blob: 79c8dbe59b5474bdc3e23ba8adc227e1bee016a4 [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
Joe Perchesafd465032012-03-12 07:03:32 +000080#define pr_fmt(fmt) "UDP: " fmt
81
Linus Torvalds1da177e2005-04-16 15:20:36 -070082#include <asm/uaccess.h>
83#include <asm/ioctls.h>
Hideo Aoki95766ff2007-12-31 00:29:24 -080084#include <linux/bootmem.h>
Eric Dumazet8203efb2008-10-29 02:32:32 -070085#include <linux/highmem.h>
86#include <linux/swap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070087#include <linux/types.h>
88#include <linux/fcntl.h>
89#include <linux/module.h>
90#include <linux/socket.h>
91#include <linux/sockios.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020092#include <linux/igmp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070093#include <linux/in.h>
94#include <linux/errno.h>
95#include <linux/timer.h>
96#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070097#include <linux/inet.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include <linux/netdevice.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090099#include <linux/slab.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -0700100#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101#include <linux/skbuff.h>
102#include <linux/proc_fs.h>
103#include <linux/seq_file.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200104#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#include <net/icmp.h>
106#include <net/route.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#include <net/checksum.h>
108#include <net/xfrm.h>
Satoru Moriya296f7ea2011-06-17 11:58:39 +0000109#include <trace/events/udp.h>
Eric Dumazet447167b2012-04-11 23:05:28 +0000110#include <linux/static_key.h>
Eric Dumazet22911fc2012-06-27 00:23:44 +0000111#include <trace/events/skb.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800112#include "udp_impl.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000114struct udp_table udp_table __read_mostly;
Eric Dumazet645ca702008-10-29 01:41:45 -0700115EXPORT_SYMBOL(udp_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116
Eric Dumazet8d987e52010-11-09 23:24:26 +0000117long sysctl_udp_mem[3] __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800118EXPORT_SYMBOL(sysctl_udp_mem);
Eric Dumazetc482c562009-07-17 00:26:32 +0000119
120int sysctl_udp_rmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800121EXPORT_SYMBOL(sysctl_udp_rmem_min);
Eric Dumazetc482c562009-07-17 00:26:32 +0000122
123int sysctl_udp_wmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800124EXPORT_SYMBOL(sysctl_udp_wmem_min);
125
Eric Dumazet8d987e52010-11-09 23:24:26 +0000126atomic_long_t udp_memory_allocated;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800127EXPORT_SYMBOL(udp_memory_allocated);
128
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000129#define MAX_UDP_PORTS 65536
130#define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN)
Eric Dumazet98322f22009-01-26 21:35:35 -0800131
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700132static int udp_lib_lport_inuse(struct net *net, __u16 num,
Eric Dumazet645ca702008-10-29 01:41:45 -0700133 const struct udp_hslot *hslot,
Eric Dumazet98322f22009-01-26 21:35:35 -0800134 unsigned long *bitmap,
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700135 struct sock *sk,
136 int (*saddr_comp)(const struct sock *sk1,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000137 const struct sock *sk2),
138 unsigned int log)
Gerrit Renker25030a72006-08-26 20:06:05 -0700139{
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700140 struct sock *sk2;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800141 struct hlist_nulls_node *node;
Gerrit Renker25030a72006-08-26 20:06:05 -0700142
Eric Dumazet88ab1932008-11-16 19:39:21 -0800143 sk_nulls_for_each(sk2, node, &hslot->head)
Joe Perches9d4fb272009-11-23 10:41:23 -0800144 if (net_eq(sock_net(sk2), net) &&
145 sk2 != sk &&
Eric Dumazetd4cada42009-11-08 10:17:30 +0000146 (bitmap || udp_sk(sk2)->udp_port_hash == num) &&
Joe Perches9d4fb272009-11-23 10:41:23 -0800147 (!sk2->sk_reuse || !sk->sk_reuse) &&
148 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
149 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Eric Dumazet98322f22009-01-26 21:35:35 -0800150 (*saddr_comp)(sk, sk2)) {
151 if (bitmap)
Eric Dumazetd4cada42009-11-08 10:17:30 +0000152 __set_bit(udp_sk(sk2)->udp_port_hash >> log,
153 bitmap);
Eric Dumazet98322f22009-01-26 21:35:35 -0800154 else
155 return 1;
156 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700157 return 0;
158}
159
Eric Dumazet30fff922009-11-09 05:26:33 +0000160/*
161 * Note: we still hold spinlock of primary hash chain, so no other writer
162 * can insert/delete a socket with local_port == num
163 */
164static int udp_lib_lport_inuse2(struct net *net, __u16 num,
165 struct udp_hslot *hslot2,
166 struct sock *sk,
167 int (*saddr_comp)(const struct sock *sk1,
168 const struct sock *sk2))
169{
170 struct sock *sk2;
171 struct hlist_nulls_node *node;
172 int res = 0;
173
174 spin_lock(&hslot2->lock);
175 udp_portaddr_for_each_entry(sk2, node, &hslot2->head)
Joe Perches9d4fb272009-11-23 10:41:23 -0800176 if (net_eq(sock_net(sk2), net) &&
177 sk2 != sk &&
178 (udp_sk(sk2)->udp_port_hash == num) &&
179 (!sk2->sk_reuse || !sk->sk_reuse) &&
180 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
181 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Eric Dumazet30fff922009-11-09 05:26:33 +0000182 (*saddr_comp)(sk, sk2)) {
183 res = 1;
184 break;
185 }
186 spin_unlock(&hslot2->lock);
187 return res;
188}
189
Gerrit Renker25030a72006-08-26 20:06:05 -0700190/**
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700191 * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
Gerrit Renker25030a72006-08-26 20:06:05 -0700192 *
193 * @sk: socket struct in question
194 * @snum: port number to look up
David S. Millerdf2bc452007-06-05 15:18:43 -0700195 * @saddr_comp: AF-dependent comparison of bound local IP addresses
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300196 * @hash2_nulladdr: AF-dependent hash value in secondary hash chains,
Eric Dumazet30fff922009-11-09 05:26:33 +0000197 * with NULL address
Gerrit Renker25030a72006-08-26 20:06:05 -0700198 */
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700199int udp_lib_get_port(struct sock *sk, unsigned short snum,
David S. Millerdf2bc452007-06-05 15:18:43 -0700200 int (*saddr_comp)(const struct sock *sk1,
Eric Dumazet30fff922009-11-09 05:26:33 +0000201 const struct sock *sk2),
202 unsigned int hash2_nulladdr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203{
Eric Dumazet512615b2009-11-08 10:17:58 +0000204 struct udp_hslot *hslot, *hslot2;
Eric Dumazet645ca702008-10-29 01:41:45 -0700205 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Gerrit Renker25030a72006-08-26 20:06:05 -0700206 int error = 1;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900207 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700209 if (!snum) {
Eric Dumazet9088c562008-10-08 11:44:17 -0700210 int low, high, remaining;
Eric Dumazet95c96172012-04-15 05:58:06 +0000211 unsigned int rand;
Eric Dumazet98322f22009-01-26 21:35:35 -0800212 unsigned short first, last;
213 DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700215 inet_get_local_port_range(&low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700216 remaining = (high - low) + 1;
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700217
Eric Dumazet9088c562008-10-08 11:44:17 -0700218 rand = net_random();
Eric Dumazet98322f22009-01-26 21:35:35 -0800219 first = (((u64)rand * remaining) >> 32) + low;
220 /*
221 * force rand to be an odd multiple of UDP_HTABLE_SIZE
222 */
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000223 rand = (rand | 1) * (udptable->mask + 1);
Eric Dumazet5781b232009-12-13 19:32:39 -0800224 last = first + udptable->mask + 1;
225 do {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000226 hslot = udp_hashslot(udptable, net, first);
Eric Dumazet98322f22009-01-26 21:35:35 -0800227 bitmap_zero(bitmap, PORTS_PER_CHAIN);
Eric Dumazet645ca702008-10-29 01:41:45 -0700228 spin_lock_bh(&hslot->lock);
Eric Dumazet98322f22009-01-26 21:35:35 -0800229 udp_lib_lport_inuse(net, snum, hslot, bitmap, sk,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000230 saddr_comp, udptable->log);
Eric Dumazet98322f22009-01-26 21:35:35 -0800231
232 snum = first;
233 /*
234 * Iterate on all possible values of snum for this hash.
235 * Using steps of an odd multiple of UDP_HTABLE_SIZE
236 * give us randomization and full range coverage.
237 */
Eric Dumazet9088c562008-10-08 11:44:17 -0700238 do {
Eric Dumazet98322f22009-01-26 21:35:35 -0800239 if (low <= snum && snum <= high &&
Amerigo Wange3826f12010-05-05 00:27:06 +0000240 !test_bit(snum >> udptable->log, bitmap) &&
241 !inet_is_reserved_local_port(snum))
Eric Dumazet98322f22009-01-26 21:35:35 -0800242 goto found;
243 snum += rand;
244 } while (snum != first);
245 spin_unlock_bh(&hslot->lock);
Eric Dumazet5781b232009-12-13 19:32:39 -0800246 } while (++first != last);
Eric Dumazet98322f22009-01-26 21:35:35 -0800247 goto fail;
Eric Dumazet645ca702008-10-29 01:41:45 -0700248 } else {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000249 hslot = udp_hashslot(udptable, net, snum);
Eric Dumazet645ca702008-10-29 01:41:45 -0700250 spin_lock_bh(&hslot->lock);
Eric Dumazet30fff922009-11-09 05:26:33 +0000251 if (hslot->count > 10) {
252 int exist;
253 unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum;
254
255 slot2 &= udptable->mask;
256 hash2_nulladdr &= udptable->mask;
257
258 hslot2 = udp_hashslot2(udptable, slot2);
259 if (hslot->count < hslot2->count)
260 goto scan_primary_hash;
261
262 exist = udp_lib_lport_inuse2(net, snum, hslot2,
263 sk, saddr_comp);
264 if (!exist && (hash2_nulladdr != slot2)) {
265 hslot2 = udp_hashslot2(udptable, hash2_nulladdr);
266 exist = udp_lib_lport_inuse2(net, snum, hslot2,
267 sk, saddr_comp);
268 }
269 if (exist)
270 goto fail_unlock;
271 else
272 goto found;
273 }
274scan_primary_hash:
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000275 if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk,
276 saddr_comp, 0))
Eric Dumazet645ca702008-10-29 01:41:45 -0700277 goto fail_unlock;
278 }
Eric Dumazet98322f22009-01-26 21:35:35 -0800279found:
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000280 inet_sk(sk)->inet_num = snum;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000281 udp_sk(sk)->udp_port_hash = snum;
282 udp_sk(sk)->udp_portaddr_hash ^= snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 if (sk_unhashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800284 sk_nulls_add_node_rcu(sk, &hslot->head);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +0000285 hslot->count++;
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700286 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet512615b2009-11-08 10:17:58 +0000287
288 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
289 spin_lock(&hslot2->lock);
290 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
291 &hslot2->head);
292 hslot2->count++;
293 spin_unlock(&hslot2->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700295 error = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -0700296fail_unlock:
297 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298fail:
Gerrit Renker25030a72006-08-26 20:06:05 -0700299 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300}
Eric Dumazetc482c562009-07-17 00:26:32 +0000301EXPORT_SYMBOL(udp_lib_get_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302
Vlad Yasevich499923c2009-04-09 17:37:33 +0000303static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
David S. Millerdb8dac22008-03-06 16:22:02 -0800304{
305 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
306
Eric Dumazetc482c562009-07-17 00:26:32 +0000307 return (!ipv6_only_sock(sk2) &&
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000308 (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr ||
309 inet1->inet_rcv_saddr == inet2->inet_rcv_saddr));
David S. Millerdb8dac22008-03-06 16:22:02 -0800310}
311
Eric Dumazetd4cada42009-11-08 10:17:30 +0000312static unsigned int udp4_portaddr_hash(struct net *net, __be32 saddr,
313 unsigned int port)
314{
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700315 return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000316}
317
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700318int udp_v4_get_port(struct sock *sk, unsigned short snum)
David S. Millerdb8dac22008-03-06 16:22:02 -0800319{
Eric Dumazet30fff922009-11-09 05:26:33 +0000320 unsigned int hash2_nulladdr =
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700321 udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum);
Eric Dumazet30fff922009-11-09 05:26:33 +0000322 unsigned int hash2_partial =
323 udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0);
324
Eric Dumazetd4cada42009-11-08 10:17:30 +0000325 /* precompute partial secondary hash */
Eric Dumazet30fff922009-11-09 05:26:33 +0000326 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
327 return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr);
David S. Millerdb8dac22008-03-06 16:22:02 -0800328}
329
Eric Dumazet645ca702008-10-29 01:41:45 -0700330static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr,
331 unsigned short hnum,
332 __be16 sport, __be32 daddr, __be16 dport, int dif)
333{
334 int score = -1;
335
Eric Dumazetd4cada42009-11-08 10:17:30 +0000336 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
Eric Dumazet645ca702008-10-29 01:41:45 -0700337 !ipv6_only_sock(sk)) {
338 struct inet_sock *inet = inet_sk(sk);
339
340 score = (sk->sk_family == PF_INET ? 1 : 0);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000341 if (inet->inet_rcv_saddr) {
342 if (inet->inet_rcv_saddr != daddr)
Eric Dumazet645ca702008-10-29 01:41:45 -0700343 return -1;
344 score += 2;
345 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000346 if (inet->inet_daddr) {
347 if (inet->inet_daddr != saddr)
Eric Dumazet645ca702008-10-29 01:41:45 -0700348 return -1;
349 score += 2;
350 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000351 if (inet->inet_dport) {
352 if (inet->inet_dport != sport)
Eric Dumazet645ca702008-10-29 01:41:45 -0700353 return -1;
354 score += 2;
355 }
356 if (sk->sk_bound_dev_if) {
357 if (sk->sk_bound_dev_if != dif)
358 return -1;
359 score += 2;
360 }
361 }
362 return score;
363}
364
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000365/*
366 * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num)
367 */
368#define SCORE2_MAX (1 + 2 + 2 + 2)
369static inline int compute_score2(struct sock *sk, struct net *net,
370 __be32 saddr, __be16 sport,
371 __be32 daddr, unsigned int hnum, int dif)
372{
373 int score = -1;
374
375 if (net_eq(sock_net(sk), net) && !ipv6_only_sock(sk)) {
376 struct inet_sock *inet = inet_sk(sk);
377
378 if (inet->inet_rcv_saddr != daddr)
379 return -1;
380 if (inet->inet_num != hnum)
381 return -1;
382
383 score = (sk->sk_family == PF_INET ? 1 : 0);
384 if (inet->inet_daddr) {
385 if (inet->inet_daddr != saddr)
386 return -1;
387 score += 2;
388 }
389 if (inet->inet_dport) {
390 if (inet->inet_dport != sport)
391 return -1;
392 score += 2;
393 }
394 if (sk->sk_bound_dev_if) {
395 if (sk->sk_bound_dev_if != dif)
396 return -1;
397 score += 2;
398 }
399 }
400 return score;
401}
402
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000403
404/* called with read_rcu_lock() */
405static struct sock *udp4_lib_lookup2(struct net *net,
406 __be32 saddr, __be16 sport,
407 __be32 daddr, unsigned int hnum, int dif,
408 struct udp_hslot *hslot2, unsigned int slot2)
409{
410 struct sock *sk, *result;
411 struct hlist_nulls_node *node;
412 int score, badness;
413
414begin:
415 result = NULL;
416 badness = -1;
417 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
418 score = compute_score2(sk, net, saddr, sport,
419 daddr, hnum, dif);
420 if (score > badness) {
421 result = sk;
422 badness = score;
423 if (score == SCORE2_MAX)
424 goto exact_match;
425 }
426 }
427 /*
428 * if the nulls value we got at the end of this lookup is
429 * not the expected one, we must restart lookup.
430 * We probably met an item that was moved to another chain.
431 */
432 if (get_nulls_value(node) != slot2)
433 goto begin;
434
435 if (result) {
436exact_match:
Eric Dumazetc31504d2010-11-15 19:58:26 +0000437 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000438 result = NULL;
439 else if (unlikely(compute_score2(result, net, saddr, sport,
440 daddr, hnum, dif) < badness)) {
441 sock_put(result);
442 goto begin;
443 }
444 }
445 return result;
446}
447
David S. Millerdb8dac22008-03-06 16:22:02 -0800448/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
449 * harder than this. -DaveM
450 */
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000451struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -0800452 __be16 sport, __be32 daddr, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700453 int dif, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800454{
Eric Dumazet271b72c2008-10-29 02:11:14 -0700455 struct sock *sk, *result;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800456 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800457 unsigned short hnum = ntohs(dport);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000458 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
459 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
Eric Dumazet271b72c2008-10-29 02:11:14 -0700460 int score, badness;
David S. Millerdb8dac22008-03-06 16:22:02 -0800461
Eric Dumazet271b72c2008-10-29 02:11:14 -0700462 rcu_read_lock();
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000463 if (hslot->count > 10) {
464 hash2 = udp4_portaddr_hash(net, daddr, hnum);
465 slot2 = hash2 & udptable->mask;
466 hslot2 = &udptable->hash2[slot2];
467 if (hslot->count < hslot2->count)
468 goto begin;
469
470 result = udp4_lib_lookup2(net, saddr, sport,
471 daddr, hnum, dif,
472 hslot2, slot2);
473 if (!result) {
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700474 hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000475 slot2 = hash2 & udptable->mask;
476 hslot2 = &udptable->hash2[slot2];
477 if (hslot->count < hslot2->count)
478 goto begin;
479
Jorge Boncompte [DTI2]1223c672010-04-08 04:56:48 +0000480 result = udp4_lib_lookup2(net, saddr, sport,
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700481 htonl(INADDR_ANY), hnum, dif,
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000482 hslot2, slot2);
483 }
484 rcu_read_unlock();
485 return result;
486 }
Eric Dumazet271b72c2008-10-29 02:11:14 -0700487begin:
488 result = NULL;
489 badness = -1;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800490 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
Eric Dumazet645ca702008-10-29 01:41:45 -0700491 score = compute_score(sk, net, saddr, hnum, sport,
492 daddr, dport, dif);
493 if (score > badness) {
494 result = sk;
495 badness = score;
David S. Millerdb8dac22008-03-06 16:22:02 -0800496 }
497 }
Eric Dumazet88ab1932008-11-16 19:39:21 -0800498 /*
499 * if the nulls value we got at the end of this lookup is
500 * not the expected one, we must restart lookup.
501 * We probably met an item that was moved to another chain.
502 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000503 if (get_nulls_value(node) != slot)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800504 goto begin;
505
Eric Dumazet271b72c2008-10-29 02:11:14 -0700506 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000507 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet271b72c2008-10-29 02:11:14 -0700508 result = NULL;
509 else if (unlikely(compute_score(result, net, saddr, hnum, sport,
510 daddr, dport, dif) < badness)) {
511 sock_put(result);
512 goto begin;
513 }
514 }
515 rcu_read_unlock();
David S. Millerdb8dac22008-03-06 16:22:02 -0800516 return result;
517}
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000518EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
David S. Millerdb8dac22008-03-06 16:22:02 -0800519
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700520static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
521 __be16 sport, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700522 struct udp_table *udptable)
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700523{
KOVACS Krisztian23542612008-10-07 12:41:01 -0700524 struct sock *sk;
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700525 const struct iphdr *iph = ip_hdr(skb);
526
KOVACS Krisztian23542612008-10-07 12:41:01 -0700527 if (unlikely(sk = skb_steal_sock(skb)))
528 return sk;
529 else
Eric Dumazetadf30902009-06-02 05:19:30 +0000530 return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport,
KOVACS Krisztian23542612008-10-07 12:41:01 -0700531 iph->daddr, dport, inet_iif(skb),
532 udptable);
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700533}
534
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700535struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
536 __be32 daddr, __be16 dport, int dif)
537{
Eric Dumazet645ca702008-10-29 01:41:45 -0700538 return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700539}
540EXPORT_SYMBOL_GPL(udp4_lib_lookup);
541
Eric Dumazet920a4612008-11-01 21:22:23 -0700542static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk,
David S. Millerdb8dac22008-03-06 16:22:02 -0800543 __be16 loc_port, __be32 loc_addr,
544 __be16 rmt_port, __be32 rmt_addr,
545 int dif)
546{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800547 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800548 struct sock *s = sk;
549 unsigned short hnum = ntohs(loc_port);
550
Eric Dumazet88ab1932008-11-16 19:39:21 -0800551 sk_nulls_for_each_from(s, node) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800552 struct inet_sock *inet = inet_sk(s);
553
Joe Perches9d4fb272009-11-23 10:41:23 -0800554 if (!net_eq(sock_net(s), net) ||
555 udp_sk(s)->udp_port_hash != hnum ||
556 (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
557 (inet->inet_dport != rmt_port && inet->inet_dport) ||
558 (inet->inet_rcv_saddr &&
559 inet->inet_rcv_saddr != loc_addr) ||
560 ipv6_only_sock(s) ||
David S. Millerdb8dac22008-03-06 16:22:02 -0800561 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
562 continue;
563 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
564 continue;
565 goto found;
566 }
567 s = NULL;
568found:
569 return s;
570}
571
572/*
573 * This routine is called by the ICMP module when it gets some
574 * sort of error condition. If err < 0 then the socket should
575 * be closed and the error returned to the user. If err > 0
576 * it's just the icmp type << 8 | icmp code.
577 * Header points to the ip header of the error packet. We move
578 * on past this. Then (as it used to claim before adjustment)
579 * header points to the first 8 bytes of the udp header. We need
580 * to find the appropriate port.
581 */
582
Eric Dumazet645ca702008-10-29 01:41:45 -0700583void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800584{
585 struct inet_sock *inet;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000586 const struct iphdr *iph = (const struct iphdr *)skb->data;
Eric Dumazetc482c562009-07-17 00:26:32 +0000587 struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
David S. Millerdb8dac22008-03-06 16:22:02 -0800588 const int type = icmp_hdr(skb)->type;
589 const int code = icmp_hdr(skb)->code;
590 struct sock *sk;
591 int harderr;
592 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700593 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800594
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700595 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
David S. Millerdb8dac22008-03-06 16:22:02 -0800596 iph->saddr, uh->source, skb->dev->ifindex, udptable);
597 if (sk == NULL) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700598 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800599 return; /* No socket for error */
600 }
601
602 err = 0;
603 harderr = 0;
604 inet = inet_sk(sk);
605
606 switch (type) {
607 default:
608 case ICMP_TIME_EXCEEDED:
609 err = EHOSTUNREACH;
610 break;
611 case ICMP_SOURCE_QUENCH:
612 goto out;
613 case ICMP_PARAMETERPROB:
614 err = EPROTO;
615 harderr = 1;
616 break;
617 case ICMP_DEST_UNREACH:
618 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
David S. Miller36393392012-06-14 22:21:46 -0700619 ipv4_sk_update_pmtu(skb, sk, info);
David S. Millerdb8dac22008-03-06 16:22:02 -0800620 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
621 err = EMSGSIZE;
622 harderr = 1;
623 break;
624 }
625 goto out;
626 }
627 err = EHOSTUNREACH;
628 if (code <= NR_ICMP_UNREACH) {
629 harderr = icmp_err_convert[code].fatal;
630 err = icmp_err_convert[code].errno;
631 }
632 break;
David S. Miller55be7a92012-07-11 21:27:49 -0700633 case ICMP_REDIRECT:
634 ipv4_sk_redirect(skb, sk);
635 break;
David S. Millerdb8dac22008-03-06 16:22:02 -0800636 }
637
638 /*
639 * RFC1122: OK. Passes ICMP errors back to application, as per
640 * 4.1.3.3.
641 */
642 if (!inet->recverr) {
643 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
644 goto out;
Eric Dumazetb1faf562010-05-31 23:44:05 -0700645 } else
Eric Dumazetc482c562009-07-17 00:26:32 +0000646 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
Eric Dumazetb1faf562010-05-31 23:44:05 -0700647
David S. Millerdb8dac22008-03-06 16:22:02 -0800648 sk->sk_err = err;
649 sk->sk_error_report(sk);
650out:
651 sock_put(sk);
652}
653
654void udp_err(struct sk_buff *skb, u32 info)
655{
Eric Dumazet645ca702008-10-29 01:41:45 -0700656 __udp4_lib_err(skb, info, &udp_table);
David S. Millerdb8dac22008-03-06 16:22:02 -0800657}
658
659/*
660 * Throw away all pending data and cancel the corking. Socket is locked.
661 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400662void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800663{
664 struct udp_sock *up = udp_sk(sk);
665
666 if (up->pending) {
667 up->len = 0;
668 up->pending = 0;
669 ip_flush_pending_frames(sk);
670 }
671}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400672EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800673
674/**
Herbert Xuf6b96642011-03-01 02:36:48 +0000675 * udp4_hwcsum - handle outgoing HW checksumming
David S. Millerdb8dac22008-03-06 16:22:02 -0800676 * @skb: sk_buff containing the filled-in UDP header
677 * (checksum field must be zeroed out)
Herbert Xuf6b96642011-03-01 02:36:48 +0000678 * @src: source IP address
679 * @dst: destination IP address
David S. Millerdb8dac22008-03-06 16:22:02 -0800680 */
Herbert Xuf6b96642011-03-01 02:36:48 +0000681static void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
David S. Millerdb8dac22008-03-06 16:22:02 -0800682{
David S. Millerdb8dac22008-03-06 16:22:02 -0800683 struct udphdr *uh = udp_hdr(skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000684 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
685 int offset = skb_transport_offset(skb);
686 int len = skb->len - offset;
687 int hlen = len;
David S. Millerdb8dac22008-03-06 16:22:02 -0800688 __wsum csum = 0;
689
Herbert Xuf6b96642011-03-01 02:36:48 +0000690 if (!frags) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800691 /*
692 * Only one fragment on the socket.
693 */
694 skb->csum_start = skb_transport_header(skb) - skb->head;
695 skb->csum_offset = offsetof(struct udphdr, check);
Herbert Xuf6b96642011-03-01 02:36:48 +0000696 uh->check = ~csum_tcpudp_magic(src, dst, len,
697 IPPROTO_UDP, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800698 } else {
699 /*
700 * HW-checksum won't work as there are two or more
701 * fragments on the socket so that all csums of sk_buffs
702 * should be together
703 */
Herbert Xuf6b96642011-03-01 02:36:48 +0000704 do {
705 csum = csum_add(csum, frags->csum);
706 hlen -= frags->len;
707 } while ((frags = frags->next));
David S. Millerdb8dac22008-03-06 16:22:02 -0800708
Herbert Xuf6b96642011-03-01 02:36:48 +0000709 csum = skb_checksum(skb, offset, hlen, csum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800710 skb->ip_summed = CHECKSUM_NONE;
711
David S. Millerdb8dac22008-03-06 16:22:02 -0800712 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
713 if (uh->check == 0)
714 uh->check = CSUM_MANGLED_0;
715 }
716}
717
David S. Miller79ab0532011-05-09 13:31:04 -0700718static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
Herbert Xuf6b96642011-03-01 02:36:48 +0000719{
720 struct sock *sk = skb->sk;
721 struct inet_sock *inet = inet_sk(sk);
722 struct udphdr *uh;
Herbert Xuf6b96642011-03-01 02:36:48 +0000723 int err = 0;
724 int is_udplite = IS_UDPLITE(sk);
725 int offset = skb_transport_offset(skb);
726 int len = skb->len - offset;
727 __wsum csum = 0;
728
729 /*
730 * Create a UDP header
731 */
732 uh = udp_hdr(skb);
733 uh->source = inet->inet_sport;
David S. Miller79ab0532011-05-09 13:31:04 -0700734 uh->dest = fl4->fl4_dport;
Herbert Xuf6b96642011-03-01 02:36:48 +0000735 uh->len = htons(len);
736 uh->check = 0;
737
738 if (is_udplite) /* UDP-Lite */
739 csum = udplite_csum(skb);
740
741 else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */
742
743 skb->ip_summed = CHECKSUM_NONE;
744 goto send;
745
746 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
747
David S. Miller79ab0532011-05-09 13:31:04 -0700748 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
Herbert Xuf6b96642011-03-01 02:36:48 +0000749 goto send;
750
751 } else
752 csum = udp_csum(skb);
753
754 /* add protocol-dependent pseudo-header */
David S. Miller79ab0532011-05-09 13:31:04 -0700755 uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
Herbert Xuf6b96642011-03-01 02:36:48 +0000756 sk->sk_protocol, csum);
757 if (uh->check == 0)
758 uh->check = CSUM_MANGLED_0;
759
760send:
Eric Dumazetb5ec8ee2012-08-10 02:22:47 +0000761 err = ip_send_skb(sock_net(sk), skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000762 if (err) {
763 if (err == -ENOBUFS && !inet->recverr) {
764 UDP_INC_STATS_USER(sock_net(sk),
765 UDP_MIB_SNDBUFERRORS, is_udplite);
766 err = 0;
767 }
768 } else
769 UDP_INC_STATS_USER(sock_net(sk),
770 UDP_MIB_OUTDATAGRAMS, is_udplite);
771 return err;
772}
773
David S. Millerdb8dac22008-03-06 16:22:02 -0800774/*
775 * Push out all pending data as one UDP datagram. Socket is locked.
776 */
777static int udp_push_pending_frames(struct sock *sk)
778{
779 struct udp_sock *up = udp_sk(sk);
780 struct inet_sock *inet = inet_sk(sk);
David S. Millerb6f21b22011-03-12 02:09:18 -0500781 struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800782 struct sk_buff *skb;
David S. Millerdb8dac22008-03-06 16:22:02 -0800783 int err = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800784
David S. Miller77968b72011-05-08 17:12:19 -0700785 skb = ip_finish_skb(sk, fl4);
Herbert Xuf6b96642011-03-01 02:36:48 +0000786 if (!skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800787 goto out;
788
David S. Miller79ab0532011-05-09 13:31:04 -0700789 err = udp_send_skb(skb, fl4);
David S. Millerdb8dac22008-03-06 16:22:02 -0800790
David S. Millerdb8dac22008-03-06 16:22:02 -0800791out:
792 up->len = 0;
793 up->pending = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800794 return err;
795}
796
797int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
798 size_t len)
799{
800 struct inet_sock *inet = inet_sk(sk);
801 struct udp_sock *up = udp_sk(sk);
David S. Millere4749952011-05-08 16:38:45 -0700802 struct flowi4 fl4_stack;
David S. Millerb6f21b22011-03-12 02:09:18 -0500803 struct flowi4 *fl4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800804 int ulen = len;
805 struct ipcm_cookie ipc;
806 struct rtable *rt = NULL;
807 int free = 0;
808 int connected = 0;
809 __be32 daddr, faddr, saddr;
810 __be16 dport;
811 u8 tos;
812 int err, is_udplite = IS_UDPLITE(sk);
813 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
814 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
Herbert Xu903ab862011-03-01 02:36:48 +0000815 struct sk_buff *skb;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000816 struct ip_options_data opt_copy;
David S. Millerdb8dac22008-03-06 16:22:02 -0800817
818 if (len > 0xFFFF)
819 return -EMSGSIZE;
820
821 /*
822 * Check the flags.
823 */
824
Eric Dumazetc482c562009-07-17 00:26:32 +0000825 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
David S. Millerdb8dac22008-03-06 16:22:02 -0800826 return -EOPNOTSUPP;
827
828 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000829 ipc.tx_flags = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800830
Herbert Xu903ab862011-03-01 02:36:48 +0000831 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
832
David S. Millerf5fca602011-05-08 17:24:10 -0700833 fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800834 if (up->pending) {
835 /*
836 * There are pending frames.
837 * The socket lock must be held while it's corked.
838 */
839 lock_sock(sk);
840 if (likely(up->pending)) {
841 if (unlikely(up->pending != AF_INET)) {
842 release_sock(sk);
843 return -EINVAL;
844 }
845 goto do_append_data;
846 }
847 release_sock(sk);
848 }
849 ulen += sizeof(struct udphdr);
850
851 /*
852 * Get and verify the address.
853 */
854 if (msg->msg_name) {
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000855 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
David S. Millerdb8dac22008-03-06 16:22:02 -0800856 if (msg->msg_namelen < sizeof(*usin))
857 return -EINVAL;
858 if (usin->sin_family != AF_INET) {
859 if (usin->sin_family != AF_UNSPEC)
860 return -EAFNOSUPPORT;
861 }
862
863 daddr = usin->sin_addr.s_addr;
864 dport = usin->sin_port;
865 if (dport == 0)
866 return -EINVAL;
867 } else {
868 if (sk->sk_state != TCP_ESTABLISHED)
869 return -EDESTADDRREQ;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000870 daddr = inet->inet_daddr;
871 dport = inet->inet_dport;
David S. Millerdb8dac22008-03-06 16:22:02 -0800872 /* Open fast path for connected socket.
873 Route will not be used, if at least one option is set.
874 */
875 connected = 1;
876 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000877 ipc.addr = inet->inet_saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800878
879 ipc.oif = sk->sk_bound_dev_if;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000880 err = sock_tx_timestamp(sk, &ipc.tx_flags);
Patrick Ohly51f31ca2009-02-12 05:03:39 +0000881 if (err)
882 return err;
David S. Millerdb8dac22008-03-06 16:22:02 -0800883 if (msg->msg_controllen) {
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900884 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
David S. Millerdb8dac22008-03-06 16:22:02 -0800885 if (err)
886 return err;
887 if (ipc.opt)
888 free = 1;
889 connected = 0;
890 }
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000891 if (!ipc.opt) {
892 struct ip_options_rcu *inet_opt;
893
894 rcu_read_lock();
895 inet_opt = rcu_dereference(inet->inet_opt);
896 if (inet_opt) {
897 memcpy(&opt_copy, inet_opt,
898 sizeof(*inet_opt) + inet_opt->opt.optlen);
899 ipc.opt = &opt_copy.opt;
900 }
901 rcu_read_unlock();
902 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800903
904 saddr = ipc.addr;
905 ipc.addr = faddr = daddr;
906
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000907 if (ipc.opt && ipc.opt->opt.srr) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800908 if (!daddr)
909 return -EINVAL;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000910 faddr = ipc.opt->opt.faddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800911 connected = 0;
912 }
913 tos = RT_TOS(inet->tos);
914 if (sock_flag(sk, SOCK_LOCALROUTE) ||
915 (msg->msg_flags & MSG_DONTROUTE) ||
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000916 (ipc.opt && ipc.opt->opt.is_strictroute)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800917 tos |= RTO_ONLINK;
918 connected = 0;
919 }
920
921 if (ipv4_is_multicast(daddr)) {
922 if (!ipc.oif)
923 ipc.oif = inet->mc_index;
924 if (!saddr)
925 saddr = inet->mc_addr;
926 connected = 0;
Erich E. Hoover76e21052012-02-08 09:11:07 +0000927 } else if (!ipc.oif)
928 ipc.oif = inet->uc_index;
David S. Millerdb8dac22008-03-06 16:22:02 -0800929
930 if (connected)
Eric Dumazetc482c562009-07-17 00:26:32 +0000931 rt = (struct rtable *)sk_dst_check(sk, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800932
933 if (rt == NULL) {
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -0700934 struct net *net = sock_net(sk);
935
David S. Millere4749952011-05-08 16:38:45 -0700936 fl4 = &fl4_stack;
937 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
David S. Millerc0951cb2011-03-31 04:54:27 -0700938 RT_SCOPE_UNIVERSE, sk->sk_protocol,
939 inet_sk_flowi_flags(sk)|FLOWI_FLAG_CAN_SLEEP,
940 faddr, saddr, dport, inet->inet_sport);
941
David S. Millere4749952011-05-08 16:38:45 -0700942 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
943 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800944 if (IS_ERR(rt)) {
945 err = PTR_ERR(rt);
David S. Miller06dc94b2011-03-03 10:38:01 -0800946 rt = NULL;
David S. Millerdb8dac22008-03-06 16:22:02 -0800947 if (err == -ENETUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700948 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -0800949 goto out;
950 }
951
952 err = -EACCES;
953 if ((rt->rt_flags & RTCF_BROADCAST) &&
954 !sock_flag(sk, SOCK_BROADCAST))
955 goto out;
956 if (connected)
Changli Gaod8d1f302010-06-10 23:31:35 -0700957 sk_dst_set(sk, dst_clone(&rt->dst));
David S. Millerdb8dac22008-03-06 16:22:02 -0800958 }
959
960 if (msg->msg_flags&MSG_CONFIRM)
961 goto do_confirm;
962back_from_confirm:
963
David S. Millere4749952011-05-08 16:38:45 -0700964 saddr = fl4->saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800965 if (!ipc.addr)
David S. Millere4749952011-05-08 16:38:45 -0700966 daddr = ipc.addr = fl4->daddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800967
Herbert Xu903ab862011-03-01 02:36:48 +0000968 /* Lockless fast path for the non-corking case. */
969 if (!corkreq) {
David S. Miller77968b72011-05-08 17:12:19 -0700970 skb = ip_make_skb(sk, fl4, getfrag, msg->msg_iov, ulen,
Herbert Xu903ab862011-03-01 02:36:48 +0000971 sizeof(struct udphdr), &ipc, &rt,
972 msg->msg_flags);
973 err = PTR_ERR(skb);
974 if (skb && !IS_ERR(skb))
David S. Miller79ab0532011-05-09 13:31:04 -0700975 err = udp_send_skb(skb, fl4);
Herbert Xu903ab862011-03-01 02:36:48 +0000976 goto out;
977 }
978
David S. Millerdb8dac22008-03-06 16:22:02 -0800979 lock_sock(sk);
980 if (unlikely(up->pending)) {
981 /* The socket is already corked while preparing it. */
982 /* ... which is an evident application bug. --ANK */
983 release_sock(sk);
984
Joe Perchesafd465032012-03-12 07:03:32 +0000985 LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("cork app bug 2\n"));
David S. Millerdb8dac22008-03-06 16:22:02 -0800986 err = -EINVAL;
987 goto out;
988 }
989 /*
990 * Now cork the socket to pend data.
991 */
David S. Millerb6f21b22011-03-12 02:09:18 -0500992 fl4 = &inet->cork.fl.u.ip4;
993 fl4->daddr = daddr;
994 fl4->saddr = saddr;
David S. Miller9cce96d2011-03-12 03:00:33 -0500995 fl4->fl4_dport = dport;
996 fl4->fl4_sport = inet->inet_sport;
David S. Millerdb8dac22008-03-06 16:22:02 -0800997 up->pending = AF_INET;
998
999do_append_data:
1000 up->len += ulen;
David S. Millerf5fca602011-05-08 17:24:10 -07001001 err = ip_append_data(sk, fl4, getfrag, msg->msg_iov, ulen,
1002 sizeof(struct udphdr), &ipc, &rt,
1003 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001004 if (err)
1005 udp_flush_pending_frames(sk);
1006 else if (!corkreq)
1007 err = udp_push_pending_frames(sk);
1008 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1009 up->pending = 0;
1010 release_sock(sk);
1011
1012out:
1013 ip_rt_put(rt);
1014 if (free)
1015 kfree(ipc.opt);
1016 if (!err)
1017 return len;
1018 /*
1019 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1020 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1021 * we don't have a good statistic (IpOutDiscards but it can be too many
1022 * things). We could add another new stat but at least for now that
1023 * seems like overkill.
1024 */
1025 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001026 UDP_INC_STATS_USER(sock_net(sk),
1027 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001028 }
1029 return err;
1030
1031do_confirm:
Changli Gaod8d1f302010-06-10 23:31:35 -07001032 dst_confirm(&rt->dst);
David S. Millerdb8dac22008-03-06 16:22:02 -08001033 if (!(msg->msg_flags&MSG_PROBE) || len)
1034 goto back_from_confirm;
1035 err = 0;
1036 goto out;
1037}
Eric Dumazetc482c562009-07-17 00:26:32 +00001038EXPORT_SYMBOL(udp_sendmsg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001039
1040int udp_sendpage(struct sock *sk, struct page *page, int offset,
1041 size_t size, int flags)
1042{
David S. Millerf5fca602011-05-08 17:24:10 -07001043 struct inet_sock *inet = inet_sk(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001044 struct udp_sock *up = udp_sk(sk);
1045 int ret;
1046
1047 if (!up->pending) {
1048 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
1049
1050 /* Call udp_sendmsg to specify destination address which
1051 * sendpage interface can't pass.
1052 * This will succeed only when the socket is connected.
1053 */
1054 ret = udp_sendmsg(NULL, sk, &msg, 0);
1055 if (ret < 0)
1056 return ret;
1057 }
1058
1059 lock_sock(sk);
1060
1061 if (unlikely(!up->pending)) {
1062 release_sock(sk);
1063
Joe Perchesafd465032012-03-12 07:03:32 +00001064 LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("udp cork app bug 3\n"));
David S. Millerdb8dac22008-03-06 16:22:02 -08001065 return -EINVAL;
1066 }
1067
David S. Millerf5fca602011-05-08 17:24:10 -07001068 ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1069 page, offset, size, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001070 if (ret == -EOPNOTSUPP) {
1071 release_sock(sk);
1072 return sock_no_sendpage(sk->sk_socket, page, offset,
1073 size, flags);
1074 }
1075 if (ret < 0) {
1076 udp_flush_pending_frames(sk);
1077 goto out;
1078 }
1079
1080 up->len += size;
1081 if (!(up->corkflag || (flags&MSG_MORE)))
1082 ret = udp_push_pending_frames(sk);
1083 if (!ret)
1084 ret = size;
1085out:
1086 release_sock(sk);
1087 return ret;
1088}
1089
Eric Dumazet85584672009-10-09 04:43:40 +00001090
1091/**
1092 * first_packet_length - return length of first packet in receive queue
1093 * @sk: socket
1094 *
1095 * Drops all bad checksum frames, until a valid one is found.
1096 * Returns the length of found skb, or 0 if none is found.
1097 */
1098static unsigned int first_packet_length(struct sock *sk)
1099{
1100 struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1101 struct sk_buff *skb;
1102 unsigned int res;
1103
1104 __skb_queue_head_init(&list_kill);
1105
1106 spin_lock_bh(&rcvq->lock);
1107 while ((skb = skb_peek(rcvq)) != NULL &&
1108 udp_lib_checksum_complete(skb)) {
1109 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1110 IS_UDPLITE(sk));
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001111 atomic_inc(&sk->sk_drops);
Eric Dumazet85584672009-10-09 04:43:40 +00001112 __skb_unlink(skb, rcvq);
1113 __skb_queue_tail(&list_kill, skb);
1114 }
1115 res = skb ? skb->len : 0;
1116 spin_unlock_bh(&rcvq->lock);
1117
1118 if (!skb_queue_empty(&list_kill)) {
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001119 bool slow = lock_sock_fast(sk);
1120
Eric Dumazet85584672009-10-09 04:43:40 +00001121 __skb_queue_purge(&list_kill);
1122 sk_mem_reclaim_partial(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001123 unlock_sock_fast(sk, slow);
Eric Dumazet85584672009-10-09 04:43:40 +00001124 }
1125 return res;
1126}
1127
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128/*
1129 * IOCTL requests applicable to the UDP protocol
1130 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001131
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1133{
Stephen Hemminger6516c652007-03-08 20:41:55 -08001134 switch (cmd) {
1135 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136 {
Eric Dumazet31e6d362009-06-17 19:05:41 -07001137 int amount = sk_wmem_alloc_get(sk);
1138
Stephen Hemminger6516c652007-03-08 20:41:55 -08001139 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 }
Stephen Hemminger6516c652007-03-08 20:41:55 -08001141
1142 case SIOCINQ:
1143 {
Eric Dumazet85584672009-10-09 04:43:40 +00001144 unsigned int amount = first_packet_length(sk);
Stephen Hemminger6516c652007-03-08 20:41:55 -08001145
Eric Dumazet85584672009-10-09 04:43:40 +00001146 if (amount)
Stephen Hemminger6516c652007-03-08 20:41:55 -08001147 /*
1148 * We will only return the amount
1149 * of this packet since that is all
1150 * that will be read.
1151 */
Eric Dumazet85584672009-10-09 04:43:40 +00001152 amount -= sizeof(struct udphdr);
1153
Stephen Hemminger6516c652007-03-08 20:41:55 -08001154 return put_user(amount, (int __user *)arg);
1155 }
1156
1157 default:
1158 return -ENOIOCTLCMD;
1159 }
1160
1161 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162}
Eric Dumazetc482c562009-07-17 00:26:32 +00001163EXPORT_SYMBOL(udp_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164
David S. Millerdb8dac22008-03-06 16:22:02 -08001165/*
1166 * This should be easy, if there is something there we
1167 * return it, otherwise we block.
1168 */
1169
1170int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
1171 size_t len, int noblock, int flags, int *addr_len)
1172{
1173 struct inet_sock *inet = inet_sk(sk);
1174 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
1175 struct sk_buff *skb;
David S. Miller59c2cda2011-12-01 14:12:55 -05001176 unsigned int ulen, copied;
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001177 int peeked, off = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001178 int err;
1179 int is_udplite = IS_UDPLITE(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001180 bool slow;
David S. Millerdb8dac22008-03-06 16:22:02 -08001181
1182 /*
1183 * Check any passed addresses
1184 */
1185 if (addr_len)
Eric Dumazetc482c562009-07-17 00:26:32 +00001186 *addr_len = sizeof(*sin);
David S. Millerdb8dac22008-03-06 16:22:02 -08001187
1188 if (flags & MSG_ERRQUEUE)
1189 return ip_recv_error(sk, msg, len);
1190
1191try_again:
1192 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001193 &peeked, &off, &err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001194 if (!skb)
1195 goto out;
1196
1197 ulen = skb->len - sizeof(struct udphdr);
David S. Miller59c2cda2011-12-01 14:12:55 -05001198 copied = len;
1199 if (copied > ulen)
1200 copied = ulen;
1201 else if (copied < ulen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001202 msg->msg_flags |= MSG_TRUNC;
1203
1204 /*
1205 * If checksum is needed at all, try to do it while copying the
1206 * data. If the data is truncated, or if we only want a partial
1207 * coverage checksum (UDP-Lite), do it before the copy.
1208 */
1209
David S. Miller59c2cda2011-12-01 14:12:55 -05001210 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001211 if (udp_lib_checksum_complete(skb))
1212 goto csum_copy_err;
1213 }
1214
1215 if (skb_csum_unnecessary(skb))
1216 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
David S. Miller59c2cda2011-12-01 14:12:55 -05001217 msg->msg_iov, copied);
David S. Millerdb8dac22008-03-06 16:22:02 -08001218 else {
Eric Dumazetc482c562009-07-17 00:26:32 +00001219 err = skb_copy_and_csum_datagram_iovec(skb,
1220 sizeof(struct udphdr),
1221 msg->msg_iov);
David S. Millerdb8dac22008-03-06 16:22:02 -08001222
1223 if (err == -EINVAL)
1224 goto csum_copy_err;
1225 }
1226
Eric Dumazet22911fc2012-06-27 00:23:44 +00001227 if (unlikely(err)) {
1228 trace_kfree_skb(skb, udp_recvmsg);
Eric Dumazet979402b2012-09-05 23:34:44 +00001229 if (!peeked) {
1230 atomic_inc(&sk->sk_drops);
1231 UDP_INC_STATS_USER(sock_net(sk),
1232 UDP_MIB_INERRORS, is_udplite);
1233 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001234 goto out_free;
Eric Dumazet22911fc2012-06-27 00:23:44 +00001235 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001236
1237 if (!peeked)
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001238 UDP_INC_STATS_USER(sock_net(sk),
1239 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001240
Neil Horman3b885782009-10-12 13:26:31 -07001241 sock_recv_ts_and_drops(msg, sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001242
1243 /* Copy the address. */
Eric Dumazetc482c562009-07-17 00:26:32 +00001244 if (sin) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001245 sin->sin_family = AF_INET;
1246 sin->sin_port = udp_hdr(skb)->source;
1247 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1248 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
1249 }
1250 if (inet->cmsg_flags)
1251 ip_cmsg_recv(msg, skb);
1252
David S. Miller59c2cda2011-12-01 14:12:55 -05001253 err = copied;
David S. Millerdb8dac22008-03-06 16:22:02 -08001254 if (flags & MSG_TRUNC)
1255 err = ulen;
1256
1257out_free:
Eric Dumazet9d410c72009-10-30 05:03:53 +00001258 skb_free_datagram_locked(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001259out:
1260 return err;
1261
1262csum_copy_err:
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001263 slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001264 if (!skb_kill_datagram(sk, skb, flags))
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001265 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001266 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001267
1268 if (noblock)
1269 return -EAGAIN;
Xufeng Zhang9cfaa8d2011-06-21 10:43:40 +00001270
1271 /* starting over for a new packet */
1272 msg->msg_flags &= ~MSG_TRUNC;
David S. Millerdb8dac22008-03-06 16:22:02 -08001273 goto try_again;
1274}
1275
1276
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277int udp_disconnect(struct sock *sk, int flags)
1278{
1279 struct inet_sock *inet = inet_sk(sk);
1280 /*
1281 * 1003.1g - break association.
1282 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001283
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 sk->sk_state = TCP_CLOSE;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001285 inet->inet_daddr = 0;
1286 inet->inet_dport = 0;
Tom Herbertbdeab992011-08-14 19:45:55 +00001287 sock_rps_reset_rxhash(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 sk->sk_bound_dev_if = 0;
1289 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1290 inet_reset_saddr(sk);
1291
1292 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1293 sk->sk_prot->unhash(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001294 inet->inet_sport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 }
1296 sk_dst_reset(sk);
1297 return 0;
1298}
Eric Dumazetc482c562009-07-17 00:26:32 +00001299EXPORT_SYMBOL(udp_disconnect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300
Eric Dumazet645ca702008-10-29 01:41:45 -07001301void udp_lib_unhash(struct sock *sk)
1302{
Eric Dumazet723b4612008-11-25 13:55:15 -08001303 if (sk_hashed(sk)) {
1304 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Eric Dumazet512615b2009-11-08 10:17:58 +00001305 struct udp_hslot *hslot, *hslot2;
1306
1307 hslot = udp_hashslot(udptable, sock_net(sk),
1308 udp_sk(sk)->udp_port_hash);
1309 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
Eric Dumazet645ca702008-10-29 01:41:45 -07001310
Eric Dumazet723b4612008-11-25 13:55:15 -08001311 spin_lock_bh(&hslot->lock);
1312 if (sk_nulls_del_node_init_rcu(sk)) {
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00001313 hslot->count--;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001314 inet_sk(sk)->inet_num = 0;
Eric Dumazet723b4612008-11-25 13:55:15 -08001315 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet512615b2009-11-08 10:17:58 +00001316
1317 spin_lock(&hslot2->lock);
1318 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1319 hslot2->count--;
1320 spin_unlock(&hslot2->lock);
Eric Dumazet723b4612008-11-25 13:55:15 -08001321 }
1322 spin_unlock_bh(&hslot->lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001323 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001324}
1325EXPORT_SYMBOL(udp_lib_unhash);
1326
Eric Dumazet719f8352010-09-08 05:08:44 +00001327/*
1328 * inet_rcv_saddr was changed, we must rehash secondary hash
1329 */
1330void udp_lib_rehash(struct sock *sk, u16 newhash)
1331{
1332 if (sk_hashed(sk)) {
1333 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1334 struct udp_hslot *hslot, *hslot2, *nhslot2;
1335
1336 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1337 nhslot2 = udp_hashslot2(udptable, newhash);
1338 udp_sk(sk)->udp_portaddr_hash = newhash;
1339 if (hslot2 != nhslot2) {
1340 hslot = udp_hashslot(udptable, sock_net(sk),
1341 udp_sk(sk)->udp_port_hash);
1342 /* we must lock primary chain too */
1343 spin_lock_bh(&hslot->lock);
1344
1345 spin_lock(&hslot2->lock);
1346 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1347 hslot2->count--;
1348 spin_unlock(&hslot2->lock);
1349
1350 spin_lock(&nhslot2->lock);
1351 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
1352 &nhslot2->head);
1353 nhslot2->count++;
1354 spin_unlock(&nhslot2->lock);
1355
1356 spin_unlock_bh(&hslot->lock);
1357 }
1358 }
1359}
1360EXPORT_SYMBOL(udp_lib_rehash);
1361
1362static void udp_v4_rehash(struct sock *sk)
1363{
1364 u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1365 inet_sk(sk)->inet_rcv_saddr,
1366 inet_sk(sk)->inet_num);
1367 udp_lib_rehash(sk, new_hash);
1368}
1369
Herbert Xu93821772008-09-15 11:48:46 -07001370static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1371{
Tom Herbertfec5e652010-04-16 16:01:27 -07001372 int rc;
Herbert Xu93821772008-09-15 11:48:46 -07001373
Tom Herbertfec5e652010-04-16 16:01:27 -07001374 if (inet_sk(sk)->inet_daddr)
Tom Herbertbdeab992011-08-14 19:45:55 +00001375 sock_rps_save_rxhash(sk, skb);
Tom Herbertfec5e652010-04-16 16:01:27 -07001376
Eric Dumazetd826eb12011-11-09 07:24:35 +00001377 rc = sock_queue_rcv_skb(sk, skb);
Eric Dumazet766e90372009-10-14 20:40:11 -07001378 if (rc < 0) {
1379 int is_udplite = IS_UDPLITE(sk);
1380
Herbert Xu93821772008-09-15 11:48:46 -07001381 /* Note that an ENOMEM error is charged twice */
Eric Dumazet766e90372009-10-14 20:40:11 -07001382 if (rc == -ENOMEM)
Herbert Xu93821772008-09-15 11:48:46 -07001383 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1384 is_udplite);
Eric Dumazet766e90372009-10-14 20:40:11 -07001385 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1386 kfree_skb(skb);
Satoru Moriya296f7ea2011-06-17 11:58:39 +00001387 trace_udp_fail_queue_rcv_skb(rc, sk);
Eric Dumazet766e90372009-10-14 20:40:11 -07001388 return -1;
Herbert Xu93821772008-09-15 11:48:46 -07001389 }
1390
1391 return 0;
1392
Herbert Xu93821772008-09-15 11:48:46 -07001393}
1394
Eric Dumazet447167b2012-04-11 23:05:28 +00001395static struct static_key udp_encap_needed __read_mostly;
1396void udp_encap_enable(void)
1397{
1398 if (!static_key_enabled(&udp_encap_needed))
1399 static_key_slow_inc(&udp_encap_needed);
1400}
1401EXPORT_SYMBOL(udp_encap_enable);
1402
David S. Millerdb8dac22008-03-06 16:22:02 -08001403/* returns:
1404 * -1: error
1405 * 0: success
1406 * >0: "udp encap" protocol resubmission
1407 *
1408 * Note that in the success and error cases, the skb is assumed to
1409 * have either been requeued or freed.
1410 */
Eric Dumazetc482c562009-07-17 00:26:32 +00001411int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -08001412{
1413 struct udp_sock *up = udp_sk(sk);
1414 int rc;
1415 int is_udplite = IS_UDPLITE(sk);
1416
1417 /*
1418 * Charge it to the socket, dropping if the queue is full.
1419 */
1420 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1421 goto drop;
1422 nf_reset(skb);
1423
Eric Dumazet447167b2012-04-11 23:05:28 +00001424 if (static_key_false(&udp_encap_needed) && up->encap_type) {
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001425 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1426
David S. Millerdb8dac22008-03-06 16:22:02 -08001427 /*
1428 * This is an encapsulation socket so pass the skb to
1429 * the socket's udp_encap_rcv() hook. Otherwise, just
1430 * fall through and pass this up the UDP socket.
1431 * up->encap_rcv() returns the following value:
1432 * =0 if skb was successfully passed to the encap
1433 * handler or was discarded by it.
1434 * >0 if skb should be passed on to UDP.
1435 * <0 if skb should be resubmitted as proto -N
1436 */
1437
1438 /* if we're overly short, let UDP handle it */
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001439 encap_rcv = ACCESS_ONCE(up->encap_rcv);
1440 if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001441 int ret;
1442
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001443 ret = encap_rcv(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001444 if (ret <= 0) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001445 UDP_INC_STATS_BH(sock_net(sk),
1446 UDP_MIB_INDATAGRAMS,
David S. Millerdb8dac22008-03-06 16:22:02 -08001447 is_udplite);
1448 return -ret;
1449 }
1450 }
1451
1452 /* FALLTHROUGH -- it's a UDP Packet */
1453 }
1454
1455 /*
1456 * UDP-Lite specific tests, ignored on UDP sockets
1457 */
1458 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1459
1460 /*
1461 * MIB statistics other than incrementing the error count are
1462 * disabled for the following two types of errors: these depend
1463 * on the application settings, not on the functioning of the
1464 * protocol stack as such.
1465 *
1466 * RFC 3828 here recommends (sec 3.3): "There should also be a
1467 * way ... to ... at least let the receiving application block
1468 * delivery of packets with coverage values less than a value
1469 * provided by the application."
1470 */
1471 if (up->pcrlen == 0) { /* full coverage was set */
Joe Perchesafd465032012-03-12 07:03:32 +00001472 LIMIT_NETDEBUG(KERN_WARNING "UDPLite: partial coverage %d while full coverage %d requested\n",
1473 UDP_SKB_CB(skb)->cscov, skb->len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001474 goto drop;
1475 }
1476 /* The next case involves violating the min. coverage requested
1477 * by the receiver. This is subtle: if receiver wants x and x is
1478 * greater than the buffersize/MTU then receiver will complain
1479 * that it wants x while sender emits packets of smaller size y.
1480 * Therefore the above ...()->partial_cov statement is essential.
1481 */
1482 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
Joe Perchesafd465032012-03-12 07:03:32 +00001483 LIMIT_NETDEBUG(KERN_WARNING "UDPLite: coverage %d too small, need min %d\n",
1484 UDP_SKB_CB(skb)->cscov, up->pcrlen);
David S. Millerdb8dac22008-03-06 16:22:02 -08001485 goto drop;
1486 }
1487 }
1488
Eric Dumazet33d480c2011-08-11 19:30:52 +00001489 if (rcu_access_pointer(sk->sk_filter) &&
1490 udp_lib_checksum_complete(skb))
1491 goto drop;
David S. Millerdb8dac22008-03-06 16:22:02 -08001492
Eric Dumazetc3774112010-04-27 15:13:20 -07001493
Eric Dumazetf545a382012-04-22 23:34:26 +00001494 if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf))
Eric Dumazetc3774112010-04-27 15:13:20 -07001495 goto drop;
1496
Herbert Xu93821772008-09-15 11:48:46 -07001497 rc = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001498
Eric Dumazetd826eb12011-11-09 07:24:35 +00001499 ipv4_pktinfo_prepare(skb);
Herbert Xu93821772008-09-15 11:48:46 -07001500 bh_lock_sock(sk);
1501 if (!sock_owned_by_user(sk))
1502 rc = __udp_queue_rcv_skb(sk, skb);
Eric Dumazetf545a382012-04-22 23:34:26 +00001503 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
Zhu Yi55349792010-03-04 18:01:42 +00001504 bh_unlock_sock(sk);
1505 goto drop;
1506 }
Herbert Xu93821772008-09-15 11:48:46 -07001507 bh_unlock_sock(sk);
1508
1509 return rc;
David S. Millerdb8dac22008-03-06 16:22:02 -08001510
1511drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001512 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001513 atomic_inc(&sk->sk_drops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001514 kfree_skb(skb);
1515 return -1;
1516}
1517
Eric Dumazet1240d132009-11-08 10:18:44 +00001518
1519static void flush_stack(struct sock **stack, unsigned int count,
1520 struct sk_buff *skb, unsigned int final)
1521{
1522 unsigned int i;
1523 struct sk_buff *skb1 = NULL;
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001524 struct sock *sk;
Eric Dumazet1240d132009-11-08 10:18:44 +00001525
1526 for (i = 0; i < count; i++) {
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001527 sk = stack[i];
Eric Dumazet1240d132009-11-08 10:18:44 +00001528 if (likely(skb1 == NULL))
1529 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
1530
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001531 if (!skb1) {
1532 atomic_inc(&sk->sk_drops);
1533 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1534 IS_UDPLITE(sk));
1535 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1536 IS_UDPLITE(sk));
1537 }
1538
1539 if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0)
Eric Dumazet1240d132009-11-08 10:18:44 +00001540 skb1 = NULL;
1541 }
1542 if (unlikely(skb1))
1543 kfree_skb(skb1);
1544}
1545
David S. Millerdb8dac22008-03-06 16:22:02 -08001546/*
1547 * Multicasts and broadcasts go to each listener.
1548 *
Eric Dumazet1240d132009-11-08 10:18:44 +00001549 * Note: called only from the BH handler context.
David S. Millerdb8dac22008-03-06 16:22:02 -08001550 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001551static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001552 struct udphdr *uh,
1553 __be32 saddr, __be32 daddr,
Eric Dumazet645ca702008-10-29 01:41:45 -07001554 struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -08001555{
Eric Dumazet1240d132009-11-08 10:18:44 +00001556 struct sock *sk, *stack[256 / sizeof(struct sock *)];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00001557 struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001558 int dif;
Eric Dumazet1240d132009-11-08 10:18:44 +00001559 unsigned int i, count = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001560
Eric Dumazet645ca702008-10-29 01:41:45 -07001561 spin_lock(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08001562 sk = sk_nulls_head(&hslot->head);
David S. Millerdb8dac22008-03-06 16:22:02 -08001563 dif = skb->dev->ifindex;
Eric Dumazet920a4612008-11-01 21:22:23 -07001564 sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
Eric Dumazet1240d132009-11-08 10:18:44 +00001565 while (sk) {
1566 stack[count++] = sk;
1567 sk = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest,
1568 daddr, uh->source, saddr, dif);
1569 if (unlikely(count == ARRAY_SIZE(stack))) {
1570 if (!sk)
1571 break;
1572 flush_stack(stack, count, skb, ~0);
1573 count = 0;
1574 }
1575 }
1576 /*
1577 * before releasing chain lock, we must take a reference on sockets
1578 */
1579 for (i = 0; i < count; i++)
1580 sock_hold(stack[i]);
David S. Millerdb8dac22008-03-06 16:22:02 -08001581
Eric Dumazet645ca702008-10-29 01:41:45 -07001582 spin_unlock(&hslot->lock);
Eric Dumazet1240d132009-11-08 10:18:44 +00001583
1584 /*
1585 * do the slow work with no lock held
1586 */
1587 if (count) {
1588 flush_stack(stack, count, skb, count - 1);
1589
1590 for (i = 0; i < count; i++)
1591 sock_put(stack[i]);
1592 } else {
1593 kfree_skb(skb);
1594 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001595 return 0;
1596}
1597
1598/* Initialize UDP checksum. If exited with zero value (success),
1599 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1600 * Otherwise, csum completion requires chacksumming packet body,
1601 * including udp header and folding it to skb->csum.
1602 */
1603static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1604 int proto)
1605{
1606 const struct iphdr *iph;
1607 int err;
1608
1609 UDP_SKB_CB(skb)->partial_cov = 0;
1610 UDP_SKB_CB(skb)->cscov = skb->len;
1611
1612 if (proto == IPPROTO_UDPLITE) {
1613 err = udplite_checksum_init(skb, uh);
1614 if (err)
1615 return err;
1616 }
1617
1618 iph = ip_hdr(skb);
1619 if (uh->check == 0) {
1620 skb->ip_summed = CHECKSUM_UNNECESSARY;
1621 } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
Eric Dumazetc482c562009-07-17 00:26:32 +00001622 if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len,
David S. Millerdb8dac22008-03-06 16:22:02 -08001623 proto, skb->csum))
1624 skb->ip_summed = CHECKSUM_UNNECESSARY;
1625 }
1626 if (!skb_csum_unnecessary(skb))
1627 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
1628 skb->len, proto, 0);
1629 /* Probably, we should checksum udp header (it should be in cache
1630 * in any case) and data in tiny packets (< rx copybreak).
1631 */
1632
1633 return 0;
1634}
1635
1636/*
1637 * All we need to do is get the socket, and then do a checksum.
1638 */
1639
Eric Dumazet645ca702008-10-29 01:41:45 -07001640int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
David S. Millerdb8dac22008-03-06 16:22:02 -08001641 int proto)
1642{
1643 struct sock *sk;
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001644 struct udphdr *uh;
David S. Millerdb8dac22008-03-06 16:22:02 -08001645 unsigned short ulen;
Eric Dumazetadf30902009-06-02 05:19:30 +00001646 struct rtable *rt = skb_rtable(skb);
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001647 __be32 saddr, daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001648 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001649
1650 /*
1651 * Validate the packet.
1652 */
1653 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1654 goto drop; /* No space for header. */
1655
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001656 uh = udp_hdr(skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001657 ulen = ntohs(uh->len);
Bjørn Morkccc2d972010-05-06 03:44:34 +00001658 saddr = ip_hdr(skb)->saddr;
1659 daddr = ip_hdr(skb)->daddr;
1660
David S. Millerdb8dac22008-03-06 16:22:02 -08001661 if (ulen > skb->len)
1662 goto short_packet;
1663
1664 if (proto == IPPROTO_UDP) {
1665 /* UDP validates ulen. */
1666 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1667 goto short_packet;
1668 uh = udp_hdr(skb);
1669 }
1670
1671 if (udp4_csum_init(skb, uh, proto))
1672 goto csum_error;
1673
1674 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
Pavel Emelyanove3163492008-06-16 17:12:11 -07001675 return __udp4_lib_mcast_deliver(net, skb, uh,
1676 saddr, daddr, udptable);
David S. Millerdb8dac22008-03-06 16:22:02 -08001677
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -07001678 sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
David S. Millerdb8dac22008-03-06 16:22:02 -08001679
1680 if (sk != NULL) {
Herbert Xu93821772008-09-15 11:48:46 -07001681 int ret = udp_queue_rcv_skb(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001682 sock_put(sk);
1683
1684 /* a return value > 0 means to resubmit the input, but
1685 * it wants the return to be -protocol, or 0
1686 */
1687 if (ret > 0)
1688 return -ret;
1689 return 0;
1690 }
1691
1692 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1693 goto drop;
1694 nf_reset(skb);
1695
1696 /* No socket. Drop packet silently, if checksum is wrong */
1697 if (udp_lib_checksum_complete(skb))
1698 goto csum_error;
1699
Pavel Emelyanov02833282008-07-05 21:18:48 -07001700 UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001701 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1702
1703 /*
1704 * Hmm. We got an UDP packet to a port to which we
1705 * don't wanna listen. Ignore it.
1706 */
1707 kfree_skb(skb);
1708 return 0;
1709
1710short_packet:
Harvey Harrison673d57e2008-10-31 00:53:57 -07001711 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
Joe Perchesafd465032012-03-12 07:03:32 +00001712 proto == IPPROTO_UDPLITE ? "Lite" : "",
1713 &saddr, ntohs(uh->source),
1714 ulen, skb->len,
1715 &daddr, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001716 goto drop;
1717
1718csum_error:
1719 /*
1720 * RFC1122: OK. Discards the bad packet silently (as far as
1721 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1722 */
Harvey Harrison673d57e2008-10-31 00:53:57 -07001723 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
Joe Perchesafd465032012-03-12 07:03:32 +00001724 proto == IPPROTO_UDPLITE ? "Lite" : "",
1725 &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
David S. Millerdb8dac22008-03-06 16:22:02 -08001726 ulen);
1727drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001728 UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001729 kfree_skb(skb);
1730 return 0;
1731}
1732
1733int udp_rcv(struct sk_buff *skb)
1734{
Eric Dumazet645ca702008-10-29 01:41:45 -07001735 return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
David S. Millerdb8dac22008-03-06 16:22:02 -08001736}
1737
Brian Haley7d06b2e2008-06-14 17:04:49 -07001738void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08001739{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001740 bool slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001741 udp_flush_pending_frames(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001742 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001743}
1744
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745/*
1746 * Socket option code for UDP
1747 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001748int udp_lib_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001749 char __user *optval, unsigned int optlen,
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001750 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751{
1752 struct udp_sock *up = udp_sk(sk);
1753 int val;
1754 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11001755 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756
Eric Dumazetc482c562009-07-17 00:26:32 +00001757 if (optlen < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758 return -EINVAL;
1759
1760 if (get_user(val, (int __user *)optval))
1761 return -EFAULT;
1762
Stephen Hemminger6516c652007-03-08 20:41:55 -08001763 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764 case UDP_CORK:
1765 if (val != 0) {
1766 up->corkflag = 1;
1767 } else {
1768 up->corkflag = 0;
1769 lock_sock(sk);
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001770 (*push_pending_frames)(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 release_sock(sk);
1772 }
1773 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001774
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775 case UDP_ENCAP:
1776 switch (val) {
1777 case 0:
1778 case UDP_ENCAP_ESPINUDP:
1779 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07001780 up->encap_rcv = xfrm4_udp_encap_rcv;
1781 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07001782 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783 up->encap_type = val;
Eric Dumazet447167b2012-04-11 23:05:28 +00001784 udp_encap_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 break;
1786 default:
1787 err = -ENOPROTOOPT;
1788 break;
1789 }
1790 break;
1791
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001792 /*
1793 * UDP-Lite's partial checksum coverage (RFC 3828).
1794 */
1795 /* The sender sets actual checksum coverage length via this option.
1796 * The case coverage > packet length is handled by send module. */
1797 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11001798 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001799 return -ENOPROTOOPT;
1800 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
1801 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07001802 else if (val > USHRT_MAX)
1803 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001804 up->pcslen = val;
1805 up->pcflag |= UDPLITE_SEND_CC;
1806 break;
1807
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001808 /* The receiver specifies a minimum checksum coverage value. To make
1809 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001810 * used, this again means full checksum coverage. */
1811 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11001812 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001813 return -ENOPROTOOPT;
1814 if (val != 0 && val < 8) /* Avoid silly minimal values. */
1815 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07001816 else if (val > USHRT_MAX)
1817 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001818 up->pcrlen = val;
1819 up->pcflag |= UDPLITE_RECV_CC;
1820 break;
1821
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822 default:
1823 err = -ENOPROTOOPT;
1824 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001825 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826
1827 return err;
1828}
Eric Dumazetc482c562009-07-17 00:26:32 +00001829EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830
David S. Millerdb8dac22008-03-06 16:22:02 -08001831int udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001832 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001833{
1834 if (level == SOL_UDP || level == SOL_UDPLITE)
1835 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1836 udp_push_pending_frames);
1837 return ip_setsockopt(sk, level, optname, optval, optlen);
1838}
1839
1840#ifdef CONFIG_COMPAT
1841int compat_udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001842 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001843{
1844 if (level == SOL_UDP || level == SOL_UDPLITE)
1845 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1846 udp_push_pending_frames);
1847 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
1848}
1849#endif
1850
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001851int udp_lib_getsockopt(struct sock *sk, int level, int optname,
1852 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853{
1854 struct udp_sock *up = udp_sk(sk);
1855 int val, len;
1856
Eric Dumazetc482c562009-07-17 00:26:32 +00001857 if (get_user(len, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 return -EFAULT;
1859
1860 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001861
Stephen Hemminger6516c652007-03-08 20:41:55 -08001862 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 return -EINVAL;
1864
Stephen Hemminger6516c652007-03-08 20:41:55 -08001865 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866 case UDP_CORK:
1867 val = up->corkflag;
1868 break;
1869
1870 case UDP_ENCAP:
1871 val = up->encap_type;
1872 break;
1873
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001874 /* The following two cannot be changed on UDP sockets, the return is
1875 * always 0 (which corresponds to the full checksum coverage of UDP). */
1876 case UDPLITE_SEND_CSCOV:
1877 val = up->pcslen;
1878 break;
1879
1880 case UDPLITE_RECV_CSCOV:
1881 val = up->pcrlen;
1882 break;
1883
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884 default:
1885 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001886 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887
Stephen Hemminger6516c652007-03-08 20:41:55 -08001888 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001889 return -EFAULT;
Eric Dumazetc482c562009-07-17 00:26:32 +00001890 if (copy_to_user(optval, &val, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001892 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893}
Eric Dumazetc482c562009-07-17 00:26:32 +00001894EXPORT_SYMBOL(udp_lib_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895
David S. Millerdb8dac22008-03-06 16:22:02 -08001896int udp_getsockopt(struct sock *sk, int level, int optname,
1897 char __user *optval, int __user *optlen)
1898{
1899 if (level == SOL_UDP || level == SOL_UDPLITE)
1900 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1901 return ip_getsockopt(sk, level, optname, optval, optlen);
1902}
1903
1904#ifdef CONFIG_COMPAT
1905int compat_udp_getsockopt(struct sock *sk, int level, int optname,
1906 char __user *optval, int __user *optlen)
1907{
1908 if (level == SOL_UDP || level == SOL_UDPLITE)
1909 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1910 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
1911}
1912#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913/**
1914 * udp_poll - wait for a UDP event.
1915 * @file - file struct
1916 * @sock - socket
1917 * @wait - poll table
1918 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001919 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920 * blocking sockets. If application is using a blocking fd
1921 * and a packet with checksum error is in the queue;
1922 * then it could get return from select indicating data available
1923 * but then block when reading it. Add special case code
1924 * to work around these arguably broken applications.
1925 */
1926unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1927{
1928 unsigned int mask = datagram_poll(file, sock, wait);
1929 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001930
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931 /* Check for false positives due to checksum errors */
Eric Dumazet85584672009-10-09 04:43:40 +00001932 if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
1933 !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk))
1934 mask &= ~(POLLIN | POLLRDNORM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935
1936 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001937
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938}
Eric Dumazetc482c562009-07-17 00:26:32 +00001939EXPORT_SYMBOL(udp_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940
David S. Millerdb8dac22008-03-06 16:22:02 -08001941struct proto udp_prot = {
1942 .name = "UDP",
1943 .owner = THIS_MODULE,
1944 .close = udp_lib_close,
1945 .connect = ip4_datagram_connect,
1946 .disconnect = udp_disconnect,
1947 .ioctl = udp_ioctl,
1948 .destroy = udp_destroy_sock,
1949 .setsockopt = udp_setsockopt,
1950 .getsockopt = udp_getsockopt,
1951 .sendmsg = udp_sendmsg,
1952 .recvmsg = udp_recvmsg,
1953 .sendpage = udp_sendpage,
Herbert Xu93821772008-09-15 11:48:46 -07001954 .backlog_rcv = __udp_queue_rcv_skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001955 .hash = udp_lib_hash,
1956 .unhash = udp_lib_unhash,
Eric Dumazet719f8352010-09-08 05:08:44 +00001957 .rehash = udp_v4_rehash,
David S. Millerdb8dac22008-03-06 16:22:02 -08001958 .get_port = udp_v4_get_port,
1959 .memory_allocated = &udp_memory_allocated,
1960 .sysctl_mem = sysctl_udp_mem,
1961 .sysctl_wmem = &sysctl_udp_wmem_min,
1962 .sysctl_rmem = &sysctl_udp_rmem_min,
1963 .obj_size = sizeof(struct udp_sock),
Eric Dumazet271b72c2008-10-29 02:11:14 -07001964 .slab_flags = SLAB_DESTROY_BY_RCU,
Eric Dumazet645ca702008-10-29 01:41:45 -07001965 .h.udp_table = &udp_table,
David S. Millerdb8dac22008-03-06 16:22:02 -08001966#ifdef CONFIG_COMPAT
1967 .compat_setsockopt = compat_udp_setsockopt,
1968 .compat_getsockopt = compat_udp_getsockopt,
1969#endif
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001970 .clear_sk = sk_prot_clear_portaddr_nulls,
David S. Millerdb8dac22008-03-06 16:22:02 -08001971};
Eric Dumazetc482c562009-07-17 00:26:32 +00001972EXPORT_SYMBOL(udp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973
1974/* ------------------------------------------------------------------------ */
1975#ifdef CONFIG_PROC_FS
1976
Eric Dumazet645ca702008-10-29 01:41:45 -07001977static struct sock *udp_get_first(struct seq_file *seq, int start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978{
1979 struct sock *sk;
1980 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07001981 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982
Eric Dumazetf86dcc52009-10-07 00:37:59 +00001983 for (state->bucket = start; state->bucket <= state->udp_table->mask;
1984 ++state->bucket) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08001985 struct hlist_nulls_node *node;
Eric Dumazet645ca702008-10-29 01:41:45 -07001986 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00001987
1988 if (hlist_nulls_empty(&hslot->head))
1989 continue;
1990
Eric Dumazet645ca702008-10-29 01:41:45 -07001991 spin_lock_bh(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08001992 sk_nulls_for_each(sk, node, &hslot->head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09001993 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07001994 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 if (sk->sk_family == state->family)
1996 goto found;
1997 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001998 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999 }
2000 sk = NULL;
2001found:
2002 return sk;
2003}
2004
2005static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
2006{
2007 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002008 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009
2010 do {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002011 sk = sk_nulls_next(sk);
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002012 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013
Eric Dumazet645ca702008-10-29 01:41:45 -07002014 if (!sk) {
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002015 if (state->bucket <= state->udp_table->mask)
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002016 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07002017 return udp_get_first(seq, state->bucket + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 }
2019 return sk;
2020}
2021
2022static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2023{
Eric Dumazet645ca702008-10-29 01:41:45 -07002024 struct sock *sk = udp_get_first(seq, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025
2026 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08002027 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 --pos;
2029 return pos ? NULL : sk;
2030}
2031
2032static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2033{
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002034 struct udp_iter_state *state = seq->private;
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002035 state->bucket = MAX_UDP_PORTS;
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002036
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002037 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038}
2039
2040static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2041{
2042 struct sock *sk;
2043
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002044 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 sk = udp_get_idx(seq, 0);
2046 else
2047 sk = udp_get_next(seq, v);
2048
2049 ++*pos;
2050 return sk;
2051}
2052
2053static void udp_seq_stop(struct seq_file *seq, void *v)
2054{
Eric Dumazet645ca702008-10-29 01:41:45 -07002055 struct udp_iter_state *state = seq->private;
2056
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002057 if (state->bucket <= state->udp_table->mask)
Eric Dumazet645ca702008-10-29 01:41:45 -07002058 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059}
2060
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002061int udp_seq_open(struct inode *inode, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002062{
2063 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002064 struct udp_iter_state *s;
2065 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002067 err = seq_open_net(inode, file, &afinfo->seq_ops,
2068 sizeof(struct udp_iter_state));
2069 if (err < 0)
2070 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002071
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002072 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073 s->family = afinfo->family;
Eric Dumazet645ca702008-10-29 01:41:45 -07002074 s->udp_table = afinfo->udp_table;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002075 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002076}
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002077EXPORT_SYMBOL(udp_seq_open);
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002078
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002080int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081{
2082 struct proc_dir_entry *p;
2083 int rc = 0;
2084
Denis V. Lunevdda61922008-03-28 18:24:26 -07002085 afinfo->seq_ops.start = udp_seq_start;
2086 afinfo->seq_ops.next = udp_seq_next;
2087 afinfo->seq_ops.stop = udp_seq_stop;
2088
Denis V. Lunev84841c32008-05-02 04:10:08 -07002089 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002090 afinfo->seq_fops, afinfo);
Denis V. Lunev84841c32008-05-02 04:10:08 -07002091 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092 rc = -ENOMEM;
2093 return rc;
2094}
Eric Dumazetc482c562009-07-17 00:26:32 +00002095EXPORT_SYMBOL(udp_proc_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002097void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098{
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002099 proc_net_remove(net, afinfo->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100}
Eric Dumazetc482c562009-07-17 00:26:32 +00002101EXPORT_SYMBOL(udp_proc_unregister);
David S. Millerdb8dac22008-03-06 16:22:02 -08002102
2103/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002104static void udp4_format_sock(struct sock *sp, struct seq_file *f,
2105 int bucket, int *len)
David S. Millerdb8dac22008-03-06 16:22:02 -08002106{
2107 struct inet_sock *inet = inet_sk(sp);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002108 __be32 dest = inet->inet_daddr;
2109 __be32 src = inet->inet_rcv_saddr;
2110 __u16 destp = ntohs(inet->inet_dport);
2111 __u16 srcp = ntohs(inet->inet_sport);
David S. Millerdb8dac22008-03-06 16:22:02 -08002112
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002113 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
Dan Rosenberg71338aa2011-05-23 12:17:35 +00002114 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
David S. Millerdb8dac22008-03-06 16:22:02 -08002115 bucket, src, srcp, dest, destp, sp->sk_state,
Eric Dumazet31e6d362009-06-17 19:05:41 -07002116 sk_wmem_alloc_get(sp),
2117 sk_rmem_alloc_get(sp),
Eric W. Biedermana7cb5a42012-05-24 01:10:10 -06002118 0, 0L, 0,
2119 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
2120 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002121 atomic_read(&sp->sk_refcnt), sp,
2122 atomic_read(&sp->sk_drops), len);
David S. Millerdb8dac22008-03-06 16:22:02 -08002123}
2124
2125int udp4_seq_show(struct seq_file *seq, void *v)
2126{
2127 if (v == SEQ_START_TOKEN)
2128 seq_printf(seq, "%-127s\n",
2129 " sl local_address rem_address st tx_queue "
2130 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002131 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08002132 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08002133 struct udp_iter_state *state = seq->private;
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002134 int len;
David S. Millerdb8dac22008-03-06 16:22:02 -08002135
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002136 udp4_format_sock(v, seq, state->bucket, &len);
Eric Dumazetc482c562009-07-17 00:26:32 +00002137 seq_printf(seq, "%*s\n", 127 - len, "");
David S. Millerdb8dac22008-03-06 16:22:02 -08002138 }
2139 return 0;
2140}
2141
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002142static const struct file_operations udp_afinfo_seq_fops = {
2143 .owner = THIS_MODULE,
2144 .open = udp_seq_open,
2145 .read = seq_read,
2146 .llseek = seq_lseek,
2147 .release = seq_release_net
2148};
2149
David S. Millerdb8dac22008-03-06 16:22:02 -08002150/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08002151static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08002152 .name = "udp",
2153 .family = AF_INET,
Eric Dumazet645ca702008-10-29 01:41:45 -07002154 .udp_table = &udp_table,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002155 .seq_fops = &udp_afinfo_seq_fops,
Denis V. Lunevdda61922008-03-28 18:24:26 -07002156 .seq_ops = {
2157 .show = udp4_seq_show,
2158 },
David S. Millerdb8dac22008-03-06 16:22:02 -08002159};
2160
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002161static int __net_init udp4_proc_init_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002162{
2163 return udp_proc_register(net, &udp4_seq_afinfo);
2164}
2165
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002166static void __net_exit udp4_proc_exit_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002167{
2168 udp_proc_unregister(net, &udp4_seq_afinfo);
2169}
2170
2171static struct pernet_operations udp4_net_ops = {
2172 .init = udp4_proc_init_net,
2173 .exit = udp4_proc_exit_net,
2174};
2175
David S. Millerdb8dac22008-03-06 16:22:02 -08002176int __init udp4_proc_init(void)
2177{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002178 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002179}
2180
2181void udp4_proc_exit(void)
2182{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002183 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002184}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002185#endif /* CONFIG_PROC_FS */
2186
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002187static __initdata unsigned long uhash_entries;
2188static int __init set_uhash_entries(char *str)
Eric Dumazet645ca702008-10-29 01:41:45 -07002189{
Eldad Zack413c27d2012-05-19 14:13:18 +00002190 ssize_t ret;
2191
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002192 if (!str)
2193 return 0;
Eldad Zack413c27d2012-05-19 14:13:18 +00002194
2195 ret = kstrtoul(str, 0, &uhash_entries);
2196 if (ret)
2197 return 0;
2198
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002199 if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2200 uhash_entries = UDP_HTABLE_SIZE_MIN;
2201 return 1;
2202}
2203__setup("uhash_entries=", set_uhash_entries);
Eric Dumazet645ca702008-10-29 01:41:45 -07002204
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002205void __init udp_table_init(struct udp_table *table, const char *name)
2206{
2207 unsigned int i;
2208
Tim Bird31fe62b2012-05-23 13:33:35 +00002209 table->hash = alloc_large_system_hash(name,
2210 2 * sizeof(struct udp_hslot),
2211 uhash_entries,
2212 21, /* one slot per 2 MB */
2213 0,
2214 &table->log,
2215 &table->mask,
2216 UDP_HTABLE_SIZE_MIN,
2217 64 * 1024);
2218
Eric Dumazet512615b2009-11-08 10:17:58 +00002219 table->hash2 = table->hash + (table->mask + 1);
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002220 for (i = 0; i <= table->mask; i++) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002221 INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00002222 table->hash[i].count = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -07002223 spin_lock_init(&table->hash[i].lock);
2224 }
Eric Dumazet512615b2009-11-08 10:17:58 +00002225 for (i = 0; i <= table->mask; i++) {
2226 INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i);
2227 table->hash2[i].count = 0;
2228 spin_lock_init(&table->hash2[i].lock);
2229 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002230}
2231
Hideo Aoki95766ff2007-12-31 00:29:24 -08002232void __init udp_init(void)
2233{
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002234 unsigned long limit;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002235
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002236 udp_table_init(&udp_table, "UDP");
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002237 limit = nr_free_buffer_pages() / 8;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002238 limit = max(limit, 128UL);
2239 sysctl_udp_mem[0] = limit / 4 * 3;
2240 sysctl_udp_mem[1] = limit;
2241 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2242
2243 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2244 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2245}
2246
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00002247int udp4_ufo_send_check(struct sk_buff *skb)
2248{
2249 const struct iphdr *iph;
2250 struct udphdr *uh;
2251
2252 if (!pskb_may_pull(skb, sizeof(*uh)))
2253 return -EINVAL;
2254
2255 iph = ip_hdr(skb);
2256 uh = udp_hdr(skb);
2257
2258 uh->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len,
2259 IPPROTO_UDP, 0);
2260 skb->csum_start = skb_transport_header(skb) - skb->head;
2261 skb->csum_offset = offsetof(struct udphdr, check);
2262 skb->ip_summed = CHECKSUM_PARTIAL;
2263 return 0;
2264}
2265
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002266struct sk_buff *udp4_ufo_fragment(struct sk_buff *skb,
2267 netdev_features_t features)
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00002268{
2269 struct sk_buff *segs = ERR_PTR(-EINVAL);
2270 unsigned int mss;
2271 int offset;
2272 __wsum csum;
2273
2274 mss = skb_shinfo(skb)->gso_size;
2275 if (unlikely(skb->len <= mss))
2276 goto out;
2277
2278 if (skb_gso_ok(skb, features | NETIF_F_GSO_ROBUST)) {
2279 /* Packet is from an untrusted source, reset gso_segs. */
2280 int type = skb_shinfo(skb)->gso_type;
2281
2282 if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY) ||
2283 !(type & (SKB_GSO_UDP))))
2284 goto out;
2285
2286 skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss);
2287
2288 segs = NULL;
2289 goto out;
2290 }
2291
2292 /* Do software UFO. Complete and fill in the UDP checksum as HW cannot
2293 * do checksum of UDP packets sent as multiple IP fragments.
2294 */
Michał Mirosław55508d62010-12-14 15:24:08 +00002295 offset = skb_checksum_start_offset(skb);
Eric Dumazetc482c562009-07-17 00:26:32 +00002296 csum = skb_checksum(skb, offset, skb->len - offset, 0);
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00002297 offset += skb->csum_offset;
2298 *(__sum16 *)(skb->data + offset) = csum_fold(csum);
2299 skb->ip_summed = CHECKSUM_NONE;
2300
2301 /* Fragment the skb. IP headers of the fragments are updated in
2302 * inet_gso_segment()
2303 */
2304 segs = skb_segment(skb, features);
2305out:
2306 return segs;
2307}
2308