blob: d6f5feeb3eaf9e780f25acbdd360467beddc2536 [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/system.h>
83#include <asm/uaccess.h>
84#include <asm/ioctls.h>
Hideo Aoki95766ff2007-12-31 00:29:24 -080085#include <linux/bootmem.h>
Eric Dumazet8203efb2008-10-29 02:32:32 -070086#include <linux/highmem.h>
87#include <linux/swap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070088#include <linux/types.h>
89#include <linux/fcntl.h>
90#include <linux/module.h>
91#include <linux/socket.h>
92#include <linux/sockios.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020093#include <linux/igmp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094#include <linux/in.h>
95#include <linux/errno.h>
96#include <linux/timer.h>
97#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include <linux/inet.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099#include <linux/netdevice.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +0900100#include <linux/slab.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -0700101#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102#include <linux/skbuff.h>
103#include <linux/proc_fs.h>
104#include <linux/seq_file.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200105#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106#include <net/icmp.h>
107#include <net/route.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108#include <net/checksum.h>
109#include <net/xfrm.h>
Satoru Moriya296f7ea2011-06-17 11:58:39 +0000110#include <trace/events/udp.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800111#include "udp_impl.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000113struct udp_table udp_table __read_mostly;
Eric Dumazet645ca702008-10-29 01:41:45 -0700114EXPORT_SYMBOL(udp_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115
Eric Dumazet8d987e52010-11-09 23:24:26 +0000116long sysctl_udp_mem[3] __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800117EXPORT_SYMBOL(sysctl_udp_mem);
Eric Dumazetc482c562009-07-17 00:26:32 +0000118
119int sysctl_udp_rmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800120EXPORT_SYMBOL(sysctl_udp_rmem_min);
Eric Dumazetc482c562009-07-17 00:26:32 +0000121
122int sysctl_udp_wmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800123EXPORT_SYMBOL(sysctl_udp_wmem_min);
124
Eric Dumazet8d987e52010-11-09 23:24:26 +0000125atomic_long_t udp_memory_allocated;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800126EXPORT_SYMBOL(udp_memory_allocated);
127
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000128#define MAX_UDP_PORTS 65536
129#define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN)
Eric Dumazet98322f22009-01-26 21:35:35 -0800130
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700131static int udp_lib_lport_inuse(struct net *net, __u16 num,
Eric Dumazet645ca702008-10-29 01:41:45 -0700132 const struct udp_hslot *hslot,
Eric Dumazet98322f22009-01-26 21:35:35 -0800133 unsigned long *bitmap,
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700134 struct sock *sk,
135 int (*saddr_comp)(const struct sock *sk1,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000136 const struct sock *sk2),
137 unsigned int log)
Gerrit Renker25030a72006-08-26 20:06:05 -0700138{
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700139 struct sock *sk2;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800140 struct hlist_nulls_node *node;
Gerrit Renker25030a72006-08-26 20:06:05 -0700141
Eric Dumazet88ab1932008-11-16 19:39:21 -0800142 sk_nulls_for_each(sk2, node, &hslot->head)
Joe Perches9d4fb272009-11-23 10:41:23 -0800143 if (net_eq(sock_net(sk2), net) &&
144 sk2 != sk &&
Eric Dumazetd4cada42009-11-08 10:17:30 +0000145 (bitmap || udp_sk(sk2)->udp_port_hash == num) &&
Joe Perches9d4fb272009-11-23 10:41:23 -0800146 (!sk2->sk_reuse || !sk->sk_reuse) &&
147 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
148 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Eric Dumazet98322f22009-01-26 21:35:35 -0800149 (*saddr_comp)(sk, sk2)) {
150 if (bitmap)
Eric Dumazetd4cada42009-11-08 10:17:30 +0000151 __set_bit(udp_sk(sk2)->udp_port_hash >> log,
152 bitmap);
Eric Dumazet98322f22009-01-26 21:35:35 -0800153 else
154 return 1;
155 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700156 return 0;
157}
158
Eric Dumazet30fff922009-11-09 05:26:33 +0000159/*
160 * Note: we still hold spinlock of primary hash chain, so no other writer
161 * can insert/delete a socket with local_port == num
162 */
163static int udp_lib_lport_inuse2(struct net *net, __u16 num,
164 struct udp_hslot *hslot2,
165 struct sock *sk,
166 int (*saddr_comp)(const struct sock *sk1,
167 const struct sock *sk2))
168{
169 struct sock *sk2;
170 struct hlist_nulls_node *node;
171 int res = 0;
172
173 spin_lock(&hslot2->lock);
174 udp_portaddr_for_each_entry(sk2, node, &hslot2->head)
Joe Perches9d4fb272009-11-23 10:41:23 -0800175 if (net_eq(sock_net(sk2), net) &&
176 sk2 != sk &&
177 (udp_sk(sk2)->udp_port_hash == num) &&
178 (!sk2->sk_reuse || !sk->sk_reuse) &&
179 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
180 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Eric Dumazet30fff922009-11-09 05:26:33 +0000181 (*saddr_comp)(sk, sk2)) {
182 res = 1;
183 break;
184 }
185 spin_unlock(&hslot2->lock);
186 return res;
187}
188
Gerrit Renker25030a72006-08-26 20:06:05 -0700189/**
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700190 * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
Gerrit Renker25030a72006-08-26 20:06:05 -0700191 *
192 * @sk: socket struct in question
193 * @snum: port number to look up
David S. Millerdf2bc452007-06-05 15:18:43 -0700194 * @saddr_comp: AF-dependent comparison of bound local IP addresses
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300195 * @hash2_nulladdr: AF-dependent hash value in secondary hash chains,
Eric Dumazet30fff922009-11-09 05:26:33 +0000196 * with NULL address
Gerrit Renker25030a72006-08-26 20:06:05 -0700197 */
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700198int udp_lib_get_port(struct sock *sk, unsigned short snum,
David S. Millerdf2bc452007-06-05 15:18:43 -0700199 int (*saddr_comp)(const struct sock *sk1,
Eric Dumazet30fff922009-11-09 05:26:33 +0000200 const struct sock *sk2),
201 unsigned int hash2_nulladdr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202{
Eric Dumazet512615b2009-11-08 10:17:58 +0000203 struct udp_hslot *hslot, *hslot2;
Eric Dumazet645ca702008-10-29 01:41:45 -0700204 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Gerrit Renker25030a72006-08-26 20:06:05 -0700205 int error = 1;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900206 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700208 if (!snum) {
Eric Dumazet9088c562008-10-08 11:44:17 -0700209 int low, high, remaining;
210 unsigned rand;
Eric Dumazet98322f22009-01-26 21:35:35 -0800211 unsigned short first, last;
212 DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700214 inet_get_local_port_range(&low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700215 remaining = (high - low) + 1;
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700216
Eric Dumazet9088c562008-10-08 11:44:17 -0700217 rand = net_random();
Eric Dumazet98322f22009-01-26 21:35:35 -0800218 first = (((u64)rand * remaining) >> 32) + low;
219 /*
220 * force rand to be an odd multiple of UDP_HTABLE_SIZE
221 */
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000222 rand = (rand | 1) * (udptable->mask + 1);
Eric Dumazet5781b232009-12-13 19:32:39 -0800223 last = first + udptable->mask + 1;
224 do {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000225 hslot = udp_hashslot(udptable, net, first);
Eric Dumazet98322f22009-01-26 21:35:35 -0800226 bitmap_zero(bitmap, PORTS_PER_CHAIN);
Eric Dumazet645ca702008-10-29 01:41:45 -0700227 spin_lock_bh(&hslot->lock);
Eric Dumazet98322f22009-01-26 21:35:35 -0800228 udp_lib_lport_inuse(net, snum, hslot, bitmap, sk,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000229 saddr_comp, udptable->log);
Eric Dumazet98322f22009-01-26 21:35:35 -0800230
231 snum = first;
232 /*
233 * Iterate on all possible values of snum for this hash.
234 * Using steps of an odd multiple of UDP_HTABLE_SIZE
235 * give us randomization and full range coverage.
236 */
Eric Dumazet9088c562008-10-08 11:44:17 -0700237 do {
Eric Dumazet98322f22009-01-26 21:35:35 -0800238 if (low <= snum && snum <= high &&
Amerigo Wange3826f12010-05-05 00:27:06 +0000239 !test_bit(snum >> udptable->log, bitmap) &&
240 !inet_is_reserved_local_port(snum))
Eric Dumazet98322f22009-01-26 21:35:35 -0800241 goto found;
242 snum += rand;
243 } while (snum != first);
244 spin_unlock_bh(&hslot->lock);
Eric Dumazet5781b232009-12-13 19:32:39 -0800245 } while (++first != last);
Eric Dumazet98322f22009-01-26 21:35:35 -0800246 goto fail;
Eric Dumazet645ca702008-10-29 01:41:45 -0700247 } else {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000248 hslot = udp_hashslot(udptable, net, snum);
Eric Dumazet645ca702008-10-29 01:41:45 -0700249 spin_lock_bh(&hslot->lock);
Eric Dumazet30fff922009-11-09 05:26:33 +0000250 if (hslot->count > 10) {
251 int exist;
252 unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum;
253
254 slot2 &= udptable->mask;
255 hash2_nulladdr &= udptable->mask;
256
257 hslot2 = udp_hashslot2(udptable, slot2);
258 if (hslot->count < hslot2->count)
259 goto scan_primary_hash;
260
261 exist = udp_lib_lport_inuse2(net, snum, hslot2,
262 sk, saddr_comp);
263 if (!exist && (hash2_nulladdr != slot2)) {
264 hslot2 = udp_hashslot2(udptable, hash2_nulladdr);
265 exist = udp_lib_lport_inuse2(net, snum, hslot2,
266 sk, saddr_comp);
267 }
268 if (exist)
269 goto fail_unlock;
270 else
271 goto found;
272 }
273scan_primary_hash:
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000274 if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk,
275 saddr_comp, 0))
Eric Dumazet645ca702008-10-29 01:41:45 -0700276 goto fail_unlock;
277 }
Eric Dumazet98322f22009-01-26 21:35:35 -0800278found:
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000279 inet_sk(sk)->inet_num = snum;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000280 udp_sk(sk)->udp_port_hash = snum;
281 udp_sk(sk)->udp_portaddr_hash ^= snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 if (sk_unhashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800283 sk_nulls_add_node_rcu(sk, &hslot->head);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +0000284 hslot->count++;
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700285 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet512615b2009-11-08 10:17:58 +0000286
287 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
288 spin_lock(&hslot2->lock);
289 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
290 &hslot2->head);
291 hslot2->count++;
292 spin_unlock(&hslot2->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700294 error = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -0700295fail_unlock:
296 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297fail:
Gerrit Renker25030a72006-08-26 20:06:05 -0700298 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299}
Eric Dumazetc482c562009-07-17 00:26:32 +0000300EXPORT_SYMBOL(udp_lib_get_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301
Vlad Yasevich499923c2009-04-09 17:37:33 +0000302static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
David S. Millerdb8dac22008-03-06 16:22:02 -0800303{
304 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
305
Eric Dumazetc482c562009-07-17 00:26:32 +0000306 return (!ipv6_only_sock(sk2) &&
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000307 (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr ||
308 inet1->inet_rcv_saddr == inet2->inet_rcv_saddr));
David S. Millerdb8dac22008-03-06 16:22:02 -0800309}
310
Eric Dumazetd4cada42009-11-08 10:17:30 +0000311static unsigned int udp4_portaddr_hash(struct net *net, __be32 saddr,
312 unsigned int port)
313{
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700314 return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000315}
316
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700317int udp_v4_get_port(struct sock *sk, unsigned short snum)
David S. Millerdb8dac22008-03-06 16:22:02 -0800318{
Eric Dumazet30fff922009-11-09 05:26:33 +0000319 unsigned int hash2_nulladdr =
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700320 udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum);
Eric Dumazet30fff922009-11-09 05:26:33 +0000321 unsigned int hash2_partial =
322 udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0);
323
Eric Dumazetd4cada42009-11-08 10:17:30 +0000324 /* precompute partial secondary hash */
Eric Dumazet30fff922009-11-09 05:26:33 +0000325 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
326 return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr);
David S. Millerdb8dac22008-03-06 16:22:02 -0800327}
328
Eric Dumazet645ca702008-10-29 01:41:45 -0700329static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr,
330 unsigned short hnum,
331 __be16 sport, __be32 daddr, __be16 dport, int dif)
332{
333 int score = -1;
334
Eric Dumazetd4cada42009-11-08 10:17:30 +0000335 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
Eric Dumazet645ca702008-10-29 01:41:45 -0700336 !ipv6_only_sock(sk)) {
337 struct inet_sock *inet = inet_sk(sk);
338
339 score = (sk->sk_family == PF_INET ? 1 : 0);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000340 if (inet->inet_rcv_saddr) {
341 if (inet->inet_rcv_saddr != daddr)
Eric Dumazet645ca702008-10-29 01:41:45 -0700342 return -1;
343 score += 2;
344 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000345 if (inet->inet_daddr) {
346 if (inet->inet_daddr != saddr)
Eric Dumazet645ca702008-10-29 01:41:45 -0700347 return -1;
348 score += 2;
349 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000350 if (inet->inet_dport) {
351 if (inet->inet_dport != sport)
Eric Dumazet645ca702008-10-29 01:41:45 -0700352 return -1;
353 score += 2;
354 }
355 if (sk->sk_bound_dev_if) {
356 if (sk->sk_bound_dev_if != dif)
357 return -1;
358 score += 2;
359 }
360 }
361 return score;
362}
363
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000364/*
365 * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num)
366 */
367#define SCORE2_MAX (1 + 2 + 2 + 2)
368static inline int compute_score2(struct sock *sk, struct net *net,
369 __be32 saddr, __be16 sport,
370 __be32 daddr, unsigned int hnum, int dif)
371{
372 int score = -1;
373
374 if (net_eq(sock_net(sk), net) && !ipv6_only_sock(sk)) {
375 struct inet_sock *inet = inet_sk(sk);
376
377 if (inet->inet_rcv_saddr != daddr)
378 return -1;
379 if (inet->inet_num != hnum)
380 return -1;
381
382 score = (sk->sk_family == PF_INET ? 1 : 0);
383 if (inet->inet_daddr) {
384 if (inet->inet_daddr != saddr)
385 return -1;
386 score += 2;
387 }
388 if (inet->inet_dport) {
389 if (inet->inet_dport != sport)
390 return -1;
391 score += 2;
392 }
393 if (sk->sk_bound_dev_if) {
394 if (sk->sk_bound_dev_if != dif)
395 return -1;
396 score += 2;
397 }
398 }
399 return score;
400}
401
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000402
403/* called with read_rcu_lock() */
404static struct sock *udp4_lib_lookup2(struct net *net,
405 __be32 saddr, __be16 sport,
406 __be32 daddr, unsigned int hnum, int dif,
407 struct udp_hslot *hslot2, unsigned int slot2)
408{
409 struct sock *sk, *result;
410 struct hlist_nulls_node *node;
411 int score, badness;
412
413begin:
414 result = NULL;
415 badness = -1;
416 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
417 score = compute_score2(sk, net, saddr, sport,
418 daddr, hnum, dif);
419 if (score > badness) {
420 result = sk;
421 badness = score;
422 if (score == SCORE2_MAX)
423 goto exact_match;
424 }
425 }
426 /*
427 * if the nulls value we got at the end of this lookup is
428 * not the expected one, we must restart lookup.
429 * We probably met an item that was moved to another chain.
430 */
431 if (get_nulls_value(node) != slot2)
432 goto begin;
433
434 if (result) {
435exact_match:
Eric Dumazetc31504d2010-11-15 19:58:26 +0000436 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000437 result = NULL;
438 else if (unlikely(compute_score2(result, net, saddr, sport,
439 daddr, hnum, dif) < badness)) {
440 sock_put(result);
441 goto begin;
442 }
443 }
444 return result;
445}
446
David S. Millerdb8dac22008-03-06 16:22:02 -0800447/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
448 * harder than this. -DaveM
449 */
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000450struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -0800451 __be16 sport, __be32 daddr, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700452 int dif, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800453{
Eric Dumazet271b72c2008-10-29 02:11:14 -0700454 struct sock *sk, *result;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800455 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800456 unsigned short hnum = ntohs(dport);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000457 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
458 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
Eric Dumazet271b72c2008-10-29 02:11:14 -0700459 int score, badness;
David S. Millerdb8dac22008-03-06 16:22:02 -0800460
Eric Dumazet271b72c2008-10-29 02:11:14 -0700461 rcu_read_lock();
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000462 if (hslot->count > 10) {
463 hash2 = udp4_portaddr_hash(net, daddr, hnum);
464 slot2 = hash2 & udptable->mask;
465 hslot2 = &udptable->hash2[slot2];
466 if (hslot->count < hslot2->count)
467 goto begin;
468
469 result = udp4_lib_lookup2(net, saddr, sport,
470 daddr, hnum, dif,
471 hslot2, slot2);
472 if (!result) {
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700473 hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000474 slot2 = hash2 & udptable->mask;
475 hslot2 = &udptable->hash2[slot2];
476 if (hslot->count < hslot2->count)
477 goto begin;
478
Jorge Boncompte [DTI2]1223c672010-04-08 04:56:48 +0000479 result = udp4_lib_lookup2(net, saddr, sport,
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700480 htonl(INADDR_ANY), hnum, dif,
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000481 hslot2, slot2);
482 }
483 rcu_read_unlock();
484 return result;
485 }
Eric Dumazet271b72c2008-10-29 02:11:14 -0700486begin:
487 result = NULL;
488 badness = -1;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800489 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
Eric Dumazet645ca702008-10-29 01:41:45 -0700490 score = compute_score(sk, net, saddr, hnum, sport,
491 daddr, dport, dif);
492 if (score > badness) {
493 result = sk;
494 badness = score;
David S. Millerdb8dac22008-03-06 16:22:02 -0800495 }
496 }
Eric Dumazet88ab1932008-11-16 19:39:21 -0800497 /*
498 * if the nulls value we got at the end of this lookup is
499 * not the expected one, we must restart lookup.
500 * We probably met an item that was moved to another chain.
501 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000502 if (get_nulls_value(node) != slot)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800503 goto begin;
504
Eric Dumazet271b72c2008-10-29 02:11:14 -0700505 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000506 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet271b72c2008-10-29 02:11:14 -0700507 result = NULL;
508 else if (unlikely(compute_score(result, net, saddr, hnum, sport,
509 daddr, dport, dif) < badness)) {
510 sock_put(result);
511 goto begin;
512 }
513 }
514 rcu_read_unlock();
David S. Millerdb8dac22008-03-06 16:22:02 -0800515 return result;
516}
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000517EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
David S. Millerdb8dac22008-03-06 16:22:02 -0800518
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700519static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
520 __be16 sport, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700521 struct udp_table *udptable)
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700522{
KOVACS Krisztian23542612008-10-07 12:41:01 -0700523 struct sock *sk;
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700524 const struct iphdr *iph = ip_hdr(skb);
525
KOVACS Krisztian23542612008-10-07 12:41:01 -0700526 if (unlikely(sk = skb_steal_sock(skb)))
527 return sk;
528 else
Eric Dumazetadf30902009-06-02 05:19:30 +0000529 return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport,
KOVACS Krisztian23542612008-10-07 12:41:01 -0700530 iph->daddr, dport, inet_iif(skb),
531 udptable);
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700532}
533
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700534struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
535 __be32 daddr, __be16 dport, int dif)
536{
Eric Dumazet645ca702008-10-29 01:41:45 -0700537 return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700538}
539EXPORT_SYMBOL_GPL(udp4_lib_lookup);
540
Eric Dumazet920a4612008-11-01 21:22:23 -0700541static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk,
David S. Millerdb8dac22008-03-06 16:22:02 -0800542 __be16 loc_port, __be32 loc_addr,
543 __be16 rmt_port, __be32 rmt_addr,
544 int dif)
545{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800546 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800547 struct sock *s = sk;
548 unsigned short hnum = ntohs(loc_port);
549
Eric Dumazet88ab1932008-11-16 19:39:21 -0800550 sk_nulls_for_each_from(s, node) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800551 struct inet_sock *inet = inet_sk(s);
552
Joe Perches9d4fb272009-11-23 10:41:23 -0800553 if (!net_eq(sock_net(s), net) ||
554 udp_sk(s)->udp_port_hash != hnum ||
555 (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
556 (inet->inet_dport != rmt_port && inet->inet_dport) ||
557 (inet->inet_rcv_saddr &&
558 inet->inet_rcv_saddr != loc_addr) ||
559 ipv6_only_sock(s) ||
David S. Millerdb8dac22008-03-06 16:22:02 -0800560 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
561 continue;
562 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
563 continue;
564 goto found;
565 }
566 s = NULL;
567found:
568 return s;
569}
570
571/*
572 * This routine is called by the ICMP module when it gets some
573 * sort of error condition. If err < 0 then the socket should
574 * be closed and the error returned to the user. If err > 0
575 * it's just the icmp type << 8 | icmp code.
576 * Header points to the ip header of the error packet. We move
577 * on past this. Then (as it used to claim before adjustment)
578 * header points to the first 8 bytes of the udp header. We need
579 * to find the appropriate port.
580 */
581
Eric Dumazet645ca702008-10-29 01:41:45 -0700582void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800583{
584 struct inet_sock *inet;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000585 const struct iphdr *iph = (const struct iphdr *)skb->data;
Eric Dumazetc482c562009-07-17 00:26:32 +0000586 struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
David S. Millerdb8dac22008-03-06 16:22:02 -0800587 const int type = icmp_hdr(skb)->type;
588 const int code = icmp_hdr(skb)->code;
589 struct sock *sk;
590 int harderr;
591 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700592 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800593
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700594 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
David S. Millerdb8dac22008-03-06 16:22:02 -0800595 iph->saddr, uh->source, skb->dev->ifindex, udptable);
596 if (sk == NULL) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700597 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800598 return; /* No socket for error */
599 }
600
601 err = 0;
602 harderr = 0;
603 inet = inet_sk(sk);
604
605 switch (type) {
606 default:
607 case ICMP_TIME_EXCEEDED:
608 err = EHOSTUNREACH;
609 break;
610 case ICMP_SOURCE_QUENCH:
611 goto out;
612 case ICMP_PARAMETERPROB:
613 err = EPROTO;
614 harderr = 1;
615 break;
616 case ICMP_DEST_UNREACH:
617 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
618 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
619 err = EMSGSIZE;
620 harderr = 1;
621 break;
622 }
623 goto out;
624 }
625 err = EHOSTUNREACH;
626 if (code <= NR_ICMP_UNREACH) {
627 harderr = icmp_err_convert[code].fatal;
628 err = icmp_err_convert[code].errno;
629 }
630 break;
631 }
632
633 /*
634 * RFC1122: OK. Passes ICMP errors back to application, as per
635 * 4.1.3.3.
636 */
637 if (!inet->recverr) {
638 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
639 goto out;
Eric Dumazetb1faf562010-05-31 23:44:05 -0700640 } else
Eric Dumazetc482c562009-07-17 00:26:32 +0000641 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
Eric Dumazetb1faf562010-05-31 23:44:05 -0700642
David S. Millerdb8dac22008-03-06 16:22:02 -0800643 sk->sk_err = err;
644 sk->sk_error_report(sk);
645out:
646 sock_put(sk);
647}
648
649void udp_err(struct sk_buff *skb, u32 info)
650{
Eric Dumazet645ca702008-10-29 01:41:45 -0700651 __udp4_lib_err(skb, info, &udp_table);
David S. Millerdb8dac22008-03-06 16:22:02 -0800652}
653
654/*
655 * Throw away all pending data and cancel the corking. Socket is locked.
656 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400657void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800658{
659 struct udp_sock *up = udp_sk(sk);
660
661 if (up->pending) {
662 up->len = 0;
663 up->pending = 0;
664 ip_flush_pending_frames(sk);
665 }
666}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400667EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800668
669/**
Herbert Xuf6b96642011-03-01 02:36:48 +0000670 * udp4_hwcsum - handle outgoing HW checksumming
David S. Millerdb8dac22008-03-06 16:22:02 -0800671 * @skb: sk_buff containing the filled-in UDP header
672 * (checksum field must be zeroed out)
Herbert Xuf6b96642011-03-01 02:36:48 +0000673 * @src: source IP address
674 * @dst: destination IP address
David S. Millerdb8dac22008-03-06 16:22:02 -0800675 */
Herbert Xuf6b96642011-03-01 02:36:48 +0000676static void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
David S. Millerdb8dac22008-03-06 16:22:02 -0800677{
David S. Millerdb8dac22008-03-06 16:22:02 -0800678 struct udphdr *uh = udp_hdr(skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000679 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
680 int offset = skb_transport_offset(skb);
681 int len = skb->len - offset;
682 int hlen = len;
David S. Millerdb8dac22008-03-06 16:22:02 -0800683 __wsum csum = 0;
684
Herbert Xuf6b96642011-03-01 02:36:48 +0000685 if (!frags) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800686 /*
687 * Only one fragment on the socket.
688 */
689 skb->csum_start = skb_transport_header(skb) - skb->head;
690 skb->csum_offset = offsetof(struct udphdr, check);
Herbert Xuf6b96642011-03-01 02:36:48 +0000691 uh->check = ~csum_tcpudp_magic(src, dst, len,
692 IPPROTO_UDP, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800693 } else {
694 /*
695 * HW-checksum won't work as there are two or more
696 * fragments on the socket so that all csums of sk_buffs
697 * should be together
698 */
Herbert Xuf6b96642011-03-01 02:36:48 +0000699 do {
700 csum = csum_add(csum, frags->csum);
701 hlen -= frags->len;
702 } while ((frags = frags->next));
David S. Millerdb8dac22008-03-06 16:22:02 -0800703
Herbert Xuf6b96642011-03-01 02:36:48 +0000704 csum = skb_checksum(skb, offset, hlen, csum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800705 skb->ip_summed = CHECKSUM_NONE;
706
David S. Millerdb8dac22008-03-06 16:22:02 -0800707 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
708 if (uh->check == 0)
709 uh->check = CSUM_MANGLED_0;
710 }
711}
712
David S. Miller79ab0532011-05-09 13:31:04 -0700713static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
Herbert Xuf6b96642011-03-01 02:36:48 +0000714{
715 struct sock *sk = skb->sk;
716 struct inet_sock *inet = inet_sk(sk);
717 struct udphdr *uh;
Herbert Xuf6b96642011-03-01 02:36:48 +0000718 int err = 0;
719 int is_udplite = IS_UDPLITE(sk);
720 int offset = skb_transport_offset(skb);
721 int len = skb->len - offset;
722 __wsum csum = 0;
723
724 /*
725 * Create a UDP header
726 */
727 uh = udp_hdr(skb);
728 uh->source = inet->inet_sport;
David S. Miller79ab0532011-05-09 13:31:04 -0700729 uh->dest = fl4->fl4_dport;
Herbert Xuf6b96642011-03-01 02:36:48 +0000730 uh->len = htons(len);
731 uh->check = 0;
732
733 if (is_udplite) /* UDP-Lite */
734 csum = udplite_csum(skb);
735
736 else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */
737
738 skb->ip_summed = CHECKSUM_NONE;
739 goto send;
740
741 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
742
David S. Miller79ab0532011-05-09 13:31:04 -0700743 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
Herbert Xuf6b96642011-03-01 02:36:48 +0000744 goto send;
745
746 } else
747 csum = udp_csum(skb);
748
749 /* add protocol-dependent pseudo-header */
David S. Miller79ab0532011-05-09 13:31:04 -0700750 uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
Herbert Xuf6b96642011-03-01 02:36:48 +0000751 sk->sk_protocol, csum);
752 if (uh->check == 0)
753 uh->check = CSUM_MANGLED_0;
754
755send:
756 err = ip_send_skb(skb);
757 if (err) {
758 if (err == -ENOBUFS && !inet->recverr) {
759 UDP_INC_STATS_USER(sock_net(sk),
760 UDP_MIB_SNDBUFERRORS, is_udplite);
761 err = 0;
762 }
763 } else
764 UDP_INC_STATS_USER(sock_net(sk),
765 UDP_MIB_OUTDATAGRAMS, is_udplite);
766 return err;
767}
768
David S. Millerdb8dac22008-03-06 16:22:02 -0800769/*
770 * Push out all pending data as one UDP datagram. Socket is locked.
771 */
772static int udp_push_pending_frames(struct sock *sk)
773{
774 struct udp_sock *up = udp_sk(sk);
775 struct inet_sock *inet = inet_sk(sk);
David S. Millerb6f21b22011-03-12 02:09:18 -0500776 struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800777 struct sk_buff *skb;
David S. Millerdb8dac22008-03-06 16:22:02 -0800778 int err = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800779
David S. Miller77968b72011-05-08 17:12:19 -0700780 skb = ip_finish_skb(sk, fl4);
Herbert Xuf6b96642011-03-01 02:36:48 +0000781 if (!skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800782 goto out;
783
David S. Miller79ab0532011-05-09 13:31:04 -0700784 err = udp_send_skb(skb, fl4);
David S. Millerdb8dac22008-03-06 16:22:02 -0800785
David S. Millerdb8dac22008-03-06 16:22:02 -0800786out:
787 up->len = 0;
788 up->pending = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800789 return err;
790}
791
792int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
793 size_t len)
794{
795 struct inet_sock *inet = inet_sk(sk);
796 struct udp_sock *up = udp_sk(sk);
David S. Millere4749952011-05-08 16:38:45 -0700797 struct flowi4 fl4_stack;
David S. Millerb6f21b22011-03-12 02:09:18 -0500798 struct flowi4 *fl4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800799 int ulen = len;
800 struct ipcm_cookie ipc;
801 struct rtable *rt = NULL;
802 int free = 0;
803 int connected = 0;
804 __be32 daddr, faddr, saddr;
805 __be16 dport;
806 u8 tos;
807 int err, is_udplite = IS_UDPLITE(sk);
808 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
809 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
Herbert Xu903ab862011-03-01 02:36:48 +0000810 struct sk_buff *skb;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000811 struct ip_options_data opt_copy;
David S. Millerdb8dac22008-03-06 16:22:02 -0800812
813 if (len > 0xFFFF)
814 return -EMSGSIZE;
815
816 /*
817 * Check the flags.
818 */
819
Eric Dumazetc482c562009-07-17 00:26:32 +0000820 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
David S. Millerdb8dac22008-03-06 16:22:02 -0800821 return -EOPNOTSUPP;
822
823 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000824 ipc.tx_flags = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800825
Herbert Xu903ab862011-03-01 02:36:48 +0000826 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
827
David S. Millerf5fca602011-05-08 17:24:10 -0700828 fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800829 if (up->pending) {
830 /*
831 * There are pending frames.
832 * The socket lock must be held while it's corked.
833 */
834 lock_sock(sk);
835 if (likely(up->pending)) {
836 if (unlikely(up->pending != AF_INET)) {
837 release_sock(sk);
838 return -EINVAL;
839 }
840 goto do_append_data;
841 }
842 release_sock(sk);
843 }
844 ulen += sizeof(struct udphdr);
845
846 /*
847 * Get and verify the address.
848 */
849 if (msg->msg_name) {
Eric Dumazetc482c562009-07-17 00:26:32 +0000850 struct sockaddr_in * usin = (struct sockaddr_in *)msg->msg_name;
David S. Millerdb8dac22008-03-06 16:22:02 -0800851 if (msg->msg_namelen < sizeof(*usin))
852 return -EINVAL;
853 if (usin->sin_family != AF_INET) {
854 if (usin->sin_family != AF_UNSPEC)
855 return -EAFNOSUPPORT;
856 }
857
858 daddr = usin->sin_addr.s_addr;
859 dport = usin->sin_port;
860 if (dport == 0)
861 return -EINVAL;
862 } else {
863 if (sk->sk_state != TCP_ESTABLISHED)
864 return -EDESTADDRREQ;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000865 daddr = inet->inet_daddr;
866 dport = inet->inet_dport;
David S. Millerdb8dac22008-03-06 16:22:02 -0800867 /* Open fast path for connected socket.
868 Route will not be used, if at least one option is set.
869 */
870 connected = 1;
871 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000872 ipc.addr = inet->inet_saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800873
874 ipc.oif = sk->sk_bound_dev_if;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000875 err = sock_tx_timestamp(sk, &ipc.tx_flags);
Patrick Ohly51f31ca2009-02-12 05:03:39 +0000876 if (err)
877 return err;
David S. Millerdb8dac22008-03-06 16:22:02 -0800878 if (msg->msg_controllen) {
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900879 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
David S. Millerdb8dac22008-03-06 16:22:02 -0800880 if (err)
881 return err;
882 if (ipc.opt)
883 free = 1;
884 connected = 0;
885 }
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000886 if (!ipc.opt) {
887 struct ip_options_rcu *inet_opt;
888
889 rcu_read_lock();
890 inet_opt = rcu_dereference(inet->inet_opt);
891 if (inet_opt) {
892 memcpy(&opt_copy, inet_opt,
893 sizeof(*inet_opt) + inet_opt->opt.optlen);
894 ipc.opt = &opt_copy.opt;
895 }
896 rcu_read_unlock();
897 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800898
899 saddr = ipc.addr;
900 ipc.addr = faddr = daddr;
901
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000902 if (ipc.opt && ipc.opt->opt.srr) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800903 if (!daddr)
904 return -EINVAL;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000905 faddr = ipc.opt->opt.faddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800906 connected = 0;
907 }
908 tos = RT_TOS(inet->tos);
909 if (sock_flag(sk, SOCK_LOCALROUTE) ||
910 (msg->msg_flags & MSG_DONTROUTE) ||
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000911 (ipc.opt && ipc.opt->opt.is_strictroute)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800912 tos |= RTO_ONLINK;
913 connected = 0;
914 }
915
916 if (ipv4_is_multicast(daddr)) {
917 if (!ipc.oif)
918 ipc.oif = inet->mc_index;
919 if (!saddr)
920 saddr = inet->mc_addr;
921 connected = 0;
Erich E. Hoover76e21052012-02-08 09:11:07 +0000922 } else if (!ipc.oif)
923 ipc.oif = inet->uc_index;
David S. Millerdb8dac22008-03-06 16:22:02 -0800924
925 if (connected)
Eric Dumazetc482c562009-07-17 00:26:32 +0000926 rt = (struct rtable *)sk_dst_check(sk, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800927
928 if (rt == NULL) {
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -0700929 struct net *net = sock_net(sk);
930
David S. Millere4749952011-05-08 16:38:45 -0700931 fl4 = &fl4_stack;
932 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
David S. Millerc0951cb2011-03-31 04:54:27 -0700933 RT_SCOPE_UNIVERSE, sk->sk_protocol,
934 inet_sk_flowi_flags(sk)|FLOWI_FLAG_CAN_SLEEP,
935 faddr, saddr, dport, inet->inet_sport);
936
David S. Millere4749952011-05-08 16:38:45 -0700937 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
938 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800939 if (IS_ERR(rt)) {
940 err = PTR_ERR(rt);
David S. Miller06dc94b2011-03-03 10:38:01 -0800941 rt = NULL;
David S. Millerdb8dac22008-03-06 16:22:02 -0800942 if (err == -ENETUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700943 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -0800944 goto out;
945 }
946
947 err = -EACCES;
948 if ((rt->rt_flags & RTCF_BROADCAST) &&
949 !sock_flag(sk, SOCK_BROADCAST))
950 goto out;
951 if (connected)
Changli Gaod8d1f302010-06-10 23:31:35 -0700952 sk_dst_set(sk, dst_clone(&rt->dst));
David S. Millerdb8dac22008-03-06 16:22:02 -0800953 }
954
955 if (msg->msg_flags&MSG_CONFIRM)
956 goto do_confirm;
957back_from_confirm:
958
David S. Millere4749952011-05-08 16:38:45 -0700959 saddr = fl4->saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800960 if (!ipc.addr)
David S. Millere4749952011-05-08 16:38:45 -0700961 daddr = ipc.addr = fl4->daddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800962
Herbert Xu903ab862011-03-01 02:36:48 +0000963 /* Lockless fast path for the non-corking case. */
964 if (!corkreq) {
David S. Miller77968b72011-05-08 17:12:19 -0700965 skb = ip_make_skb(sk, fl4, getfrag, msg->msg_iov, ulen,
Herbert Xu903ab862011-03-01 02:36:48 +0000966 sizeof(struct udphdr), &ipc, &rt,
967 msg->msg_flags);
968 err = PTR_ERR(skb);
969 if (skb && !IS_ERR(skb))
David S. Miller79ab0532011-05-09 13:31:04 -0700970 err = udp_send_skb(skb, fl4);
Herbert Xu903ab862011-03-01 02:36:48 +0000971 goto out;
972 }
973
David S. Millerdb8dac22008-03-06 16:22:02 -0800974 lock_sock(sk);
975 if (unlikely(up->pending)) {
976 /* The socket is already corked while preparing it. */
977 /* ... which is an evident application bug. --ANK */
978 release_sock(sk);
979
Joe Perchesafd465032012-03-12 07:03:32 +0000980 LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("cork app bug 2\n"));
David S. Millerdb8dac22008-03-06 16:22:02 -0800981 err = -EINVAL;
982 goto out;
983 }
984 /*
985 * Now cork the socket to pend data.
986 */
David S. Millerb6f21b22011-03-12 02:09:18 -0500987 fl4 = &inet->cork.fl.u.ip4;
988 fl4->daddr = daddr;
989 fl4->saddr = saddr;
David S. Miller9cce96d2011-03-12 03:00:33 -0500990 fl4->fl4_dport = dport;
991 fl4->fl4_sport = inet->inet_sport;
David S. Millerdb8dac22008-03-06 16:22:02 -0800992 up->pending = AF_INET;
993
994do_append_data:
995 up->len += ulen;
David S. Millerf5fca602011-05-08 17:24:10 -0700996 err = ip_append_data(sk, fl4, getfrag, msg->msg_iov, ulen,
997 sizeof(struct udphdr), &ipc, &rt,
998 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
David S. Millerdb8dac22008-03-06 16:22:02 -0800999 if (err)
1000 udp_flush_pending_frames(sk);
1001 else if (!corkreq)
1002 err = udp_push_pending_frames(sk);
1003 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1004 up->pending = 0;
1005 release_sock(sk);
1006
1007out:
1008 ip_rt_put(rt);
1009 if (free)
1010 kfree(ipc.opt);
1011 if (!err)
1012 return len;
1013 /*
1014 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1015 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1016 * we don't have a good statistic (IpOutDiscards but it can be too many
1017 * things). We could add another new stat but at least for now that
1018 * seems like overkill.
1019 */
1020 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001021 UDP_INC_STATS_USER(sock_net(sk),
1022 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001023 }
1024 return err;
1025
1026do_confirm:
Changli Gaod8d1f302010-06-10 23:31:35 -07001027 dst_confirm(&rt->dst);
David S. Millerdb8dac22008-03-06 16:22:02 -08001028 if (!(msg->msg_flags&MSG_PROBE) || len)
1029 goto back_from_confirm;
1030 err = 0;
1031 goto out;
1032}
Eric Dumazetc482c562009-07-17 00:26:32 +00001033EXPORT_SYMBOL(udp_sendmsg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001034
1035int udp_sendpage(struct sock *sk, struct page *page, int offset,
1036 size_t size, int flags)
1037{
David S. Millerf5fca602011-05-08 17:24:10 -07001038 struct inet_sock *inet = inet_sk(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001039 struct udp_sock *up = udp_sk(sk);
1040 int ret;
1041
1042 if (!up->pending) {
1043 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
1044
1045 /* Call udp_sendmsg to specify destination address which
1046 * sendpage interface can't pass.
1047 * This will succeed only when the socket is connected.
1048 */
1049 ret = udp_sendmsg(NULL, sk, &msg, 0);
1050 if (ret < 0)
1051 return ret;
1052 }
1053
1054 lock_sock(sk);
1055
1056 if (unlikely(!up->pending)) {
1057 release_sock(sk);
1058
Joe Perchesafd465032012-03-12 07:03:32 +00001059 LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("udp cork app bug 3\n"));
David S. Millerdb8dac22008-03-06 16:22:02 -08001060 return -EINVAL;
1061 }
1062
David S. Millerf5fca602011-05-08 17:24:10 -07001063 ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1064 page, offset, size, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001065 if (ret == -EOPNOTSUPP) {
1066 release_sock(sk);
1067 return sock_no_sendpage(sk->sk_socket, page, offset,
1068 size, flags);
1069 }
1070 if (ret < 0) {
1071 udp_flush_pending_frames(sk);
1072 goto out;
1073 }
1074
1075 up->len += size;
1076 if (!(up->corkflag || (flags&MSG_MORE)))
1077 ret = udp_push_pending_frames(sk);
1078 if (!ret)
1079 ret = size;
1080out:
1081 release_sock(sk);
1082 return ret;
1083}
1084
Eric Dumazet85584672009-10-09 04:43:40 +00001085
1086/**
1087 * first_packet_length - return length of first packet in receive queue
1088 * @sk: socket
1089 *
1090 * Drops all bad checksum frames, until a valid one is found.
1091 * Returns the length of found skb, or 0 if none is found.
1092 */
1093static unsigned int first_packet_length(struct sock *sk)
1094{
1095 struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1096 struct sk_buff *skb;
1097 unsigned int res;
1098
1099 __skb_queue_head_init(&list_kill);
1100
1101 spin_lock_bh(&rcvq->lock);
1102 while ((skb = skb_peek(rcvq)) != NULL &&
1103 udp_lib_checksum_complete(skb)) {
1104 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1105 IS_UDPLITE(sk));
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001106 atomic_inc(&sk->sk_drops);
Eric Dumazet85584672009-10-09 04:43:40 +00001107 __skb_unlink(skb, rcvq);
1108 __skb_queue_tail(&list_kill, skb);
1109 }
1110 res = skb ? skb->len : 0;
1111 spin_unlock_bh(&rcvq->lock);
1112
1113 if (!skb_queue_empty(&list_kill)) {
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001114 bool slow = lock_sock_fast(sk);
1115
Eric Dumazet85584672009-10-09 04:43:40 +00001116 __skb_queue_purge(&list_kill);
1117 sk_mem_reclaim_partial(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001118 unlock_sock_fast(sk, slow);
Eric Dumazet85584672009-10-09 04:43:40 +00001119 }
1120 return res;
1121}
1122
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123/*
1124 * IOCTL requests applicable to the UDP protocol
1125 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001126
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1128{
Stephen Hemminger6516c652007-03-08 20:41:55 -08001129 switch (cmd) {
1130 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 {
Eric Dumazet31e6d362009-06-17 19:05:41 -07001132 int amount = sk_wmem_alloc_get(sk);
1133
Stephen Hemminger6516c652007-03-08 20:41:55 -08001134 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 }
Stephen Hemminger6516c652007-03-08 20:41:55 -08001136
1137 case SIOCINQ:
1138 {
Eric Dumazet85584672009-10-09 04:43:40 +00001139 unsigned int amount = first_packet_length(sk);
Stephen Hemminger6516c652007-03-08 20:41:55 -08001140
Eric Dumazet85584672009-10-09 04:43:40 +00001141 if (amount)
Stephen Hemminger6516c652007-03-08 20:41:55 -08001142 /*
1143 * We will only return the amount
1144 * of this packet since that is all
1145 * that will be read.
1146 */
Eric Dumazet85584672009-10-09 04:43:40 +00001147 amount -= sizeof(struct udphdr);
1148
Stephen Hemminger6516c652007-03-08 20:41:55 -08001149 return put_user(amount, (int __user *)arg);
1150 }
1151
1152 default:
1153 return -ENOIOCTLCMD;
1154 }
1155
1156 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157}
Eric Dumazetc482c562009-07-17 00:26:32 +00001158EXPORT_SYMBOL(udp_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159
David S. Millerdb8dac22008-03-06 16:22:02 -08001160/*
1161 * This should be easy, if there is something there we
1162 * return it, otherwise we block.
1163 */
1164
1165int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
1166 size_t len, int noblock, int flags, int *addr_len)
1167{
1168 struct inet_sock *inet = inet_sk(sk);
1169 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
1170 struct sk_buff *skb;
David S. Miller59c2cda2011-12-01 14:12:55 -05001171 unsigned int ulen, copied;
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001172 int peeked, off = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001173 int err;
1174 int is_udplite = IS_UDPLITE(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001175 bool slow;
David S. Millerdb8dac22008-03-06 16:22:02 -08001176
1177 /*
1178 * Check any passed addresses
1179 */
1180 if (addr_len)
Eric Dumazetc482c562009-07-17 00:26:32 +00001181 *addr_len = sizeof(*sin);
David S. Millerdb8dac22008-03-06 16:22:02 -08001182
1183 if (flags & MSG_ERRQUEUE)
1184 return ip_recv_error(sk, msg, len);
1185
1186try_again:
1187 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001188 &peeked, &off, &err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001189 if (!skb)
1190 goto out;
1191
1192 ulen = skb->len - sizeof(struct udphdr);
David S. Miller59c2cda2011-12-01 14:12:55 -05001193 copied = len;
1194 if (copied > ulen)
1195 copied = ulen;
1196 else if (copied < ulen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001197 msg->msg_flags |= MSG_TRUNC;
1198
1199 /*
1200 * If checksum is needed at all, try to do it while copying the
1201 * data. If the data is truncated, or if we only want a partial
1202 * coverage checksum (UDP-Lite), do it before the copy.
1203 */
1204
David S. Miller59c2cda2011-12-01 14:12:55 -05001205 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001206 if (udp_lib_checksum_complete(skb))
1207 goto csum_copy_err;
1208 }
1209
1210 if (skb_csum_unnecessary(skb))
1211 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
David S. Miller59c2cda2011-12-01 14:12:55 -05001212 msg->msg_iov, copied);
David S. Millerdb8dac22008-03-06 16:22:02 -08001213 else {
Eric Dumazetc482c562009-07-17 00:26:32 +00001214 err = skb_copy_and_csum_datagram_iovec(skb,
1215 sizeof(struct udphdr),
1216 msg->msg_iov);
David S. Millerdb8dac22008-03-06 16:22:02 -08001217
1218 if (err == -EINVAL)
1219 goto csum_copy_err;
1220 }
1221
1222 if (err)
1223 goto out_free;
1224
1225 if (!peeked)
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001226 UDP_INC_STATS_USER(sock_net(sk),
1227 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001228
Neil Horman3b885782009-10-12 13:26:31 -07001229 sock_recv_ts_and_drops(msg, sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001230
1231 /* Copy the address. */
Eric Dumazetc482c562009-07-17 00:26:32 +00001232 if (sin) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001233 sin->sin_family = AF_INET;
1234 sin->sin_port = udp_hdr(skb)->source;
1235 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1236 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
1237 }
1238 if (inet->cmsg_flags)
1239 ip_cmsg_recv(msg, skb);
1240
David S. Miller59c2cda2011-12-01 14:12:55 -05001241 err = copied;
David S. Millerdb8dac22008-03-06 16:22:02 -08001242 if (flags & MSG_TRUNC)
1243 err = ulen;
1244
1245out_free:
Eric Dumazet9d410c72009-10-30 05:03:53 +00001246 skb_free_datagram_locked(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001247out:
1248 return err;
1249
1250csum_copy_err:
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001251 slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001252 if (!skb_kill_datagram(sk, skb, flags))
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001253 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001254 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001255
1256 if (noblock)
1257 return -EAGAIN;
Xufeng Zhang9cfaa8d2011-06-21 10:43:40 +00001258
1259 /* starting over for a new packet */
1260 msg->msg_flags &= ~MSG_TRUNC;
David S. Millerdb8dac22008-03-06 16:22:02 -08001261 goto try_again;
1262}
1263
1264
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265int udp_disconnect(struct sock *sk, int flags)
1266{
1267 struct inet_sock *inet = inet_sk(sk);
1268 /*
1269 * 1003.1g - break association.
1270 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001271
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272 sk->sk_state = TCP_CLOSE;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001273 inet->inet_daddr = 0;
1274 inet->inet_dport = 0;
Tom Herbertbdeab992011-08-14 19:45:55 +00001275 sock_rps_reset_rxhash(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 sk->sk_bound_dev_if = 0;
1277 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1278 inet_reset_saddr(sk);
1279
1280 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1281 sk->sk_prot->unhash(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001282 inet->inet_sport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 }
1284 sk_dst_reset(sk);
1285 return 0;
1286}
Eric Dumazetc482c562009-07-17 00:26:32 +00001287EXPORT_SYMBOL(udp_disconnect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288
Eric Dumazet645ca702008-10-29 01:41:45 -07001289void udp_lib_unhash(struct sock *sk)
1290{
Eric Dumazet723b4612008-11-25 13:55:15 -08001291 if (sk_hashed(sk)) {
1292 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Eric Dumazet512615b2009-11-08 10:17:58 +00001293 struct udp_hslot *hslot, *hslot2;
1294
1295 hslot = udp_hashslot(udptable, sock_net(sk),
1296 udp_sk(sk)->udp_port_hash);
1297 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
Eric Dumazet645ca702008-10-29 01:41:45 -07001298
Eric Dumazet723b4612008-11-25 13:55:15 -08001299 spin_lock_bh(&hslot->lock);
1300 if (sk_nulls_del_node_init_rcu(sk)) {
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00001301 hslot->count--;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001302 inet_sk(sk)->inet_num = 0;
Eric Dumazet723b4612008-11-25 13:55:15 -08001303 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet512615b2009-11-08 10:17:58 +00001304
1305 spin_lock(&hslot2->lock);
1306 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1307 hslot2->count--;
1308 spin_unlock(&hslot2->lock);
Eric Dumazet723b4612008-11-25 13:55:15 -08001309 }
1310 spin_unlock_bh(&hslot->lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001311 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001312}
1313EXPORT_SYMBOL(udp_lib_unhash);
1314
Eric Dumazet719f8352010-09-08 05:08:44 +00001315/*
1316 * inet_rcv_saddr was changed, we must rehash secondary hash
1317 */
1318void udp_lib_rehash(struct sock *sk, u16 newhash)
1319{
1320 if (sk_hashed(sk)) {
1321 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1322 struct udp_hslot *hslot, *hslot2, *nhslot2;
1323
1324 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1325 nhslot2 = udp_hashslot2(udptable, newhash);
1326 udp_sk(sk)->udp_portaddr_hash = newhash;
1327 if (hslot2 != nhslot2) {
1328 hslot = udp_hashslot(udptable, sock_net(sk),
1329 udp_sk(sk)->udp_port_hash);
1330 /* we must lock primary chain too */
1331 spin_lock_bh(&hslot->lock);
1332
1333 spin_lock(&hslot2->lock);
1334 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1335 hslot2->count--;
1336 spin_unlock(&hslot2->lock);
1337
1338 spin_lock(&nhslot2->lock);
1339 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
1340 &nhslot2->head);
1341 nhslot2->count++;
1342 spin_unlock(&nhslot2->lock);
1343
1344 spin_unlock_bh(&hslot->lock);
1345 }
1346 }
1347}
1348EXPORT_SYMBOL(udp_lib_rehash);
1349
1350static void udp_v4_rehash(struct sock *sk)
1351{
1352 u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1353 inet_sk(sk)->inet_rcv_saddr,
1354 inet_sk(sk)->inet_num);
1355 udp_lib_rehash(sk, new_hash);
1356}
1357
Herbert Xu93821772008-09-15 11:48:46 -07001358static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1359{
Tom Herbertfec5e652010-04-16 16:01:27 -07001360 int rc;
Herbert Xu93821772008-09-15 11:48:46 -07001361
Tom Herbertfec5e652010-04-16 16:01:27 -07001362 if (inet_sk(sk)->inet_daddr)
Tom Herbertbdeab992011-08-14 19:45:55 +00001363 sock_rps_save_rxhash(sk, skb);
Tom Herbertfec5e652010-04-16 16:01:27 -07001364
Eric Dumazetd826eb12011-11-09 07:24:35 +00001365 rc = sock_queue_rcv_skb(sk, skb);
Eric Dumazet766e90372009-10-14 20:40:11 -07001366 if (rc < 0) {
1367 int is_udplite = IS_UDPLITE(sk);
1368
Herbert Xu93821772008-09-15 11:48:46 -07001369 /* Note that an ENOMEM error is charged twice */
Eric Dumazet766e90372009-10-14 20:40:11 -07001370 if (rc == -ENOMEM)
Herbert Xu93821772008-09-15 11:48:46 -07001371 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1372 is_udplite);
Eric Dumazet766e90372009-10-14 20:40:11 -07001373 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1374 kfree_skb(skb);
Satoru Moriya296f7ea2011-06-17 11:58:39 +00001375 trace_udp_fail_queue_rcv_skb(rc, sk);
Eric Dumazet766e90372009-10-14 20:40:11 -07001376 return -1;
Herbert Xu93821772008-09-15 11:48:46 -07001377 }
1378
1379 return 0;
1380
Herbert Xu93821772008-09-15 11:48:46 -07001381}
1382
David S. Millerdb8dac22008-03-06 16:22:02 -08001383/* returns:
1384 * -1: error
1385 * 0: success
1386 * >0: "udp encap" protocol resubmission
1387 *
1388 * Note that in the success and error cases, the skb is assumed to
1389 * have either been requeued or freed.
1390 */
Eric Dumazetc482c562009-07-17 00:26:32 +00001391int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -08001392{
1393 struct udp_sock *up = udp_sk(sk);
1394 int rc;
1395 int is_udplite = IS_UDPLITE(sk);
1396
1397 /*
1398 * Charge it to the socket, dropping if the queue is full.
1399 */
1400 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1401 goto drop;
1402 nf_reset(skb);
1403
1404 if (up->encap_type) {
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001405 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1406
David S. Millerdb8dac22008-03-06 16:22:02 -08001407 /*
1408 * This is an encapsulation socket so pass the skb to
1409 * the socket's udp_encap_rcv() hook. Otherwise, just
1410 * fall through and pass this up the UDP socket.
1411 * up->encap_rcv() returns the following value:
1412 * =0 if skb was successfully passed to the encap
1413 * handler or was discarded by it.
1414 * >0 if skb should be passed on to UDP.
1415 * <0 if skb should be resubmitted as proto -N
1416 */
1417
1418 /* if we're overly short, let UDP handle it */
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001419 encap_rcv = ACCESS_ONCE(up->encap_rcv);
1420 if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001421 int ret;
1422
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001423 ret = encap_rcv(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001424 if (ret <= 0) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001425 UDP_INC_STATS_BH(sock_net(sk),
1426 UDP_MIB_INDATAGRAMS,
David S. Millerdb8dac22008-03-06 16:22:02 -08001427 is_udplite);
1428 return -ret;
1429 }
1430 }
1431
1432 /* FALLTHROUGH -- it's a UDP Packet */
1433 }
1434
1435 /*
1436 * UDP-Lite specific tests, ignored on UDP sockets
1437 */
1438 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1439
1440 /*
1441 * MIB statistics other than incrementing the error count are
1442 * disabled for the following two types of errors: these depend
1443 * on the application settings, not on the functioning of the
1444 * protocol stack as such.
1445 *
1446 * RFC 3828 here recommends (sec 3.3): "There should also be a
1447 * way ... to ... at least let the receiving application block
1448 * delivery of packets with coverage values less than a value
1449 * provided by the application."
1450 */
1451 if (up->pcrlen == 0) { /* full coverage was set */
Joe Perchesafd465032012-03-12 07:03:32 +00001452 LIMIT_NETDEBUG(KERN_WARNING "UDPLite: partial coverage %d while full coverage %d requested\n",
1453 UDP_SKB_CB(skb)->cscov, skb->len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001454 goto drop;
1455 }
1456 /* The next case involves violating the min. coverage requested
1457 * by the receiver. This is subtle: if receiver wants x and x is
1458 * greater than the buffersize/MTU then receiver will complain
1459 * that it wants x while sender emits packets of smaller size y.
1460 * Therefore the above ...()->partial_cov statement is essential.
1461 */
1462 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
Joe Perchesafd465032012-03-12 07:03:32 +00001463 LIMIT_NETDEBUG(KERN_WARNING "UDPLite: coverage %d too small, need min %d\n",
1464 UDP_SKB_CB(skb)->cscov, up->pcrlen);
David S. Millerdb8dac22008-03-06 16:22:02 -08001465 goto drop;
1466 }
1467 }
1468
Eric Dumazet33d480c2011-08-11 19:30:52 +00001469 if (rcu_access_pointer(sk->sk_filter) &&
1470 udp_lib_checksum_complete(skb))
1471 goto drop;
David S. Millerdb8dac22008-03-06 16:22:02 -08001472
Eric Dumazetc3774112010-04-27 15:13:20 -07001473
1474 if (sk_rcvqueues_full(sk, skb))
1475 goto drop;
1476
Herbert Xu93821772008-09-15 11:48:46 -07001477 rc = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001478
Eric Dumazetd826eb12011-11-09 07:24:35 +00001479 ipv4_pktinfo_prepare(skb);
Herbert Xu93821772008-09-15 11:48:46 -07001480 bh_lock_sock(sk);
1481 if (!sock_owned_by_user(sk))
1482 rc = __udp_queue_rcv_skb(sk, skb);
Zhu Yia3a858f2010-03-04 18:01:47 +00001483 else if (sk_add_backlog(sk, skb)) {
Zhu Yi55349792010-03-04 18:01:42 +00001484 bh_unlock_sock(sk);
1485 goto drop;
1486 }
Herbert Xu93821772008-09-15 11:48:46 -07001487 bh_unlock_sock(sk);
1488
1489 return rc;
David S. Millerdb8dac22008-03-06 16:22:02 -08001490
1491drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001492 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001493 atomic_inc(&sk->sk_drops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001494 kfree_skb(skb);
1495 return -1;
1496}
1497
Eric Dumazet1240d132009-11-08 10:18:44 +00001498
1499static void flush_stack(struct sock **stack, unsigned int count,
1500 struct sk_buff *skb, unsigned int final)
1501{
1502 unsigned int i;
1503 struct sk_buff *skb1 = NULL;
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001504 struct sock *sk;
Eric Dumazet1240d132009-11-08 10:18:44 +00001505
1506 for (i = 0; i < count; i++) {
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001507 sk = stack[i];
Eric Dumazet1240d132009-11-08 10:18:44 +00001508 if (likely(skb1 == NULL))
1509 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
1510
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001511 if (!skb1) {
1512 atomic_inc(&sk->sk_drops);
1513 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1514 IS_UDPLITE(sk));
1515 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1516 IS_UDPLITE(sk));
1517 }
1518
1519 if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0)
Eric Dumazet1240d132009-11-08 10:18:44 +00001520 skb1 = NULL;
1521 }
1522 if (unlikely(skb1))
1523 kfree_skb(skb1);
1524}
1525
David S. Millerdb8dac22008-03-06 16:22:02 -08001526/*
1527 * Multicasts and broadcasts go to each listener.
1528 *
Eric Dumazet1240d132009-11-08 10:18:44 +00001529 * Note: called only from the BH handler context.
David S. Millerdb8dac22008-03-06 16:22:02 -08001530 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001531static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001532 struct udphdr *uh,
1533 __be32 saddr, __be32 daddr,
Eric Dumazet645ca702008-10-29 01:41:45 -07001534 struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -08001535{
Eric Dumazet1240d132009-11-08 10:18:44 +00001536 struct sock *sk, *stack[256 / sizeof(struct sock *)];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00001537 struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001538 int dif;
Eric Dumazet1240d132009-11-08 10:18:44 +00001539 unsigned int i, count = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001540
Eric Dumazet645ca702008-10-29 01:41:45 -07001541 spin_lock(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08001542 sk = sk_nulls_head(&hslot->head);
David S. Millerdb8dac22008-03-06 16:22:02 -08001543 dif = skb->dev->ifindex;
Eric Dumazet920a4612008-11-01 21:22:23 -07001544 sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
Eric Dumazet1240d132009-11-08 10:18:44 +00001545 while (sk) {
1546 stack[count++] = sk;
1547 sk = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest,
1548 daddr, uh->source, saddr, dif);
1549 if (unlikely(count == ARRAY_SIZE(stack))) {
1550 if (!sk)
1551 break;
1552 flush_stack(stack, count, skb, ~0);
1553 count = 0;
1554 }
1555 }
1556 /*
1557 * before releasing chain lock, we must take a reference on sockets
1558 */
1559 for (i = 0; i < count; i++)
1560 sock_hold(stack[i]);
David S. Millerdb8dac22008-03-06 16:22:02 -08001561
Eric Dumazet645ca702008-10-29 01:41:45 -07001562 spin_unlock(&hslot->lock);
Eric Dumazet1240d132009-11-08 10:18:44 +00001563
1564 /*
1565 * do the slow work with no lock held
1566 */
1567 if (count) {
1568 flush_stack(stack, count, skb, count - 1);
1569
1570 for (i = 0; i < count; i++)
1571 sock_put(stack[i]);
1572 } else {
1573 kfree_skb(skb);
1574 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001575 return 0;
1576}
1577
1578/* Initialize UDP checksum. If exited with zero value (success),
1579 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1580 * Otherwise, csum completion requires chacksumming packet body,
1581 * including udp header and folding it to skb->csum.
1582 */
1583static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1584 int proto)
1585{
1586 const struct iphdr *iph;
1587 int err;
1588
1589 UDP_SKB_CB(skb)->partial_cov = 0;
1590 UDP_SKB_CB(skb)->cscov = skb->len;
1591
1592 if (proto == IPPROTO_UDPLITE) {
1593 err = udplite_checksum_init(skb, uh);
1594 if (err)
1595 return err;
1596 }
1597
1598 iph = ip_hdr(skb);
1599 if (uh->check == 0) {
1600 skb->ip_summed = CHECKSUM_UNNECESSARY;
1601 } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
Eric Dumazetc482c562009-07-17 00:26:32 +00001602 if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len,
David S. Millerdb8dac22008-03-06 16:22:02 -08001603 proto, skb->csum))
1604 skb->ip_summed = CHECKSUM_UNNECESSARY;
1605 }
1606 if (!skb_csum_unnecessary(skb))
1607 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
1608 skb->len, proto, 0);
1609 /* Probably, we should checksum udp header (it should be in cache
1610 * in any case) and data in tiny packets (< rx copybreak).
1611 */
1612
1613 return 0;
1614}
1615
1616/*
1617 * All we need to do is get the socket, and then do a checksum.
1618 */
1619
Eric Dumazet645ca702008-10-29 01:41:45 -07001620int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
David S. Millerdb8dac22008-03-06 16:22:02 -08001621 int proto)
1622{
1623 struct sock *sk;
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001624 struct udphdr *uh;
David S. Millerdb8dac22008-03-06 16:22:02 -08001625 unsigned short ulen;
Eric Dumazetadf30902009-06-02 05:19:30 +00001626 struct rtable *rt = skb_rtable(skb);
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001627 __be32 saddr, daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001628 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001629
1630 /*
1631 * Validate the packet.
1632 */
1633 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1634 goto drop; /* No space for header. */
1635
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001636 uh = udp_hdr(skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001637 ulen = ntohs(uh->len);
Bjørn Morkccc2d972010-05-06 03:44:34 +00001638 saddr = ip_hdr(skb)->saddr;
1639 daddr = ip_hdr(skb)->daddr;
1640
David S. Millerdb8dac22008-03-06 16:22:02 -08001641 if (ulen > skb->len)
1642 goto short_packet;
1643
1644 if (proto == IPPROTO_UDP) {
1645 /* UDP validates ulen. */
1646 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1647 goto short_packet;
1648 uh = udp_hdr(skb);
1649 }
1650
1651 if (udp4_csum_init(skb, uh, proto))
1652 goto csum_error;
1653
1654 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
Pavel Emelyanove3163492008-06-16 17:12:11 -07001655 return __udp4_lib_mcast_deliver(net, skb, uh,
1656 saddr, daddr, udptable);
David S. Millerdb8dac22008-03-06 16:22:02 -08001657
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -07001658 sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
David S. Millerdb8dac22008-03-06 16:22:02 -08001659
1660 if (sk != NULL) {
Herbert Xu93821772008-09-15 11:48:46 -07001661 int ret = udp_queue_rcv_skb(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001662 sock_put(sk);
1663
1664 /* a return value > 0 means to resubmit the input, but
1665 * it wants the return to be -protocol, or 0
1666 */
1667 if (ret > 0)
1668 return -ret;
1669 return 0;
1670 }
1671
1672 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1673 goto drop;
1674 nf_reset(skb);
1675
1676 /* No socket. Drop packet silently, if checksum is wrong */
1677 if (udp_lib_checksum_complete(skb))
1678 goto csum_error;
1679
Pavel Emelyanov02833282008-07-05 21:18:48 -07001680 UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001681 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1682
1683 /*
1684 * Hmm. We got an UDP packet to a port to which we
1685 * don't wanna listen. Ignore it.
1686 */
1687 kfree_skb(skb);
1688 return 0;
1689
1690short_packet:
Harvey Harrison673d57e2008-10-31 00:53:57 -07001691 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
Joe Perchesafd465032012-03-12 07:03:32 +00001692 proto == IPPROTO_UDPLITE ? "Lite" : "",
1693 &saddr, ntohs(uh->source),
1694 ulen, skb->len,
1695 &daddr, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001696 goto drop;
1697
1698csum_error:
1699 /*
1700 * RFC1122: OK. Discards the bad packet silently (as far as
1701 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1702 */
Harvey Harrison673d57e2008-10-31 00:53:57 -07001703 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
Joe Perchesafd465032012-03-12 07:03:32 +00001704 proto == IPPROTO_UDPLITE ? "Lite" : "",
1705 &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
David S. Millerdb8dac22008-03-06 16:22:02 -08001706 ulen);
1707drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001708 UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001709 kfree_skb(skb);
1710 return 0;
1711}
1712
1713int udp_rcv(struct sk_buff *skb)
1714{
Eric Dumazet645ca702008-10-29 01:41:45 -07001715 return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
David S. Millerdb8dac22008-03-06 16:22:02 -08001716}
1717
Brian Haley7d06b2e2008-06-14 17:04:49 -07001718void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08001719{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001720 bool slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001721 udp_flush_pending_frames(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001722 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001723}
1724
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725/*
1726 * Socket option code for UDP
1727 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001728int udp_lib_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001729 char __user *optval, unsigned int optlen,
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001730 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731{
1732 struct udp_sock *up = udp_sk(sk);
1733 int val;
1734 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11001735 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736
Eric Dumazetc482c562009-07-17 00:26:32 +00001737 if (optlen < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738 return -EINVAL;
1739
1740 if (get_user(val, (int __user *)optval))
1741 return -EFAULT;
1742
Stephen Hemminger6516c652007-03-08 20:41:55 -08001743 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 case UDP_CORK:
1745 if (val != 0) {
1746 up->corkflag = 1;
1747 } else {
1748 up->corkflag = 0;
1749 lock_sock(sk);
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001750 (*push_pending_frames)(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 release_sock(sk);
1752 }
1753 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001754
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 case UDP_ENCAP:
1756 switch (val) {
1757 case 0:
1758 case UDP_ENCAP_ESPINUDP:
1759 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07001760 up->encap_rcv = xfrm4_udp_encap_rcv;
1761 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07001762 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 up->encap_type = val;
1764 break;
1765 default:
1766 err = -ENOPROTOOPT;
1767 break;
1768 }
1769 break;
1770
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001771 /*
1772 * UDP-Lite's partial checksum coverage (RFC 3828).
1773 */
1774 /* The sender sets actual checksum coverage length via this option.
1775 * The case coverage > packet length is handled by send module. */
1776 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11001777 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001778 return -ENOPROTOOPT;
1779 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
1780 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07001781 else if (val > USHRT_MAX)
1782 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001783 up->pcslen = val;
1784 up->pcflag |= UDPLITE_SEND_CC;
1785 break;
1786
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001787 /* The receiver specifies a minimum checksum coverage value. To make
1788 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001789 * used, this again means full checksum coverage. */
1790 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11001791 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001792 return -ENOPROTOOPT;
1793 if (val != 0 && val < 8) /* Avoid silly minimal values. */
1794 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07001795 else if (val > USHRT_MAX)
1796 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001797 up->pcrlen = val;
1798 up->pcflag |= UDPLITE_RECV_CC;
1799 break;
1800
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 default:
1802 err = -ENOPROTOOPT;
1803 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001804 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805
1806 return err;
1807}
Eric Dumazetc482c562009-07-17 00:26:32 +00001808EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809
David S. Millerdb8dac22008-03-06 16:22:02 -08001810int udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001811 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001812{
1813 if (level == SOL_UDP || level == SOL_UDPLITE)
1814 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1815 udp_push_pending_frames);
1816 return ip_setsockopt(sk, level, optname, optval, optlen);
1817}
1818
1819#ifdef CONFIG_COMPAT
1820int compat_udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001821 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001822{
1823 if (level == SOL_UDP || level == SOL_UDPLITE)
1824 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1825 udp_push_pending_frames);
1826 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
1827}
1828#endif
1829
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001830int udp_lib_getsockopt(struct sock *sk, int level, int optname,
1831 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832{
1833 struct udp_sock *up = udp_sk(sk);
1834 int val, len;
1835
Eric Dumazetc482c562009-07-17 00:26:32 +00001836 if (get_user(len, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837 return -EFAULT;
1838
1839 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001840
Stephen Hemminger6516c652007-03-08 20:41:55 -08001841 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842 return -EINVAL;
1843
Stephen Hemminger6516c652007-03-08 20:41:55 -08001844 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845 case UDP_CORK:
1846 val = up->corkflag;
1847 break;
1848
1849 case UDP_ENCAP:
1850 val = up->encap_type;
1851 break;
1852
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001853 /* The following two cannot be changed on UDP sockets, the return is
1854 * always 0 (which corresponds to the full checksum coverage of UDP). */
1855 case UDPLITE_SEND_CSCOV:
1856 val = up->pcslen;
1857 break;
1858
1859 case UDPLITE_RECV_CSCOV:
1860 val = up->pcrlen;
1861 break;
1862
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 default:
1864 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001865 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866
Stephen Hemminger6516c652007-03-08 20:41:55 -08001867 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001868 return -EFAULT;
Eric Dumazetc482c562009-07-17 00:26:32 +00001869 if (copy_to_user(optval, &val, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001871 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872}
Eric Dumazetc482c562009-07-17 00:26:32 +00001873EXPORT_SYMBOL(udp_lib_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874
David S. Millerdb8dac22008-03-06 16:22:02 -08001875int udp_getsockopt(struct sock *sk, int level, int optname,
1876 char __user *optval, int __user *optlen)
1877{
1878 if (level == SOL_UDP || level == SOL_UDPLITE)
1879 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1880 return ip_getsockopt(sk, level, optname, optval, optlen);
1881}
1882
1883#ifdef CONFIG_COMPAT
1884int compat_udp_getsockopt(struct sock *sk, int level, int optname,
1885 char __user *optval, int __user *optlen)
1886{
1887 if (level == SOL_UDP || level == SOL_UDPLITE)
1888 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1889 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
1890}
1891#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892/**
1893 * udp_poll - wait for a UDP event.
1894 * @file - file struct
1895 * @sock - socket
1896 * @wait - poll table
1897 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001898 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 * blocking sockets. If application is using a blocking fd
1900 * and a packet with checksum error is in the queue;
1901 * then it could get return from select indicating data available
1902 * but then block when reading it. Add special case code
1903 * to work around these arguably broken applications.
1904 */
1905unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1906{
1907 unsigned int mask = datagram_poll(file, sock, wait);
1908 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001909
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 /* Check for false positives due to checksum errors */
Eric Dumazet85584672009-10-09 04:43:40 +00001911 if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
1912 !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk))
1913 mask &= ~(POLLIN | POLLRDNORM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914
1915 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001916
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917}
Eric Dumazetc482c562009-07-17 00:26:32 +00001918EXPORT_SYMBOL(udp_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919
David S. Millerdb8dac22008-03-06 16:22:02 -08001920struct proto udp_prot = {
1921 .name = "UDP",
1922 .owner = THIS_MODULE,
1923 .close = udp_lib_close,
1924 .connect = ip4_datagram_connect,
1925 .disconnect = udp_disconnect,
1926 .ioctl = udp_ioctl,
1927 .destroy = udp_destroy_sock,
1928 .setsockopt = udp_setsockopt,
1929 .getsockopt = udp_getsockopt,
1930 .sendmsg = udp_sendmsg,
1931 .recvmsg = udp_recvmsg,
1932 .sendpage = udp_sendpage,
Herbert Xu93821772008-09-15 11:48:46 -07001933 .backlog_rcv = __udp_queue_rcv_skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001934 .hash = udp_lib_hash,
1935 .unhash = udp_lib_unhash,
Eric Dumazet719f8352010-09-08 05:08:44 +00001936 .rehash = udp_v4_rehash,
David S. Millerdb8dac22008-03-06 16:22:02 -08001937 .get_port = udp_v4_get_port,
1938 .memory_allocated = &udp_memory_allocated,
1939 .sysctl_mem = sysctl_udp_mem,
1940 .sysctl_wmem = &sysctl_udp_wmem_min,
1941 .sysctl_rmem = &sysctl_udp_rmem_min,
1942 .obj_size = sizeof(struct udp_sock),
Eric Dumazet271b72c2008-10-29 02:11:14 -07001943 .slab_flags = SLAB_DESTROY_BY_RCU,
Eric Dumazet645ca702008-10-29 01:41:45 -07001944 .h.udp_table = &udp_table,
David S. Millerdb8dac22008-03-06 16:22:02 -08001945#ifdef CONFIG_COMPAT
1946 .compat_setsockopt = compat_udp_setsockopt,
1947 .compat_getsockopt = compat_udp_getsockopt,
1948#endif
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001949 .clear_sk = sk_prot_clear_portaddr_nulls,
David S. Millerdb8dac22008-03-06 16:22:02 -08001950};
Eric Dumazetc482c562009-07-17 00:26:32 +00001951EXPORT_SYMBOL(udp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952
1953/* ------------------------------------------------------------------------ */
1954#ifdef CONFIG_PROC_FS
1955
Eric Dumazet645ca702008-10-29 01:41:45 -07001956static struct sock *udp_get_first(struct seq_file *seq, int start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957{
1958 struct sock *sk;
1959 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07001960 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961
Eric Dumazetf86dcc52009-10-07 00:37:59 +00001962 for (state->bucket = start; state->bucket <= state->udp_table->mask;
1963 ++state->bucket) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08001964 struct hlist_nulls_node *node;
Eric Dumazet645ca702008-10-29 01:41:45 -07001965 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00001966
1967 if (hlist_nulls_empty(&hslot->head))
1968 continue;
1969
Eric Dumazet645ca702008-10-29 01:41:45 -07001970 spin_lock_bh(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08001971 sk_nulls_for_each(sk, node, &hslot->head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09001972 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07001973 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974 if (sk->sk_family == state->family)
1975 goto found;
1976 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001977 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 }
1979 sk = NULL;
1980found:
1981 return sk;
1982}
1983
1984static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1985{
1986 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07001987 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988
1989 do {
Eric Dumazet88ab1932008-11-16 19:39:21 -08001990 sk = sk_nulls_next(sk);
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09001991 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992
Eric Dumazet645ca702008-10-29 01:41:45 -07001993 if (!sk) {
Eric Dumazetf86dcc52009-10-07 00:37:59 +00001994 if (state->bucket <= state->udp_table->mask)
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07001995 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001996 return udp_get_first(seq, state->bucket + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997 }
1998 return sk;
1999}
2000
2001static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2002{
Eric Dumazet645ca702008-10-29 01:41:45 -07002003 struct sock *sk = udp_get_first(seq, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004
2005 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08002006 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 --pos;
2008 return pos ? NULL : sk;
2009}
2010
2011static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2012{
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002013 struct udp_iter_state *state = seq->private;
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002014 state->bucket = MAX_UDP_PORTS;
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002015
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002016 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017}
2018
2019static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2020{
2021 struct sock *sk;
2022
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002023 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002024 sk = udp_get_idx(seq, 0);
2025 else
2026 sk = udp_get_next(seq, v);
2027
2028 ++*pos;
2029 return sk;
2030}
2031
2032static void udp_seq_stop(struct seq_file *seq, void *v)
2033{
Eric Dumazet645ca702008-10-29 01:41:45 -07002034 struct udp_iter_state *state = seq->private;
2035
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002036 if (state->bucket <= state->udp_table->mask)
Eric Dumazet645ca702008-10-29 01:41:45 -07002037 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038}
2039
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002040int udp_seq_open(struct inode *inode, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041{
2042 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002043 struct udp_iter_state *s;
2044 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002046 err = seq_open_net(inode, file, &afinfo->seq_ops,
2047 sizeof(struct udp_iter_state));
2048 if (err < 0)
2049 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002050
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002051 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052 s->family = afinfo->family;
Eric Dumazet645ca702008-10-29 01:41:45 -07002053 s->udp_table = afinfo->udp_table;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002054 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002055}
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002056EXPORT_SYMBOL(udp_seq_open);
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002057
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002059int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060{
2061 struct proc_dir_entry *p;
2062 int rc = 0;
2063
Denis V. Lunevdda61922008-03-28 18:24:26 -07002064 afinfo->seq_ops.start = udp_seq_start;
2065 afinfo->seq_ops.next = udp_seq_next;
2066 afinfo->seq_ops.stop = udp_seq_stop;
2067
Denis V. Lunev84841c32008-05-02 04:10:08 -07002068 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002069 afinfo->seq_fops, afinfo);
Denis V. Lunev84841c32008-05-02 04:10:08 -07002070 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071 rc = -ENOMEM;
2072 return rc;
2073}
Eric Dumazetc482c562009-07-17 00:26:32 +00002074EXPORT_SYMBOL(udp_proc_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002076void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002077{
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002078 proc_net_remove(net, afinfo->name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079}
Eric Dumazetc482c562009-07-17 00:26:32 +00002080EXPORT_SYMBOL(udp_proc_unregister);
David S. Millerdb8dac22008-03-06 16:22:02 -08002081
2082/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002083static void udp4_format_sock(struct sock *sp, struct seq_file *f,
2084 int bucket, int *len)
David S. Millerdb8dac22008-03-06 16:22:02 -08002085{
2086 struct inet_sock *inet = inet_sk(sp);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002087 __be32 dest = inet->inet_daddr;
2088 __be32 src = inet->inet_rcv_saddr;
2089 __u16 destp = ntohs(inet->inet_dport);
2090 __u16 srcp = ntohs(inet->inet_sport);
David S. Millerdb8dac22008-03-06 16:22:02 -08002091
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002092 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
Dan Rosenberg71338aa2011-05-23 12:17:35 +00002093 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
David S. Millerdb8dac22008-03-06 16:22:02 -08002094 bucket, src, srcp, dest, destp, sp->sk_state,
Eric Dumazet31e6d362009-06-17 19:05:41 -07002095 sk_wmem_alloc_get(sp),
2096 sk_rmem_alloc_get(sp),
David S. Millerdb8dac22008-03-06 16:22:02 -08002097 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002098 atomic_read(&sp->sk_refcnt), sp,
2099 atomic_read(&sp->sk_drops), len);
David S. Millerdb8dac22008-03-06 16:22:02 -08002100}
2101
2102int udp4_seq_show(struct seq_file *seq, void *v)
2103{
2104 if (v == SEQ_START_TOKEN)
2105 seq_printf(seq, "%-127s\n",
2106 " sl local_address rem_address st tx_queue "
2107 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002108 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08002109 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08002110 struct udp_iter_state *state = seq->private;
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002111 int len;
David S. Millerdb8dac22008-03-06 16:22:02 -08002112
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002113 udp4_format_sock(v, seq, state->bucket, &len);
Eric Dumazetc482c562009-07-17 00:26:32 +00002114 seq_printf(seq, "%*s\n", 127 - len, "");
David S. Millerdb8dac22008-03-06 16:22:02 -08002115 }
2116 return 0;
2117}
2118
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002119static const struct file_operations udp_afinfo_seq_fops = {
2120 .owner = THIS_MODULE,
2121 .open = udp_seq_open,
2122 .read = seq_read,
2123 .llseek = seq_lseek,
2124 .release = seq_release_net
2125};
2126
David S. Millerdb8dac22008-03-06 16:22:02 -08002127/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08002128static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08002129 .name = "udp",
2130 .family = AF_INET,
Eric Dumazet645ca702008-10-29 01:41:45 -07002131 .udp_table = &udp_table,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002132 .seq_fops = &udp_afinfo_seq_fops,
Denis V. Lunevdda61922008-03-28 18:24:26 -07002133 .seq_ops = {
2134 .show = udp4_seq_show,
2135 },
David S. Millerdb8dac22008-03-06 16:22:02 -08002136};
2137
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002138static int __net_init udp4_proc_init_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002139{
2140 return udp_proc_register(net, &udp4_seq_afinfo);
2141}
2142
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002143static void __net_exit udp4_proc_exit_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002144{
2145 udp_proc_unregister(net, &udp4_seq_afinfo);
2146}
2147
2148static struct pernet_operations udp4_net_ops = {
2149 .init = udp4_proc_init_net,
2150 .exit = udp4_proc_exit_net,
2151};
2152
David S. Millerdb8dac22008-03-06 16:22:02 -08002153int __init udp4_proc_init(void)
2154{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002155 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002156}
2157
2158void udp4_proc_exit(void)
2159{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002160 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002161}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002162#endif /* CONFIG_PROC_FS */
2163
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002164static __initdata unsigned long uhash_entries;
2165static int __init set_uhash_entries(char *str)
Eric Dumazet645ca702008-10-29 01:41:45 -07002166{
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002167 if (!str)
2168 return 0;
2169 uhash_entries = simple_strtoul(str, &str, 0);
2170 if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2171 uhash_entries = UDP_HTABLE_SIZE_MIN;
2172 return 1;
2173}
2174__setup("uhash_entries=", set_uhash_entries);
Eric Dumazet645ca702008-10-29 01:41:45 -07002175
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002176void __init udp_table_init(struct udp_table *table, const char *name)
2177{
2178 unsigned int i;
2179
2180 if (!CONFIG_BASE_SMALL)
2181 table->hash = alloc_large_system_hash(name,
Eric Dumazet512615b2009-11-08 10:17:58 +00002182 2 * sizeof(struct udp_hslot),
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002183 uhash_entries,
2184 21, /* one slot per 2 MB */
2185 0,
2186 &table->log,
2187 &table->mask,
2188 64 * 1024);
2189 /*
2190 * Make sure hash table has the minimum size
2191 */
2192 if (CONFIG_BASE_SMALL || table->mask < UDP_HTABLE_SIZE_MIN - 1) {
2193 table->hash = kmalloc(UDP_HTABLE_SIZE_MIN *
Eric Dumazet512615b2009-11-08 10:17:58 +00002194 2 * sizeof(struct udp_hslot), GFP_KERNEL);
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002195 if (!table->hash)
2196 panic(name);
2197 table->log = ilog2(UDP_HTABLE_SIZE_MIN);
2198 table->mask = UDP_HTABLE_SIZE_MIN - 1;
2199 }
Eric Dumazet512615b2009-11-08 10:17:58 +00002200 table->hash2 = table->hash + (table->mask + 1);
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002201 for (i = 0; i <= table->mask; i++) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002202 INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00002203 table->hash[i].count = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -07002204 spin_lock_init(&table->hash[i].lock);
2205 }
Eric Dumazet512615b2009-11-08 10:17:58 +00002206 for (i = 0; i <= table->mask; i++) {
2207 INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i);
2208 table->hash2[i].count = 0;
2209 spin_lock_init(&table->hash2[i].lock);
2210 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002211}
2212
Hideo Aoki95766ff2007-12-31 00:29:24 -08002213void __init udp_init(void)
2214{
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002215 unsigned long limit;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002216
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002217 udp_table_init(&udp_table, "UDP");
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002218 limit = nr_free_buffer_pages() / 8;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002219 limit = max(limit, 128UL);
2220 sysctl_udp_mem[0] = limit / 4 * 3;
2221 sysctl_udp_mem[1] = limit;
2222 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2223
2224 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2225 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2226}
2227
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00002228int udp4_ufo_send_check(struct sk_buff *skb)
2229{
2230 const struct iphdr *iph;
2231 struct udphdr *uh;
2232
2233 if (!pskb_may_pull(skb, sizeof(*uh)))
2234 return -EINVAL;
2235
2236 iph = ip_hdr(skb);
2237 uh = udp_hdr(skb);
2238
2239 uh->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len,
2240 IPPROTO_UDP, 0);
2241 skb->csum_start = skb_transport_header(skb) - skb->head;
2242 skb->csum_offset = offsetof(struct udphdr, check);
2243 skb->ip_summed = CHECKSUM_PARTIAL;
2244 return 0;
2245}
2246
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002247struct sk_buff *udp4_ufo_fragment(struct sk_buff *skb,
2248 netdev_features_t features)
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00002249{
2250 struct sk_buff *segs = ERR_PTR(-EINVAL);
2251 unsigned int mss;
2252 int offset;
2253 __wsum csum;
2254
2255 mss = skb_shinfo(skb)->gso_size;
2256 if (unlikely(skb->len <= mss))
2257 goto out;
2258
2259 if (skb_gso_ok(skb, features | NETIF_F_GSO_ROBUST)) {
2260 /* Packet is from an untrusted source, reset gso_segs. */
2261 int type = skb_shinfo(skb)->gso_type;
2262
2263 if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY) ||
2264 !(type & (SKB_GSO_UDP))))
2265 goto out;
2266
2267 skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss);
2268
2269 segs = NULL;
2270 goto out;
2271 }
2272
2273 /* Do software UFO. Complete and fill in the UDP checksum as HW cannot
2274 * do checksum of UDP packets sent as multiple IP fragments.
2275 */
Michał Mirosław55508d62010-12-14 15:24:08 +00002276 offset = skb_checksum_start_offset(skb);
Eric Dumazetc482c562009-07-17 00:26:32 +00002277 csum = skb_checksum(skb, offset, skb->len - offset, 0);
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00002278 offset += skb->csum_offset;
2279 *(__sum16 *)(skb->data + offset) = csum_fold(csum);
2280 skb->ip_summed = CHECKSUM_NONE;
2281
2282 /* Fragment the skb. IP headers of the fragments are updated in
2283 * inet_gso_segment()
2284 */
2285 segs = skb_segment(skb, features);
2286out:
2287 return segs;
2288}
2289