCTLCMD;
865 }
866
867 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868}
869
David S. Millerdb8dac22008-03-06 16:22:02 -0800870/*
871 * This should be easy, if there is something there we
872 * return it, otherwise we block.
873 */
874
875int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
876 size_t len, int noblock, int flags, int *addr_len)
877{
878 struct inet_sock *inet = inet_sk(sk);
879 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
880 struct sk_buff *skb;
881 unsigned int ulen, copied;
882 int peeked;
883 int err;
884 int is_udplite = IS_UDPLITE(sk);
885
886 /*
887 * Check any passed addresses
888 */
889 if (addr_len)
890 *addr_len=sizeof(*sin);
891
892 if (flags & MSG_ERRQUEUE)
893 return ip_recv_error(sk, msg, len);
894
895try_again:
896 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
897 &peeked, &err);
898 if (!skb)
899 goto out;
900
901 ulen = skb->len - sizeof(struct udphdr);
902 copied = len;
903 if (copied > ulen)
904 copied = ulen;
905 else if (copied < ulen)
906 msg->msg_flags |= MSG_TRUNC;
907
908 /*
909 * If checksum is needed at all, try to do it while copying the
910 * data. If the data is truncated, or if we only want a partial
911 * coverage checksum (UDP-Lite), do it before the copy.
912 */
913
914 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
915 if (udp_lib_checksum_complete(skb))
916 goto csum_copy_err;
917 }
918
919 if (skb_csum_unnecessary(skb))
920 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
921 msg->msg_iov, copied );
922 else {
923 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
924
925 if (err == -EINVAL)
926 goto csum_copy_err;
927 }
928
929 if (err)
930 goto out_free;
931
932 if (!peeked)
Pavel Emelyanov629ca232008-07-05 21:18:07 -0700933 UDP_INC_STATS_USER(sock_net(sk),
934 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -0800935
936 sock_recv_timestamp(msg, sk, skb);
937
938 /* Copy the address. */
939 if (sin)
940 {
941 sin->sin_family = AF_INET;
942 sin->sin_port = udp_hdr(skb)->source;
943 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
944 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
945 }
946 if (inet->cmsg_flags)
947 ip_cmsg_recv(msg, skb);
948
949 err = copied;
950 if (flags & MSG_TRUNC)
951 err = ulen;
952
953out_free:
954 lock_sock(sk);
955 skb_free_datagram(sk, skb);
956 release_sock(sk);
957out:
958 return err;
959
960csum_copy_err:
961 lock_sock(sk);
962 if (!skb_kill_datagram(sk, skb, flags))
Pavel Emelyanov629ca232008-07-05 21:18:07 -0700963 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -0800964 release_sock(sk);
965
966 if (noblock)
967 return -EAGAIN;
968 goto try_again;
969}
970
971
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972int udp_disconnect(struct sock *sk, int flags)
973{
974 struct inet_sock *inet = inet_sk(sk);
975 /*
976 * 1003.1g - break association.
977 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900978
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 sk->sk_state = TCP_CLOSE;
980 inet->daddr = 0;
981 inet->dport = 0;
982 sk->sk_bound_dev_if = 0;
983 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
984 inet_reset_saddr(sk);
985
986 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
987 sk->sk_prot->unhash(sk);
988 inet->sport = 0;
989 }
990 sk_dst_reset(sk);
991 return 0;
992}
993
Eric Dumazet645ca702008-10-29 01:41:45 -0700994void udp_lib_unhash(struct sock *sk)
995{
Eric Dumazet723b4612008-11-25 13:55:15 -0800996 if (sk_hashed(sk)) {
997 struct udp_table *udptable = sk->sk_prot->h.udp_table;
998 unsigned int hash = udp_hashfn(sock_net(sk), sk->sk_hash);
999 struct udp_hslot *hslot = &udptable->hash[hash];
Eric Dumazet645ca702008-10-29 01:41:45 -07001000
Eric Dumazet723b4612008-11-25 13:55:15 -08001001 spin_lock_bh(&hslot->lock);
1002 if (sk_nulls_del_node_init_rcu(sk)) {
1003 inet_sk(sk)->num = 0;
1004 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
1005 }
1006 spin_unlock_bh(&hslot->lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001007 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001008}
1009EXPORT_SYMBOL(udp_lib_unhash);
1010
Herbert Xu93821772008-09-15 11:48:46 -07001011static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1012{
1013 int is_udplite = IS_UDPLITE(sk);
1014 int rc;
1015
1016 if ((rc = sock_queue_rcv_skb(sk, skb)) < 0) {
1017 /* Note that an ENOMEM error is charged twice */
Eric Dumazete408b8d2009-02-02 13:41:57 -08001018 if (rc == -ENOMEM) {
Herbert Xu93821772008-09-15 11:48:46 -07001019 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1020 is_udplite);
Eric Dumazete408b8d2009-02-02 13:41:57 -08001021 atomic_inc(&sk->sk_drops);
1022 }
Herbert Xu93821772008-09-15 11:48:46 -07001023 goto drop;
1024 }
1025
1026 return 0;
1027
1028drop:
1029 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1030 kfree_skb(skb);
1031 return -1;
1032}
1033
David S. Millerdb8dac22008-03-06 16:22:02 -08001034/* returns:
1035 * -1: error
1036 * 0: success
1037 * >0: "udp encap" protocol resubmission
1038 *
1039 * Note that in the success and error cases, the skb is assumed to
1040 * have either been requeued or freed.
1041 */
1042int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
1043{
1044 struct udp_sock *up = udp_sk(sk);
1045 int rc;
1046 int is_udplite = IS_UDPLITE(sk);
1047
1048 /*
1049 * Charge it to the socket, dropping if the queue is full.
1050 */
1051 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1052 goto drop;
1053 nf_reset(skb);
1054
1055 if (up->encap_type) {
1056 /*
1057 * This is an encapsulation socket so pass the skb to
1058 * the socket's udp_encap_rcv() hook. Otherwise, just
1059 * fall through and pass this up the UDP socket.
1060 * up->encap_rcv() returns the following value:
1061 * =0 if skb was successfully passed to the encap
1062 * handler or was discarded by it.
1063 * >0 if skb should be passed on to UDP.
1064 * <0 if skb should be resubmitted as proto -N
1065 */
1066
1067 /* if we're overly short, let UDP handle it */
1068 if (skb->len > sizeof(struct udphdr) &&
1069 up->encap_rcv != NULL) {
1070 int ret;
1071
1072 ret = (*up->encap_rcv)(sk, skb);
1073 if (ret <= 0) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001074 UDP_INC_STATS_BH(sock_net(sk),
1075 UDP_MIB_INDATAGRAMS,
David S. Millerdb8dac22008-03-06 16:22:02 -08001076 is_udplite);
1077 return -ret;
1078 }
1079 }
1080
1081 /* FALLTHROUGH -- it's a UDP Packet */
1082 }
1083
1084 /*
1085 * UDP-Lite specific tests, ignored on UDP sockets
1086 */
1087 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1088
1089 /*
1090 * MIB statistics other than incrementing the error count are
1091 * disabled for the following two types of errors: these depend
1092 * on the application settings, not on the functioning of the
1093 * protocol stack as such.
1094 *
1095 * RFC 3828 here recommends (sec 3.3): "There should also be a
1096 * way ... to ... at least let the receiving application block
1097 * delivery of packets with coverage values less than a value
1098 * provided by the application."
1099 */
1100 if (up->pcrlen == 0) { /* full coverage was set */
1101 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage "
1102 "%d while full coverage %d requested\n",
1103 UDP_SKB_CB(skb)->cscov, skb->len);
1104 goto drop;
1105 }
1106 /* The next case involves violating the min. coverage requested
1107 * by the receiver. This is subtle: if receiver wants x and x is
1108 * greater than the buffersize/MTU then receiver will complain
1109 * that it wants x while sender emits packets of smaller size y.
1110 * Therefore the above ...()->partial_cov statement is essential.
1111 */
1112 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
1113 LIMIT_NETDEBUG(KERN_WARNING
1114 "UDPLITE: coverage %d too small, need min %d\n",
1115 UDP_SKB_CB(skb)->cscov, up->pcrlen);
1116 goto drop;
1117 }
1118 }
1119
1120 if (sk->sk_filter) {
1121 if (udp_lib_checksum_complete(skb))
1122 goto drop;
1123 }
1124
Herbert Xu93821772008-09-15 11:48:46 -07001125 rc = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001126
Herbert Xu93821772008-09-15 11:48:46 -07001127 bh_lock_sock(sk);
1128 if (!sock_owned_by_user(sk))
1129 rc = __udp_queue_rcv_skb(sk, skb);
1130 else
1131 sk_add_backlog(sk, skb);
1132 bh_unlock_sock(sk);
1133
1134 return rc;
David S. Millerdb8dac22008-03-06 16:22:02 -08001135
1136drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001137 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001138 kfree_skb(skb);
1139 return -1;
1140}
1141
1142/*
1143 * Multicasts and broadcasts go to each listener.
1144 *
1145 * Note: called only from the BH handler context,
1146 * so we don't need to lock the hashes.
1147 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001148static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001149 struct udphdr *uh,
1150 __be32 saddr, __be32 daddr,
Eric Dumazet645ca702008-10-29 01:41:45 -07001151 struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -08001152{
1153 struct sock *sk;
Eric Dumazet645ca702008-10-29 01:41:45 -07001154 struct udp_hslot *hslot = &udptable->hash[udp_hashfn(net, ntohs(uh->dest))];
David S. Millerdb8dac22008-03-06 16:22:02 -08001155 int dif;
1156
Eric Dumazet645ca702008-10-29 01:41:45 -07001157 spin_lock(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08001158 sk = sk_nulls_head(&hslot->head);
David S. Millerdb8dac22008-03-06 16:22:02 -08001159 dif = skb->dev->ifindex;
Eric Dumazet920a4612008-11-01 21:22:23 -07001160 sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
David S. Millerdb8dac22008-03-06 16:22:02 -08001161 if (sk) {
1162 struct sock *sknext = NULL;
1163
1164 do {
1165 struct sk_buff *skb1 = skb;
1166
Eric Dumazet88ab1932008-11-16 19:39:21 -08001167 sknext = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest,
Eric Dumazet920a4612008-11-01 21:22:23 -07001168 daddr, uh->source, saddr,
1169 dif);
David S. Millerdb8dac22008-03-06 16:22:02 -08001170 if (sknext)
1171 skb1 = skb_clone(skb, GFP_ATOMIC);
1172
1173 if (skb1) {
Herbert Xu93821772008-09-15 11:48:46 -07001174 int ret = udp_queue_rcv_skb(sk, skb1);
David S. Millerdb8dac22008-03-06 16:22:02 -08001175 if (ret > 0)
1176 /* we should probably re-process instead
1177 * of dropping packets here. */
1178 kfree_skb(skb1);
1179 }
1180 sk = sknext;
1181 } while (sknext);
1182 } else
1183 kfree_skb(skb);
Eric Dumazet645ca702008-10-29 01:41:45 -07001184 spin_unlock(&hslot->lock);
David S. Millerdb8dac22008-03-06 16:22:02 -08001185 return 0;
1186}
1187
1188/* Initialize UDP checksum. If exited with zero value (success),
1189 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1190 * Otherwise, csum completion requires chacksumming packet body,
1191 * including udp header and folding it to skb->csum.
1192 */
1193static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1194 int proto)
1195{
1196 const struct iphdr *iph;
1197 int err;
1198
1199 UDP_SKB_CB(skb)->partial_cov = 0;
1200 UDP_SKB_CB(skb)->cscov = skb->len;
1201
1202 if (proto == IPPROTO_UDPLITE) {
1203 err = udplite_checksum_init(skb, uh);
1204 if (err)
1205 return err;
1206 }
1207
1208 iph = ip_hdr(skb);
1209 if (uh->check == 0) {
1210 skb->ip_summed = CHECKSUM_UNNECESSARY;
1211 } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
1212 if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len,
1213 proto, skb->csum))
1214 skb->ip_summed = CHECKSUM_UNNECESSARY;
1215 }
1216 if (!skb_csum_unnecessary(skb))
1217 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
1218 skb->len, proto, 0);
1219 /* Probably, we should checksum udp header (it should be in cache
1220 * in any case) and data in tiny packets (< rx copybreak).
1221 */
1222
1223 return 0;
1224}
1225
1226/*
1227 * All we need to do is get the socket, and then do a checksum.
1228 */
1229
Eric Dumazet645ca702008-10-29 01:41:45 -07001230int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
David S. Millerdb8dac22008-03-06 16:22:02 -08001231 int proto)
1232{
1233 struct sock *sk;
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001234 struct udphdr *uh;
David S. Millerdb8dac22008-03-06 16:22:02 -08001235 unsigned short ulen;
1236 struct rtable *rt = (struct rtable*)skb->dst;
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001237 __be32 saddr, daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001238 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001239
1240 /*
1241 * Validate the packet.
1242 */
1243 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1244 goto drop; /* No space for header. */
1245
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001246 uh = udp_hdr(skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001247 ulen = ntohs(uh->len);
1248 if (ulen > skb->len)
1249 goto short_packet;
1250
1251 if (proto == IPPROTO_UDP) {
1252 /* UDP validates ulen. */
1253 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1254 goto short_packet;
1255 uh = udp_hdr(skb);
1256 }
1257
1258 if (udp4_csum_init(skb, uh, proto))
1259 goto csum_error;
1260
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001261 saddr = ip_hdr(skb)->saddr;
1262 daddr = ip_hdr(skb)->daddr;
1263
David S. Millerdb8dac22008-03-06 16:22:02 -08001264 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
Pavel Emelyanove3163492008-06-16 17:12:11 -07001265 return __udp4_lib_mcast_deliver(net, skb, uh,
1266 saddr, daddr, udptable);
David S. Millerdb8dac22008-03-06 16:22:02 -08001267
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -07001268 sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
David S. Millerdb8dac22008-03-06 16:22:02 -08001269
1270 if (sk != NULL) {
Herbert Xu93821772008-09-15 11:48:46 -07001271 int ret = udp_queue_rcv_skb(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001272 sock_put(sk);
1273
1274 /* a return value > 0 means to resubmit the input, but
1275 * it wants the return to be -protocol, or 0
1276 */
1277 if (ret > 0)
1278 return -ret;
1279 return 0;
1280 }
1281
1282 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1283 goto drop;
1284 nf_reset(skb);
1285
1286 /* No socket. Drop packet silently, if checksum is wrong */
1287 if (udp_lib_checksum_complete(skb))
1288 goto csum_error;
1289
Pavel Emelyanov02833282008-07-05 21:18:48 -07001290 UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001291 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1292
1293 /*
1294 * Hmm. We got an UDP packet to a port to which we
1295 * don't wanna listen. Ignore it.
1296 */
1297 kfree_skb(skb);
1298 return 0;
1299
1300short_packet:
Harvey Harrison673d57e2008-10-31 00:53:57 -07001301 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
David S. Millerdb8dac22008-03-06 16:22:02 -08001302 proto == IPPROTO_UDPLITE ? "-Lite" : "",
Harvey Harrison673d57e2008-10-31 00:53:57 -07001303 &saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -08001304 ntohs(uh->source),
1305 ulen,
1306 skb->len,
Harvey Harrison673d57e2008-10-31 00:53:57 -07001307 &daddr,
David S. Millerdb8dac22008-03-06 16:22:02 -08001308 ntohs(uh->dest));
1309 goto drop;
1310
1311csum_error:
1312 /*
1313 * RFC1122: OK. Discards the bad packet silently (as far as
1314 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1315 */
Harvey Harrison673d57e2008-10-31 00:53:57 -07001316 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
David S. Millerdb8dac22008-03-06 16:22:02 -08001317 proto == IPPROTO_UDPLITE ? "-Lite" : "",
Harvey Harrison673d57e2008-10-31 00:53:57 -07001318 &saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -08001319 ntohs(uh->source),
Harvey Harrison673d57e2008-10-31 00:53:57 -07001320 &daddr,
David S. Millerdb8dac22008-03-06 16:22:02 -08001321 ntohs(uh->dest),
1322 ulen);
1323drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001324 UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001325 kfree_skb(skb);
1326 return 0;
1327}
1328
1329int udp_rcv(struct sk_buff *skb)
1330{
Eric Dumazet645ca702008-10-29 01:41:45 -07001331 return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
David S. Millerdb8dac22008-03-06 16:22:02 -08001332}
1333
Brian Haley7d06b2e2008-06-14 17:04:49 -07001334void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08001335{
1336 lock_sock(sk);
1337 udp_flush_pending_frames(sk);
1338 release_sock(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001339}
1340
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341/*
1342 * Socket option code for UDP
1343 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001344int udp_lib_setsockopt(struct sock *sk, int level, int optname,
1345 char __user *optval, int optlen,
1346 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347{
1348 struct udp_sock *up = udp_sk(sk);
1349 int val;
1350 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11001351 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352
Stephen Hemminger6516c652007-03-08 20:41:55 -08001353 if (optlen<sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 return -EINVAL;
1355
1356 if (get_user(val, (int __user *)optval))
1357 return -EFAULT;
1358
Stephen Hemminger6516c652007-03-08 20:41:55 -08001359 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 case UDP_CORK:
1361 if (val != 0) {
1362 up->corkflag = 1;
1363 } else {
1364 up->corkflag = 0;
1365 lock_sock(sk);
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001366 (*push_pending_frames)(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 release_sock(sk);
1368 }
1369 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001370
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 case UDP_ENCAP:
1372 switch (val) {
1373 case 0:
1374 case UDP_ENCAP_ESPINUDP:
1375 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07001376 up->encap_rcv = xfrm4_udp_encap_rcv;
1377 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07001378 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 up->encap_type = val;
1380 break;
1381 default:
1382 err = -ENOPROTOOPT;
1383 break;
1384 }
1385 break;
1386
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001387 /*
1388 * UDP-Lite's partial checksum coverage (RFC 3828).
1389 */
1390 /* The sender sets actual checksum coverage length via this option.
1391 * The case coverage > packet length is handled by send module. */
1392 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11001393 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001394 return -ENOPROTOOPT;
1395 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
1396 val = 8;
Gerrit Renker47112e22008-07-21 13:35:08 -07001397 else if (val > USHORT_MAX)
1398 val = USHORT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001399 up->pcslen = val;
1400 up->pcflag |= UDPLITE_SEND_CC;
1401 break;
1402
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001403 /* The receiver specifies a minimum checksum coverage value. To make
1404 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001405 * used, this again means full checksum coverage. */
1406 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11001407 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001408 return -ENOPROTOOPT;
1409 if (val != 0 && val < 8) /* Avoid silly minimal values. */
1410 val = 8;
Gerrit Renker47112e22008-07-21 13:35:08 -07001411 else if (val > USHORT_MAX)
1412 val = USHORT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001413 up->pcrlen = val;
1414 up->pcflag |= UDPLITE_RECV_CC;
1415 break;
1416
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417 default:
1418 err = -ENOPROTOOPT;
1419 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001420 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421
1422 return err;
1423}
1424
David S. Millerdb8dac22008-03-06 16:22:02 -08001425int udp_setsockopt(struct sock *sk, int level, int optname,
1426 char __user *optval, int optlen)
1427{
1428 if (level == SOL_UDP || level == SOL_UDPLITE)
1429 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1430 udp_push_pending_frames);
1431 return ip_setsockopt(sk, level, optname, optval, optlen);
1432}
1433
1434#ifdef CONFIG_COMPAT
1435int compat_udp_setsockopt(struct sock *sk, int level, int optname,
1436 char __user *optval, int optlen)
1437{
1438 if (level == SOL_UDP || level == SOL_UDPLITE)
1439 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1440 udp_push_pending_frames);
1441 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
1442}
1443#endif
1444
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001445int udp_lib_getsockopt(struct sock *sk, int level, int optname,
1446 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447{
1448 struct udp_sock *up = udp_sk(sk);
1449 int val, len;
1450
Stephen Hemminger6516c652007-03-08 20:41:55 -08001451 if (get_user(len,optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 return -EFAULT;
1453
1454 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001455
Stephen Hemminger6516c652007-03-08 20:41:55 -08001456 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 return -EINVAL;
1458
Stephen Hemminger6516c652007-03-08 20:41:55 -08001459 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 case UDP_CORK:
1461 val = up->corkflag;
1462 break;
1463
1464 case UDP_ENCAP:
1465 val = up->encap_type;
1466 break;
1467
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001468 /* The following two cannot be changed on UDP sockets, the return is
1469 * always 0 (which corresponds to the full checksum coverage of UDP). */
1470 case UDPLITE_SEND_CSCOV:
1471 val = up->pcslen;
1472 break;
1473
1474 case UDPLITE_RECV_CSCOV:
1475 val = up->pcrlen;
1476 break;
1477
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 default:
1479 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001480 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481
Stephen Hemminger6516c652007-03-08 20:41:55 -08001482 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001483 return -EFAULT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001484 if (copy_to_user(optval, &val,len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001486 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487}
1488
David S. Millerdb8dac22008-03-06 16:22:02 -08001489int udp_getsockopt(struct sock *sk, int level, int optname,
1490 char __user *optval, int __user *optlen)
1491{
1492 if (level == SOL_UDP || level == SOL_UDPLITE)
1493 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1494 return ip_getsockopt(sk, level, optname, optval, optlen);
1495}
1496
1497#ifdef CONFIG_COMPAT
1498int compat_udp_getsockopt(struct sock *sk, int level, int optname,
1499 char __user *optval, int __user *optlen)
1500{
1501 if (level == SOL_UDP || level == SOL_UDPLITE)
1502 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1503 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
1504}
1505#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506/**
1507 * udp_poll - wait for a UDP event.
1508 * @file - file struct
1509 * @sock - socket
1510 * @wait - poll table
1511 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001512 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 * blocking sockets. If application is using a blocking fd
1514 * and a packet with checksum error is in the queue;
1515 * then it could get return from select indicating data available
1516 * but then block when reading it. Add special case code
1517 * to work around these arguably broken applications.
1518 */
1519unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1520{
1521 unsigned int mask = datagram_poll(file, sock, wait);
1522 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001523 int is_lite = IS_UDPLITE(sk);
1524
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 /* Check for false positives due to checksum errors */
1526 if ( (mask & POLLRDNORM) &&
1527 !(file->f_flags & O_NONBLOCK) &&
1528 !(sk->sk_shutdown & RCV_SHUTDOWN)){
1529 struct sk_buff_head *rcvq = &sk->sk_receive_queue;
1530 struct sk_buff *skb;
1531
Herbert Xu208d8982005-05-30 15:50:15 -07001532 spin_lock_bh(&rcvq->lock);
Herbert Xu759e5d02007-03-25 20:10:56 -07001533 while ((skb = skb_peek(rcvq)) != NULL &&
1534 udp_lib_checksum_complete(skb)) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001535 UDP_INC_STATS_BH(sock_net(sk),
1536 UDP_MIB_INERRORS, is_lite);
Herbert Xu759e5d02007-03-25 20:10:56 -07001537 __skb_unlink(skb, rcvq);
1538 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 }
Herbert Xu208d8982005-05-30 15:50:15 -07001540 spin_unlock_bh(&rcvq->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541
1542 /* nothing to see, move along */
1543 if (skb == NULL)
1544 mask &= ~(POLLIN | POLLRDNORM);
1545 }
1546
1547 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001548
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549}
1550
David S. Millerdb8dac22008-03-06 16:22:02 -08001551struct proto udp_prot = {
1552 .name = "UDP",
1553 .owner = THIS_MODULE,
1554 .close = udp_lib_close,
1555 .connect = ip4_datagram_connect,
1556 .disconnect = udp_disconnect,
1557 .ioctl = udp_ioctl,
1558 .destroy = udp_destroy_sock,
1559 .setsockopt = udp_setsockopt,
1560 .getsockopt = udp_getsockopt,
1561 .sendmsg = udp_sendmsg,
1562 .recvmsg = udp_recvmsg,
1563 .sendpage = udp_sendpage,
Herbert Xu93821772008-09-15 11:48:46 -07001564 .backlog_rcv = __udp_queue_rcv_skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001565 .hash = udp_lib_hash,
1566 .unhash = udp_lib_unhash,
1567 .get_port = udp_v4_get_port,
1568 .memory_allocated = &udp_memory_allocated,
1569 .sysctl_mem = sysctl_udp_mem,
1570 .sysctl_wmem = &sysctl_udp_wmem_min,
1571 .sysctl_rmem = &sysctl_udp_rmem_min,
1572 .obj_size = sizeof(struct udp_sock),
Eric Dumazet271b72c2008-10-29 02:11:14 -07001573 .slab_flags = SLAB_DESTROY_BY_RCU,
Eric Dumazet645ca702008-10-29 01:41:45 -07001574 .h.udp_table = &udp_table,
David S. Millerdb8dac22008-03-06 16:22:02 -08001575#ifdef CONFIG_COMPAT
1576 .compat_setsockopt = compat_udp_setsockopt,
1577 .compat_getsockopt = compat_udp_getsockopt,
1578#endif
David S. Millerdb8dac22008-03-06 16:22:02 -08001579};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580
1581/* ------------------------------------------------------------------------ */
1582#ifdef CONFIG_PROC_FS
1583
Eric Dumazet645ca702008-10-29 01:41:45 -07001584static struct sock *udp_get_first(struct seq_file *seq, int start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585{
1586 struct sock *sk;
1587 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07001588 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589
Eric Dumazet645ca702008-10-29 01:41:45 -07001590 for (state->bucket = start; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08001591 struct hlist_nulls_node *node;
Eric Dumazet645ca702008-10-29 01:41:45 -07001592 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
1593 spin_lock_bh(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08001594 sk_nulls_for_each(sk, node, &hslot->head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09001595 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07001596 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 if (sk->sk_family == state->family)
1598 goto found;
1599 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001600 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 }
1602 sk = NULL;
1603found:
1604 return sk;
1605}
1606
1607static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1608{
1609 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07001610 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611
1612 do {
Eric Dumazet88ab1932008-11-16 19:39:21 -08001613 sk = sk_nulls_next(sk);
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09001614 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615
Eric Dumazet645ca702008-10-29 01:41:45 -07001616 if (!sk) {
Eric Dumazetf52b5052008-10-29 11:19:11 -07001617 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001618 return udp_get_first(seq, state->bucket + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 }
1620 return sk;
1621}
1622
1623static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
1624{
Eric Dumazet645ca702008-10-29 01:41:45 -07001625 struct sock *sk = udp_get_first(seq, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626
1627 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08001628 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 --pos;
1630 return pos ? NULL : sk;
1631}
1632
1633static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
1634{
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07001635 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636}
1637
1638static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1639{
1640 struct sock *sk;
1641
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07001642 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 sk = udp_get_idx(seq, 0);
1644 else
1645 sk = udp_get_next(seq, v);
1646
1647 ++*pos;
1648 return sk;
1649}
1650
1651static void udp_seq_stop(struct seq_file *seq, void *v)
1652{
Eric Dumazet645ca702008-10-29 01:41:45 -07001653 struct udp_iter_state *state = seq->private;
1654
1655 if (state->bucket < UDP_HTABLE_SIZE)
1656 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657}
1658
1659static int udp_seq_open(struct inode *inode, struct file *file)
1660{
1661 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07001662 struct udp_iter_state *s;
1663 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664
Denis V. Luneva2be75c2008-03-28 18:25:06 -07001665 err = seq_open_net(inode, file, &afinfo->seq_ops,
1666 sizeof(struct udp_iter_state));
1667 if (err < 0)
1668 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07001669
Denis V. Luneva2be75c2008-03-28 18:25:06 -07001670 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 s->family = afinfo->family;
Eric Dumazet645ca702008-10-29 01:41:45 -07001672 s->udp_table = afinfo->udp_table;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07001673 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07001674}
1675
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07001677int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678{
1679 struct proc_dir_entry *p;
1680 int rc = 0;
1681
Denis V. Lunev3ba94412008-03-28 18:25:32 -07001682 afinfo->seq_fops.open = udp_seq_open;
1683 afinfo->seq_fops.read = seq_read;
1684 afinfo->seq_fops.llseek = seq_lseek;
1685 afinfo->seq_fops.release = seq_release_net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686
Denis V. Lunevdda61922008-03-28 18:24:26 -07001687 afinfo->seq_ops.start = udp_seq_start;
1688 afinfo->seq_ops.next = udp_seq_next;
1689 afinfo->seq_ops.stop = udp_seq_stop;
1690
Denis V. Lunev84841c32008-05-02 04:10:08 -07001691 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1692 &afinfo->seq_fops, afinfo);
1693 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694 rc = -ENOMEM;
1695 return rc;
1696}
1697
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07001698void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699{
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07001700 proc_net_remove(net, afinfo->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701}
David S. Millerdb8dac22008-03-06 16:22:02 -08001702
1703/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07001704static void udp4_format_sock(struct sock *sp, struct seq_file *f,
1705 int bucket, int *len)
David S. Millerdb8dac22008-03-06 16:22:02 -08001706{
1707 struct inet_sock *inet = inet_sk(sp);
1708 __be32 dest = inet->daddr;
1709 __be32 src = inet->rcv_saddr;
1710 __u16 destp = ntohs(inet->dport);
1711 __u16 srcp = ntohs(inet->sport);
1712
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07001713 seq_printf(f, "%4d: %08X:%04X %08X:%04X"
Eric Dumazetcb61cb92008-06-17 21:04:56 -07001714 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d%n",
David S. Millerdb8dac22008-03-06 16:22:02 -08001715 bucket, src, srcp, dest, destp, sp->sk_state,
1716 atomic_read(&sp->sk_wmem_alloc),
1717 atomic_read(&sp->sk_rmem_alloc),
1718 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07001719 atomic_read(&sp->sk_refcnt), sp,
1720 atomic_read(&sp->sk_drops), len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001721}
1722
1723int udp4_seq_show(struct seq_file *seq, void *v)
1724{
1725 if (v == SEQ_START_TOKEN)
1726 seq_printf(seq, "%-127s\n",
1727 " sl local_address rem_address st tx_queue "
1728 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07001729 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08001730 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08001731 struct udp_iter_state *state = seq->private;
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07001732 int len;
David S. Millerdb8dac22008-03-06 16:22:02 -08001733
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07001734 udp4_format_sock(v, seq, state->bucket, &len);
1735 seq_printf(seq, "%*s\n", 127 - len ,"");
David S. Millerdb8dac22008-03-06 16:22:02 -08001736 }
1737 return 0;
1738}
1739
1740/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08001741static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08001742 .name = "udp",
1743 .family = AF_INET,
Eric Dumazet645ca702008-10-29 01:41:45 -07001744 .udp_table = &udp_table,
Denis V. Lunev4ad96d32008-03-28 18:25:53 -07001745 .seq_fops = {
1746 .owner = THIS_MODULE,
1747 },
Denis V. Lunevdda61922008-03-28 18:24:26 -07001748 .seq_ops = {
1749 .show = udp4_seq_show,
1750 },
David S. Millerdb8dac22008-03-06 16:22:02 -08001751};
1752
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07001753static int udp4_proc_init_net(struct net *net)
1754{
1755 return udp_proc_register(net, &udp4_seq_afinfo);
1756}
1757
1758static void udp4_proc_exit_net(struct net *net)
1759{
1760 udp_proc_unregister(net, &udp4_seq_afinfo);
1761}
1762
1763static struct pernet_operations udp4_net_ops = {
1764 .init = udp4_proc_init_net,
1765 .exit = udp4_proc_exit_net,
1766};
1767
David S. Millerdb8dac22008-03-06 16:22:02 -08001768int __init udp4_proc_init(void)
1769{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07001770 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001771}
1772
1773void udp4_proc_exit(void)
1774{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07001775 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001776}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777#endif /* CONFIG_PROC_FS */
1778
Eric Dumazet645ca702008-10-29 01:41:45 -07001779void __init udp_table_init(struct udp_table *table)
1780{
1781 int i;
1782
1783 for (i = 0; i < UDP_HTABLE_SIZE; i++) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08001784 INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i);
Eric Dumazet645ca702008-10-29 01:41:45 -07001785 spin_lock_init(&table->hash[i].lock);
1786 }
1787}
1788
Hideo Aoki95766ff2007-12-31 00:29:24 -08001789void __init udp_init(void)
1790{
Eric Dumazet8203efb2008-10-29 02:32:32 -07001791 unsigned long nr_pages, limit;
Hideo Aoki95766ff2007-12-31 00:29:24 -08001792
Eric Dumazet645ca702008-10-29 01:41:45 -07001793 udp_table_init(&udp_table);
Hideo Aoki95766ff2007-12-31 00:29:24 -08001794 /* Set the pressure threshold up by the same strategy of TCP. It is a
1795 * fraction of global memory that is up to 1/2 at 256 MB, decreasing
1796 * toward zero with the amount of memory, with a floor of 128 pages.
1797 */
Eric Dumazet8203efb2008-10-29 02:32:32 -07001798 nr_pages = totalram_pages - totalhigh_pages;
1799 limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
1800 limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
Hideo Aoki95766ff2007-12-31 00:29:24 -08001801 limit = max(limit, 128UL);
1802 sysctl_udp_mem[0] = limit / 4 * 3;
1803 sysctl_udp_mem[1] = limit;
1804 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
1805
1806 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
1807 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
1808}
1809
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810EXPORT_SYMBOL(udp_disconnect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811EXPORT_SYMBOL(udp_ioctl);
David S. Millerdb8dac22008-03-06 16:22:02 -08001812EXPORT_SYMBOL(udp_prot);
1813EXPORT_SYMBOL(udp_sendmsg);
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001814EXPORT_SYMBOL(udp_lib_getsockopt);
1815EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816EXPORT_SYMBOL(udp_poll);
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -07001817EXPORT_SYMBOL(udp_lib_get_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818
1819#ifdef CONFIG_PROC_FS
1820EXPORT_SYMBOL(udp_proc_register);
1821EXPORT_SYMBOL(udp_proc_unregister);
1822#endif