blob: 0ca44df51ee94a2875427e5c9d3b1c77434f5f40 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * The User Datagram Protocol (UDP).
7 *
Jesper Juhl02c30a82005-05-05 16:16:16 -07008 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
Alan Cox113aa832008-10-13 19:01:08 -070011 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012 * Hirokazu Takahashi, <taka@valinux.co.jp>
13 *
14 * Fixes:
15 * Alan Cox : verify_area() calls
16 * Alan Cox : stopped close while in use off icmp
17 * messages. Not a fix but a botch that
18 * for udp at least is 'valid'.
19 * Alan Cox : Fixed icmp handling properly
20 * Alan Cox : Correct error for oversized datagrams
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090021 * Alan Cox : Tidied select() semantics.
22 * Alan Cox : udp_err() fixed properly, also now
Linus Torvalds1da177e2005-04-16 15:20:36 -070023 * select and read wake correctly on errors
24 * Alan Cox : udp_send verify_area moved to avoid mem leak
25 * Alan Cox : UDP can count its memory
26 * Alan Cox : send to an unknown connection causes
27 * an ECONNREFUSED off the icmp, but
28 * does NOT close.
29 * Alan Cox : Switched to new sk_buff handlers. No more backlog!
30 * Alan Cox : Using generic datagram code. Even smaller and the PEEK
31 * bug no longer crashes it.
32 * Fred Van Kempen : Net2e support for sk->broadcast.
33 * Alan Cox : Uses skb_free_datagram
34 * Alan Cox : Added get/set sockopt support.
35 * Alan Cox : Broadcasting without option set returns EACCES.
36 * Alan Cox : No wakeup calls. Instead we now use the callbacks.
37 * Alan Cox : Use ip_tos and ip_ttl
38 * Alan Cox : SNMP Mibs
39 * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support.
40 * Matt Dillon : UDP length checks.
41 * Alan Cox : Smarter af_inet used properly.
42 * Alan Cox : Use new kernel side addressing.
43 * Alan Cox : Incorrect return on truncated datagram receive.
44 * Arnt Gulbrandsen : New udp_send and stuff
45 * Alan Cox : Cache last socket
46 * Alan Cox : Route cache
47 * Jon Peatfield : Minor efficiency fix to sendto().
48 * Mike Shaver : RFC1122 checks.
49 * Alan Cox : Nonblocking error fix.
50 * Willy Konynenberg : Transparent proxying support.
51 * Mike McLagan : Routing by source
52 * David S. Miller : New socket lookup architecture.
53 * Last socket cache retained as it
54 * does have a high hit rate.
55 * Olaf Kirch : Don't linearise iovec on sendmsg.
56 * Andi Kleen : Some cleanups, cache destination entry
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090057 * for connect.
Linus Torvalds1da177e2005-04-16 15:20:36 -070058 * Vitaly E. Lavrov : Transparent proxy revived after year coma.
59 * Melvin Smith : Check msg_name not msg_namelen in sendto(),
60 * return ENOTCONN for unconnected sockets (POSIX)
61 * Janos Farkas : don't deliver multi/broadcasts to a different
62 * bound-to-device socket
63 * Hirokazu Takahashi : HW checksumming for outgoing UDP
64 * datagrams.
65 * Hirokazu Takahashi : sendfile() on UDP works now.
66 * Arnaldo C. Melo : convert /proc/net/udp to seq_file
67 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
68 * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind
69 * a single port at the same time.
70 * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support
James Chapman342f0232007-06-27 15:37:46 -070071 * James Chapman : Add L2TP encapsulation type.
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 *
73 *
74 * This program is free software; you can redistribute it and/or
75 * modify it under the terms of the GNU General Public License
76 * as published by the Free Software Foundation; either version
77 * 2 of the License, or (at your option) any later version.
78 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090079
Joe Perchesafd465032012-03-12 07:03:32 +000080#define pr_fmt(fmt) "UDP: " fmt
81
Linus Torvalds1da177e2005-04-16 15:20:36 -070082#include <asm/uaccess.h>
83#include <asm/ioctls.h>
Hideo Aoki95766ff2007-12-31 00:29:24 -080084#include <linux/bootmem.h>
Eric Dumazet8203efb2008-10-29 02:32:32 -070085#include <linux/highmem.h>
86#include <linux/swap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070087#include <linux/types.h>
88#include <linux/fcntl.h>
89#include <linux/module.h>
90#include <linux/socket.h>
91#include <linux/sockios.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020092#include <linux/igmp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070093#include <linux/in.h>
94#include <linux/errno.h>
95#include <linux/timer.h>
96#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070097#include <linux/inet.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include <linux/netdevice.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090099#include <linux/slab.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -0700100#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101#include <linux/skbuff.h>
102#include <linux/proc_fs.h>
103#include <linux/seq_file.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200104#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#include <net/icmp.h>
106#include <net/route.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#include <net/checksum.h>
108#include <net/xfrm.h>
Satoru Moriya296f7ea2011-06-17 11:58:39 +0000109#include <trace/events/udp.h>
Eric Dumazet447167b2012-04-11 23:05:28 +0000110#include <linux/static_key.h>
Eric Dumazet22911fc2012-06-27 00:23:44 +0000111#include <trace/events/skb.h>
Eliezer Tamir076bb0c2013-07-10 17:13:17 +0300112#include <net/busy_poll.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800113#include "udp_impl.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000115struct udp_table udp_table __read_mostly;
Eric Dumazet645ca702008-10-29 01:41:45 -0700116EXPORT_SYMBOL(udp_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
Eric Dumazet8d987e52010-11-09 23:24:26 +0000118long sysctl_udp_mem[3] __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800119EXPORT_SYMBOL(sysctl_udp_mem);
Eric Dumazetc482c562009-07-17 00:26:32 +0000120
121int sysctl_udp_rmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800122EXPORT_SYMBOL(sysctl_udp_rmem_min);
Eric Dumazetc482c562009-07-17 00:26:32 +0000123
124int sysctl_udp_wmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800125EXPORT_SYMBOL(sysctl_udp_wmem_min);
126
Eric Dumazet8d987e52010-11-09 23:24:26 +0000127atomic_long_t udp_memory_allocated;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800128EXPORT_SYMBOL(udp_memory_allocated);
129
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000130#define MAX_UDP_PORTS 65536
131#define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN)
Eric Dumazet98322f22009-01-26 21:35:35 -0800132
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700133static int udp_lib_lport_inuse(struct net *net, __u16 num,
Eric Dumazet645ca702008-10-29 01:41:45 -0700134 const struct udp_hslot *hslot,
Eric Dumazet98322f22009-01-26 21:35:35 -0800135 unsigned long *bitmap,
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700136 struct sock *sk,
137 int (*saddr_comp)(const struct sock *sk1,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000138 const struct sock *sk2),
139 unsigned int log)
Gerrit Renker25030a72006-08-26 20:06:05 -0700140{
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700141 struct sock *sk2;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800142 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000143 kuid_t uid = sock_i_uid(sk);
Gerrit Renker25030a72006-08-26 20:06:05 -0700144
Eric Dumazet88ab1932008-11-16 19:39:21 -0800145 sk_nulls_for_each(sk2, node, &hslot->head)
Joe Perches9d4fb272009-11-23 10:41:23 -0800146 if (net_eq(sock_net(sk2), net) &&
147 sk2 != sk &&
Eric Dumazetd4cada42009-11-08 10:17:30 +0000148 (bitmap || udp_sk(sk2)->udp_port_hash == num) &&
Joe Perches9d4fb272009-11-23 10:41:23 -0800149 (!sk2->sk_reuse || !sk->sk_reuse) &&
150 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
151 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000152 (!sk2->sk_reuseport || !sk->sk_reuseport ||
153 !uid_eq(uid, sock_i_uid(sk2))) &&
Eric Dumazet98322f22009-01-26 21:35:35 -0800154 (*saddr_comp)(sk, sk2)) {
155 if (bitmap)
Eric Dumazetd4cada42009-11-08 10:17:30 +0000156 __set_bit(udp_sk(sk2)->udp_port_hash >> log,
157 bitmap);
Eric Dumazet98322f22009-01-26 21:35:35 -0800158 else
159 return 1;
160 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700161 return 0;
162}
163
Eric Dumazet30fff922009-11-09 05:26:33 +0000164/*
165 * Note: we still hold spinlock of primary hash chain, so no other writer
166 * can insert/delete a socket with local_port == num
167 */
168static int udp_lib_lport_inuse2(struct net *net, __u16 num,
169 struct udp_hslot *hslot2,
170 struct sock *sk,
171 int (*saddr_comp)(const struct sock *sk1,
172 const struct sock *sk2))
173{
174 struct sock *sk2;
175 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000176 kuid_t uid = sock_i_uid(sk);
Eric Dumazet30fff922009-11-09 05:26:33 +0000177 int res = 0;
178
179 spin_lock(&hslot2->lock);
180 udp_portaddr_for_each_entry(sk2, node, &hslot2->head)
Joe Perches9d4fb272009-11-23 10:41:23 -0800181 if (net_eq(sock_net(sk2), net) &&
182 sk2 != sk &&
183 (udp_sk(sk2)->udp_port_hash == num) &&
184 (!sk2->sk_reuse || !sk->sk_reuse) &&
185 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
186 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000187 (!sk2->sk_reuseport || !sk->sk_reuseport ||
188 !uid_eq(uid, sock_i_uid(sk2))) &&
Eric Dumazet30fff922009-11-09 05:26:33 +0000189 (*saddr_comp)(sk, sk2)) {
190 res = 1;
191 break;
192 }
193 spin_unlock(&hslot2->lock);
194 return res;
195}
196
Gerrit Renker25030a72006-08-26 20:06:05 -0700197/**
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700198 * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
Gerrit Renker25030a72006-08-26 20:06:05 -0700199 *
200 * @sk: socket struct in question
201 * @snum: port number to look up
David S. Millerdf2bc452007-06-05 15:18:43 -0700202 * @saddr_comp: AF-dependent comparison of bound local IP addresses
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300203 * @hash2_nulladdr: AF-dependent hash value in secondary hash chains,
Eric Dumazet30fff922009-11-09 05:26:33 +0000204 * with NULL address
Gerrit Renker25030a72006-08-26 20:06:05 -0700205 */
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700206int udp_lib_get_port(struct sock *sk, unsigned short snum,
David S. Millerdf2bc452007-06-05 15:18:43 -0700207 int (*saddr_comp)(const struct sock *sk1,
Eric Dumazet30fff922009-11-09 05:26:33 +0000208 const struct sock *sk2),
209 unsigned int hash2_nulladdr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210{
Eric Dumazet512615b2009-11-08 10:17:58 +0000211 struct udp_hslot *hslot, *hslot2;
Eric Dumazet645ca702008-10-29 01:41:45 -0700212 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Gerrit Renker25030a72006-08-26 20:06:05 -0700213 int error = 1;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900214 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700216 if (!snum) {
Eric Dumazet9088c562008-10-08 11:44:17 -0700217 int low, high, remaining;
Eric Dumazet95c96172012-04-15 05:58:06 +0000218 unsigned int rand;
Eric Dumazet98322f22009-01-26 21:35:35 -0800219 unsigned short first, last;
220 DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700222 inet_get_local_port_range(&low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700223 remaining = (high - low) + 1;
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700224
Eric Dumazet9088c562008-10-08 11:44:17 -0700225 rand = net_random();
Eric Dumazet98322f22009-01-26 21:35:35 -0800226 first = (((u64)rand * remaining) >> 32) + low;
227 /*
228 * force rand to be an odd multiple of UDP_HTABLE_SIZE
229 */
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000230 rand = (rand | 1) * (udptable->mask + 1);
Eric Dumazet5781b232009-12-13 19:32:39 -0800231 last = first + udptable->mask + 1;
232 do {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000233 hslot = udp_hashslot(udptable, net, first);
Eric Dumazet98322f22009-01-26 21:35:35 -0800234 bitmap_zero(bitmap, PORTS_PER_CHAIN);
Eric Dumazet645ca702008-10-29 01:41:45 -0700235 spin_lock_bh(&hslot->lock);
Eric Dumazet98322f22009-01-26 21:35:35 -0800236 udp_lib_lport_inuse(net, snum, hslot, bitmap, sk,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000237 saddr_comp, udptable->log);
Eric Dumazet98322f22009-01-26 21:35:35 -0800238
239 snum = first;
240 /*
241 * Iterate on all possible values of snum for this hash.
242 * Using steps of an odd multiple of UDP_HTABLE_SIZE
243 * give us randomization and full range coverage.
244 */
Eric Dumazet9088c562008-10-08 11:44:17 -0700245 do {
Eric Dumazet98322f22009-01-26 21:35:35 -0800246 if (low <= snum && snum <= high &&
Amerigo Wange3826f12010-05-05 00:27:06 +0000247 !test_bit(snum >> udptable->log, bitmap) &&
248 !inet_is_reserved_local_port(snum))
Eric Dumazet98322f22009-01-26 21:35:35 -0800249 goto found;
250 snum += rand;
251 } while (snum != first);
252 spin_unlock_bh(&hslot->lock);
Eric Dumazet5781b232009-12-13 19:32:39 -0800253 } while (++first != last);
Eric Dumazet98322f22009-01-26 21:35:35 -0800254 goto fail;
Eric Dumazet645ca702008-10-29 01:41:45 -0700255 } else {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000256 hslot = udp_hashslot(udptable, net, snum);
Eric Dumazet645ca702008-10-29 01:41:45 -0700257 spin_lock_bh(&hslot->lock);
Eric Dumazet30fff922009-11-09 05:26:33 +0000258 if (hslot->count > 10) {
259 int exist;
260 unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum;
261
262 slot2 &= udptable->mask;
263 hash2_nulladdr &= udptable->mask;
264
265 hslot2 = udp_hashslot2(udptable, slot2);
266 if (hslot->count < hslot2->count)
267 goto scan_primary_hash;
268
269 exist = udp_lib_lport_inuse2(net, snum, hslot2,
270 sk, saddr_comp);
271 if (!exist && (hash2_nulladdr != slot2)) {
272 hslot2 = udp_hashslot2(udptable, hash2_nulladdr);
273 exist = udp_lib_lport_inuse2(net, snum, hslot2,
274 sk, saddr_comp);
275 }
276 if (exist)
277 goto fail_unlock;
278 else
279 goto found;
280 }
281scan_primary_hash:
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000282 if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk,
283 saddr_comp, 0))
Eric Dumazet645ca702008-10-29 01:41:45 -0700284 goto fail_unlock;
285 }
Eric Dumazet98322f22009-01-26 21:35:35 -0800286found:
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000287 inet_sk(sk)->inet_num = snum;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000288 udp_sk(sk)->udp_port_hash = snum;
289 udp_sk(sk)->udp_portaddr_hash ^= snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 if (sk_unhashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800291 sk_nulls_add_node_rcu(sk, &hslot->head);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +0000292 hslot->count++;
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700293 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet512615b2009-11-08 10:17:58 +0000294
295 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
296 spin_lock(&hslot2->lock);
297 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
298 &hslot2->head);
299 hslot2->count++;
300 spin_unlock(&hslot2->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700302 error = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -0700303fail_unlock:
304 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305fail:
Gerrit Renker25030a72006-08-26 20:06:05 -0700306 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307}
Eric Dumazetc482c562009-07-17 00:26:32 +0000308EXPORT_SYMBOL(udp_lib_get_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309
Vlad Yasevich499923c2009-04-09 17:37:33 +0000310static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
David S. Millerdb8dac22008-03-06 16:22:02 -0800311{
312 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
313
Eric Dumazetc482c562009-07-17 00:26:32 +0000314 return (!ipv6_only_sock(sk2) &&
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000315 (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr ||
316 inet1->inet_rcv_saddr == inet2->inet_rcv_saddr));
David S. Millerdb8dac22008-03-06 16:22:02 -0800317}
318
Eric Dumazetd4cada42009-11-08 10:17:30 +0000319static unsigned int udp4_portaddr_hash(struct net *net, __be32 saddr,
320 unsigned int port)
321{
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700322 return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000323}
324
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700325int udp_v4_get_port(struct sock *sk, unsigned short snum)
David S. Millerdb8dac22008-03-06 16:22:02 -0800326{
Eric Dumazet30fff922009-11-09 05:26:33 +0000327 unsigned int hash2_nulladdr =
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700328 udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum);
Eric Dumazet30fff922009-11-09 05:26:33 +0000329 unsigned int hash2_partial =
330 udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0);
331
Eric Dumazetd4cada42009-11-08 10:17:30 +0000332 /* precompute partial secondary hash */
Eric Dumazet30fff922009-11-09 05:26:33 +0000333 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
334 return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr);
David S. Millerdb8dac22008-03-06 16:22:02 -0800335}
336
Eric Dumazet645ca702008-10-29 01:41:45 -0700337static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr,
338 unsigned short hnum,
339 __be16 sport, __be32 daddr, __be16 dport, int dif)
340{
341 int score = -1;
342
Eric Dumazetd4cada42009-11-08 10:17:30 +0000343 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
Eric Dumazet645ca702008-10-29 01:41:45 -0700344 !ipv6_only_sock(sk)) {
345 struct inet_sock *inet = inet_sk(sk);
346
Tom Herbertba418fa2013-01-22 09:50:32 +0000347 score = (sk->sk_family == PF_INET ? 2 : 1);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000348 if (inet->inet_rcv_saddr) {
349 if (inet->inet_rcv_saddr != daddr)
Eric Dumazet645ca702008-10-29 01:41:45 -0700350 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000351 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700352 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000353 if (inet->inet_daddr) {
354 if (inet->inet_daddr != saddr)
Eric Dumazet645ca702008-10-29 01:41:45 -0700355 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000356 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700357 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000358 if (inet->inet_dport) {
359 if (inet->inet_dport != sport)
Eric Dumazet645ca702008-10-29 01:41:45 -0700360 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000361 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700362 }
363 if (sk->sk_bound_dev_if) {
364 if (sk->sk_bound_dev_if != dif)
365 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000366 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700367 }
368 }
369 return score;
370}
371
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000372/*
373 * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num)
374 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000375static inline int compute_score2(struct sock *sk, struct net *net,
376 __be32 saddr, __be16 sport,
377 __be32 daddr, unsigned int hnum, int dif)
378{
379 int score = -1;
380
381 if (net_eq(sock_net(sk), net) && !ipv6_only_sock(sk)) {
382 struct inet_sock *inet = inet_sk(sk);
383
384 if (inet->inet_rcv_saddr != daddr)
385 return -1;
386 if (inet->inet_num != hnum)
387 return -1;
388
Tom Herbertba418fa2013-01-22 09:50:32 +0000389 score = (sk->sk_family == PF_INET ? 2 : 1);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000390 if (inet->inet_daddr) {
391 if (inet->inet_daddr != saddr)
392 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000393 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000394 }
395 if (inet->inet_dport) {
396 if (inet->inet_dport != sport)
397 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000398 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000399 }
400 if (sk->sk_bound_dev_if) {
401 if (sk->sk_bound_dev_if != dif)
402 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000403 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000404 }
405 }
406 return score;
407}
408
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000409
410/* called with read_rcu_lock() */
411static struct sock *udp4_lib_lookup2(struct net *net,
412 __be32 saddr, __be16 sport,
413 __be32 daddr, unsigned int hnum, int dif,
414 struct udp_hslot *hslot2, unsigned int slot2)
415{
416 struct sock *sk, *result;
417 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000418 int score, badness, matches = 0, reuseport = 0;
419 u32 hash = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000420
421begin:
422 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000423 badness = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000424 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
425 score = compute_score2(sk, net, saddr, sport,
426 daddr, hnum, dif);
427 if (score > badness) {
428 result = sk;
429 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000430 reuseport = sk->sk_reuseport;
431 if (reuseport) {
432 hash = inet_ehashfn(net, daddr, hnum,
Eric Dumazet7c0cadc2013-06-12 14:31:39 -0700433 saddr, sport);
Tom Herbertba418fa2013-01-22 09:50:32 +0000434 matches = 1;
435 }
436 } else if (score == badness && reuseport) {
437 matches++;
438 if (((u64)hash * matches) >> 32 == 0)
439 result = sk;
440 hash = next_pseudo_random32(hash);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000441 }
442 }
443 /*
444 * if the nulls value we got at the end of this lookup is
445 * not the expected one, we must restart lookup.
446 * We probably met an item that was moved to another chain.
447 */
448 if (get_nulls_value(node) != slot2)
449 goto begin;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000450 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000451 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000452 result = NULL;
453 else if (unlikely(compute_score2(result, net, saddr, sport,
454 daddr, hnum, dif) < badness)) {
455 sock_put(result);
456 goto begin;
457 }
458 }
459 return result;
460}
461
David S. Millerdb8dac22008-03-06 16:22:02 -0800462/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
463 * harder than this. -DaveM
464 */
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000465struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -0800466 __be16 sport, __be32 daddr, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700467 int dif, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800468{
Eric Dumazet271b72c2008-10-29 02:11:14 -0700469 struct sock *sk, *result;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800470 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800471 unsigned short hnum = ntohs(dport);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000472 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
473 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
Tom Herbertba418fa2013-01-22 09:50:32 +0000474 int score, badness, matches = 0, reuseport = 0;
475 u32 hash = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800476
Eric Dumazet271b72c2008-10-29 02:11:14 -0700477 rcu_read_lock();
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000478 if (hslot->count > 10) {
479 hash2 = udp4_portaddr_hash(net, daddr, hnum);
480 slot2 = hash2 & udptable->mask;
481 hslot2 = &udptable->hash2[slot2];
482 if (hslot->count < hslot2->count)
483 goto begin;
484
485 result = udp4_lib_lookup2(net, saddr, sport,
486 daddr, hnum, dif,
487 hslot2, slot2);
488 if (!result) {
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700489 hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000490 slot2 = hash2 & udptable->mask;
491 hslot2 = &udptable->hash2[slot2];
492 if (hslot->count < hslot2->count)
493 goto begin;
494
Jorge Boncompte [DTI2]1223c672010-04-08 04:56:48 +0000495 result = udp4_lib_lookup2(net, saddr, sport,
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700496 htonl(INADDR_ANY), hnum, dif,
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000497 hslot2, slot2);
498 }
499 rcu_read_unlock();
500 return result;
501 }
Eric Dumazet271b72c2008-10-29 02:11:14 -0700502begin:
503 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000504 badness = 0;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800505 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
Eric Dumazet645ca702008-10-29 01:41:45 -0700506 score = compute_score(sk, net, saddr, hnum, sport,
507 daddr, dport, dif);
508 if (score > badness) {
509 result = sk;
510 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000511 reuseport = sk->sk_reuseport;
512 if (reuseport) {
513 hash = inet_ehashfn(net, daddr, hnum,
Eric Dumazet7c0cadc2013-06-12 14:31:39 -0700514 saddr, sport);
Tom Herbertba418fa2013-01-22 09:50:32 +0000515 matches = 1;
516 }
517 } else if (score == badness && reuseport) {
518 matches++;
519 if (((u64)hash * matches) >> 32 == 0)
520 result = sk;
521 hash = next_pseudo_random32(hash);
David S. Millerdb8dac22008-03-06 16:22:02 -0800522 }
523 }
Eric Dumazet88ab1932008-11-16 19:39:21 -0800524 /*
525 * if the nulls value we got at the end of this lookup is
526 * not the expected one, we must restart lookup.
527 * We probably met an item that was moved to another chain.
528 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000529 if (get_nulls_value(node) != slot)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800530 goto begin;
531
Eric Dumazet271b72c2008-10-29 02:11:14 -0700532 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000533 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet271b72c2008-10-29 02:11:14 -0700534 result = NULL;
535 else if (unlikely(compute_score(result, net, saddr, hnum, sport,
536 daddr, dport, dif) < badness)) {
537 sock_put(result);
538 goto begin;
539 }
540 }
541 rcu_read_unlock();
David S. Millerdb8dac22008-03-06 16:22:02 -0800542 return result;
543}
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000544EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
David S. Millerdb8dac22008-03-06 16:22:02 -0800545
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700546static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
547 __be16 sport, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700548 struct udp_table *udptable)
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700549{
KOVACS Krisztian23542612008-10-07 12:41:01 -0700550 struct sock *sk;
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700551 const struct iphdr *iph = ip_hdr(skb);
552
KOVACS Krisztian23542612008-10-07 12:41:01 -0700553 if (unlikely(sk = skb_steal_sock(skb)))
554 return sk;
555 else
Eric Dumazetadf30902009-06-02 05:19:30 +0000556 return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport,
KOVACS Krisztian23542612008-10-07 12:41:01 -0700557 iph->daddr, dport, inet_iif(skb),
558 udptable);
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700559}
560
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700561struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
562 __be32 daddr, __be16 dport, int dif)
563{
Eric Dumazet645ca702008-10-29 01:41:45 -0700564 return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700565}
566EXPORT_SYMBOL_GPL(udp4_lib_lookup);
567
Eric Dumazet920a4612008-11-01 21:22:23 -0700568static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk,
David S. Millerdb8dac22008-03-06 16:22:02 -0800569 __be16 loc_port, __be32 loc_addr,
570 __be16 rmt_port, __be32 rmt_addr,
571 int dif)
572{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800573 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800574 struct sock *s = sk;
575 unsigned short hnum = ntohs(loc_port);
576
Eric Dumazet88ab1932008-11-16 19:39:21 -0800577 sk_nulls_for_each_from(s, node) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800578 struct inet_sock *inet = inet_sk(s);
579
Joe Perches9d4fb272009-11-23 10:41:23 -0800580 if (!net_eq(sock_net(s), net) ||
581 udp_sk(s)->udp_port_hash != hnum ||
582 (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
583 (inet->inet_dport != rmt_port && inet->inet_dport) ||
584 (inet->inet_rcv_saddr &&
585 inet->inet_rcv_saddr != loc_addr) ||
586 ipv6_only_sock(s) ||
David S. Millerdb8dac22008-03-06 16:22:02 -0800587 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
588 continue;
589 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
590 continue;
591 goto found;
592 }
593 s = NULL;
594found:
595 return s;
596}
597
598/*
599 * This routine is called by the ICMP module when it gets some
600 * sort of error condition. If err < 0 then the socket should
601 * be closed and the error returned to the user. If err > 0
602 * it's just the icmp type << 8 | icmp code.
603 * Header points to the ip header of the error packet. We move
604 * on past this. Then (as it used to claim before adjustment)
605 * header points to the first 8 bytes of the udp header. We need
606 * to find the appropriate port.
607 */
608
Eric Dumazet645ca702008-10-29 01:41:45 -0700609void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800610{
611 struct inet_sock *inet;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000612 const struct iphdr *iph = (const struct iphdr *)skb->data;
Eric Dumazetc482c562009-07-17 00:26:32 +0000613 struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
David S. Millerdb8dac22008-03-06 16:22:02 -0800614 const int type = icmp_hdr(skb)->type;
615 const int code = icmp_hdr(skb)->code;
616 struct sock *sk;
617 int harderr;
618 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700619 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800620
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700621 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
David S. Millerdb8dac22008-03-06 16:22:02 -0800622 iph->saddr, uh->source, skb->dev->ifindex, udptable);
623 if (sk == NULL) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700624 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800625 return; /* No socket for error */
626 }
627
628 err = 0;
629 harderr = 0;
630 inet = inet_sk(sk);
631
632 switch (type) {
633 default:
634 case ICMP_TIME_EXCEEDED:
635 err = EHOSTUNREACH;
636 break;
637 case ICMP_SOURCE_QUENCH:
638 goto out;
639 case ICMP_PARAMETERPROB:
640 err = EPROTO;
641 harderr = 1;
642 break;
643 case ICMP_DEST_UNREACH:
644 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
David S. Miller36393392012-06-14 22:21:46 -0700645 ipv4_sk_update_pmtu(skb, sk, info);
David S. Millerdb8dac22008-03-06 16:22:02 -0800646 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
647 err = EMSGSIZE;
648 harderr = 1;
649 break;
650 }
651 goto out;
652 }
653 err = EHOSTUNREACH;
654 if (code <= NR_ICMP_UNREACH) {
655 harderr = icmp_err_convert[code].fatal;
656 err = icmp_err_convert[code].errno;
657 }
658 break;
David S. Miller55be7a92012-07-11 21:27:49 -0700659 case ICMP_REDIRECT:
660 ipv4_sk_redirect(skb, sk);
Duan Jiong1a462d12013-09-20 18:20:28 +0800661 goto out;
David S. Millerdb8dac22008-03-06 16:22:02 -0800662 }
663
664 /*
665 * RFC1122: OK. Passes ICMP errors back to application, as per
666 * 4.1.3.3.
667 */
668 if (!inet->recverr) {
669 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
670 goto out;
Eric Dumazetb1faf562010-05-31 23:44:05 -0700671 } else
Eric Dumazetc482c562009-07-17 00:26:32 +0000672 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
Eric Dumazetb1faf562010-05-31 23:44:05 -0700673
David S. Millerdb8dac22008-03-06 16:22:02 -0800674 sk->sk_err = err;
675 sk->sk_error_report(sk);
676out:
677 sock_put(sk);
678}
679
680void udp_err(struct sk_buff *skb, u32 info)
681{
Eric Dumazet645ca702008-10-29 01:41:45 -0700682 __udp4_lib_err(skb, info, &udp_table);
David S. Millerdb8dac22008-03-06 16:22:02 -0800683}
684
685/*
686 * Throw away all pending data and cancel the corking. Socket is locked.
687 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400688void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800689{
690 struct udp_sock *up = udp_sk(sk);
691
692 if (up->pending) {
693 up->len = 0;
694 up->pending = 0;
695 ip_flush_pending_frames(sk);
696 }
697}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400698EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800699
700/**
Herbert Xuf6b96642011-03-01 02:36:48 +0000701 * udp4_hwcsum - handle outgoing HW checksumming
David S. Millerdb8dac22008-03-06 16:22:02 -0800702 * @skb: sk_buff containing the filled-in UDP header
703 * (checksum field must be zeroed out)
Herbert Xuf6b96642011-03-01 02:36:48 +0000704 * @src: source IP address
705 * @dst: destination IP address
David S. Millerdb8dac22008-03-06 16:22:02 -0800706 */
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200707void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
David S. Millerdb8dac22008-03-06 16:22:02 -0800708{
David S. Millerdb8dac22008-03-06 16:22:02 -0800709 struct udphdr *uh = udp_hdr(skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000710 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
711 int offset = skb_transport_offset(skb);
712 int len = skb->len - offset;
713 int hlen = len;
David S. Millerdb8dac22008-03-06 16:22:02 -0800714 __wsum csum = 0;
715
Herbert Xuf6b96642011-03-01 02:36:48 +0000716 if (!frags) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800717 /*
718 * Only one fragment on the socket.
719 */
720 skb->csum_start = skb_transport_header(skb) - skb->head;
721 skb->csum_offset = offsetof(struct udphdr, check);
Herbert Xuf6b96642011-03-01 02:36:48 +0000722 uh->check = ~csum_tcpudp_magic(src, dst, len,
723 IPPROTO_UDP, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800724 } else {
725 /*
726 * HW-checksum won't work as there are two or more
727 * fragments on the socket so that all csums of sk_buffs
728 * should be together
729 */
Herbert Xuf6b96642011-03-01 02:36:48 +0000730 do {
731 csum = csum_add(csum, frags->csum);
732 hlen -= frags->len;
733 } while ((frags = frags->next));
David S. Millerdb8dac22008-03-06 16:22:02 -0800734
Herbert Xuf6b96642011-03-01 02:36:48 +0000735 csum = skb_checksum(skb, offset, hlen, csum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800736 skb->ip_summed = CHECKSUM_NONE;
737
David S. Millerdb8dac22008-03-06 16:22:02 -0800738 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
739 if (uh->check == 0)
740 uh->check = CSUM_MANGLED_0;
741 }
742}
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200743EXPORT_SYMBOL_GPL(udp4_hwcsum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800744
David S. Miller79ab0532011-05-09 13:31:04 -0700745static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
Herbert Xuf6b96642011-03-01 02:36:48 +0000746{
747 struct sock *sk = skb->sk;
748 struct inet_sock *inet = inet_sk(sk);
749 struct udphdr *uh;
Herbert Xuf6b96642011-03-01 02:36:48 +0000750 int err = 0;
751 int is_udplite = IS_UDPLITE(sk);
752 int offset = skb_transport_offset(skb);
753 int len = skb->len - offset;
754 __wsum csum = 0;
755
756 /*
757 * Create a UDP header
758 */
759 uh = udp_hdr(skb);
760 uh->source = inet->inet_sport;
David S. Miller79ab0532011-05-09 13:31:04 -0700761 uh->dest = fl4->fl4_dport;
Herbert Xuf6b96642011-03-01 02:36:48 +0000762 uh->len = htons(len);
763 uh->check = 0;
764
765 if (is_udplite) /* UDP-Lite */
766 csum = udplite_csum(skb);
767
768 else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */
769
770 skb->ip_summed = CHECKSUM_NONE;
771 goto send;
772
773 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
774
David S. Miller79ab0532011-05-09 13:31:04 -0700775 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
Herbert Xuf6b96642011-03-01 02:36:48 +0000776 goto send;
777
778 } else
779 csum = udp_csum(skb);
780
781 /* add protocol-dependent pseudo-header */
David S. Miller79ab0532011-05-09 13:31:04 -0700782 uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
Herbert Xuf6b96642011-03-01 02:36:48 +0000783 sk->sk_protocol, csum);
784 if (uh->check == 0)
785 uh->check = CSUM_MANGLED_0;
786
787send:
Eric Dumazetb5ec8ee2012-08-10 02:22:47 +0000788 err = ip_send_skb(sock_net(sk), skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000789 if (err) {
790 if (err == -ENOBUFS && !inet->recverr) {
791 UDP_INC_STATS_USER(sock_net(sk),
792 UDP_MIB_SNDBUFERRORS, is_udplite);
793 err = 0;
794 }
795 } else
796 UDP_INC_STATS_USER(sock_net(sk),
797 UDP_MIB_OUTDATAGRAMS, is_udplite);
798 return err;
799}
800
David S. Millerdb8dac22008-03-06 16:22:02 -0800801/*
802 * Push out all pending data as one UDP datagram. Socket is locked.
803 */
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200804int udp_push_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800805{
806 struct udp_sock *up = udp_sk(sk);
807 struct inet_sock *inet = inet_sk(sk);
David S. Millerb6f21b22011-03-12 02:09:18 -0500808 struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800809 struct sk_buff *skb;
David S. Millerdb8dac22008-03-06 16:22:02 -0800810 int err = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800811
David S. Miller77968b72011-05-08 17:12:19 -0700812 skb = ip_finish_skb(sk, fl4);
Herbert Xuf6b96642011-03-01 02:36:48 +0000813 if (!skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800814 goto out;
815
David S. Miller79ab0532011-05-09 13:31:04 -0700816 err = udp_send_skb(skb, fl4);
David S. Millerdb8dac22008-03-06 16:22:02 -0800817
David S. Millerdb8dac22008-03-06 16:22:02 -0800818out:
819 up->len = 0;
820 up->pending = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800821 return err;
822}
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200823EXPORT_SYMBOL(udp_push_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800824
825int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
826 size_t len)
827{
828 struct inet_sock *inet = inet_sk(sk);
829 struct udp_sock *up = udp_sk(sk);
David S. Millere4749952011-05-08 16:38:45 -0700830 struct flowi4 fl4_stack;
David S. Millerb6f21b22011-03-12 02:09:18 -0500831 struct flowi4 *fl4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800832 int ulen = len;
833 struct ipcm_cookie ipc;
834 struct rtable *rt = NULL;
835 int free = 0;
836 int connected = 0;
837 __be32 daddr, faddr, saddr;
838 __be16 dport;
839 u8 tos;
840 int err, is_udplite = IS_UDPLITE(sk);
841 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
842 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
Herbert Xu903ab862011-03-01 02:36:48 +0000843 struct sk_buff *skb;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000844 struct ip_options_data opt_copy;
David S. Millerdb8dac22008-03-06 16:22:02 -0800845
846 if (len > 0xFFFF)
847 return -EMSGSIZE;
848
849 /*
850 * Check the flags.
851 */
852
Eric Dumazetc482c562009-07-17 00:26:32 +0000853 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
David S. Millerdb8dac22008-03-06 16:22:02 -0800854 return -EOPNOTSUPP;
855
856 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000857 ipc.tx_flags = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800858
Herbert Xu903ab862011-03-01 02:36:48 +0000859 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
860
David S. Millerf5fca602011-05-08 17:24:10 -0700861 fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800862 if (up->pending) {
863 /*
864 * There are pending frames.
865 * The socket lock must be held while it's corked.
866 */
867 lock_sock(sk);
868 if (likely(up->pending)) {
869 if (unlikely(up->pending != AF_INET)) {
870 release_sock(sk);
871 return -EINVAL;
872 }
873 goto do_append_data;
874 }
875 release_sock(sk);
876 }
877 ulen += sizeof(struct udphdr);
878
879 /*
880 * Get and verify the address.
881 */
882 if (msg->msg_name) {
Daniel Baluta5e73ea12012-04-15 01:34:41 +0000883 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
David S. Millerdb8dac22008-03-06 16:22:02 -0800884 if (msg->msg_namelen < sizeof(*usin))
885 return -EINVAL;
886 if (usin->sin_family != AF_INET) {
887 if (usin->sin_family != AF_UNSPEC)
888 return -EAFNOSUPPORT;
889 }
890
891 daddr = usin->sin_addr.s_addr;
892 dport = usin->sin_port;
893 if (dport == 0)
894 return -EINVAL;
895 } else {
896 if (sk->sk_state != TCP_ESTABLISHED)
897 return -EDESTADDRREQ;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000898 daddr = inet->inet_daddr;
899 dport = inet->inet_dport;
David S. Millerdb8dac22008-03-06 16:22:02 -0800900 /* Open fast path for connected socket.
901 Route will not be used, if at least one option is set.
902 */
903 connected = 1;
904 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000905 ipc.addr = inet->inet_saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800906
907 ipc.oif = sk->sk_bound_dev_if;
Daniel Borkmannbf84a012013-04-14 08:08:13 +0000908
909 sock_tx_timestamp(sk, &ipc.tx_flags);
910
David S. Millerdb8dac22008-03-06 16:22:02 -0800911 if (msg->msg_controllen) {
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900912 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
David S. Millerdb8dac22008-03-06 16:22:02 -0800913 if (err)
914 return err;
915 if (ipc.opt)
916 free = 1;
917 connected = 0;
918 }
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000919 if (!ipc.opt) {
920 struct ip_options_rcu *inet_opt;
921
922 rcu_read_lock();
923 inet_opt = rcu_dereference(inet->inet_opt);
924 if (inet_opt) {
925 memcpy(&opt_copy, inet_opt,
926 sizeof(*inet_opt) + inet_opt->opt.optlen);
927 ipc.opt = &opt_copy.opt;
928 }
929 rcu_read_unlock();
930 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800931
932 saddr = ipc.addr;
933 ipc.addr = faddr = daddr;
934
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000935 if (ipc.opt && ipc.opt->opt.srr) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800936 if (!daddr)
937 return -EINVAL;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000938 faddr = ipc.opt->opt.faddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800939 connected = 0;
940 }
941 tos = RT_TOS(inet->tos);
942 if (sock_flag(sk, SOCK_LOCALROUTE) ||
943 (msg->msg_flags & MSG_DONTROUTE) ||
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000944 (ipc.opt && ipc.opt->opt.is_strictroute)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800945 tos |= RTO_ONLINK;
946 connected = 0;
947 }
948
949 if (ipv4_is_multicast(daddr)) {
950 if (!ipc.oif)
951 ipc.oif = inet->mc_index;
952 if (!saddr)
953 saddr = inet->mc_addr;
954 connected = 0;
Erich E. Hoover76e21052012-02-08 09:11:07 +0000955 } else if (!ipc.oif)
956 ipc.oif = inet->uc_index;
David S. Millerdb8dac22008-03-06 16:22:02 -0800957
958 if (connected)
Eric Dumazetc482c562009-07-17 00:26:32 +0000959 rt = (struct rtable *)sk_dst_check(sk, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800960
961 if (rt == NULL) {
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -0700962 struct net *net = sock_net(sk);
963
David S. Millere4749952011-05-08 16:38:45 -0700964 fl4 = &fl4_stack;
965 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
David S. Millerc0951cb2011-03-31 04:54:27 -0700966 RT_SCOPE_UNIVERSE, sk->sk_protocol,
967 inet_sk_flowi_flags(sk)|FLOWI_FLAG_CAN_SLEEP,
968 faddr, saddr, dport, inet->inet_sport);
969
David S. Millere4749952011-05-08 16:38:45 -0700970 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
971 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800972 if (IS_ERR(rt)) {
973 err = PTR_ERR(rt);
David S. Miller06dc94b2011-03-03 10:38:01 -0800974 rt = NULL;
David S. Millerdb8dac22008-03-06 16:22:02 -0800975 if (err == -ENETUNREACH)
Pavel Emelyanov7c73a6f2008-07-16 20:20:11 -0700976 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -0800977 goto out;
978 }
979
980 err = -EACCES;
981 if ((rt->rt_flags & RTCF_BROADCAST) &&
982 !sock_flag(sk, SOCK_BROADCAST))
983 goto out;
984 if (connected)
Changli Gaod8d1f302010-06-10 23:31:35 -0700985 sk_dst_set(sk, dst_clone(&rt->dst));
David S. Millerdb8dac22008-03-06 16:22:02 -0800986 }
987
988 if (msg->msg_flags&MSG_CONFIRM)
989 goto do_confirm;
990back_from_confirm:
991
David S. Millere4749952011-05-08 16:38:45 -0700992 saddr = fl4->saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800993 if (!ipc.addr)
David S. Millere4749952011-05-08 16:38:45 -0700994 daddr = ipc.addr = fl4->daddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800995
Herbert Xu903ab862011-03-01 02:36:48 +0000996 /* Lockless fast path for the non-corking case. */
997 if (!corkreq) {
David S. Miller77968b72011-05-08 17:12:19 -0700998 skb = ip_make_skb(sk, fl4, getfrag, msg->msg_iov, ulen,
Herbert Xu903ab862011-03-01 02:36:48 +0000999 sizeof(struct udphdr), &ipc, &rt,
1000 msg->msg_flags);
1001 err = PTR_ERR(skb);
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001002 if (!IS_ERR_OR_NULL(skb))
David S. Miller79ab0532011-05-09 13:31:04 -07001003 err = udp_send_skb(skb, fl4);
Herbert Xu903ab862011-03-01 02:36:48 +00001004 goto out;
1005 }
1006
David S. Millerdb8dac22008-03-06 16:22:02 -08001007 lock_sock(sk);
1008 if (unlikely(up->pending)) {
1009 /* The socket is already corked while preparing it. */
1010 /* ... which is an evident application bug. --ANK */
1011 release_sock(sk);
1012
Joe Perchesafd465032012-03-12 07:03:32 +00001013 LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("cork app bug 2\n"));
David S. Millerdb8dac22008-03-06 16:22:02 -08001014 err = -EINVAL;
1015 goto out;
1016 }
1017 /*
1018 * Now cork the socket to pend data.
1019 */
David S. Millerb6f21b22011-03-12 02:09:18 -05001020 fl4 = &inet->cork.fl.u.ip4;
1021 fl4->daddr = daddr;
1022 fl4->saddr = saddr;
David S. Miller9cce96d2011-03-12 03:00:33 -05001023 fl4->fl4_dport = dport;
1024 fl4->fl4_sport = inet->inet_sport;
David S. Millerdb8dac22008-03-06 16:22:02 -08001025 up->pending = AF_INET;
1026
1027do_append_data:
1028 up->len += ulen;
David S. Millerf5fca602011-05-08 17:24:10 -07001029 err = ip_append_data(sk, fl4, getfrag, msg->msg_iov, ulen,
1030 sizeof(struct udphdr), &ipc, &rt,
1031 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001032 if (err)
1033 udp_flush_pending_frames(sk);
1034 else if (!corkreq)
1035 err = udp_push_pending_frames(sk);
1036 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1037 up->pending = 0;
1038 release_sock(sk);
1039
1040out:
1041 ip_rt_put(rt);
1042 if (free)
1043 kfree(ipc.opt);
1044 if (!err)
1045 return len;
1046 /*
1047 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1048 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1049 * we don't have a good statistic (IpOutDiscards but it can be too many
1050 * things). We could add another new stat but at least for now that
1051 * seems like overkill.
1052 */
1053 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001054 UDP_INC_STATS_USER(sock_net(sk),
1055 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001056 }
1057 return err;
1058
1059do_confirm:
Changli Gaod8d1f302010-06-10 23:31:35 -07001060 dst_confirm(&rt->dst);
David S. Millerdb8dac22008-03-06 16:22:02 -08001061 if (!(msg->msg_flags&MSG_PROBE) || len)
1062 goto back_from_confirm;
1063 err = 0;
1064 goto out;
1065}
Eric Dumazetc482c562009-07-17 00:26:32 +00001066EXPORT_SYMBOL(udp_sendmsg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001067
1068int udp_sendpage(struct sock *sk, struct page *page, int offset,
1069 size_t size, int flags)
1070{
David S. Millerf5fca602011-05-08 17:24:10 -07001071 struct inet_sock *inet = inet_sk(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001072 struct udp_sock *up = udp_sk(sk);
1073 int ret;
1074
1075 if (!up->pending) {
1076 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
1077
1078 /* Call udp_sendmsg to specify destination address which
1079 * sendpage interface can't pass.
1080 * This will succeed only when the socket is connected.
1081 */
1082 ret = udp_sendmsg(NULL, sk, &msg, 0);
1083 if (ret < 0)
1084 return ret;
1085 }
1086
1087 lock_sock(sk);
1088
1089 if (unlikely(!up->pending)) {
1090 release_sock(sk);
1091
Joe Perchesafd465032012-03-12 07:03:32 +00001092 LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("udp cork app bug 3\n"));
David S. Millerdb8dac22008-03-06 16:22:02 -08001093 return -EINVAL;
1094 }
1095
David S. Millerf5fca602011-05-08 17:24:10 -07001096 ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1097 page, offset, size, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001098 if (ret == -EOPNOTSUPP) {
1099 release_sock(sk);
1100 return sock_no_sendpage(sk->sk_socket, page, offset,
1101 size, flags);
1102 }
1103 if (ret < 0) {
1104 udp_flush_pending_frames(sk);
1105 goto out;
1106 }
1107
1108 up->len += size;
1109 if (!(up->corkflag || (flags&MSG_MORE)))
1110 ret = udp_push_pending_frames(sk);
1111 if (!ret)
1112 ret = size;
1113out:
1114 release_sock(sk);
1115 return ret;
1116}
1117
Eric Dumazet85584672009-10-09 04:43:40 +00001118
1119/**
1120 * first_packet_length - return length of first packet in receive queue
1121 * @sk: socket
1122 *
1123 * Drops all bad checksum frames, until a valid one is found.
1124 * Returns the length of found skb, or 0 if none is found.
1125 */
1126static unsigned int first_packet_length(struct sock *sk)
1127{
1128 struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1129 struct sk_buff *skb;
1130 unsigned int res;
1131
1132 __skb_queue_head_init(&list_kill);
1133
1134 spin_lock_bh(&rcvq->lock);
1135 while ((skb = skb_peek(rcvq)) != NULL &&
1136 udp_lib_checksum_complete(skb)) {
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001137 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS,
1138 IS_UDPLITE(sk));
Eric Dumazet85584672009-10-09 04:43:40 +00001139 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1140 IS_UDPLITE(sk));
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001141 atomic_inc(&sk->sk_drops);
Eric Dumazet85584672009-10-09 04:43:40 +00001142 __skb_unlink(skb, rcvq);
1143 __skb_queue_tail(&list_kill, skb);
1144 }
1145 res = skb ? skb->len : 0;
1146 spin_unlock_bh(&rcvq->lock);
1147
1148 if (!skb_queue_empty(&list_kill)) {
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001149 bool slow = lock_sock_fast(sk);
1150
Eric Dumazet85584672009-10-09 04:43:40 +00001151 __skb_queue_purge(&list_kill);
1152 sk_mem_reclaim_partial(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001153 unlock_sock_fast(sk, slow);
Eric Dumazet85584672009-10-09 04:43:40 +00001154 }
1155 return res;
1156}
1157
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158/*
1159 * IOCTL requests applicable to the UDP protocol
1160 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001161
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1163{
Stephen Hemminger6516c652007-03-08 20:41:55 -08001164 switch (cmd) {
1165 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 {
Eric Dumazet31e6d362009-06-17 19:05:41 -07001167 int amount = sk_wmem_alloc_get(sk);
1168
Stephen Hemminger6516c652007-03-08 20:41:55 -08001169 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170 }
Stephen Hemminger6516c652007-03-08 20:41:55 -08001171
1172 case SIOCINQ:
1173 {
Eric Dumazet85584672009-10-09 04:43:40 +00001174 unsigned int amount = first_packet_length(sk);
Stephen Hemminger6516c652007-03-08 20:41:55 -08001175
Eric Dumazet85584672009-10-09 04:43:40 +00001176 if (amount)
Stephen Hemminger6516c652007-03-08 20:41:55 -08001177 /*
1178 * We will only return the amount
1179 * of this packet since that is all
1180 * that will be read.
1181 */
Eric Dumazet85584672009-10-09 04:43:40 +00001182 amount -= sizeof(struct udphdr);
1183
Stephen Hemminger6516c652007-03-08 20:41:55 -08001184 return put_user(amount, (int __user *)arg);
1185 }
1186
1187 default:
1188 return -ENOIOCTLCMD;
1189 }
1190
1191 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192}
Eric Dumazetc482c562009-07-17 00:26:32 +00001193EXPORT_SYMBOL(udp_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194
David S. Millerdb8dac22008-03-06 16:22:02 -08001195/*
1196 * This should be easy, if there is something there we
1197 * return it, otherwise we block.
1198 */
1199
1200int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
1201 size_t len, int noblock, int flags, int *addr_len)
1202{
1203 struct inet_sock *inet = inet_sk(sk);
1204 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
1205 struct sk_buff *skb;
David S. Miller59c2cda2011-12-01 14:12:55 -05001206 unsigned int ulen, copied;
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001207 int peeked, off = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001208 int err;
1209 int is_udplite = IS_UDPLITE(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001210 bool slow;
David S. Millerdb8dac22008-03-06 16:22:02 -08001211
1212 /*
1213 * Check any passed addresses
1214 */
1215 if (addr_len)
Eric Dumazetc482c562009-07-17 00:26:32 +00001216 *addr_len = sizeof(*sin);
David S. Millerdb8dac22008-03-06 16:22:02 -08001217
1218 if (flags & MSG_ERRQUEUE)
1219 return ip_recv_error(sk, msg, len);
1220
1221try_again:
1222 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001223 &peeked, &off, &err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001224 if (!skb)
1225 goto out;
1226
1227 ulen = skb->len - sizeof(struct udphdr);
David S. Miller59c2cda2011-12-01 14:12:55 -05001228 copied = len;
1229 if (copied > ulen)
1230 copied = ulen;
1231 else if (copied < ulen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001232 msg->msg_flags |= MSG_TRUNC;
1233
1234 /*
1235 * If checksum is needed at all, try to do it while copying the
1236 * data. If the data is truncated, or if we only want a partial
1237 * coverage checksum (UDP-Lite), do it before the copy.
1238 */
1239
David S. Miller59c2cda2011-12-01 14:12:55 -05001240 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001241 if (udp_lib_checksum_complete(skb))
1242 goto csum_copy_err;
1243 }
1244
1245 if (skb_csum_unnecessary(skb))
1246 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
David S. Miller59c2cda2011-12-01 14:12:55 -05001247 msg->msg_iov, copied);
David S. Millerdb8dac22008-03-06 16:22:02 -08001248 else {
Eric Dumazetc482c562009-07-17 00:26:32 +00001249 err = skb_copy_and_csum_datagram_iovec(skb,
1250 sizeof(struct udphdr),
1251 msg->msg_iov);
David S. Millerdb8dac22008-03-06 16:22:02 -08001252
1253 if (err == -EINVAL)
1254 goto csum_copy_err;
1255 }
1256
Eric Dumazet22911fc2012-06-27 00:23:44 +00001257 if (unlikely(err)) {
1258 trace_kfree_skb(skb, udp_recvmsg);
Eric Dumazet979402b2012-09-05 23:34:44 +00001259 if (!peeked) {
1260 atomic_inc(&sk->sk_drops);
1261 UDP_INC_STATS_USER(sock_net(sk),
1262 UDP_MIB_INERRORS, is_udplite);
1263 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001264 goto out_free;
Eric Dumazet22911fc2012-06-27 00:23:44 +00001265 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001266
1267 if (!peeked)
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001268 UDP_INC_STATS_USER(sock_net(sk),
1269 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001270
Neil Horman3b885782009-10-12 13:26:31 -07001271 sock_recv_ts_and_drops(msg, sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001272
1273 /* Copy the address. */
Eric Dumazetc482c562009-07-17 00:26:32 +00001274 if (sin) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001275 sin->sin_family = AF_INET;
1276 sin->sin_port = udp_hdr(skb)->source;
1277 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1278 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
1279 }
1280 if (inet->cmsg_flags)
1281 ip_cmsg_recv(msg, skb);
1282
David S. Miller59c2cda2011-12-01 14:12:55 -05001283 err = copied;
David S. Millerdb8dac22008-03-06 16:22:02 -08001284 if (flags & MSG_TRUNC)
1285 err = ulen;
1286
1287out_free:
Eric Dumazet9d410c72009-10-30 05:03:53 +00001288 skb_free_datagram_locked(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001289out:
1290 return err;
1291
1292csum_copy_err:
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001293 slow = lock_sock_fast(sk);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001294 if (!skb_kill_datagram(sk, skb, flags)) {
1295 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001296 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001297 }
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001298 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001299
1300 if (noblock)
1301 return -EAGAIN;
Xufeng Zhang9cfaa8d2011-06-21 10:43:40 +00001302
1303 /* starting over for a new packet */
1304 msg->msg_flags &= ~MSG_TRUNC;
David S. Millerdb8dac22008-03-06 16:22:02 -08001305 goto try_again;
1306}
1307
1308
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309int udp_disconnect(struct sock *sk, int flags)
1310{
1311 struct inet_sock *inet = inet_sk(sk);
1312 /*
1313 * 1003.1g - break association.
1314 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001315
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 sk->sk_state = TCP_CLOSE;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001317 inet->inet_daddr = 0;
1318 inet->inet_dport = 0;
Tom Herbertbdeab992011-08-14 19:45:55 +00001319 sock_rps_reset_rxhash(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 sk->sk_bound_dev_if = 0;
1321 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1322 inet_reset_saddr(sk);
1323
1324 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1325 sk->sk_prot->unhash(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001326 inet->inet_sport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 }
1328 sk_dst_reset(sk);
1329 return 0;
1330}
Eric Dumazetc482c562009-07-17 00:26:32 +00001331EXPORT_SYMBOL(udp_disconnect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332
Eric Dumazet645ca702008-10-29 01:41:45 -07001333void udp_lib_unhash(struct sock *sk)
1334{
Eric Dumazet723b4612008-11-25 13:55:15 -08001335 if (sk_hashed(sk)) {
1336 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Eric Dumazet512615b2009-11-08 10:17:58 +00001337 struct udp_hslot *hslot, *hslot2;
1338
1339 hslot = udp_hashslot(udptable, sock_net(sk),
1340 udp_sk(sk)->udp_port_hash);
1341 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
Eric Dumazet645ca702008-10-29 01:41:45 -07001342
Eric Dumazet723b4612008-11-25 13:55:15 -08001343 spin_lock_bh(&hslot->lock);
1344 if (sk_nulls_del_node_init_rcu(sk)) {
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00001345 hslot->count--;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001346 inet_sk(sk)->inet_num = 0;
Eric Dumazet723b4612008-11-25 13:55:15 -08001347 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet512615b2009-11-08 10:17:58 +00001348
1349 spin_lock(&hslot2->lock);
1350 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1351 hslot2->count--;
1352 spin_unlock(&hslot2->lock);
Eric Dumazet723b4612008-11-25 13:55:15 -08001353 }
1354 spin_unlock_bh(&hslot->lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001355 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001356}
1357EXPORT_SYMBOL(udp_lib_unhash);
1358
Eric Dumazet719f8352010-09-08 05:08:44 +00001359/*
1360 * inet_rcv_saddr was changed, we must rehash secondary hash
1361 */
1362void udp_lib_rehash(struct sock *sk, u16 newhash)
1363{
1364 if (sk_hashed(sk)) {
1365 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1366 struct udp_hslot *hslot, *hslot2, *nhslot2;
1367
1368 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1369 nhslot2 = udp_hashslot2(udptable, newhash);
1370 udp_sk(sk)->udp_portaddr_hash = newhash;
1371 if (hslot2 != nhslot2) {
1372 hslot = udp_hashslot(udptable, sock_net(sk),
1373 udp_sk(sk)->udp_port_hash);
1374 /* we must lock primary chain too */
1375 spin_lock_bh(&hslot->lock);
1376
1377 spin_lock(&hslot2->lock);
1378 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1379 hslot2->count--;
1380 spin_unlock(&hslot2->lock);
1381
1382 spin_lock(&nhslot2->lock);
1383 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
1384 &nhslot2->head);
1385 nhslot2->count++;
1386 spin_unlock(&nhslot2->lock);
1387
1388 spin_unlock_bh(&hslot->lock);
1389 }
1390 }
1391}
1392EXPORT_SYMBOL(udp_lib_rehash);
1393
1394static void udp_v4_rehash(struct sock *sk)
1395{
1396 u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1397 inet_sk(sk)->inet_rcv_saddr,
1398 inet_sk(sk)->inet_num);
1399 udp_lib_rehash(sk, new_hash);
1400}
1401
Herbert Xu93821772008-09-15 11:48:46 -07001402static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1403{
Tom Herbertfec5e652010-04-16 16:01:27 -07001404 int rc;
Herbert Xu93821772008-09-15 11:48:46 -07001405
Tom Herbertfec5e652010-04-16 16:01:27 -07001406 if (inet_sk(sk)->inet_daddr)
Tom Herbertbdeab992011-08-14 19:45:55 +00001407 sock_rps_save_rxhash(sk, skb);
Tom Herbertfec5e652010-04-16 16:01:27 -07001408
Eric Dumazetd826eb12011-11-09 07:24:35 +00001409 rc = sock_queue_rcv_skb(sk, skb);
Eric Dumazet766e90372009-10-14 20:40:11 -07001410 if (rc < 0) {
1411 int is_udplite = IS_UDPLITE(sk);
1412
Herbert Xu93821772008-09-15 11:48:46 -07001413 /* Note that an ENOMEM error is charged twice */
Eric Dumazet766e90372009-10-14 20:40:11 -07001414 if (rc == -ENOMEM)
Herbert Xu93821772008-09-15 11:48:46 -07001415 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1416 is_udplite);
Eric Dumazet766e90372009-10-14 20:40:11 -07001417 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1418 kfree_skb(skb);
Satoru Moriya296f7ea2011-06-17 11:58:39 +00001419 trace_udp_fail_queue_rcv_skb(rc, sk);
Eric Dumazet766e90372009-10-14 20:40:11 -07001420 return -1;
Herbert Xu93821772008-09-15 11:48:46 -07001421 }
1422
1423 return 0;
1424
Herbert Xu93821772008-09-15 11:48:46 -07001425}
1426
Eric Dumazet447167b2012-04-11 23:05:28 +00001427static struct static_key udp_encap_needed __read_mostly;
1428void udp_encap_enable(void)
1429{
1430 if (!static_key_enabled(&udp_encap_needed))
1431 static_key_slow_inc(&udp_encap_needed);
1432}
1433EXPORT_SYMBOL(udp_encap_enable);
1434
David S. Millerdb8dac22008-03-06 16:22:02 -08001435/* returns:
1436 * -1: error
1437 * 0: success
1438 * >0: "udp encap" protocol resubmission
1439 *
1440 * Note that in the success and error cases, the skb is assumed to
1441 * have either been requeued or freed.
1442 */
Eric Dumazetc482c562009-07-17 00:26:32 +00001443int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -08001444{
1445 struct udp_sock *up = udp_sk(sk);
1446 int rc;
1447 int is_udplite = IS_UDPLITE(sk);
1448
1449 /*
1450 * Charge it to the socket, dropping if the queue is full.
1451 */
1452 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1453 goto drop;
1454 nf_reset(skb);
1455
Eric Dumazet447167b2012-04-11 23:05:28 +00001456 if (static_key_false(&udp_encap_needed) && up->encap_type) {
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001457 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1458
David S. Millerdb8dac22008-03-06 16:22:02 -08001459 /*
1460 * This is an encapsulation socket so pass the skb to
1461 * the socket's udp_encap_rcv() hook. Otherwise, just
1462 * fall through and pass this up the UDP socket.
1463 * up->encap_rcv() returns the following value:
1464 * =0 if skb was successfully passed to the encap
1465 * handler or was discarded by it.
1466 * >0 if skb should be passed on to UDP.
1467 * <0 if skb should be resubmitted as proto -N
1468 */
1469
1470 /* if we're overly short, let UDP handle it */
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001471 encap_rcv = ACCESS_ONCE(up->encap_rcv);
1472 if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001473 int ret;
1474
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001475 ret = encap_rcv(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001476 if (ret <= 0) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001477 UDP_INC_STATS_BH(sock_net(sk),
1478 UDP_MIB_INDATAGRAMS,
David S. Millerdb8dac22008-03-06 16:22:02 -08001479 is_udplite);
1480 return -ret;
1481 }
1482 }
1483
1484 /* FALLTHROUGH -- it's a UDP Packet */
1485 }
1486
1487 /*
1488 * UDP-Lite specific tests, ignored on UDP sockets
1489 */
1490 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1491
1492 /*
1493 * MIB statistics other than incrementing the error count are
1494 * disabled for the following two types of errors: these depend
1495 * on the application settings, not on the functioning of the
1496 * protocol stack as such.
1497 *
1498 * RFC 3828 here recommends (sec 3.3): "There should also be a
1499 * way ... to ... at least let the receiving application block
1500 * delivery of packets with coverage values less than a value
1501 * provided by the application."
1502 */
1503 if (up->pcrlen == 0) { /* full coverage was set */
Joe Perchesafd465032012-03-12 07:03:32 +00001504 LIMIT_NETDEBUG(KERN_WARNING "UDPLite: partial coverage %d while full coverage %d requested\n",
1505 UDP_SKB_CB(skb)->cscov, skb->len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001506 goto drop;
1507 }
1508 /* The next case involves violating the min. coverage requested
1509 * by the receiver. This is subtle: if receiver wants x and x is
1510 * greater than the buffersize/MTU then receiver will complain
1511 * that it wants x while sender emits packets of smaller size y.
1512 * Therefore the above ...()->partial_cov statement is essential.
1513 */
1514 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
Joe Perchesafd465032012-03-12 07:03:32 +00001515 LIMIT_NETDEBUG(KERN_WARNING "UDPLite: coverage %d too small, need min %d\n",
1516 UDP_SKB_CB(skb)->cscov, up->pcrlen);
David S. Millerdb8dac22008-03-06 16:22:02 -08001517 goto drop;
1518 }
1519 }
1520
Eric Dumazet33d480c2011-08-11 19:30:52 +00001521 if (rcu_access_pointer(sk->sk_filter) &&
1522 udp_lib_checksum_complete(skb))
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001523 goto csum_error;
David S. Millerdb8dac22008-03-06 16:22:02 -08001524
Eric Dumazetc3774112010-04-27 15:13:20 -07001525
Eric Dumazetf545a382012-04-22 23:34:26 +00001526 if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf))
Eric Dumazetc3774112010-04-27 15:13:20 -07001527 goto drop;
1528
Herbert Xu93821772008-09-15 11:48:46 -07001529 rc = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001530
Eric Dumazetd826eb12011-11-09 07:24:35 +00001531 ipv4_pktinfo_prepare(skb);
Herbert Xu93821772008-09-15 11:48:46 -07001532 bh_lock_sock(sk);
1533 if (!sock_owned_by_user(sk))
1534 rc = __udp_queue_rcv_skb(sk, skb);
Eric Dumazetf545a382012-04-22 23:34:26 +00001535 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
Zhu Yi55349792010-03-04 18:01:42 +00001536 bh_unlock_sock(sk);
1537 goto drop;
1538 }
Herbert Xu93821772008-09-15 11:48:46 -07001539 bh_unlock_sock(sk);
1540
1541 return rc;
David S. Millerdb8dac22008-03-06 16:22:02 -08001542
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001543csum_error:
1544 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001545drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001546 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001547 atomic_inc(&sk->sk_drops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001548 kfree_skb(skb);
1549 return -1;
1550}
1551
Eric Dumazet1240d132009-11-08 10:18:44 +00001552
1553static void flush_stack(struct sock **stack, unsigned int count,
1554 struct sk_buff *skb, unsigned int final)
1555{
1556 unsigned int i;
1557 struct sk_buff *skb1 = NULL;
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001558 struct sock *sk;
Eric Dumazet1240d132009-11-08 10:18:44 +00001559
1560 for (i = 0; i < count; i++) {
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001561 sk = stack[i];
Eric Dumazet1240d132009-11-08 10:18:44 +00001562 if (likely(skb1 == NULL))
1563 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
1564
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001565 if (!skb1) {
1566 atomic_inc(&sk->sk_drops);
1567 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1568 IS_UDPLITE(sk));
1569 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1570 IS_UDPLITE(sk));
1571 }
1572
1573 if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0)
Eric Dumazet1240d132009-11-08 10:18:44 +00001574 skb1 = NULL;
1575 }
1576 if (unlikely(skb1))
1577 kfree_skb(skb1);
1578}
1579
David S. Millerdb8dac22008-03-06 16:22:02 -08001580/*
1581 * Multicasts and broadcasts go to each listener.
1582 *
Eric Dumazet1240d132009-11-08 10:18:44 +00001583 * Note: called only from the BH handler context.
David S. Millerdb8dac22008-03-06 16:22:02 -08001584 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001585static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001586 struct udphdr *uh,
1587 __be32 saddr, __be32 daddr,
Eric Dumazet645ca702008-10-29 01:41:45 -07001588 struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -08001589{
Eric Dumazet1240d132009-11-08 10:18:44 +00001590 struct sock *sk, *stack[256 / sizeof(struct sock *)];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00001591 struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001592 int dif;
Eric Dumazet1240d132009-11-08 10:18:44 +00001593 unsigned int i, count = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001594
Eric Dumazet645ca702008-10-29 01:41:45 -07001595 spin_lock(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08001596 sk = sk_nulls_head(&hslot->head);
David S. Millerdb8dac22008-03-06 16:22:02 -08001597 dif = skb->dev->ifindex;
Eric Dumazet920a4612008-11-01 21:22:23 -07001598 sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
Eric Dumazet1240d132009-11-08 10:18:44 +00001599 while (sk) {
1600 stack[count++] = sk;
1601 sk = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest,
1602 daddr, uh->source, saddr, dif);
1603 if (unlikely(count == ARRAY_SIZE(stack))) {
1604 if (!sk)
1605 break;
1606 flush_stack(stack, count, skb, ~0);
1607 count = 0;
1608 }
1609 }
1610 /*
1611 * before releasing chain lock, we must take a reference on sockets
1612 */
1613 for (i = 0; i < count; i++)
1614 sock_hold(stack[i]);
David S. Millerdb8dac22008-03-06 16:22:02 -08001615
Eric Dumazet645ca702008-10-29 01:41:45 -07001616 spin_unlock(&hslot->lock);
Eric Dumazet1240d132009-11-08 10:18:44 +00001617
1618 /*
1619 * do the slow work with no lock held
1620 */
1621 if (count) {
1622 flush_stack(stack, count, skb, count - 1);
1623
1624 for (i = 0; i < count; i++)
1625 sock_put(stack[i]);
1626 } else {
1627 kfree_skb(skb);
1628 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001629 return 0;
1630}
1631
1632/* Initialize UDP checksum. If exited with zero value (success),
1633 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1634 * Otherwise, csum completion requires chacksumming packet body,
1635 * including udp header and folding it to skb->csum.
1636 */
1637static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1638 int proto)
1639{
1640 const struct iphdr *iph;
1641 int err;
1642
1643 UDP_SKB_CB(skb)->partial_cov = 0;
1644 UDP_SKB_CB(skb)->cscov = skb->len;
1645
1646 if (proto == IPPROTO_UDPLITE) {
1647 err = udplite_checksum_init(skb, uh);
1648 if (err)
1649 return err;
1650 }
1651
1652 iph = ip_hdr(skb);
1653 if (uh->check == 0) {
1654 skb->ip_summed = CHECKSUM_UNNECESSARY;
1655 } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
Eric Dumazetc482c562009-07-17 00:26:32 +00001656 if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len,
David S. Millerdb8dac22008-03-06 16:22:02 -08001657 proto, skb->csum))
1658 skb->ip_summed = CHECKSUM_UNNECESSARY;
1659 }
1660 if (!skb_csum_unnecessary(skb))
1661 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
1662 skb->len, proto, 0);
1663 /* Probably, we should checksum udp header (it should be in cache
1664 * in any case) and data in tiny packets (< rx copybreak).
1665 */
1666
1667 return 0;
1668}
1669
1670/*
1671 * All we need to do is get the socket, and then do a checksum.
1672 */
1673
Eric Dumazet645ca702008-10-29 01:41:45 -07001674int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
David S. Millerdb8dac22008-03-06 16:22:02 -08001675 int proto)
1676{
1677 struct sock *sk;
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001678 struct udphdr *uh;
David S. Millerdb8dac22008-03-06 16:22:02 -08001679 unsigned short ulen;
Eric Dumazetadf30902009-06-02 05:19:30 +00001680 struct rtable *rt = skb_rtable(skb);
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001681 __be32 saddr, daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001682 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001683
1684 /*
1685 * Validate the packet.
1686 */
1687 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1688 goto drop; /* No space for header. */
1689
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001690 uh = udp_hdr(skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001691 ulen = ntohs(uh->len);
Bjørn Morkccc2d972010-05-06 03:44:34 +00001692 saddr = ip_hdr(skb)->saddr;
1693 daddr = ip_hdr(skb)->daddr;
1694
David S. Millerdb8dac22008-03-06 16:22:02 -08001695 if (ulen > skb->len)
1696 goto short_packet;
1697
1698 if (proto == IPPROTO_UDP) {
1699 /* UDP validates ulen. */
1700 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1701 goto short_packet;
1702 uh = udp_hdr(skb);
1703 }
1704
1705 if (udp4_csum_init(skb, uh, proto))
1706 goto csum_error;
1707
1708 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
Pavel Emelyanove3163492008-06-16 17:12:11 -07001709 return __udp4_lib_mcast_deliver(net, skb, uh,
1710 saddr, daddr, udptable);
David S. Millerdb8dac22008-03-06 16:22:02 -08001711
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -07001712 sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
David S. Millerdb8dac22008-03-06 16:22:02 -08001713
1714 if (sk != NULL) {
Eliezer Tamira5b50472013-06-10 11:40:00 +03001715 int ret;
1716
Eliezer Tamir8b80cda2013-07-10 17:13:26 +03001717 sk_mark_napi_id(sk, skb);
Eliezer Tamira5b50472013-06-10 11:40:00 +03001718 ret = udp_queue_rcv_skb(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001719 sock_put(sk);
1720
1721 /* a return value > 0 means to resubmit the input, but
1722 * it wants the return to be -protocol, or 0
1723 */
1724 if (ret > 0)
1725 return -ret;
1726 return 0;
1727 }
1728
1729 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1730 goto drop;
1731 nf_reset(skb);
1732
1733 /* No socket. Drop packet silently, if checksum is wrong */
1734 if (udp_lib_checksum_complete(skb))
1735 goto csum_error;
1736
Pavel Emelyanov02833282008-07-05 21:18:48 -07001737 UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001738 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1739
1740 /*
1741 * Hmm. We got an UDP packet to a port to which we
1742 * don't wanna listen. Ignore it.
1743 */
1744 kfree_skb(skb);
1745 return 0;
1746
1747short_packet:
Harvey Harrison673d57e2008-10-31 00:53:57 -07001748 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
Joe Perchesafd465032012-03-12 07:03:32 +00001749 proto == IPPROTO_UDPLITE ? "Lite" : "",
1750 &saddr, ntohs(uh->source),
1751 ulen, skb->len,
1752 &daddr, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001753 goto drop;
1754
1755csum_error:
1756 /*
1757 * RFC1122: OK. Discards the bad packet silently (as far as
1758 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1759 */
Harvey Harrison673d57e2008-10-31 00:53:57 -07001760 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
Joe Perchesafd465032012-03-12 07:03:32 +00001761 proto == IPPROTO_UDPLITE ? "Lite" : "",
1762 &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
David S. Millerdb8dac22008-03-06 16:22:02 -08001763 ulen);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001764 UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001765drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001766 UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001767 kfree_skb(skb);
1768 return 0;
1769}
1770
1771int udp_rcv(struct sk_buff *skb)
1772{
Eric Dumazet645ca702008-10-29 01:41:45 -07001773 return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
David S. Millerdb8dac22008-03-06 16:22:02 -08001774}
1775
Brian Haley7d06b2e2008-06-14 17:04:49 -07001776void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08001777{
Tom Parkin44046a52013-03-19 06:11:12 +00001778 struct udp_sock *up = udp_sk(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001779 bool slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001780 udp_flush_pending_frames(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001781 unlock_sock_fast(sk, slow);
Tom Parkin44046a52013-03-19 06:11:12 +00001782 if (static_key_false(&udp_encap_needed) && up->encap_type) {
1783 void (*encap_destroy)(struct sock *sk);
1784 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1785 if (encap_destroy)
1786 encap_destroy(sk);
1787 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001788}
1789
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790/*
1791 * Socket option code for UDP
1792 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001793int udp_lib_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001794 char __user *optval, unsigned int optlen,
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001795 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796{
1797 struct udp_sock *up = udp_sk(sk);
1798 int val;
1799 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11001800 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801
Eric Dumazetc482c562009-07-17 00:26:32 +00001802 if (optlen < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803 return -EINVAL;
1804
1805 if (get_user(val, (int __user *)optval))
1806 return -EFAULT;
1807
Stephen Hemminger6516c652007-03-08 20:41:55 -08001808 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 case UDP_CORK:
1810 if (val != 0) {
1811 up->corkflag = 1;
1812 } else {
1813 up->corkflag = 0;
1814 lock_sock(sk);
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001815 (*push_pending_frames)(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816 release_sock(sk);
1817 }
1818 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001819
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 case UDP_ENCAP:
1821 switch (val) {
1822 case 0:
1823 case UDP_ENCAP_ESPINUDP:
1824 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07001825 up->encap_rcv = xfrm4_udp_encap_rcv;
1826 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07001827 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828 up->encap_type = val;
Eric Dumazet447167b2012-04-11 23:05:28 +00001829 udp_encap_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 break;
1831 default:
1832 err = -ENOPROTOOPT;
1833 break;
1834 }
1835 break;
1836
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001837 /*
1838 * UDP-Lite's partial checksum coverage (RFC 3828).
1839 */
1840 /* The sender sets actual checksum coverage length via this option.
1841 * The case coverage > packet length is handled by send module. */
1842 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11001843 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001844 return -ENOPROTOOPT;
1845 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
1846 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07001847 else if (val > USHRT_MAX)
1848 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001849 up->pcslen = val;
1850 up->pcflag |= UDPLITE_SEND_CC;
1851 break;
1852
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001853 /* The receiver specifies a minimum checksum coverage value. To make
1854 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001855 * used, this again means full checksum coverage. */
1856 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11001857 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001858 return -ENOPROTOOPT;
1859 if (val != 0 && val < 8) /* Avoid silly minimal values. */
1860 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07001861 else if (val > USHRT_MAX)
1862 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001863 up->pcrlen = val;
1864 up->pcflag |= UDPLITE_RECV_CC;
1865 break;
1866
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867 default:
1868 err = -ENOPROTOOPT;
1869 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001870 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871
1872 return err;
1873}
Eric Dumazetc482c562009-07-17 00:26:32 +00001874EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875
David S. Millerdb8dac22008-03-06 16:22:02 -08001876int udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001877 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001878{
1879 if (level == SOL_UDP || level == SOL_UDPLITE)
1880 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1881 udp_push_pending_frames);
1882 return ip_setsockopt(sk, level, optname, optval, optlen);
1883}
1884
1885#ifdef CONFIG_COMPAT
1886int compat_udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001887 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001888{
1889 if (level == SOL_UDP || level == SOL_UDPLITE)
1890 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1891 udp_push_pending_frames);
1892 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
1893}
1894#endif
1895
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08001896int udp_lib_getsockopt(struct sock *sk, int level, int optname,
1897 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898{
1899 struct udp_sock *up = udp_sk(sk);
1900 int val, len;
1901
Eric Dumazetc482c562009-07-17 00:26:32 +00001902 if (get_user(len, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 return -EFAULT;
1904
1905 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001906
Stephen Hemminger6516c652007-03-08 20:41:55 -08001907 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908 return -EINVAL;
1909
Stephen Hemminger6516c652007-03-08 20:41:55 -08001910 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 case UDP_CORK:
1912 val = up->corkflag;
1913 break;
1914
1915 case UDP_ENCAP:
1916 val = up->encap_type;
1917 break;
1918
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001919 /* The following two cannot be changed on UDP sockets, the return is
1920 * always 0 (which corresponds to the full checksum coverage of UDP). */
1921 case UDPLITE_SEND_CSCOV:
1922 val = up->pcslen;
1923 break;
1924
1925 case UDPLITE_RECV_CSCOV:
1926 val = up->pcrlen;
1927 break;
1928
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929 default:
1930 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08001931 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932
Stephen Hemminger6516c652007-03-08 20:41:55 -08001933 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001934 return -EFAULT;
Eric Dumazetc482c562009-07-17 00:26:32 +00001935 if (copy_to_user(optval, &val, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001937 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938}
Eric Dumazetc482c562009-07-17 00:26:32 +00001939EXPORT_SYMBOL(udp_lib_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940
David S. Millerdb8dac22008-03-06 16:22:02 -08001941int udp_getsockopt(struct sock *sk, int level, int optname,
1942 char __user *optval, int __user *optlen)
1943{
1944 if (level == SOL_UDP || level == SOL_UDPLITE)
1945 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1946 return ip_getsockopt(sk, level, optname, optval, optlen);
1947}
1948
1949#ifdef CONFIG_COMPAT
1950int compat_udp_getsockopt(struct sock *sk, int level, int optname,
1951 char __user *optval, int __user *optlen)
1952{
1953 if (level == SOL_UDP || level == SOL_UDPLITE)
1954 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1955 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
1956}
1957#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958/**
1959 * udp_poll - wait for a UDP event.
1960 * @file - file struct
1961 * @sock - socket
1962 * @wait - poll table
1963 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001964 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 * blocking sockets. If application is using a blocking fd
1966 * and a packet with checksum error is in the queue;
1967 * then it could get return from select indicating data available
1968 * but then block when reading it. Add special case code
1969 * to work around these arguably broken applications.
1970 */
1971unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1972{
1973 unsigned int mask = datagram_poll(file, sock, wait);
1974 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001975
David Majnemerc3f1dba2013-05-31 13:15:38 +00001976 sock_rps_record_flow(sk);
1977
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 /* Check for false positives due to checksum errors */
Eric Dumazet85584672009-10-09 04:43:40 +00001979 if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
1980 !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk))
1981 mask &= ~(POLLIN | POLLRDNORM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982
1983 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001984
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985}
Eric Dumazetc482c562009-07-17 00:26:32 +00001986EXPORT_SYMBOL(udp_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987
David S. Millerdb8dac22008-03-06 16:22:02 -08001988struct proto udp_prot = {
1989 .name = "UDP",
1990 .owner = THIS_MODULE,
1991 .close = udp_lib_close,
1992 .connect = ip4_datagram_connect,
1993 .disconnect = udp_disconnect,
1994 .ioctl = udp_ioctl,
1995 .destroy = udp_destroy_sock,
1996 .setsockopt = udp_setsockopt,
1997 .getsockopt = udp_getsockopt,
1998 .sendmsg = udp_sendmsg,
1999 .recvmsg = udp_recvmsg,
2000 .sendpage = udp_sendpage,
Herbert Xu93821772008-09-15 11:48:46 -07002001 .backlog_rcv = __udp_queue_rcv_skb,
Steffen Klassert8141ed92013-01-21 02:00:03 +00002002 .release_cb = ip4_datagram_release_cb,
David S. Millerdb8dac22008-03-06 16:22:02 -08002003 .hash = udp_lib_hash,
2004 .unhash = udp_lib_unhash,
Eric Dumazet719f8352010-09-08 05:08:44 +00002005 .rehash = udp_v4_rehash,
David S. Millerdb8dac22008-03-06 16:22:02 -08002006 .get_port = udp_v4_get_port,
2007 .memory_allocated = &udp_memory_allocated,
2008 .sysctl_mem = sysctl_udp_mem,
2009 .sysctl_wmem = &sysctl_udp_wmem_min,
2010 .sysctl_rmem = &sysctl_udp_rmem_min,
2011 .obj_size = sizeof(struct udp_sock),
Eric Dumazet271b72c2008-10-29 02:11:14 -07002012 .slab_flags = SLAB_DESTROY_BY_RCU,
Eric Dumazet645ca702008-10-29 01:41:45 -07002013 .h.udp_table = &udp_table,
David S. Millerdb8dac22008-03-06 16:22:02 -08002014#ifdef CONFIG_COMPAT
2015 .compat_setsockopt = compat_udp_setsockopt,
2016 .compat_getsockopt = compat_udp_getsockopt,
2017#endif
Octavian Purdilafcbdf092010-12-16 14:26:56 -08002018 .clear_sk = sk_prot_clear_portaddr_nulls,
David S. Millerdb8dac22008-03-06 16:22:02 -08002019};
Eric Dumazetc482c562009-07-17 00:26:32 +00002020EXPORT_SYMBOL(udp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021
2022/* ------------------------------------------------------------------------ */
2023#ifdef CONFIG_PROC_FS
2024
Eric Dumazet645ca702008-10-29 01:41:45 -07002025static struct sock *udp_get_first(struct seq_file *seq, int start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026{
2027 struct sock *sk;
2028 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002029 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002031 for (state->bucket = start; state->bucket <= state->udp_table->mask;
2032 ++state->bucket) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002033 struct hlist_nulls_node *node;
Eric Dumazet645ca702008-10-29 01:41:45 -07002034 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002035
2036 if (hlist_nulls_empty(&hslot->head))
2037 continue;
2038
Eric Dumazet645ca702008-10-29 01:41:45 -07002039 spin_lock_bh(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08002040 sk_nulls_for_each(sk, node, &hslot->head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002041 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002042 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043 if (sk->sk_family == state->family)
2044 goto found;
2045 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002046 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047 }
2048 sk = NULL;
2049found:
2050 return sk;
2051}
2052
2053static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
2054{
2055 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002056 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057
2058 do {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002059 sk = sk_nulls_next(sk);
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002060 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061
Eric Dumazet645ca702008-10-29 01:41:45 -07002062 if (!sk) {
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002063 if (state->bucket <= state->udp_table->mask)
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002064 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07002065 return udp_get_first(seq, state->bucket + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066 }
2067 return sk;
2068}
2069
2070static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2071{
Eric Dumazet645ca702008-10-29 01:41:45 -07002072 struct sock *sk = udp_get_first(seq, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073
2074 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08002075 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076 --pos;
2077 return pos ? NULL : sk;
2078}
2079
2080static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2081{
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002082 struct udp_iter_state *state = seq->private;
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002083 state->bucket = MAX_UDP_PORTS;
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002084
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002085 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086}
2087
2088static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2089{
2090 struct sock *sk;
2091
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002092 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093 sk = udp_get_idx(seq, 0);
2094 else
2095 sk = udp_get_next(seq, v);
2096
2097 ++*pos;
2098 return sk;
2099}
2100
2101static void udp_seq_stop(struct seq_file *seq, void *v)
2102{
Eric Dumazet645ca702008-10-29 01:41:45 -07002103 struct udp_iter_state *state = seq->private;
2104
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002105 if (state->bucket <= state->udp_table->mask)
Eric Dumazet645ca702008-10-29 01:41:45 -07002106 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107}
2108
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002109int udp_seq_open(struct inode *inode, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110{
Al Virod9dda782013-03-31 18:16:14 -04002111 struct udp_seq_afinfo *afinfo = PDE_DATA(inode);
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002112 struct udp_iter_state *s;
2113 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002115 err = seq_open_net(inode, file, &afinfo->seq_ops,
2116 sizeof(struct udp_iter_state));
2117 if (err < 0)
2118 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002119
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002120 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121 s->family = afinfo->family;
Eric Dumazet645ca702008-10-29 01:41:45 -07002122 s->udp_table = afinfo->udp_table;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002123 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002124}
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002125EXPORT_SYMBOL(udp_seq_open);
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002126
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002128int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129{
2130 struct proc_dir_entry *p;
2131 int rc = 0;
2132
Denis V. Lunevdda61922008-03-28 18:24:26 -07002133 afinfo->seq_ops.start = udp_seq_start;
2134 afinfo->seq_ops.next = udp_seq_next;
2135 afinfo->seq_ops.stop = udp_seq_stop;
2136
Denis V. Lunev84841c32008-05-02 04:10:08 -07002137 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002138 afinfo->seq_fops, afinfo);
Denis V. Lunev84841c32008-05-02 04:10:08 -07002139 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140 rc = -ENOMEM;
2141 return rc;
2142}
Eric Dumazetc482c562009-07-17 00:26:32 +00002143EXPORT_SYMBOL(udp_proc_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002144
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002145void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146{
Gao fengece31ff2013-02-18 01:34:56 +00002147 remove_proc_entry(afinfo->name, net->proc_net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148}
Eric Dumazetc482c562009-07-17 00:26:32 +00002149EXPORT_SYMBOL(udp_proc_unregister);
David S. Millerdb8dac22008-03-06 16:22:02 -08002150
2151/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002152static void udp4_format_sock(struct sock *sp, struct seq_file *f,
2153 int bucket, int *len)
David S. Millerdb8dac22008-03-06 16:22:02 -08002154{
2155 struct inet_sock *inet = inet_sk(sp);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002156 __be32 dest = inet->inet_daddr;
2157 __be32 src = inet->inet_rcv_saddr;
2158 __u16 destp = ntohs(inet->inet_dport);
2159 __u16 srcp = ntohs(inet->inet_sport);
David S. Millerdb8dac22008-03-06 16:22:02 -08002160
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002161 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
Francesco Fuscod14c5ab2013-08-15 13:42:14 +02002162 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d%n",
David S. Millerdb8dac22008-03-06 16:22:02 -08002163 bucket, src, srcp, dest, destp, sp->sk_state,
Eric Dumazet31e6d362009-06-17 19:05:41 -07002164 sk_wmem_alloc_get(sp),
2165 sk_rmem_alloc_get(sp),
Eric W. Biedermana7cb5a42012-05-24 01:10:10 -06002166 0, 0L, 0,
2167 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
2168 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002169 atomic_read(&sp->sk_refcnt), sp,
2170 atomic_read(&sp->sk_drops), len);
David S. Millerdb8dac22008-03-06 16:22:02 -08002171}
2172
2173int udp4_seq_show(struct seq_file *seq, void *v)
2174{
2175 if (v == SEQ_START_TOKEN)
2176 seq_printf(seq, "%-127s\n",
2177 " sl local_address rem_address st tx_queue "
2178 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002179 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08002180 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08002181 struct udp_iter_state *state = seq->private;
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002182 int len;
David S. Millerdb8dac22008-03-06 16:22:02 -08002183
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002184 udp4_format_sock(v, seq, state->bucket, &len);
Eric Dumazetc482c562009-07-17 00:26:32 +00002185 seq_printf(seq, "%*s\n", 127 - len, "");
David S. Millerdb8dac22008-03-06 16:22:02 -08002186 }
2187 return 0;
2188}
2189
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002190static const struct file_operations udp_afinfo_seq_fops = {
2191 .owner = THIS_MODULE,
2192 .open = udp_seq_open,
2193 .read = seq_read,
2194 .llseek = seq_lseek,
2195 .release = seq_release_net
2196};
2197
David S. Millerdb8dac22008-03-06 16:22:02 -08002198/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08002199static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08002200 .name = "udp",
2201 .family = AF_INET,
Eric Dumazet645ca702008-10-29 01:41:45 -07002202 .udp_table = &udp_table,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002203 .seq_fops = &udp_afinfo_seq_fops,
Denis V. Lunevdda61922008-03-28 18:24:26 -07002204 .seq_ops = {
2205 .show = udp4_seq_show,
2206 },
David S. Millerdb8dac22008-03-06 16:22:02 -08002207};
2208
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002209static int __net_init udp4_proc_init_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002210{
2211 return udp_proc_register(net, &udp4_seq_afinfo);
2212}
2213
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002214static void __net_exit udp4_proc_exit_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002215{
2216 udp_proc_unregister(net, &udp4_seq_afinfo);
2217}
2218
2219static struct pernet_operations udp4_net_ops = {
2220 .init = udp4_proc_init_net,
2221 .exit = udp4_proc_exit_net,
2222};
2223
David S. Millerdb8dac22008-03-06 16:22:02 -08002224int __init udp4_proc_init(void)
2225{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002226 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002227}
2228
2229void udp4_proc_exit(void)
2230{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002231 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002232}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233#endif /* CONFIG_PROC_FS */
2234
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002235static __initdata unsigned long uhash_entries;
2236static int __init set_uhash_entries(char *str)
Eric Dumazet645ca702008-10-29 01:41:45 -07002237{
Eldad Zack413c27d2012-05-19 14:13:18 +00002238 ssize_t ret;
2239
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002240 if (!str)
2241 return 0;
Eldad Zack413c27d2012-05-19 14:13:18 +00002242
2243 ret = kstrtoul(str, 0, &uhash_entries);
2244 if (ret)
2245 return 0;
2246
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002247 if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2248 uhash_entries = UDP_HTABLE_SIZE_MIN;
2249 return 1;
2250}
2251__setup("uhash_entries=", set_uhash_entries);
Eric Dumazet645ca702008-10-29 01:41:45 -07002252
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002253void __init udp_table_init(struct udp_table *table, const char *name)
2254{
2255 unsigned int i;
2256
Tim Bird31fe62b2012-05-23 13:33:35 +00002257 table->hash = alloc_large_system_hash(name,
2258 2 * sizeof(struct udp_hslot),
2259 uhash_entries,
2260 21, /* one slot per 2 MB */
2261 0,
2262 &table->log,
2263 &table->mask,
2264 UDP_HTABLE_SIZE_MIN,
2265 64 * 1024);
2266
Eric Dumazet512615b2009-11-08 10:17:58 +00002267 table->hash2 = table->hash + (table->mask + 1);
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002268 for (i = 0; i <= table->mask; i++) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002269 INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00002270 table->hash[i].count = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -07002271 spin_lock_init(&table->hash[i].lock);
2272 }
Eric Dumazet512615b2009-11-08 10:17:58 +00002273 for (i = 0; i <= table->mask; i++) {
2274 INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i);
2275 table->hash2[i].count = 0;
2276 spin_lock_init(&table->hash2[i].lock);
2277 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002278}
2279
Hideo Aoki95766ff2007-12-31 00:29:24 -08002280void __init udp_init(void)
2281{
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002282 unsigned long limit;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002283
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002284 udp_table_init(&udp_table, "UDP");
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002285 limit = nr_free_buffer_pages() / 8;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002286 limit = max(limit, 128UL);
2287 sysctl_udp_mem[0] = limit / 4 * 3;
2288 sysctl_udp_mem[1] = limit;
2289 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2290
2291 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2292 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2293}
2294
Daniel Borkmannda5bab02013-06-08 12:56:03 +02002295struct sk_buff *skb_udp_tunnel_segment(struct sk_buff *skb,
2296 netdev_features_t features)
Pravin B Shelar73136262013-03-07 13:21:51 +00002297{
2298 struct sk_buff *segs = ERR_PTR(-EINVAL);
2299 int mac_len = skb->mac_len;
2300 int tnl_hlen = skb_inner_mac_header(skb) - skb_transport_header(skb);
Pravin B Shelar0d055352013-05-02 16:17:02 +00002301 __be16 protocol = skb->protocol;
Pravin B Shelar73136262013-03-07 13:21:51 +00002302 netdev_features_t enc_features;
Pravin B Shelar0d055352013-05-02 16:17:02 +00002303 int outer_hlen;
Pravin B Shelar73136262013-03-07 13:21:51 +00002304
2305 if (unlikely(!pskb_may_pull(skb, tnl_hlen)))
2306 goto out;
2307
2308 skb->encapsulation = 0;
2309 __skb_pull(skb, tnl_hlen);
2310 skb_reset_mac_header(skb);
2311 skb_set_network_header(skb, skb_inner_network_offset(skb));
2312 skb->mac_len = skb_inner_network_offset(skb);
Pravin B Shelar19acc322013-05-07 20:41:07 +00002313 skb->protocol = htons(ETH_P_TEB);
Pravin B Shelar73136262013-03-07 13:21:51 +00002314
2315 /* segment inner packet. */
2316 enc_features = skb->dev->hw_enc_features & netif_skb_features(skb);
2317 segs = skb_mac_gso_segment(skb, enc_features);
2318 if (!segs || IS_ERR(segs))
2319 goto out;
2320
2321 outer_hlen = skb_tnl_header_len(skb);
2322 skb = segs;
2323 do {
2324 struct udphdr *uh;
2325 int udp_offset = outer_hlen - tnl_hlen;
2326
Alexander Duyckcdbaa0b2013-07-10 17:05:06 -07002327 skb_reset_inner_headers(skb);
2328 skb->encapsulation = 1;
2329
Pravin B Shelar73136262013-03-07 13:21:51 +00002330 skb->mac_len = mac_len;
2331
2332 skb_push(skb, outer_hlen);
2333 skb_reset_mac_header(skb);
2334 skb_set_network_header(skb, mac_len);
2335 skb_set_transport_header(skb, udp_offset);
2336 uh = udp_hdr(skb);
2337 uh->len = htons(skb->len - udp_offset);
2338
2339 /* csum segment if tunnel sets skb with csum. */
Cong Wangeb3c0d82013-08-31 13:44:38 +08002340 if (protocol == htons(ETH_P_IP) && unlikely(uh->check)) {
Pravin B Shelar73136262013-03-07 13:21:51 +00002341 struct iphdr *iph = ip_hdr(skb);
2342
2343 uh->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
2344 skb->len - udp_offset,
2345 IPPROTO_UDP, 0);
2346 uh->check = csum_fold(skb_checksum(skb, udp_offset,
2347 skb->len - udp_offset, 0));
2348 if (uh->check == 0)
2349 uh->check = CSUM_MANGLED_0;
2350
Cong Wangeb3c0d82013-08-31 13:44:38 +08002351 } else if (protocol == htons(ETH_P_IPV6)) {
2352 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
2353 u32 len = skb->len - udp_offset;
2354
2355 uh->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr,
2356 len, IPPROTO_UDP, 0);
2357 uh->check = csum_fold(skb_checksum(skb, udp_offset, len, 0));
2358 if (uh->check == 0)
2359 uh->check = CSUM_MANGLED_0;
2360 skb->ip_summed = CHECKSUM_NONE;
Pravin B Shelar73136262013-03-07 13:21:51 +00002361 }
Cong Wangeb3c0d82013-08-31 13:44:38 +08002362
Pravin B Shelar0d055352013-05-02 16:17:02 +00002363 skb->protocol = protocol;
Pravin B Shelar73136262013-03-07 13:21:51 +00002364 } while ((skb = skb->next));
2365out:
2366 return segs;
2367}