blob: 13b4dcf86ef610d1fcc1b26f7f69f5a6bbd31686 [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>
Tom Herbert57c67ff2014-08-22 13:34:44 -0700102#include <linux/netdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103#include <linux/proc_fs.h>
104#include <linux/seq_file.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200105#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106#include <net/icmp.h>
Shawn Bohrer421b3882013-10-07 11:01:39 -0500107#include <net/inet_hashtables.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108#include <net/route.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109#include <net/checksum.h>
110#include <net/xfrm.h>
Satoru Moriya296f7ea2011-06-17 11:58:39 +0000111#include <trace/events/udp.h>
Eric Dumazet447167b2012-04-11 23:05:28 +0000112#include <linux/static_key.h>
Eric Dumazet22911fc2012-06-27 00:23:44 +0000113#include <trace/events/skb.h>
Eliezer Tamir076bb0c2013-07-10 17:13:17 +0300114#include <net/busy_poll.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800115#include "udp_impl.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000117struct udp_table udp_table __read_mostly;
Eric Dumazet645ca702008-10-29 01:41:45 -0700118EXPORT_SYMBOL(udp_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119
Eric Dumazet8d987e52010-11-09 23:24:26 +0000120long sysctl_udp_mem[3] __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800121EXPORT_SYMBOL(sysctl_udp_mem);
Eric Dumazetc482c562009-07-17 00:26:32 +0000122
123int sysctl_udp_rmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800124EXPORT_SYMBOL(sysctl_udp_rmem_min);
Eric Dumazetc482c562009-07-17 00:26:32 +0000125
126int sysctl_udp_wmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800127EXPORT_SYMBOL(sysctl_udp_wmem_min);
128
Eric Dumazet8d987e52010-11-09 23:24:26 +0000129atomic_long_t udp_memory_allocated;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800130EXPORT_SYMBOL(udp_memory_allocated);
131
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000132#define MAX_UDP_PORTS 65536
133#define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN)
Eric Dumazet98322f22009-01-26 21:35:35 -0800134
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700135static int udp_lib_lport_inuse(struct net *net, __u16 num,
Eric Dumazet645ca702008-10-29 01:41:45 -0700136 const struct udp_hslot *hslot,
Eric Dumazet98322f22009-01-26 21:35:35 -0800137 unsigned long *bitmap,
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700138 struct sock *sk,
139 int (*saddr_comp)(const struct sock *sk1,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000140 const struct sock *sk2),
141 unsigned int log)
Gerrit Renker25030a72006-08-26 20:06:05 -0700142{
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700143 struct sock *sk2;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800144 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000145 kuid_t uid = sock_i_uid(sk);
Gerrit Renker25030a72006-08-26 20:06:05 -0700146
Joe Perches4243cdc2014-11-11 21:59:20 -0800147 sk_nulls_for_each(sk2, node, &hslot->head) {
Joe Perches9d4fb272009-11-23 10:41:23 -0800148 if (net_eq(sock_net(sk2), net) &&
149 sk2 != sk &&
Eric Dumazetd4cada42009-11-08 10:17:30 +0000150 (bitmap || udp_sk(sk2)->udp_port_hash == num) &&
Joe Perches9d4fb272009-11-23 10:41:23 -0800151 (!sk2->sk_reuse || !sk->sk_reuse) &&
152 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
153 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000154 (!sk2->sk_reuseport || !sk->sk_reuseport ||
Joe Perches4243cdc2014-11-11 21:59:20 -0800155 !uid_eq(uid, sock_i_uid(sk2))) &&
156 saddr_comp(sk, sk2)) {
157 if (!bitmap)
Eric Dumazet98322f22009-01-26 21:35:35 -0800158 return 1;
Joe Perches4243cdc2014-11-11 21:59:20 -0800159 __set_bit(udp_sk(sk2)->udp_port_hash >> log, bitmap);
Eric Dumazet98322f22009-01-26 21:35:35 -0800160 }
Joe Perches4243cdc2014-11-11 21:59:20 -0800161 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700162 return 0;
163}
164
Eric Dumazet30fff922009-11-09 05:26:33 +0000165/*
166 * Note: we still hold spinlock of primary hash chain, so no other writer
167 * can insert/delete a socket with local_port == num
168 */
169static int udp_lib_lport_inuse2(struct net *net, __u16 num,
Joe Perches4243cdc2014-11-11 21:59:20 -0800170 struct udp_hslot *hslot2,
171 struct sock *sk,
172 int (*saddr_comp)(const struct sock *sk1,
173 const struct sock *sk2))
Eric Dumazet30fff922009-11-09 05:26:33 +0000174{
175 struct sock *sk2;
176 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000177 kuid_t uid = sock_i_uid(sk);
Eric Dumazet30fff922009-11-09 05:26:33 +0000178 int res = 0;
179
180 spin_lock(&hslot2->lock);
Joe Perches4243cdc2014-11-11 21:59:20 -0800181 udp_portaddr_for_each_entry(sk2, node, &hslot2->head) {
Joe Perches9d4fb272009-11-23 10:41:23 -0800182 if (net_eq(sock_net(sk2), net) &&
183 sk2 != sk &&
184 (udp_sk(sk2)->udp_port_hash == num) &&
185 (!sk2->sk_reuse || !sk->sk_reuse) &&
186 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
187 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000188 (!sk2->sk_reuseport || !sk->sk_reuseport ||
Joe Perches4243cdc2014-11-11 21:59:20 -0800189 !uid_eq(uid, sock_i_uid(sk2))) &&
190 saddr_comp(sk, sk2)) {
Eric Dumazet30fff922009-11-09 05:26:33 +0000191 res = 1;
192 break;
193 }
Joe Perches4243cdc2014-11-11 21:59:20 -0800194 }
Eric Dumazet30fff922009-11-09 05:26:33 +0000195 spin_unlock(&hslot2->lock);
196 return res;
197}
198
Gerrit Renker25030a72006-08-26 20:06:05 -0700199/**
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700200 * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
Gerrit Renker25030a72006-08-26 20:06:05 -0700201 *
202 * @sk: socket struct in question
203 * @snum: port number to look up
David S. Millerdf2bc452007-06-05 15:18:43 -0700204 * @saddr_comp: AF-dependent comparison of bound local IP addresses
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300205 * @hash2_nulladdr: AF-dependent hash value in secondary hash chains,
Eric Dumazet30fff922009-11-09 05:26:33 +0000206 * with NULL address
Gerrit Renker25030a72006-08-26 20:06:05 -0700207 */
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700208int udp_lib_get_port(struct sock *sk, unsigned short snum,
Joe Perches4243cdc2014-11-11 21:59:20 -0800209 int (*saddr_comp)(const struct sock *sk1,
210 const struct sock *sk2),
Eric Dumazet30fff922009-11-09 05:26:33 +0000211 unsigned int hash2_nulladdr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212{
Eric Dumazet512615b2009-11-08 10:17:58 +0000213 struct udp_hslot *hslot, *hslot2;
Eric Dumazet645ca702008-10-29 01:41:45 -0700214 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Gerrit Renker25030a72006-08-26 20:06:05 -0700215 int error = 1;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900216 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700218 if (!snum) {
Eric Dumazet9088c562008-10-08 11:44:17 -0700219 int low, high, remaining;
Eric Dumazet95c96172012-04-15 05:58:06 +0000220 unsigned int rand;
Eric Dumazet98322f22009-01-26 21:35:35 -0800221 unsigned short first, last;
222 DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
Eric W. Biederman0bbf87d2013-09-28 14:10:59 -0700224 inet_get_local_port_range(net, &low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700225 remaining = (high - low) + 1;
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700226
Aruna-Hewapathirane63862b52014-01-11 07:15:59 -0500227 rand = prandom_u32();
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200228 first = reciprocal_scale(rand, remaining) + low;
Eric Dumazet98322f22009-01-26 21:35:35 -0800229 /*
230 * force rand to be an odd multiple of UDP_HTABLE_SIZE
231 */
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000232 rand = (rand | 1) * (udptable->mask + 1);
Eric Dumazet5781b232009-12-13 19:32:39 -0800233 last = first + udptable->mask + 1;
234 do {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000235 hslot = udp_hashslot(udptable, net, first);
Eric Dumazet98322f22009-01-26 21:35:35 -0800236 bitmap_zero(bitmap, PORTS_PER_CHAIN);
Eric Dumazet645ca702008-10-29 01:41:45 -0700237 spin_lock_bh(&hslot->lock);
Eric Dumazet98322f22009-01-26 21:35:35 -0800238 udp_lib_lport_inuse(net, snum, hslot, bitmap, sk,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000239 saddr_comp, udptable->log);
Eric Dumazet98322f22009-01-26 21:35:35 -0800240
241 snum = first;
242 /*
243 * Iterate on all possible values of snum for this hash.
244 * Using steps of an odd multiple of UDP_HTABLE_SIZE
245 * give us randomization and full range coverage.
246 */
Eric Dumazet9088c562008-10-08 11:44:17 -0700247 do {
Eric Dumazet98322f22009-01-26 21:35:35 -0800248 if (low <= snum && snum <= high &&
Amerigo Wange3826f12010-05-05 00:27:06 +0000249 !test_bit(snum >> udptable->log, bitmap) &&
WANG Cong122ff242014-05-12 16:04:53 -0700250 !inet_is_local_reserved_port(net, snum))
Eric Dumazet98322f22009-01-26 21:35:35 -0800251 goto found;
252 snum += rand;
253 } while (snum != first);
254 spin_unlock_bh(&hslot->lock);
Eric Dumazet5781b232009-12-13 19:32:39 -0800255 } while (++first != last);
Eric Dumazet98322f22009-01-26 21:35:35 -0800256 goto fail;
Eric Dumazet645ca702008-10-29 01:41:45 -0700257 } else {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000258 hslot = udp_hashslot(udptable, net, snum);
Eric Dumazet645ca702008-10-29 01:41:45 -0700259 spin_lock_bh(&hslot->lock);
Eric Dumazet30fff922009-11-09 05:26:33 +0000260 if (hslot->count > 10) {
261 int exist;
262 unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum;
263
264 slot2 &= udptable->mask;
265 hash2_nulladdr &= udptable->mask;
266
267 hslot2 = udp_hashslot2(udptable, slot2);
268 if (hslot->count < hslot2->count)
269 goto scan_primary_hash;
270
271 exist = udp_lib_lport_inuse2(net, snum, hslot2,
272 sk, saddr_comp);
273 if (!exist && (hash2_nulladdr != slot2)) {
274 hslot2 = udp_hashslot2(udptable, hash2_nulladdr);
275 exist = udp_lib_lport_inuse2(net, snum, hslot2,
276 sk, saddr_comp);
277 }
278 if (exist)
279 goto fail_unlock;
280 else
281 goto found;
282 }
283scan_primary_hash:
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000284 if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk,
285 saddr_comp, 0))
Eric Dumazet645ca702008-10-29 01:41:45 -0700286 goto fail_unlock;
287 }
Eric Dumazet98322f22009-01-26 21:35:35 -0800288found:
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000289 inet_sk(sk)->inet_num = snum;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000290 udp_sk(sk)->udp_port_hash = snum;
291 udp_sk(sk)->udp_portaddr_hash ^= snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 if (sk_unhashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800293 sk_nulls_add_node_rcu(sk, &hslot->head);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +0000294 hslot->count++;
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700295 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet512615b2009-11-08 10:17:58 +0000296
297 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
298 spin_lock(&hslot2->lock);
299 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
300 &hslot2->head);
301 hslot2->count++;
302 spin_unlock(&hslot2->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700304 error = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -0700305fail_unlock:
306 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307fail:
Gerrit Renker25030a72006-08-26 20:06:05 -0700308 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309}
Eric Dumazetc482c562009-07-17 00:26:32 +0000310EXPORT_SYMBOL(udp_lib_get_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311
Vlad Yasevich499923c2009-04-09 17:37:33 +0000312static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
David S. Millerdb8dac22008-03-06 16:22:02 -0800313{
314 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
315
Eric Dumazetc482c562009-07-17 00:26:32 +0000316 return (!ipv6_only_sock(sk2) &&
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000317 (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr ||
318 inet1->inet_rcv_saddr == inet2->inet_rcv_saddr));
David S. Millerdb8dac22008-03-06 16:22:02 -0800319}
320
Eric Dumazetd4cada42009-11-08 10:17:30 +0000321static unsigned int udp4_portaddr_hash(struct net *net, __be32 saddr,
322 unsigned int port)
323{
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700324 return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000325}
326
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700327int udp_v4_get_port(struct sock *sk, unsigned short snum)
David S. Millerdb8dac22008-03-06 16:22:02 -0800328{
Eric Dumazet30fff922009-11-09 05:26:33 +0000329 unsigned int hash2_nulladdr =
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700330 udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum);
Eric Dumazet30fff922009-11-09 05:26:33 +0000331 unsigned int hash2_partial =
332 udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0);
333
Eric Dumazetd4cada42009-11-08 10:17:30 +0000334 /* precompute partial secondary hash */
Eric Dumazet30fff922009-11-09 05:26:33 +0000335 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
336 return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr);
David S. Millerdb8dac22008-03-06 16:22:02 -0800337}
338
Joe Perches60c04ae2014-12-01 20:29:06 -0800339static inline int compute_score(struct sock *sk, struct net *net,
340 __be32 saddr, unsigned short hnum, __be16 sport,
341 __be32 daddr, __be16 dport, int dif)
Eric Dumazet645ca702008-10-29 01:41:45 -0700342{
Joe Perches60c04ae2014-12-01 20:29:06 -0800343 int score;
344 struct inet_sock *inet;
Eric Dumazet645ca702008-10-29 01:41:45 -0700345
Joe Perches60c04ae2014-12-01 20:29:06 -0800346 if (!net_eq(sock_net(sk), net) ||
347 udp_sk(sk)->udp_port_hash != hnum ||
348 ipv6_only_sock(sk))
349 return -1;
Eric Dumazet645ca702008-10-29 01:41:45 -0700350
Joe Perches60c04ae2014-12-01 20:29:06 -0800351 score = (sk->sk_family == PF_INET) ? 2 : 1;
352 inet = inet_sk(sk);
353
354 if (inet->inet_rcv_saddr) {
355 if (inet->inet_rcv_saddr != daddr)
356 return -1;
357 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700358 }
Joe Perches60c04ae2014-12-01 20:29:06 -0800359
360 if (inet->inet_daddr) {
361 if (inet->inet_daddr != saddr)
362 return -1;
363 score += 4;
364 }
365
366 if (inet->inet_dport) {
367 if (inet->inet_dport != sport)
368 return -1;
369 score += 4;
370 }
371
372 if (sk->sk_bound_dev_if) {
373 if (sk->sk_bound_dev_if != dif)
374 return -1;
375 score += 4;
376 }
377
Eric Dumazet645ca702008-10-29 01:41:45 -0700378 return score;
379}
380
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000381/*
382 * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num)
383 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000384static inline int compute_score2(struct sock *sk, struct net *net,
385 __be32 saddr, __be16 sport,
386 __be32 daddr, unsigned int hnum, int dif)
387{
Joe Perches60c04ae2014-12-01 20:29:06 -0800388 int score;
389 struct inet_sock *inet;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000390
Joe Perches60c04ae2014-12-01 20:29:06 -0800391 if (!net_eq(sock_net(sk), net) ||
392 ipv6_only_sock(sk))
393 return -1;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000394
Joe Perches60c04ae2014-12-01 20:29:06 -0800395 inet = inet_sk(sk);
396
397 if (inet->inet_rcv_saddr != daddr ||
398 inet->inet_num != hnum)
399 return -1;
400
401 score = (sk->sk_family == PF_INET) ? 2 : 1;
402
403 if (inet->inet_daddr) {
404 if (inet->inet_daddr != saddr)
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000405 return -1;
Joe Perches60c04ae2014-12-01 20:29:06 -0800406 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000407 }
Joe Perches60c04ae2014-12-01 20:29:06 -0800408
409 if (inet->inet_dport) {
410 if (inet->inet_dport != sport)
411 return -1;
412 score += 4;
413 }
414
415 if (sk->sk_bound_dev_if) {
416 if (sk->sk_bound_dev_if != dif)
417 return -1;
418 score += 4;
419 }
420
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000421 return score;
422}
423
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200424static unsigned int udp_ehashfn(struct net *net, const __be32 laddr,
425 const __u16 lport, const __be32 faddr,
426 const __be16 fport)
427{
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200428 static u32 udp_ehash_secret __read_mostly;
429
430 net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret));
431
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200432 return __inet_ehashfn(laddr, lport, faddr, fport,
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200433 udp_ehash_secret + net_hash_mix(net));
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200434}
435
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000436
437/* called with read_rcu_lock() */
438static struct sock *udp4_lib_lookup2(struct net *net,
439 __be32 saddr, __be16 sport,
440 __be32 daddr, unsigned int hnum, int dif,
441 struct udp_hslot *hslot2, unsigned int slot2)
442{
443 struct sock *sk, *result;
444 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000445 int score, badness, matches = 0, reuseport = 0;
446 u32 hash = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000447
448begin:
449 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000450 badness = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000451 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
452 score = compute_score2(sk, net, saddr, sport,
453 daddr, hnum, dif);
454 if (score > badness) {
455 result = sk;
456 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000457 reuseport = sk->sk_reuseport;
458 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200459 hash = udp_ehashfn(net, daddr, hnum,
460 saddr, sport);
Tom Herbertba418fa2013-01-22 09:50:32 +0000461 matches = 1;
462 }
463 } else if (score == badness && reuseport) {
464 matches++;
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200465 if (reciprocal_scale(hash, matches) == 0)
Tom Herbertba418fa2013-01-22 09:50:32 +0000466 result = sk;
467 hash = next_pseudo_random32(hash);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000468 }
469 }
470 /*
471 * if the nulls value we got at the end of this lookup is
472 * not the expected one, we must restart lookup.
473 * We probably met an item that was moved to another chain.
474 */
475 if (get_nulls_value(node) != slot2)
476 goto begin;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000477 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000478 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000479 result = NULL;
480 else if (unlikely(compute_score2(result, net, saddr, sport,
481 daddr, hnum, dif) < badness)) {
482 sock_put(result);
483 goto begin;
484 }
485 }
486 return result;
487}
488
David S. Millerdb8dac22008-03-06 16:22:02 -0800489/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
490 * harder than this. -DaveM
491 */
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000492struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -0800493 __be16 sport, __be32 daddr, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700494 int dif, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800495{
Eric Dumazet271b72c2008-10-29 02:11:14 -0700496 struct sock *sk, *result;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800497 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800498 unsigned short hnum = ntohs(dport);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000499 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
500 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
Tom Herbertba418fa2013-01-22 09:50:32 +0000501 int score, badness, matches = 0, reuseport = 0;
502 u32 hash = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800503
Eric Dumazet271b72c2008-10-29 02:11:14 -0700504 rcu_read_lock();
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000505 if (hslot->count > 10) {
506 hash2 = udp4_portaddr_hash(net, daddr, hnum);
507 slot2 = hash2 & udptable->mask;
508 hslot2 = &udptable->hash2[slot2];
509 if (hslot->count < hslot2->count)
510 goto begin;
511
512 result = udp4_lib_lookup2(net, saddr, sport,
513 daddr, hnum, dif,
514 hslot2, slot2);
515 if (!result) {
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700516 hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000517 slot2 = hash2 & udptable->mask;
518 hslot2 = &udptable->hash2[slot2];
519 if (hslot->count < hslot2->count)
520 goto begin;
521
Jorge Boncompte [DTI2]1223c672010-04-08 04:56:48 +0000522 result = udp4_lib_lookup2(net, saddr, sport,
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700523 htonl(INADDR_ANY), hnum, dif,
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000524 hslot2, slot2);
525 }
526 rcu_read_unlock();
527 return result;
528 }
Eric Dumazet271b72c2008-10-29 02:11:14 -0700529begin:
530 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000531 badness = 0;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800532 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
Eric Dumazet645ca702008-10-29 01:41:45 -0700533 score = compute_score(sk, net, saddr, hnum, sport,
534 daddr, dport, dif);
535 if (score > badness) {
536 result = sk;
537 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000538 reuseport = sk->sk_reuseport;
539 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200540 hash = udp_ehashfn(net, daddr, hnum,
541 saddr, sport);
Tom Herbertba418fa2013-01-22 09:50:32 +0000542 matches = 1;
543 }
544 } else if (score == badness && reuseport) {
545 matches++;
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200546 if (reciprocal_scale(hash, matches) == 0)
Tom Herbertba418fa2013-01-22 09:50:32 +0000547 result = sk;
548 hash = next_pseudo_random32(hash);
David S. Millerdb8dac22008-03-06 16:22:02 -0800549 }
550 }
Eric Dumazet88ab1932008-11-16 19:39:21 -0800551 /*
552 * if the nulls value we got at the end of this lookup is
553 * not the expected one, we must restart lookup.
554 * We probably met an item that was moved to another chain.
555 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000556 if (get_nulls_value(node) != slot)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800557 goto begin;
558
Eric Dumazet271b72c2008-10-29 02:11:14 -0700559 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000560 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet271b72c2008-10-29 02:11:14 -0700561 result = NULL;
562 else if (unlikely(compute_score(result, net, saddr, hnum, sport,
563 daddr, dport, dif) < badness)) {
564 sock_put(result);
565 goto begin;
566 }
567 }
568 rcu_read_unlock();
David S. Millerdb8dac22008-03-06 16:22:02 -0800569 return result;
570}
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000571EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
David S. Millerdb8dac22008-03-06 16:22:02 -0800572
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700573static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
574 __be16 sport, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700575 struct udp_table *udptable)
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700576{
577 const struct iphdr *iph = ip_hdr(skb);
578
Eric Dumazet8afdd992013-12-10 18:07:23 -0800579 return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport,
580 iph->daddr, dport, inet_iif(skb),
581 udptable);
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700582}
583
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700584struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
585 __be32 daddr, __be16 dport, int dif)
586{
Eric Dumazet645ca702008-10-29 01:41:45 -0700587 return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700588}
589EXPORT_SYMBOL_GPL(udp4_lib_lookup);
590
Shawn Bohrer421b3882013-10-07 11:01:39 -0500591static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk,
592 __be16 loc_port, __be32 loc_addr,
593 __be16 rmt_port, __be32 rmt_addr,
594 int dif, unsigned short hnum)
595{
596 struct inet_sock *inet = inet_sk(sk);
597
598 if (!net_eq(sock_net(sk), net) ||
599 udp_sk(sk)->udp_port_hash != hnum ||
600 (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
601 (inet->inet_dport != rmt_port && inet->inet_dport) ||
602 (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) ||
603 ipv6_only_sock(sk) ||
604 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
605 return false;
606 if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif))
607 return false;
608 return true;
609}
610
David S. Millerdb8dac22008-03-06 16:22:02 -0800611/*
612 * This routine is called by the ICMP module when it gets some
613 * sort of error condition. If err < 0 then the socket should
614 * be closed and the error returned to the user. If err > 0
615 * it's just the icmp type << 8 | icmp code.
616 * Header points to the ip header of the error packet. We move
617 * on past this. Then (as it used to claim before adjustment)
618 * header points to the first 8 bytes of the udp header. We need
619 * to find the appropriate port.
620 */
621
Eric Dumazet645ca702008-10-29 01:41:45 -0700622void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800623{
624 struct inet_sock *inet;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000625 const struct iphdr *iph = (const struct iphdr *)skb->data;
Eric Dumazetc482c562009-07-17 00:26:32 +0000626 struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
David S. Millerdb8dac22008-03-06 16:22:02 -0800627 const int type = icmp_hdr(skb)->type;
628 const int code = icmp_hdr(skb)->code;
629 struct sock *sk;
630 int harderr;
631 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700632 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800633
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700634 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
David S. Millerdb8dac22008-03-06 16:22:02 -0800635 iph->saddr, uh->source, skb->dev->ifindex, udptable);
636 if (sk == NULL) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700637 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800638 return; /* No socket for error */
639 }
640
641 err = 0;
642 harderr = 0;
643 inet = inet_sk(sk);
644
645 switch (type) {
646 default:
647 case ICMP_TIME_EXCEEDED:
648 err = EHOSTUNREACH;
649 break;
650 case ICMP_SOURCE_QUENCH:
651 goto out;
652 case ICMP_PARAMETERPROB:
653 err = EPROTO;
654 harderr = 1;
655 break;
656 case ICMP_DEST_UNREACH:
657 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
David S. Miller36393392012-06-14 22:21:46 -0700658 ipv4_sk_update_pmtu(skb, sk, info);
David S. Millerdb8dac22008-03-06 16:22:02 -0800659 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
660 err = EMSGSIZE;
661 harderr = 1;
662 break;
663 }
664 goto out;
665 }
666 err = EHOSTUNREACH;
667 if (code <= NR_ICMP_UNREACH) {
668 harderr = icmp_err_convert[code].fatal;
669 err = icmp_err_convert[code].errno;
670 }
671 break;
David S. Miller55be7a92012-07-11 21:27:49 -0700672 case ICMP_REDIRECT:
673 ipv4_sk_redirect(skb, sk);
Duan Jiong1a462d12013-09-20 18:20:28 +0800674 goto out;
David S. Millerdb8dac22008-03-06 16:22:02 -0800675 }
676
677 /*
678 * RFC1122: OK. Passes ICMP errors back to application, as per
679 * 4.1.3.3.
680 */
681 if (!inet->recverr) {
682 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
683 goto out;
Eric Dumazetb1faf562010-05-31 23:44:05 -0700684 } else
Eric Dumazetc482c562009-07-17 00:26:32 +0000685 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
Eric Dumazetb1faf562010-05-31 23:44:05 -0700686
David S. Millerdb8dac22008-03-06 16:22:02 -0800687 sk->sk_err = err;
688 sk->sk_error_report(sk);
689out:
690 sock_put(sk);
691}
692
693void udp_err(struct sk_buff *skb, u32 info)
694{
Eric Dumazet645ca702008-10-29 01:41:45 -0700695 __udp4_lib_err(skb, info, &udp_table);
David S. Millerdb8dac22008-03-06 16:22:02 -0800696}
697
698/*
699 * Throw away all pending data and cancel the corking. Socket is locked.
700 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400701void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800702{
703 struct udp_sock *up = udp_sk(sk);
704
705 if (up->pending) {
706 up->len = 0;
707 up->pending = 0;
708 ip_flush_pending_frames(sk);
709 }
710}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400711EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800712
713/**
Herbert Xuf6b96642011-03-01 02:36:48 +0000714 * udp4_hwcsum - handle outgoing HW checksumming
David S. Millerdb8dac22008-03-06 16:22:02 -0800715 * @skb: sk_buff containing the filled-in UDP header
716 * (checksum field must be zeroed out)
Herbert Xuf6b96642011-03-01 02:36:48 +0000717 * @src: source IP address
718 * @dst: destination IP address
David S. Millerdb8dac22008-03-06 16:22:02 -0800719 */
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200720void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
David S. Millerdb8dac22008-03-06 16:22:02 -0800721{
David S. Millerdb8dac22008-03-06 16:22:02 -0800722 struct udphdr *uh = udp_hdr(skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000723 int offset = skb_transport_offset(skb);
724 int len = skb->len - offset;
725 int hlen = len;
David S. Millerdb8dac22008-03-06 16:22:02 -0800726 __wsum csum = 0;
727
WANG Congebbe4952014-06-02 16:12:02 -0700728 if (!skb_has_frag_list(skb)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800729 /*
730 * Only one fragment on the socket.
731 */
732 skb->csum_start = skb_transport_header(skb) - skb->head;
733 skb->csum_offset = offsetof(struct udphdr, check);
Herbert Xuf6b96642011-03-01 02:36:48 +0000734 uh->check = ~csum_tcpudp_magic(src, dst, len,
735 IPPROTO_UDP, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800736 } else {
WANG Congebbe4952014-06-02 16:12:02 -0700737 struct sk_buff *frags;
738
David S. Millerdb8dac22008-03-06 16:22:02 -0800739 /*
740 * HW-checksum won't work as there are two or more
741 * fragments on the socket so that all csums of sk_buffs
742 * should be together
743 */
WANG Congebbe4952014-06-02 16:12:02 -0700744 skb_walk_frags(skb, frags) {
Herbert Xuf6b96642011-03-01 02:36:48 +0000745 csum = csum_add(csum, frags->csum);
746 hlen -= frags->len;
WANG Congebbe4952014-06-02 16:12:02 -0700747 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800748
Herbert Xuf6b96642011-03-01 02:36:48 +0000749 csum = skb_checksum(skb, offset, hlen, csum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800750 skb->ip_summed = CHECKSUM_NONE;
751
David S. Millerdb8dac22008-03-06 16:22:02 -0800752 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
753 if (uh->check == 0)
754 uh->check = CSUM_MANGLED_0;
755 }
756}
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200757EXPORT_SYMBOL_GPL(udp4_hwcsum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800758
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700759/* Function to set UDP checksum for an IPv4 UDP packet. This is intended
760 * for the simple case like when setting the checksum for a UDP tunnel.
761 */
762void udp_set_csum(bool nocheck, struct sk_buff *skb,
763 __be32 saddr, __be32 daddr, int len)
764{
765 struct udphdr *uh = udp_hdr(skb);
766
767 if (nocheck)
768 uh->check = 0;
769 else if (skb_is_gso(skb))
770 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
771 else if (skb_dst(skb) && skb_dst(skb)->dev &&
772 (skb_dst(skb)->dev->features & NETIF_F_V4_CSUM)) {
773
774 BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL);
775
776 skb->ip_summed = CHECKSUM_PARTIAL;
777 skb->csum_start = skb_transport_header(skb) - skb->head;
778 skb->csum_offset = offsetof(struct udphdr, check);
779 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
780 } else {
781 __wsum csum;
782
783 BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL);
784
785 uh->check = 0;
786 csum = skb_checksum(skb, 0, len, 0);
787 uh->check = udp_v4_check(len, saddr, daddr, csum);
788 if (uh->check == 0)
789 uh->check = CSUM_MANGLED_0;
790
791 skb->ip_summed = CHECKSUM_UNNECESSARY;
792 }
793}
794EXPORT_SYMBOL(udp_set_csum);
795
David S. Miller79ab0532011-05-09 13:31:04 -0700796static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
Herbert Xuf6b96642011-03-01 02:36:48 +0000797{
798 struct sock *sk = skb->sk;
799 struct inet_sock *inet = inet_sk(sk);
800 struct udphdr *uh;
Herbert Xuf6b96642011-03-01 02:36:48 +0000801 int err = 0;
802 int is_udplite = IS_UDPLITE(sk);
803 int offset = skb_transport_offset(skb);
804 int len = skb->len - offset;
805 __wsum csum = 0;
806
807 /*
808 * Create a UDP header
809 */
810 uh = udp_hdr(skb);
811 uh->source = inet->inet_sport;
David S. Miller79ab0532011-05-09 13:31:04 -0700812 uh->dest = fl4->fl4_dport;
Herbert Xuf6b96642011-03-01 02:36:48 +0000813 uh->len = htons(len);
814 uh->check = 0;
815
816 if (is_udplite) /* UDP-Lite */
817 csum = udplite_csum(skb);
818
Tom Herbert28448b82014-05-23 08:47:19 -0700819 else if (sk->sk_no_check_tx) { /* UDP csum disabled */
Herbert Xuf6b96642011-03-01 02:36:48 +0000820
821 skb->ip_summed = CHECKSUM_NONE;
822 goto send;
823
824 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
825
David S. Miller79ab0532011-05-09 13:31:04 -0700826 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
Herbert Xuf6b96642011-03-01 02:36:48 +0000827 goto send;
828
829 } else
830 csum = udp_csum(skb);
831
832 /* add protocol-dependent pseudo-header */
David S. Miller79ab0532011-05-09 13:31:04 -0700833 uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
Herbert Xuf6b96642011-03-01 02:36:48 +0000834 sk->sk_protocol, csum);
835 if (uh->check == 0)
836 uh->check = CSUM_MANGLED_0;
837
838send:
Eric Dumazetb5ec8ee2012-08-10 02:22:47 +0000839 err = ip_send_skb(sock_net(sk), skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000840 if (err) {
841 if (err == -ENOBUFS && !inet->recverr) {
842 UDP_INC_STATS_USER(sock_net(sk),
843 UDP_MIB_SNDBUFERRORS, is_udplite);
844 err = 0;
845 }
846 } else
847 UDP_INC_STATS_USER(sock_net(sk),
848 UDP_MIB_OUTDATAGRAMS, is_udplite);
849 return err;
850}
851
David S. Millerdb8dac22008-03-06 16:22:02 -0800852/*
853 * Push out all pending data as one UDP datagram. Socket is locked.
854 */
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200855int udp_push_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800856{
857 struct udp_sock *up = udp_sk(sk);
858 struct inet_sock *inet = inet_sk(sk);
David S. Millerb6f21b22011-03-12 02:09:18 -0500859 struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800860 struct sk_buff *skb;
David S. Millerdb8dac22008-03-06 16:22:02 -0800861 int err = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800862
David S. Miller77968b72011-05-08 17:12:19 -0700863 skb = ip_finish_skb(sk, fl4);
Herbert Xuf6b96642011-03-01 02:36:48 +0000864 if (!skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800865 goto out;
866
David S. Miller79ab0532011-05-09 13:31:04 -0700867 err = udp_send_skb(skb, fl4);
David S. Millerdb8dac22008-03-06 16:22:02 -0800868
David S. Millerdb8dac22008-03-06 16:22:02 -0800869out:
870 up->len = 0;
871 up->pending = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800872 return err;
873}
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200874EXPORT_SYMBOL(udp_push_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800875
876int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
877 size_t len)
878{
879 struct inet_sock *inet = inet_sk(sk);
880 struct udp_sock *up = udp_sk(sk);
David S. Millere4749952011-05-08 16:38:45 -0700881 struct flowi4 fl4_stack;
David S. Millerb6f21b22011-03-12 02:09:18 -0500882 struct flowi4 *fl4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800883 int ulen = len;
884 struct ipcm_cookie ipc;
885 struct rtable *rt = NULL;
886 int free = 0;
887 int connected = 0;
888 __be32 daddr, faddr, saddr;
889 __be16 dport;
890 u8 tos;
891 int err, is_udplite = IS_UDPLITE(sk);
892 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
893 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
Herbert Xu903ab862011-03-01 02:36:48 +0000894 struct sk_buff *skb;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000895 struct ip_options_data opt_copy;
David S. Millerdb8dac22008-03-06 16:22:02 -0800896
897 if (len > 0xFFFF)
898 return -EMSGSIZE;
899
900 /*
901 * Check the flags.
902 */
903
Eric Dumazetc482c562009-07-17 00:26:32 +0000904 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
David S. Millerdb8dac22008-03-06 16:22:02 -0800905 return -EOPNOTSUPP;
906
907 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000908 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200909 ipc.ttl = 0;
910 ipc.tos = -1;
David S. Millerdb8dac22008-03-06 16:22:02 -0800911
Herbert Xu903ab862011-03-01 02:36:48 +0000912 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
913
David S. Millerf5fca602011-05-08 17:24:10 -0700914 fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800915 if (up->pending) {
916 /*
917 * There are pending frames.
918 * The socket lock must be held while it's corked.
919 */
920 lock_sock(sk);
921 if (likely(up->pending)) {
922 if (unlikely(up->pending != AF_INET)) {
923 release_sock(sk);
924 return -EINVAL;
925 }
926 goto do_append_data;
927 }
928 release_sock(sk);
929 }
930 ulen += sizeof(struct udphdr);
931
932 /*
933 * Get and verify the address.
934 */
935 if (msg->msg_name) {
Steffen Hurrle342dfc32014-01-17 22:53:15 +0100936 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -0800937 if (msg->msg_namelen < sizeof(*usin))
938 return -EINVAL;
939 if (usin->sin_family != AF_INET) {
940 if (usin->sin_family != AF_UNSPEC)
941 return -EAFNOSUPPORT;
942 }
943
944 daddr = usin->sin_addr.s_addr;
945 dport = usin->sin_port;
946 if (dport == 0)
947 return -EINVAL;
948 } else {
949 if (sk->sk_state != TCP_ESTABLISHED)
950 return -EDESTADDRREQ;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000951 daddr = inet->inet_daddr;
952 dport = inet->inet_dport;
David S. Millerdb8dac22008-03-06 16:22:02 -0800953 /* Open fast path for connected socket.
954 Route will not be used, if at least one option is set.
955 */
956 connected = 1;
957 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000958 ipc.addr = inet->inet_saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800959
960 ipc.oif = sk->sk_bound_dev_if;
Daniel Borkmannbf84a012013-04-14 08:08:13 +0000961
962 sock_tx_timestamp(sk, &ipc.tx_flags);
963
David S. Millerdb8dac22008-03-06 16:22:02 -0800964 if (msg->msg_controllen) {
Hannes Frederic Sowac8e6ad02014-02-18 21:38:08 +0100965 err = ip_cmsg_send(sock_net(sk), msg, &ipc,
966 sk->sk_family == AF_INET6);
David S. Millerdb8dac22008-03-06 16:22:02 -0800967 if (err)
968 return err;
969 if (ipc.opt)
970 free = 1;
971 connected = 0;
972 }
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000973 if (!ipc.opt) {
974 struct ip_options_rcu *inet_opt;
975
976 rcu_read_lock();
977 inet_opt = rcu_dereference(inet->inet_opt);
978 if (inet_opt) {
979 memcpy(&opt_copy, inet_opt,
980 sizeof(*inet_opt) + inet_opt->opt.optlen);
981 ipc.opt = &opt_copy.opt;
982 }
983 rcu_read_unlock();
984 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800985
986 saddr = ipc.addr;
987 ipc.addr = faddr = daddr;
988
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000989 if (ipc.opt && ipc.opt->opt.srr) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800990 if (!daddr)
991 return -EINVAL;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000992 faddr = ipc.opt->opt.faddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800993 connected = 0;
994 }
Francesco Fuscoaa661582013-09-24 15:43:09 +0200995 tos = get_rttos(&ipc, inet);
David S. Millerdb8dac22008-03-06 16:22:02 -0800996 if (sock_flag(sk, SOCK_LOCALROUTE) ||
997 (msg->msg_flags & MSG_DONTROUTE) ||
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000998 (ipc.opt && ipc.opt->opt.is_strictroute)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800999 tos |= RTO_ONLINK;
1000 connected = 0;
1001 }
1002
1003 if (ipv4_is_multicast(daddr)) {
1004 if (!ipc.oif)
1005 ipc.oif = inet->mc_index;
1006 if (!saddr)
1007 saddr = inet->mc_addr;
1008 connected = 0;
Erich E. Hoover76e21052012-02-08 09:11:07 +00001009 } else if (!ipc.oif)
1010 ipc.oif = inet->uc_index;
David S. Millerdb8dac22008-03-06 16:22:02 -08001011
1012 if (connected)
Eric Dumazetc482c562009-07-17 00:26:32 +00001013 rt = (struct rtable *)sk_dst_check(sk, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -08001014
1015 if (rt == NULL) {
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -07001016 struct net *net = sock_net(sk);
1017
David S. Millere4749952011-05-08 16:38:45 -07001018 fl4 = &fl4_stack;
1019 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
David S. Millerc0951cb2011-03-31 04:54:27 -07001020 RT_SCOPE_UNIVERSE, sk->sk_protocol,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001021 inet_sk_flowi_flags(sk),
David S. Millerc0951cb2011-03-31 04:54:27 -07001022 faddr, saddr, dport, inet->inet_sport);
1023
David S. Millere4749952011-05-08 16:38:45 -07001024 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
1025 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001026 if (IS_ERR(rt)) {
1027 err = PTR_ERR(rt);
David S. Miller06dc94b2011-03-03 10:38:01 -08001028 rt = NULL;
David S. Millerdb8dac22008-03-06 16:22:02 -08001029 if (err == -ENETUNREACH)
Eric Dumazetf1d8cba2013-11-28 09:51:22 -08001030 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -08001031 goto out;
1032 }
1033
1034 err = -EACCES;
1035 if ((rt->rt_flags & RTCF_BROADCAST) &&
1036 !sock_flag(sk, SOCK_BROADCAST))
1037 goto out;
1038 if (connected)
Changli Gaod8d1f302010-06-10 23:31:35 -07001039 sk_dst_set(sk, dst_clone(&rt->dst));
David S. Millerdb8dac22008-03-06 16:22:02 -08001040 }
1041
1042 if (msg->msg_flags&MSG_CONFIRM)
1043 goto do_confirm;
1044back_from_confirm:
1045
David S. Millere4749952011-05-08 16:38:45 -07001046 saddr = fl4->saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001047 if (!ipc.addr)
David S. Millere4749952011-05-08 16:38:45 -07001048 daddr = ipc.addr = fl4->daddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001049
Herbert Xu903ab862011-03-01 02:36:48 +00001050 /* Lockless fast path for the non-corking case. */
1051 if (!corkreq) {
Al Virof69e6d12014-11-24 13:23:40 -05001052 skb = ip_make_skb(sk, fl4, getfrag, msg, ulen,
Herbert Xu903ab862011-03-01 02:36:48 +00001053 sizeof(struct udphdr), &ipc, &rt,
1054 msg->msg_flags);
1055 err = PTR_ERR(skb);
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001056 if (!IS_ERR_OR_NULL(skb))
David S. Miller79ab0532011-05-09 13:31:04 -07001057 err = udp_send_skb(skb, fl4);
Herbert Xu903ab862011-03-01 02:36:48 +00001058 goto out;
1059 }
1060
David S. Millerdb8dac22008-03-06 16:22:02 -08001061 lock_sock(sk);
1062 if (unlikely(up->pending)) {
1063 /* The socket is already corked while preparing it. */
1064 /* ... which is an evident application bug. --ANK */
1065 release_sock(sk);
1066
Joe Perchesba7a46f2014-11-11 10:59:17 -08001067 net_dbg_ratelimited("cork app bug 2\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001068 err = -EINVAL;
1069 goto out;
1070 }
1071 /*
1072 * Now cork the socket to pend data.
1073 */
David S. Millerb6f21b22011-03-12 02:09:18 -05001074 fl4 = &inet->cork.fl.u.ip4;
1075 fl4->daddr = daddr;
1076 fl4->saddr = saddr;
David S. Miller9cce96d2011-03-12 03:00:33 -05001077 fl4->fl4_dport = dport;
1078 fl4->fl4_sport = inet->inet_sport;
David S. Millerdb8dac22008-03-06 16:22:02 -08001079 up->pending = AF_INET;
1080
1081do_append_data:
1082 up->len += ulen;
Al Virof69e6d12014-11-24 13:23:40 -05001083 err = ip_append_data(sk, fl4, getfrag, msg, ulen,
David S. Millerf5fca602011-05-08 17:24:10 -07001084 sizeof(struct udphdr), &ipc, &rt,
1085 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001086 if (err)
1087 udp_flush_pending_frames(sk);
1088 else if (!corkreq)
1089 err = udp_push_pending_frames(sk);
1090 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1091 up->pending = 0;
1092 release_sock(sk);
1093
1094out:
1095 ip_rt_put(rt);
1096 if (free)
1097 kfree(ipc.opt);
1098 if (!err)
1099 return len;
1100 /*
1101 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1102 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1103 * we don't have a good statistic (IpOutDiscards but it can be too many
1104 * things). We could add another new stat but at least for now that
1105 * seems like overkill.
1106 */
1107 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001108 UDP_INC_STATS_USER(sock_net(sk),
1109 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001110 }
1111 return err;
1112
1113do_confirm:
Changli Gaod8d1f302010-06-10 23:31:35 -07001114 dst_confirm(&rt->dst);
David S. Millerdb8dac22008-03-06 16:22:02 -08001115 if (!(msg->msg_flags&MSG_PROBE) || len)
1116 goto back_from_confirm;
1117 err = 0;
1118 goto out;
1119}
Eric Dumazetc482c562009-07-17 00:26:32 +00001120EXPORT_SYMBOL(udp_sendmsg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001121
1122int udp_sendpage(struct sock *sk, struct page *page, int offset,
1123 size_t size, int flags)
1124{
David S. Millerf5fca602011-05-08 17:24:10 -07001125 struct inet_sock *inet = inet_sk(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001126 struct udp_sock *up = udp_sk(sk);
1127 int ret;
1128
Shawn Landdend3f7d562013-11-24 22:36:28 -08001129 if (flags & MSG_SENDPAGE_NOTLAST)
1130 flags |= MSG_MORE;
1131
David S. Millerdb8dac22008-03-06 16:22:02 -08001132 if (!up->pending) {
1133 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
1134
1135 /* Call udp_sendmsg to specify destination address which
1136 * sendpage interface can't pass.
1137 * This will succeed only when the socket is connected.
1138 */
1139 ret = udp_sendmsg(NULL, sk, &msg, 0);
1140 if (ret < 0)
1141 return ret;
1142 }
1143
1144 lock_sock(sk);
1145
1146 if (unlikely(!up->pending)) {
1147 release_sock(sk);
1148
Joe Perchesba7a46f2014-11-11 10:59:17 -08001149 net_dbg_ratelimited("udp cork app bug 3\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001150 return -EINVAL;
1151 }
1152
David S. Millerf5fca602011-05-08 17:24:10 -07001153 ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1154 page, offset, size, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001155 if (ret == -EOPNOTSUPP) {
1156 release_sock(sk);
1157 return sock_no_sendpage(sk->sk_socket, page, offset,
1158 size, flags);
1159 }
1160 if (ret < 0) {
1161 udp_flush_pending_frames(sk);
1162 goto out;
1163 }
1164
1165 up->len += size;
1166 if (!(up->corkflag || (flags&MSG_MORE)))
1167 ret = udp_push_pending_frames(sk);
1168 if (!ret)
1169 ret = size;
1170out:
1171 release_sock(sk);
1172 return ret;
1173}
1174
Eric Dumazet85584672009-10-09 04:43:40 +00001175
1176/**
1177 * first_packet_length - return length of first packet in receive queue
1178 * @sk: socket
1179 *
1180 * Drops all bad checksum frames, until a valid one is found.
1181 * Returns the length of found skb, or 0 if none is found.
1182 */
1183static unsigned int first_packet_length(struct sock *sk)
1184{
1185 struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1186 struct sk_buff *skb;
1187 unsigned int res;
1188
1189 __skb_queue_head_init(&list_kill);
1190
1191 spin_lock_bh(&rcvq->lock);
1192 while ((skb = skb_peek(rcvq)) != NULL &&
1193 udp_lib_checksum_complete(skb)) {
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001194 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS,
1195 IS_UDPLITE(sk));
Eric Dumazet85584672009-10-09 04:43:40 +00001196 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1197 IS_UDPLITE(sk));
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001198 atomic_inc(&sk->sk_drops);
Eric Dumazet85584672009-10-09 04:43:40 +00001199 __skb_unlink(skb, rcvq);
1200 __skb_queue_tail(&list_kill, skb);
1201 }
1202 res = skb ? skb->len : 0;
1203 spin_unlock_bh(&rcvq->lock);
1204
1205 if (!skb_queue_empty(&list_kill)) {
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001206 bool slow = lock_sock_fast(sk);
1207
Eric Dumazet85584672009-10-09 04:43:40 +00001208 __skb_queue_purge(&list_kill);
1209 sk_mem_reclaim_partial(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001210 unlock_sock_fast(sk, slow);
Eric Dumazet85584672009-10-09 04:43:40 +00001211 }
1212 return res;
1213}
1214
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215/*
1216 * IOCTL requests applicable to the UDP protocol
1217 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001218
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1220{
Stephen Hemminger6516c652007-03-08 20:41:55 -08001221 switch (cmd) {
1222 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223 {
Eric Dumazet31e6d362009-06-17 19:05:41 -07001224 int amount = sk_wmem_alloc_get(sk);
1225
Stephen Hemminger6516c652007-03-08 20:41:55 -08001226 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 }
Stephen Hemminger6516c652007-03-08 20:41:55 -08001228
1229 case SIOCINQ:
1230 {
Eric Dumazet85584672009-10-09 04:43:40 +00001231 unsigned int amount = first_packet_length(sk);
Stephen Hemminger6516c652007-03-08 20:41:55 -08001232
Eric Dumazet85584672009-10-09 04:43:40 +00001233 if (amount)
Stephen Hemminger6516c652007-03-08 20:41:55 -08001234 /*
1235 * We will only return the amount
1236 * of this packet since that is all
1237 * that will be read.
1238 */
Eric Dumazet85584672009-10-09 04:43:40 +00001239 amount -= sizeof(struct udphdr);
1240
Stephen Hemminger6516c652007-03-08 20:41:55 -08001241 return put_user(amount, (int __user *)arg);
1242 }
1243
1244 default:
1245 return -ENOIOCTLCMD;
1246 }
1247
1248 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249}
Eric Dumazetc482c562009-07-17 00:26:32 +00001250EXPORT_SYMBOL(udp_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251
David S. Millerdb8dac22008-03-06 16:22:02 -08001252/*
1253 * This should be easy, if there is something there we
1254 * return it, otherwise we block.
1255 */
1256
1257int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
1258 size_t len, int noblock, int flags, int *addr_len)
1259{
1260 struct inet_sock *inet = inet_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001261 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -08001262 struct sk_buff *skb;
David S. Miller59c2cda2011-12-01 14:12:55 -05001263 unsigned int ulen, copied;
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001264 int peeked, off = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001265 int err;
1266 int is_udplite = IS_UDPLITE(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001267 bool slow;
David S. Millerdb8dac22008-03-06 16:22:02 -08001268
David S. Millerdb8dac22008-03-06 16:22:02 -08001269 if (flags & MSG_ERRQUEUE)
Hannes Frederic Sowa85fbaa72013-11-23 00:46:12 +01001270 return ip_recv_error(sk, msg, len, addr_len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001271
1272try_again:
1273 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001274 &peeked, &off, &err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001275 if (!skb)
1276 goto out;
1277
1278 ulen = skb->len - sizeof(struct udphdr);
David S. Miller59c2cda2011-12-01 14:12:55 -05001279 copied = len;
1280 if (copied > ulen)
1281 copied = ulen;
1282 else if (copied < ulen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001283 msg->msg_flags |= MSG_TRUNC;
1284
1285 /*
1286 * If checksum is needed at all, try to do it while copying the
1287 * data. If the data is truncated, or if we only want a partial
1288 * coverage checksum (UDP-Lite), do it before the copy.
1289 */
1290
David S. Miller59c2cda2011-12-01 14:12:55 -05001291 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001292 if (udp_lib_checksum_complete(skb))
1293 goto csum_copy_err;
1294 }
1295
1296 if (skb_csum_unnecessary(skb))
David S. Miller51f3d022014-11-05 16:46:40 -05001297 err = skb_copy_datagram_msg(skb, sizeof(struct udphdr),
1298 msg, copied);
David S. Millerdb8dac22008-03-06 16:22:02 -08001299 else {
Al Viro227158d2014-04-06 18:47:38 -04001300 err = skb_copy_and_csum_datagram_msg(skb, sizeof(struct udphdr),
1301 msg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001302
1303 if (err == -EINVAL)
1304 goto csum_copy_err;
1305 }
1306
Eric Dumazet22911fc2012-06-27 00:23:44 +00001307 if (unlikely(err)) {
1308 trace_kfree_skb(skb, udp_recvmsg);
Eric Dumazet979402b2012-09-05 23:34:44 +00001309 if (!peeked) {
1310 atomic_inc(&sk->sk_drops);
1311 UDP_INC_STATS_USER(sock_net(sk),
1312 UDP_MIB_INERRORS, is_udplite);
1313 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001314 goto out_free;
Eric Dumazet22911fc2012-06-27 00:23:44 +00001315 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001316
1317 if (!peeked)
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001318 UDP_INC_STATS_USER(sock_net(sk),
1319 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001320
Neil Horman3b885782009-10-12 13:26:31 -07001321 sock_recv_ts_and_drops(msg, sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001322
1323 /* Copy the address. */
Eric Dumazetc482c562009-07-17 00:26:32 +00001324 if (sin) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001325 sin->sin_family = AF_INET;
1326 sin->sin_port = udp_hdr(skb)->source;
1327 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1328 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
Hannes Frederic Sowabceaa902013-11-18 04:20:45 +01001329 *addr_len = sizeof(*sin);
David S. Millerdb8dac22008-03-06 16:22:02 -08001330 }
1331 if (inet->cmsg_flags)
1332 ip_cmsg_recv(msg, skb);
1333
David S. Miller59c2cda2011-12-01 14:12:55 -05001334 err = copied;
David S. Millerdb8dac22008-03-06 16:22:02 -08001335 if (flags & MSG_TRUNC)
1336 err = ulen;
1337
1338out_free:
Eric Dumazet9d410c72009-10-30 05:03:53 +00001339 skb_free_datagram_locked(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001340out:
1341 return err;
1342
1343csum_copy_err:
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001344 slow = lock_sock_fast(sk);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001345 if (!skb_kill_datagram(sk, skb, flags)) {
1346 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001347 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001348 }
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001349 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001350
1351 if (noblock)
1352 return -EAGAIN;
Xufeng Zhang9cfaa8d2011-06-21 10:43:40 +00001353
1354 /* starting over for a new packet */
1355 msg->msg_flags &= ~MSG_TRUNC;
David S. Millerdb8dac22008-03-06 16:22:02 -08001356 goto try_again;
1357}
1358
1359
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360int udp_disconnect(struct sock *sk, int flags)
1361{
1362 struct inet_sock *inet = inet_sk(sk);
1363 /*
1364 * 1003.1g - break association.
1365 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001366
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 sk->sk_state = TCP_CLOSE;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001368 inet->inet_daddr = 0;
1369 inet->inet_dport = 0;
Tom Herbertbdeab992011-08-14 19:45:55 +00001370 sock_rps_reset_rxhash(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 sk->sk_bound_dev_if = 0;
1372 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1373 inet_reset_saddr(sk);
1374
1375 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1376 sk->sk_prot->unhash(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001377 inet->inet_sport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 }
1379 sk_dst_reset(sk);
1380 return 0;
1381}
Eric Dumazetc482c562009-07-17 00:26:32 +00001382EXPORT_SYMBOL(udp_disconnect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383
Eric Dumazet645ca702008-10-29 01:41:45 -07001384void udp_lib_unhash(struct sock *sk)
1385{
Eric Dumazet723b4612008-11-25 13:55:15 -08001386 if (sk_hashed(sk)) {
1387 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Eric Dumazet512615b2009-11-08 10:17:58 +00001388 struct udp_hslot *hslot, *hslot2;
1389
1390 hslot = udp_hashslot(udptable, sock_net(sk),
1391 udp_sk(sk)->udp_port_hash);
1392 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
Eric Dumazet645ca702008-10-29 01:41:45 -07001393
Eric Dumazet723b4612008-11-25 13:55:15 -08001394 spin_lock_bh(&hslot->lock);
1395 if (sk_nulls_del_node_init_rcu(sk)) {
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00001396 hslot->count--;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001397 inet_sk(sk)->inet_num = 0;
Eric Dumazet723b4612008-11-25 13:55:15 -08001398 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet512615b2009-11-08 10:17:58 +00001399
1400 spin_lock(&hslot2->lock);
1401 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1402 hslot2->count--;
1403 spin_unlock(&hslot2->lock);
Eric Dumazet723b4612008-11-25 13:55:15 -08001404 }
1405 spin_unlock_bh(&hslot->lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001406 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001407}
1408EXPORT_SYMBOL(udp_lib_unhash);
1409
Eric Dumazet719f8352010-09-08 05:08:44 +00001410/*
1411 * inet_rcv_saddr was changed, we must rehash secondary hash
1412 */
1413void udp_lib_rehash(struct sock *sk, u16 newhash)
1414{
1415 if (sk_hashed(sk)) {
1416 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1417 struct udp_hslot *hslot, *hslot2, *nhslot2;
1418
1419 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1420 nhslot2 = udp_hashslot2(udptable, newhash);
1421 udp_sk(sk)->udp_portaddr_hash = newhash;
1422 if (hslot2 != nhslot2) {
1423 hslot = udp_hashslot(udptable, sock_net(sk),
1424 udp_sk(sk)->udp_port_hash);
1425 /* we must lock primary chain too */
1426 spin_lock_bh(&hslot->lock);
1427
1428 spin_lock(&hslot2->lock);
1429 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1430 hslot2->count--;
1431 spin_unlock(&hslot2->lock);
1432
1433 spin_lock(&nhslot2->lock);
1434 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
1435 &nhslot2->head);
1436 nhslot2->count++;
1437 spin_unlock(&nhslot2->lock);
1438
1439 spin_unlock_bh(&hslot->lock);
1440 }
1441 }
1442}
1443EXPORT_SYMBOL(udp_lib_rehash);
1444
1445static void udp_v4_rehash(struct sock *sk)
1446{
1447 u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1448 inet_sk(sk)->inet_rcv_saddr,
1449 inet_sk(sk)->inet_num);
1450 udp_lib_rehash(sk, new_hash);
1451}
1452
Herbert Xu93821772008-09-15 11:48:46 -07001453static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1454{
Tom Herbertfec5e652010-04-16 16:01:27 -07001455 int rc;
Herbert Xu93821772008-09-15 11:48:46 -07001456
Shawn Bohrer005ec972013-10-07 11:01:38 -05001457 if (inet_sk(sk)->inet_daddr) {
Tom Herbertbdeab992011-08-14 19:45:55 +00001458 sock_rps_save_rxhash(sk, skb);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001459 sk_mark_napi_id(sk, skb);
Eric Dumazet2c8c56e2014-11-11 05:54:28 -08001460 sk_incoming_cpu_update(sk);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001461 }
Tom Herbertfec5e652010-04-16 16:01:27 -07001462
Eric Dumazetd826eb12011-11-09 07:24:35 +00001463 rc = sock_queue_rcv_skb(sk, skb);
Eric Dumazet766e90372009-10-14 20:40:11 -07001464 if (rc < 0) {
1465 int is_udplite = IS_UDPLITE(sk);
1466
Herbert Xu93821772008-09-15 11:48:46 -07001467 /* Note that an ENOMEM error is charged twice */
Eric Dumazet766e90372009-10-14 20:40:11 -07001468 if (rc == -ENOMEM)
Herbert Xu93821772008-09-15 11:48:46 -07001469 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1470 is_udplite);
Eric Dumazet766e90372009-10-14 20:40:11 -07001471 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1472 kfree_skb(skb);
Satoru Moriya296f7ea2011-06-17 11:58:39 +00001473 trace_udp_fail_queue_rcv_skb(rc, sk);
Eric Dumazet766e90372009-10-14 20:40:11 -07001474 return -1;
Herbert Xu93821772008-09-15 11:48:46 -07001475 }
1476
1477 return 0;
1478
Herbert Xu93821772008-09-15 11:48:46 -07001479}
1480
Eric Dumazet447167b2012-04-11 23:05:28 +00001481static struct static_key udp_encap_needed __read_mostly;
1482void udp_encap_enable(void)
1483{
1484 if (!static_key_enabled(&udp_encap_needed))
1485 static_key_slow_inc(&udp_encap_needed);
1486}
1487EXPORT_SYMBOL(udp_encap_enable);
1488
David S. Millerdb8dac22008-03-06 16:22:02 -08001489/* returns:
1490 * -1: error
1491 * 0: success
1492 * >0: "udp encap" protocol resubmission
1493 *
1494 * Note that in the success and error cases, the skb is assumed to
1495 * have either been requeued or freed.
1496 */
Eric Dumazetc482c562009-07-17 00:26:32 +00001497int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -08001498{
1499 struct udp_sock *up = udp_sk(sk);
1500 int rc;
1501 int is_udplite = IS_UDPLITE(sk);
1502
1503 /*
1504 * Charge it to the socket, dropping if the queue is full.
1505 */
1506 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1507 goto drop;
1508 nf_reset(skb);
1509
Eric Dumazet447167b2012-04-11 23:05:28 +00001510 if (static_key_false(&udp_encap_needed) && up->encap_type) {
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001511 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1512
David S. Millerdb8dac22008-03-06 16:22:02 -08001513 /*
1514 * This is an encapsulation socket so pass the skb to
1515 * the socket's udp_encap_rcv() hook. Otherwise, just
1516 * fall through and pass this up the UDP socket.
1517 * up->encap_rcv() returns the following value:
1518 * =0 if skb was successfully passed to the encap
1519 * handler or was discarded by it.
1520 * >0 if skb should be passed on to UDP.
1521 * <0 if skb should be resubmitted as proto -N
1522 */
1523
1524 /* if we're overly short, let UDP handle it */
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001525 encap_rcv = ACCESS_ONCE(up->encap_rcv);
1526 if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001527 int ret;
1528
Tom Herbert0a809662014-05-07 16:52:39 -07001529 /* Verify checksum before giving to encap */
1530 if (udp_lib_checksum_complete(skb))
1531 goto csum_error;
1532
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001533 ret = encap_rcv(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001534 if (ret <= 0) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001535 UDP_INC_STATS_BH(sock_net(sk),
1536 UDP_MIB_INDATAGRAMS,
David S. Millerdb8dac22008-03-06 16:22:02 -08001537 is_udplite);
1538 return -ret;
1539 }
1540 }
1541
1542 /* FALLTHROUGH -- it's a UDP Packet */
1543 }
1544
1545 /*
1546 * UDP-Lite specific tests, ignored on UDP sockets
1547 */
1548 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1549
1550 /*
1551 * MIB statistics other than incrementing the error count are
1552 * disabled for the following two types of errors: these depend
1553 * on the application settings, not on the functioning of the
1554 * protocol stack as such.
1555 *
1556 * RFC 3828 here recommends (sec 3.3): "There should also be a
1557 * way ... to ... at least let the receiving application block
1558 * delivery of packets with coverage values less than a value
1559 * provided by the application."
1560 */
1561 if (up->pcrlen == 0) { /* full coverage was set */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001562 net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n",
1563 UDP_SKB_CB(skb)->cscov, skb->len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001564 goto drop;
1565 }
1566 /* The next case involves violating the min. coverage requested
1567 * by the receiver. This is subtle: if receiver wants x and x is
1568 * greater than the buffersize/MTU then receiver will complain
1569 * that it wants x while sender emits packets of smaller size y.
1570 * Therefore the above ...()->partial_cov statement is essential.
1571 */
1572 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
Joe Perchesba7a46f2014-11-11 10:59:17 -08001573 net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n",
1574 UDP_SKB_CB(skb)->cscov, up->pcrlen);
David S. Millerdb8dac22008-03-06 16:22:02 -08001575 goto drop;
1576 }
1577 }
1578
Eric Dumazet33d480c2011-08-11 19:30:52 +00001579 if (rcu_access_pointer(sk->sk_filter) &&
1580 udp_lib_checksum_complete(skb))
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001581 goto csum_error;
David S. Millerdb8dac22008-03-06 16:22:02 -08001582
Eric Dumazetc3774112010-04-27 15:13:20 -07001583
Sorin Dumitru274f4822014-07-22 21:16:51 +03001584 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
James M Leddy3e215c82014-06-25 17:38:13 -04001585 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1586 is_udplite);
Eric Dumazetc3774112010-04-27 15:13:20 -07001587 goto drop;
James M Leddy3e215c82014-06-25 17:38:13 -04001588 }
Eric Dumazetc3774112010-04-27 15:13:20 -07001589
Herbert Xu93821772008-09-15 11:48:46 -07001590 rc = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001591
Shawn Bohrerfbf88662013-10-07 11:01:40 -05001592 ipv4_pktinfo_prepare(sk, skb);
Herbert Xu93821772008-09-15 11:48:46 -07001593 bh_lock_sock(sk);
1594 if (!sock_owned_by_user(sk))
1595 rc = __udp_queue_rcv_skb(sk, skb);
Eric Dumazetf545a382012-04-22 23:34:26 +00001596 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
Zhu Yi55349792010-03-04 18:01:42 +00001597 bh_unlock_sock(sk);
1598 goto drop;
1599 }
Herbert Xu93821772008-09-15 11:48:46 -07001600 bh_unlock_sock(sk);
1601
1602 return rc;
David S. Millerdb8dac22008-03-06 16:22:02 -08001603
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001604csum_error:
1605 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001606drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001607 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001608 atomic_inc(&sk->sk_drops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001609 kfree_skb(skb);
1610 return -1;
1611}
1612
Eric Dumazet1240d132009-11-08 10:18:44 +00001613
1614static void flush_stack(struct sock **stack, unsigned int count,
1615 struct sk_buff *skb, unsigned int final)
1616{
1617 unsigned int i;
1618 struct sk_buff *skb1 = NULL;
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001619 struct sock *sk;
Eric Dumazet1240d132009-11-08 10:18:44 +00001620
1621 for (i = 0; i < count; i++) {
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001622 sk = stack[i];
Eric Dumazet1240d132009-11-08 10:18:44 +00001623 if (likely(skb1 == NULL))
1624 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
1625
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001626 if (!skb1) {
1627 atomic_inc(&sk->sk_drops);
1628 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1629 IS_UDPLITE(sk));
1630 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1631 IS_UDPLITE(sk));
1632 }
1633
1634 if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0)
Eric Dumazet1240d132009-11-08 10:18:44 +00001635 skb1 = NULL;
David Held2dc41cf2014-07-15 23:28:32 -04001636
1637 sock_put(sk);
Eric Dumazet1240d132009-11-08 10:18:44 +00001638 }
1639 if (unlikely(skb1))
1640 kfree_skb(skb1);
1641}
1642
Eric Dumazet97502232013-12-11 14:46:51 -08001643/* For TCP sockets, sk_rx_dst is protected by socket lock
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001644 * For UDP, we use xchg() to guard against concurrent changes.
Eric Dumazet97502232013-12-11 14:46:51 -08001645 */
1646static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
Shawn Bohrer421b3882013-10-07 11:01:39 -05001647{
Eric Dumazet97502232013-12-11 14:46:51 -08001648 struct dst_entry *old;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001649
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001650 dst_hold(dst);
1651 old = xchg(&sk->sk_rx_dst, dst);
1652 dst_release(old);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001653}
1654
David S. Millerdb8dac22008-03-06 16:22:02 -08001655/*
1656 * Multicasts and broadcasts go to each listener.
1657 *
Eric Dumazet1240d132009-11-08 10:18:44 +00001658 * Note: called only from the BH handler context.
David S. Millerdb8dac22008-03-06 16:22:02 -08001659 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001660static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001661 struct udphdr *uh,
1662 __be32 saddr, __be32 daddr,
Rick Jones36cbb242014-11-06 10:37:54 -08001663 struct udp_table *udptable,
1664 int proto)
David S. Millerdb8dac22008-03-06 16:22:02 -08001665{
Eric Dumazet1240d132009-11-08 10:18:44 +00001666 struct sock *sk, *stack[256 / sizeof(struct sock *)];
David Held5cf3d462014-07-15 23:28:31 -04001667 struct hlist_nulls_node *node;
1668 unsigned short hnum = ntohs(uh->dest);
1669 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
1670 int dif = skb->dev->ifindex;
David Held2dc41cf2014-07-15 23:28:32 -04001671 unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node);
1672 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
Rick Jones36cbb242014-11-06 10:37:54 -08001673 bool inner_flushed = false;
David Held2dc41cf2014-07-15 23:28:32 -04001674
1675 if (use_hash2) {
1676 hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) &
1677 udp_table.mask;
1678 hash2 = udp4_portaddr_hash(net, daddr, hnum) & udp_table.mask;
1679start_lookup:
1680 hslot = &udp_table.hash2[hash2];
1681 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
1682 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001683
Eric Dumazet645ca702008-10-29 01:41:45 -07001684 spin_lock(&hslot->lock);
David Held2dc41cf2014-07-15 23:28:32 -04001685 sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) {
David Held5cf3d462014-07-15 23:28:31 -04001686 if (__udp_is_mcast_sock(net, sk,
1687 uh->dest, daddr,
1688 uh->source, saddr,
1689 dif, hnum)) {
1690 if (unlikely(count == ARRAY_SIZE(stack))) {
1691 flush_stack(stack, count, skb, ~0);
Rick Jones36cbb242014-11-06 10:37:54 -08001692 inner_flushed = true;
David Held5cf3d462014-07-15 23:28:31 -04001693 count = 0;
1694 }
1695 stack[count++] = sk;
David Held2dc41cf2014-07-15 23:28:32 -04001696 sock_hold(sk);
Eric Dumazet1240d132009-11-08 10:18:44 +00001697 }
1698 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001699
Eric Dumazet645ca702008-10-29 01:41:45 -07001700 spin_unlock(&hslot->lock);
Eric Dumazet1240d132009-11-08 10:18:44 +00001701
David Held2dc41cf2014-07-15 23:28:32 -04001702 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
1703 if (use_hash2 && hash2 != hash2_any) {
1704 hash2 = hash2_any;
1705 goto start_lookup;
1706 }
1707
Eric Dumazet1240d132009-11-08 10:18:44 +00001708 /*
1709 * do the slow work with no lock held
1710 */
1711 if (count) {
1712 flush_stack(stack, count, skb, count - 1);
Eric Dumazet1240d132009-11-08 10:18:44 +00001713 } else {
Rick Jones36cbb242014-11-06 10:37:54 -08001714 if (!inner_flushed)
1715 UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
1716 proto == IPPROTO_UDPLITE);
1717 consume_skb(skb);
Eric Dumazet1240d132009-11-08 10:18:44 +00001718 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001719 return 0;
1720}
1721
1722/* Initialize UDP checksum. If exited with zero value (success),
1723 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1724 * Otherwise, csum completion requires chacksumming packet body,
1725 * including udp header and folding it to skb->csum.
1726 */
1727static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1728 int proto)
1729{
David S. Millerdb8dac22008-03-06 16:22:02 -08001730 int err;
1731
1732 UDP_SKB_CB(skb)->partial_cov = 0;
1733 UDP_SKB_CB(skb)->cscov = skb->len;
1734
1735 if (proto == IPPROTO_UDPLITE) {
1736 err = udplite_checksum_init(skb, uh);
1737 if (err)
1738 return err;
1739 }
1740
Tom Herberted70fcf2014-05-02 16:29:38 -07001741 return skb_checksum_init_zero_check(skb, proto, uh->check,
1742 inet_compute_pseudo);
David S. Millerdb8dac22008-03-06 16:22:02 -08001743}
1744
1745/*
1746 * All we need to do is get the socket, and then do a checksum.
1747 */
1748
Eric Dumazet645ca702008-10-29 01:41:45 -07001749int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
David S. Millerdb8dac22008-03-06 16:22:02 -08001750 int proto)
1751{
1752 struct sock *sk;
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001753 struct udphdr *uh;
David S. Millerdb8dac22008-03-06 16:22:02 -08001754 unsigned short ulen;
Eric Dumazetadf30902009-06-02 05:19:30 +00001755 struct rtable *rt = skb_rtable(skb);
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001756 __be32 saddr, daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001757 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001758
1759 /*
1760 * Validate the packet.
1761 */
1762 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1763 goto drop; /* No space for header. */
1764
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001765 uh = udp_hdr(skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001766 ulen = ntohs(uh->len);
Bjørn Morkccc2d972010-05-06 03:44:34 +00001767 saddr = ip_hdr(skb)->saddr;
1768 daddr = ip_hdr(skb)->daddr;
1769
David S. Millerdb8dac22008-03-06 16:22:02 -08001770 if (ulen > skb->len)
1771 goto short_packet;
1772
1773 if (proto == IPPROTO_UDP) {
1774 /* UDP validates ulen. */
1775 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1776 goto short_packet;
1777 uh = udp_hdr(skb);
1778 }
1779
1780 if (udp4_csum_init(skb, uh, proto))
1781 goto csum_error;
1782
Eric Dumazet8afdd992013-12-10 18:07:23 -08001783 sk = skb_steal_sock(skb);
1784 if (sk) {
Eric Dumazet97502232013-12-11 14:46:51 -08001785 struct dst_entry *dst = skb_dst(skb);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001786 int ret;
David S. Millerdb8dac22008-03-06 16:22:02 -08001787
Eric Dumazet97502232013-12-11 14:46:51 -08001788 if (unlikely(sk->sk_rx_dst != dst))
1789 udp_sk_rx_dst_set(sk, dst);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001790
1791 ret = udp_queue_rcv_skb(sk, skb);
Eric Dumazet8afdd992013-12-10 18:07:23 -08001792 sock_put(sk);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001793 /* a return value > 0 means to resubmit the input, but
1794 * it wants the return to be -protocol, or 0
1795 */
1796 if (ret > 0)
1797 return -ret;
1798 return 0;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001799 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001800
Fabian Frederickc18450a2014-11-04 20:48:41 +01001801 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1802 return __udp4_lib_mcast_deliver(net, skb, uh,
Rick Jones36cbb242014-11-06 10:37:54 -08001803 saddr, daddr, udptable, proto);
Fabian Frederickc18450a2014-11-04 20:48:41 +01001804
1805 sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
David S. Millerdb8dac22008-03-06 16:22:02 -08001806 if (sk != NULL) {
Eliezer Tamira5b50472013-06-10 11:40:00 +03001807 int ret;
1808
Tom Herbert2abb7cd2014-08-31 15:12:43 -07001809 if (udp_sk(sk)->convert_csum && uh->check && !IS_UDPLITE(sk))
1810 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
1811 inet_compute_pseudo);
1812
Eliezer Tamira5b50472013-06-10 11:40:00 +03001813 ret = udp_queue_rcv_skb(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001814 sock_put(sk);
1815
1816 /* a return value > 0 means to resubmit the input, but
1817 * it wants the return to be -protocol, or 0
1818 */
1819 if (ret > 0)
1820 return -ret;
1821 return 0;
1822 }
1823
1824 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1825 goto drop;
1826 nf_reset(skb);
1827
1828 /* No socket. Drop packet silently, if checksum is wrong */
1829 if (udp_lib_checksum_complete(skb))
1830 goto csum_error;
1831
Pavel Emelyanov02833282008-07-05 21:18:48 -07001832 UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001833 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1834
1835 /*
1836 * Hmm. We got an UDP packet to a port to which we
1837 * don't wanna listen. Ignore it.
1838 */
1839 kfree_skb(skb);
1840 return 0;
1841
1842short_packet:
Joe Perchesba7a46f2014-11-11 10:59:17 -08001843 net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
1844 proto == IPPROTO_UDPLITE ? "Lite" : "",
1845 &saddr, ntohs(uh->source),
1846 ulen, skb->len,
1847 &daddr, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001848 goto drop;
1849
1850csum_error:
1851 /*
1852 * RFC1122: OK. Discards the bad packet silently (as far as
1853 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1854 */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001855 net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
1856 proto == IPPROTO_UDPLITE ? "Lite" : "",
1857 &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
1858 ulen);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001859 UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001860drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001861 UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001862 kfree_skb(skb);
1863 return 0;
1864}
1865
Shawn Bohrer421b3882013-10-07 11:01:39 -05001866/* We can only early demux multicast if there is a single matching socket.
1867 * If more than one socket found returns NULL
1868 */
1869static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net,
1870 __be16 loc_port, __be32 loc_addr,
1871 __be16 rmt_port, __be32 rmt_addr,
1872 int dif)
1873{
1874 struct sock *sk, *result;
1875 struct hlist_nulls_node *node;
1876 unsigned short hnum = ntohs(loc_port);
1877 unsigned int count, slot = udp_hashfn(net, hnum, udp_table.mask);
1878 struct udp_hslot *hslot = &udp_table.hash[slot];
1879
Eric Dumazet63c6f812014-06-12 16:13:06 -07001880 /* Do not bother scanning a too big list */
1881 if (hslot->count > 10)
1882 return NULL;
1883
Shawn Bohrer421b3882013-10-07 11:01:39 -05001884 rcu_read_lock();
1885begin:
1886 count = 0;
1887 result = NULL;
1888 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
1889 if (__udp_is_mcast_sock(net, sk,
1890 loc_port, loc_addr,
1891 rmt_port, rmt_addr,
1892 dif, hnum)) {
1893 result = sk;
1894 ++count;
1895 }
1896 }
1897 /*
1898 * if the nulls value we got at the end of this lookup is
1899 * not the expected one, we must restart lookup.
1900 * We probably met an item that was moved to another chain.
1901 */
1902 if (get_nulls_value(node) != slot)
1903 goto begin;
1904
1905 if (result) {
1906 if (count != 1 ||
1907 unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
1908 result = NULL;
Eric Dumazetf69b9232013-10-08 21:47:29 -07001909 else if (unlikely(!__udp_is_mcast_sock(net, result,
Shawn Bohrer421b3882013-10-07 11:01:39 -05001910 loc_port, loc_addr,
1911 rmt_port, rmt_addr,
1912 dif, hnum))) {
1913 sock_put(result);
1914 result = NULL;
1915 }
1916 }
1917 rcu_read_unlock();
1918 return result;
1919}
1920
1921/* For unicast we should only early demux connected sockets or we can
1922 * break forwarding setups. The chains here can be long so only check
1923 * if the first socket is an exact match and if not move on.
1924 */
1925static struct sock *__udp4_lib_demux_lookup(struct net *net,
1926 __be16 loc_port, __be32 loc_addr,
1927 __be16 rmt_port, __be32 rmt_addr,
1928 int dif)
1929{
1930 struct sock *sk, *result;
1931 struct hlist_nulls_node *node;
1932 unsigned short hnum = ntohs(loc_port);
1933 unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum);
1934 unsigned int slot2 = hash2 & udp_table.mask;
1935 struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
Joe Perchesc7228312014-05-13 20:30:07 -07001936 INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001937 const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
1938
1939 rcu_read_lock();
1940 result = NULL;
1941 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
1942 if (INET_MATCH(sk, net, acookie,
1943 rmt_addr, loc_addr, ports, dif))
1944 result = sk;
1945 /* Only check first socket in chain */
1946 break;
1947 }
1948
1949 if (result) {
1950 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
1951 result = NULL;
1952 else if (unlikely(!INET_MATCH(sk, net, acookie,
1953 rmt_addr, loc_addr,
1954 ports, dif))) {
1955 sock_put(result);
1956 result = NULL;
1957 }
1958 }
1959 rcu_read_unlock();
1960 return result;
1961}
1962
1963void udp_v4_early_demux(struct sk_buff *skb)
1964{
Eric Dumazet610438b2013-12-11 08:10:05 -08001965 struct net *net = dev_net(skb->dev);
1966 const struct iphdr *iph;
1967 const struct udphdr *uh;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001968 struct sock *sk;
1969 struct dst_entry *dst;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001970 int dif = skb->dev->ifindex;
1971
1972 /* validate the packet */
1973 if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr)))
1974 return;
1975
Eric Dumazet610438b2013-12-11 08:10:05 -08001976 iph = ip_hdr(skb);
1977 uh = udp_hdr(skb);
1978
Shawn Bohrer421b3882013-10-07 11:01:39 -05001979 if (skb->pkt_type == PACKET_BROADCAST ||
1980 skb->pkt_type == PACKET_MULTICAST)
1981 sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr,
1982 uh->source, iph->saddr, dif);
1983 else if (skb->pkt_type == PACKET_HOST)
1984 sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr,
1985 uh->source, iph->saddr, dif);
1986 else
1987 return;
1988
1989 if (!sk)
1990 return;
1991
1992 skb->sk = sk;
Alexander Duyck82eabd92014-09-04 13:32:11 -04001993 skb->destructor = sock_efree;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001994 dst = sk->sk_rx_dst;
1995
1996 if (dst)
1997 dst = dst_check(dst, 0);
1998 if (dst)
1999 skb_dst_set_noref(skb, dst);
2000}
2001
David S. Millerdb8dac22008-03-06 16:22:02 -08002002int udp_rcv(struct sk_buff *skb)
2003{
Eric Dumazet645ca702008-10-29 01:41:45 -07002004 return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
David S. Millerdb8dac22008-03-06 16:22:02 -08002005}
2006
Brian Haley7d06b2e2008-06-14 17:04:49 -07002007void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08002008{
Tom Parkin44046a52013-03-19 06:11:12 +00002009 struct udp_sock *up = udp_sk(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00002010 bool slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08002011 udp_flush_pending_frames(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00002012 unlock_sock_fast(sk, slow);
Tom Parkin44046a52013-03-19 06:11:12 +00002013 if (static_key_false(&udp_encap_needed) && up->encap_type) {
2014 void (*encap_destroy)(struct sock *sk);
2015 encap_destroy = ACCESS_ONCE(up->encap_destroy);
2016 if (encap_destroy)
2017 encap_destroy(sk);
2018 }
David S. Millerdb8dac22008-03-06 16:22:02 -08002019}
2020
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021/*
2022 * Socket option code for UDP
2023 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002024int udp_lib_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002025 char __user *optval, unsigned int optlen,
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002026 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027{
2028 struct udp_sock *up = udp_sk(sk);
Tom Herbert1c194482014-05-23 08:47:32 -07002029 int val, valbool;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11002031 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032
Eric Dumazetc482c562009-07-17 00:26:32 +00002033 if (optlen < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034 return -EINVAL;
2035
2036 if (get_user(val, (int __user *)optval))
2037 return -EFAULT;
2038
Tom Herbert1c194482014-05-23 08:47:32 -07002039 valbool = val ? 1 : 0;
2040
Stephen Hemminger6516c652007-03-08 20:41:55 -08002041 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 case UDP_CORK:
2043 if (val != 0) {
2044 up->corkflag = 1;
2045 } else {
2046 up->corkflag = 0;
2047 lock_sock(sk);
Joe Perches4243cdc2014-11-11 21:59:20 -08002048 push_pending_frames(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 release_sock(sk);
2050 }
2051 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002052
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053 case UDP_ENCAP:
2054 switch (val) {
2055 case 0:
2056 case UDP_ENCAP_ESPINUDP:
2057 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07002058 up->encap_rcv = xfrm4_udp_encap_rcv;
2059 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07002060 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 up->encap_type = val;
Eric Dumazet447167b2012-04-11 23:05:28 +00002062 udp_encap_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063 break;
2064 default:
2065 err = -ENOPROTOOPT;
2066 break;
2067 }
2068 break;
2069
Tom Herbert1c194482014-05-23 08:47:32 -07002070 case UDP_NO_CHECK6_TX:
2071 up->no_check6_tx = valbool;
2072 break;
2073
2074 case UDP_NO_CHECK6_RX:
2075 up->no_check6_rx = valbool;
2076 break;
2077
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002078 /*
2079 * UDP-Lite's partial checksum coverage (RFC 3828).
2080 */
2081 /* The sender sets actual checksum coverage length via this option.
2082 * The case coverage > packet length is handled by send module. */
2083 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002084 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002085 return -ENOPROTOOPT;
2086 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
2087 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002088 else if (val > USHRT_MAX)
2089 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002090 up->pcslen = val;
2091 up->pcflag |= UDPLITE_SEND_CC;
2092 break;
2093
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002094 /* The receiver specifies a minimum checksum coverage value. To make
2095 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002096 * used, this again means full checksum coverage. */
2097 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002098 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002099 return -ENOPROTOOPT;
2100 if (val != 0 && val < 8) /* Avoid silly minimal values. */
2101 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002102 else if (val > USHRT_MAX)
2103 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002104 up->pcrlen = val;
2105 up->pcflag |= UDPLITE_RECV_CC;
2106 break;
2107
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108 default:
2109 err = -ENOPROTOOPT;
2110 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002111 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112
2113 return err;
2114}
Eric Dumazetc482c562009-07-17 00:26:32 +00002115EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116
David S. Millerdb8dac22008-03-06 16:22:02 -08002117int udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002118 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002119{
2120 if (level == SOL_UDP || level == SOL_UDPLITE)
2121 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2122 udp_push_pending_frames);
2123 return ip_setsockopt(sk, level, optname, optval, optlen);
2124}
2125
2126#ifdef CONFIG_COMPAT
2127int compat_udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002128 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002129{
2130 if (level == SOL_UDP || level == SOL_UDPLITE)
2131 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2132 udp_push_pending_frames);
2133 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
2134}
2135#endif
2136
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002137int udp_lib_getsockopt(struct sock *sk, int level, int optname,
2138 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002139{
2140 struct udp_sock *up = udp_sk(sk);
2141 int val, len;
2142
Eric Dumazetc482c562009-07-17 00:26:32 +00002143 if (get_user(len, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002144 return -EFAULT;
2145
2146 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002147
Stephen Hemminger6516c652007-03-08 20:41:55 -08002148 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149 return -EINVAL;
2150
Stephen Hemminger6516c652007-03-08 20:41:55 -08002151 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152 case UDP_CORK:
2153 val = up->corkflag;
2154 break;
2155
2156 case UDP_ENCAP:
2157 val = up->encap_type;
2158 break;
2159
Tom Herbert1c194482014-05-23 08:47:32 -07002160 case UDP_NO_CHECK6_TX:
2161 val = up->no_check6_tx;
2162 break;
2163
2164 case UDP_NO_CHECK6_RX:
2165 val = up->no_check6_rx;
2166 break;
2167
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002168 /* The following two cannot be changed on UDP sockets, the return is
2169 * always 0 (which corresponds to the full checksum coverage of UDP). */
2170 case UDPLITE_SEND_CSCOV:
2171 val = up->pcslen;
2172 break;
2173
2174 case UDPLITE_RECV_CSCOV:
2175 val = up->pcrlen;
2176 break;
2177
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178 default:
2179 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002180 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181
Stephen Hemminger6516c652007-03-08 20:41:55 -08002182 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002183 return -EFAULT;
Eric Dumazetc482c562009-07-17 00:26:32 +00002184 if (copy_to_user(optval, &val, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002185 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002186 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002187}
Eric Dumazetc482c562009-07-17 00:26:32 +00002188EXPORT_SYMBOL(udp_lib_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002189
David S. Millerdb8dac22008-03-06 16:22:02 -08002190int udp_getsockopt(struct sock *sk, int level, int optname,
2191 char __user *optval, int __user *optlen)
2192{
2193 if (level == SOL_UDP || level == SOL_UDPLITE)
2194 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2195 return ip_getsockopt(sk, level, optname, optval, optlen);
2196}
2197
2198#ifdef CONFIG_COMPAT
2199int compat_udp_getsockopt(struct sock *sk, int level, int optname,
2200 char __user *optval, int __user *optlen)
2201{
2202 if (level == SOL_UDP || level == SOL_UDPLITE)
2203 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2204 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
2205}
2206#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207/**
2208 * udp_poll - wait for a UDP event.
2209 * @file - file struct
2210 * @sock - socket
2211 * @wait - poll table
2212 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002213 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214 * blocking sockets. If application is using a blocking fd
2215 * and a packet with checksum error is in the queue;
2216 * then it could get return from select indicating data available
2217 * but then block when reading it. Add special case code
2218 * to work around these arguably broken applications.
2219 */
2220unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
2221{
2222 unsigned int mask = datagram_poll(file, sock, wait);
2223 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002224
David Majnemerc3f1dba2013-05-31 13:15:38 +00002225 sock_rps_record_flow(sk);
2226
Linus Torvalds1da177e2005-04-16 15:20:36 -07002227 /* Check for false positives due to checksum errors */
Eric Dumazet85584672009-10-09 04:43:40 +00002228 if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
2229 !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk))
2230 mask &= ~(POLLIN | POLLRDNORM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231
2232 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002233
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234}
Eric Dumazetc482c562009-07-17 00:26:32 +00002235EXPORT_SYMBOL(udp_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236
David S. Millerdb8dac22008-03-06 16:22:02 -08002237struct proto udp_prot = {
2238 .name = "UDP",
2239 .owner = THIS_MODULE,
2240 .close = udp_lib_close,
2241 .connect = ip4_datagram_connect,
2242 .disconnect = udp_disconnect,
2243 .ioctl = udp_ioctl,
2244 .destroy = udp_destroy_sock,
2245 .setsockopt = udp_setsockopt,
2246 .getsockopt = udp_getsockopt,
2247 .sendmsg = udp_sendmsg,
2248 .recvmsg = udp_recvmsg,
2249 .sendpage = udp_sendpage,
Herbert Xu93821772008-09-15 11:48:46 -07002250 .backlog_rcv = __udp_queue_rcv_skb,
Steffen Klassert8141ed92013-01-21 02:00:03 +00002251 .release_cb = ip4_datagram_release_cb,
David S. Millerdb8dac22008-03-06 16:22:02 -08002252 .hash = udp_lib_hash,
2253 .unhash = udp_lib_unhash,
Eric Dumazet719f8352010-09-08 05:08:44 +00002254 .rehash = udp_v4_rehash,
David S. Millerdb8dac22008-03-06 16:22:02 -08002255 .get_port = udp_v4_get_port,
2256 .memory_allocated = &udp_memory_allocated,
2257 .sysctl_mem = sysctl_udp_mem,
2258 .sysctl_wmem = &sysctl_udp_wmem_min,
2259 .sysctl_rmem = &sysctl_udp_rmem_min,
2260 .obj_size = sizeof(struct udp_sock),
Eric Dumazet271b72c2008-10-29 02:11:14 -07002261 .slab_flags = SLAB_DESTROY_BY_RCU,
Eric Dumazet645ca702008-10-29 01:41:45 -07002262 .h.udp_table = &udp_table,
David S. Millerdb8dac22008-03-06 16:22:02 -08002263#ifdef CONFIG_COMPAT
2264 .compat_setsockopt = compat_udp_setsockopt,
2265 .compat_getsockopt = compat_udp_getsockopt,
2266#endif
Octavian Purdilafcbdf092010-12-16 14:26:56 -08002267 .clear_sk = sk_prot_clear_portaddr_nulls,
David S. Millerdb8dac22008-03-06 16:22:02 -08002268};
Eric Dumazetc482c562009-07-17 00:26:32 +00002269EXPORT_SYMBOL(udp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270
2271/* ------------------------------------------------------------------------ */
2272#ifdef CONFIG_PROC_FS
2273
Eric Dumazet645ca702008-10-29 01:41:45 -07002274static struct sock *udp_get_first(struct seq_file *seq, int start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002275{
2276 struct sock *sk;
2277 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002278 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002279
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002280 for (state->bucket = start; state->bucket <= state->udp_table->mask;
2281 ++state->bucket) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002282 struct hlist_nulls_node *node;
Eric Dumazet645ca702008-10-29 01:41:45 -07002283 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002284
2285 if (hlist_nulls_empty(&hslot->head))
2286 continue;
2287
Eric Dumazet645ca702008-10-29 01:41:45 -07002288 spin_lock_bh(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08002289 sk_nulls_for_each(sk, node, &hslot->head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002290 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002291 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002292 if (sk->sk_family == state->family)
2293 goto found;
2294 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002295 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296 }
2297 sk = NULL;
2298found:
2299 return sk;
2300}
2301
2302static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
2303{
2304 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002305 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306
2307 do {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002308 sk = sk_nulls_next(sk);
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002309 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002310
Eric Dumazet645ca702008-10-29 01:41:45 -07002311 if (!sk) {
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002312 if (state->bucket <= state->udp_table->mask)
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002313 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07002314 return udp_get_first(seq, state->bucket + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315 }
2316 return sk;
2317}
2318
2319static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2320{
Eric Dumazet645ca702008-10-29 01:41:45 -07002321 struct sock *sk = udp_get_first(seq, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002322
2323 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08002324 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325 --pos;
2326 return pos ? NULL : sk;
2327}
2328
2329static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2330{
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002331 struct udp_iter_state *state = seq->private;
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002332 state->bucket = MAX_UDP_PORTS;
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002333
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002334 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002335}
2336
2337static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2338{
2339 struct sock *sk;
2340
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002341 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002342 sk = udp_get_idx(seq, 0);
2343 else
2344 sk = udp_get_next(seq, v);
2345
2346 ++*pos;
2347 return sk;
2348}
2349
2350static void udp_seq_stop(struct seq_file *seq, void *v)
2351{
Eric Dumazet645ca702008-10-29 01:41:45 -07002352 struct udp_iter_state *state = seq->private;
2353
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002354 if (state->bucket <= state->udp_table->mask)
Eric Dumazet645ca702008-10-29 01:41:45 -07002355 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356}
2357
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002358int udp_seq_open(struct inode *inode, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002359{
Al Virod9dda782013-03-31 18:16:14 -04002360 struct udp_seq_afinfo *afinfo = PDE_DATA(inode);
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002361 struct udp_iter_state *s;
2362 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002363
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002364 err = seq_open_net(inode, file, &afinfo->seq_ops,
2365 sizeof(struct udp_iter_state));
2366 if (err < 0)
2367 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002368
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002369 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002370 s->family = afinfo->family;
Eric Dumazet645ca702008-10-29 01:41:45 -07002371 s->udp_table = afinfo->udp_table;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002372 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002373}
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002374EXPORT_SYMBOL(udp_seq_open);
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002375
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002377int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002378{
2379 struct proc_dir_entry *p;
2380 int rc = 0;
2381
Denis V. Lunevdda61922008-03-28 18:24:26 -07002382 afinfo->seq_ops.start = udp_seq_start;
2383 afinfo->seq_ops.next = udp_seq_next;
2384 afinfo->seq_ops.stop = udp_seq_stop;
2385
Denis V. Lunev84841c32008-05-02 04:10:08 -07002386 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002387 afinfo->seq_fops, afinfo);
Denis V. Lunev84841c32008-05-02 04:10:08 -07002388 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389 rc = -ENOMEM;
2390 return rc;
2391}
Eric Dumazetc482c562009-07-17 00:26:32 +00002392EXPORT_SYMBOL(udp_proc_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002393
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002394void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395{
Gao fengece31ff2013-02-18 01:34:56 +00002396 remove_proc_entry(afinfo->name, net->proc_net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002397}
Eric Dumazetc482c562009-07-17 00:26:32 +00002398EXPORT_SYMBOL(udp_proc_unregister);
David S. Millerdb8dac22008-03-06 16:22:02 -08002399
2400/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002401static void udp4_format_sock(struct sock *sp, struct seq_file *f,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002402 int bucket)
David S. Millerdb8dac22008-03-06 16:22:02 -08002403{
2404 struct inet_sock *inet = inet_sk(sp);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002405 __be32 dest = inet->inet_daddr;
2406 __be32 src = inet->inet_rcv_saddr;
2407 __u16 destp = ntohs(inet->inet_dport);
2408 __u16 srcp = ntohs(inet->inet_sport);
David S. Millerdb8dac22008-03-06 16:22:02 -08002409
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002410 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
Tetsuo Handa652586d2013-11-14 14:31:57 -08002411 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
David S. Millerdb8dac22008-03-06 16:22:02 -08002412 bucket, src, srcp, dest, destp, sp->sk_state,
Eric Dumazet31e6d362009-06-17 19:05:41 -07002413 sk_wmem_alloc_get(sp),
2414 sk_rmem_alloc_get(sp),
Eric W. Biedermana7cb5a42012-05-24 01:10:10 -06002415 0, 0L, 0,
2416 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
2417 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002418 atomic_read(&sp->sk_refcnt), sp,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002419 atomic_read(&sp->sk_drops));
David S. Millerdb8dac22008-03-06 16:22:02 -08002420}
2421
2422int udp4_seq_show(struct seq_file *seq, void *v)
2423{
Tetsuo Handa652586d2013-11-14 14:31:57 -08002424 seq_setwidth(seq, 127);
David S. Millerdb8dac22008-03-06 16:22:02 -08002425 if (v == SEQ_START_TOKEN)
Tetsuo Handa652586d2013-11-14 14:31:57 -08002426 seq_puts(seq, " sl local_address rem_address st tx_queue "
David S. Millerdb8dac22008-03-06 16:22:02 -08002427 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002428 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08002429 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08002430 struct udp_iter_state *state = seq->private;
2431
Tetsuo Handa652586d2013-11-14 14:31:57 -08002432 udp4_format_sock(v, seq, state->bucket);
David S. Millerdb8dac22008-03-06 16:22:02 -08002433 }
Tetsuo Handa652586d2013-11-14 14:31:57 -08002434 seq_pad(seq, '\n');
David S. Millerdb8dac22008-03-06 16:22:02 -08002435 return 0;
2436}
2437
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002438static const struct file_operations udp_afinfo_seq_fops = {
2439 .owner = THIS_MODULE,
2440 .open = udp_seq_open,
2441 .read = seq_read,
2442 .llseek = seq_lseek,
2443 .release = seq_release_net
2444};
2445
David S. Millerdb8dac22008-03-06 16:22:02 -08002446/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08002447static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08002448 .name = "udp",
2449 .family = AF_INET,
Eric Dumazet645ca702008-10-29 01:41:45 -07002450 .udp_table = &udp_table,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002451 .seq_fops = &udp_afinfo_seq_fops,
Denis V. Lunevdda61922008-03-28 18:24:26 -07002452 .seq_ops = {
2453 .show = udp4_seq_show,
2454 },
David S. Millerdb8dac22008-03-06 16:22:02 -08002455};
2456
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002457static int __net_init udp4_proc_init_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002458{
2459 return udp_proc_register(net, &udp4_seq_afinfo);
2460}
2461
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002462static void __net_exit udp4_proc_exit_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002463{
2464 udp_proc_unregister(net, &udp4_seq_afinfo);
2465}
2466
2467static struct pernet_operations udp4_net_ops = {
2468 .init = udp4_proc_init_net,
2469 .exit = udp4_proc_exit_net,
2470};
2471
David S. Millerdb8dac22008-03-06 16:22:02 -08002472int __init udp4_proc_init(void)
2473{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002474 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002475}
2476
2477void udp4_proc_exit(void)
2478{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002479 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002480}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002481#endif /* CONFIG_PROC_FS */
2482
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002483static __initdata unsigned long uhash_entries;
2484static int __init set_uhash_entries(char *str)
Eric Dumazet645ca702008-10-29 01:41:45 -07002485{
Eldad Zack413c27d2012-05-19 14:13:18 +00002486 ssize_t ret;
2487
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002488 if (!str)
2489 return 0;
Eldad Zack413c27d2012-05-19 14:13:18 +00002490
2491 ret = kstrtoul(str, 0, &uhash_entries);
2492 if (ret)
2493 return 0;
2494
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002495 if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2496 uhash_entries = UDP_HTABLE_SIZE_MIN;
2497 return 1;
2498}
2499__setup("uhash_entries=", set_uhash_entries);
Eric Dumazet645ca702008-10-29 01:41:45 -07002500
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002501void __init udp_table_init(struct udp_table *table, const char *name)
2502{
2503 unsigned int i;
2504
Tim Bird31fe62b2012-05-23 13:33:35 +00002505 table->hash = alloc_large_system_hash(name,
2506 2 * sizeof(struct udp_hslot),
2507 uhash_entries,
2508 21, /* one slot per 2 MB */
2509 0,
2510 &table->log,
2511 &table->mask,
2512 UDP_HTABLE_SIZE_MIN,
2513 64 * 1024);
2514
Eric Dumazet512615b2009-11-08 10:17:58 +00002515 table->hash2 = table->hash + (table->mask + 1);
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002516 for (i = 0; i <= table->mask; i++) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002517 INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00002518 table->hash[i].count = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -07002519 spin_lock_init(&table->hash[i].lock);
2520 }
Eric Dumazet512615b2009-11-08 10:17:58 +00002521 for (i = 0; i <= table->mask; i++) {
2522 INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i);
2523 table->hash2[i].count = 0;
2524 spin_lock_init(&table->hash2[i].lock);
2525 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002526}
2527
Hideo Aoki95766ff2007-12-31 00:29:24 -08002528void __init udp_init(void)
2529{
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002530 unsigned long limit;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002531
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002532 udp_table_init(&udp_table, "UDP");
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002533 limit = nr_free_buffer_pages() / 8;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002534 limit = max(limit, 128UL);
2535 sysctl_udp_mem[0] = limit / 4 * 3;
2536 sysctl_udp_mem[1] = limit;
2537 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2538
2539 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2540 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2541}