blob: 4a16b9129079791e160634f4dfdcd0302beab568 [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
Eric Dumazet645ca702008-10-29 01:41:45 -0700339static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr,
340 unsigned short hnum,
341 __be16 sport, __be32 daddr, __be16 dport, int dif)
342{
343 int score = -1;
344
Eric Dumazetd4cada42009-11-08 10:17:30 +0000345 if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
Eric Dumazet645ca702008-10-29 01:41:45 -0700346 !ipv6_only_sock(sk)) {
347 struct inet_sock *inet = inet_sk(sk);
348
Tom Herbertba418fa2013-01-22 09:50:32 +0000349 score = (sk->sk_family == PF_INET ? 2 : 1);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000350 if (inet->inet_rcv_saddr) {
351 if (inet->inet_rcv_saddr != daddr)
Eric Dumazet645ca702008-10-29 01:41:45 -0700352 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000353 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700354 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000355 if (inet->inet_daddr) {
356 if (inet->inet_daddr != saddr)
Eric Dumazet645ca702008-10-29 01:41:45 -0700357 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000358 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700359 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000360 if (inet->inet_dport) {
361 if (inet->inet_dport != sport)
Eric Dumazet645ca702008-10-29 01:41:45 -0700362 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000363 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700364 }
365 if (sk->sk_bound_dev_if) {
366 if (sk->sk_bound_dev_if != dif)
367 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000368 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700369 }
370 }
371 return score;
372}
373
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000374/*
375 * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num)
376 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000377static inline int compute_score2(struct sock *sk, struct net *net,
378 __be32 saddr, __be16 sport,
379 __be32 daddr, unsigned int hnum, int dif)
380{
381 int score = -1;
382
383 if (net_eq(sock_net(sk), net) && !ipv6_only_sock(sk)) {
384 struct inet_sock *inet = inet_sk(sk);
385
386 if (inet->inet_rcv_saddr != daddr)
387 return -1;
388 if (inet->inet_num != hnum)
389 return -1;
390
Tom Herbertba418fa2013-01-22 09:50:32 +0000391 score = (sk->sk_family == PF_INET ? 2 : 1);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000392 if (inet->inet_daddr) {
393 if (inet->inet_daddr != saddr)
394 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000395 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000396 }
397 if (inet->inet_dport) {
398 if (inet->inet_dport != sport)
399 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000400 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000401 }
402 if (sk->sk_bound_dev_if) {
403 if (sk->sk_bound_dev_if != dif)
404 return -1;
Tom Herbertba418fa2013-01-22 09:50:32 +0000405 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000406 }
407 }
408 return score;
409}
410
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200411static unsigned int udp_ehashfn(struct net *net, const __be32 laddr,
412 const __u16 lport, const __be32 faddr,
413 const __be16 fport)
414{
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200415 static u32 udp_ehash_secret __read_mostly;
416
417 net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret));
418
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200419 return __inet_ehashfn(laddr, lport, faddr, fport,
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200420 udp_ehash_secret + net_hash_mix(net));
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200421}
422
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000423
424/* called with read_rcu_lock() */
425static struct sock *udp4_lib_lookup2(struct net *net,
426 __be32 saddr, __be16 sport,
427 __be32 daddr, unsigned int hnum, int dif,
428 struct udp_hslot *hslot2, unsigned int slot2)
429{
430 struct sock *sk, *result;
431 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000432 int score, badness, matches = 0, reuseport = 0;
433 u32 hash = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000434
435begin:
436 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000437 badness = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000438 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
439 score = compute_score2(sk, net, saddr, sport,
440 daddr, hnum, dif);
441 if (score > badness) {
442 result = sk;
443 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000444 reuseport = sk->sk_reuseport;
445 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200446 hash = udp_ehashfn(net, daddr, hnum,
447 saddr, sport);
Tom Herbertba418fa2013-01-22 09:50:32 +0000448 matches = 1;
449 }
450 } else if (score == badness && reuseport) {
451 matches++;
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200452 if (reciprocal_scale(hash, matches) == 0)
Tom Herbertba418fa2013-01-22 09:50:32 +0000453 result = sk;
454 hash = next_pseudo_random32(hash);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000455 }
456 }
457 /*
458 * if the nulls value we got at the end of this lookup is
459 * not the expected one, we must restart lookup.
460 * We probably met an item that was moved to another chain.
461 */
462 if (get_nulls_value(node) != slot2)
463 goto begin;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000464 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000465 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000466 result = NULL;
467 else if (unlikely(compute_score2(result, net, saddr, sport,
468 daddr, hnum, dif) < badness)) {
469 sock_put(result);
470 goto begin;
471 }
472 }
473 return result;
474}
475
David S. Millerdb8dac22008-03-06 16:22:02 -0800476/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
477 * harder than this. -DaveM
478 */
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000479struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -0800480 __be16 sport, __be32 daddr, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700481 int dif, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800482{
Eric Dumazet271b72c2008-10-29 02:11:14 -0700483 struct sock *sk, *result;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800484 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800485 unsigned short hnum = ntohs(dport);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000486 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
487 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
Tom Herbertba418fa2013-01-22 09:50:32 +0000488 int score, badness, matches = 0, reuseport = 0;
489 u32 hash = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800490
Eric Dumazet271b72c2008-10-29 02:11:14 -0700491 rcu_read_lock();
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000492 if (hslot->count > 10) {
493 hash2 = udp4_portaddr_hash(net, daddr, hnum);
494 slot2 = hash2 & udptable->mask;
495 hslot2 = &udptable->hash2[slot2];
496 if (hslot->count < hslot2->count)
497 goto begin;
498
499 result = udp4_lib_lookup2(net, saddr, sport,
500 daddr, hnum, dif,
501 hslot2, slot2);
502 if (!result) {
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700503 hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000504 slot2 = hash2 & udptable->mask;
505 hslot2 = &udptable->hash2[slot2];
506 if (hslot->count < hslot2->count)
507 goto begin;
508
Jorge Boncompte [DTI2]1223c672010-04-08 04:56:48 +0000509 result = udp4_lib_lookup2(net, saddr, sport,
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700510 htonl(INADDR_ANY), hnum, dif,
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000511 hslot2, slot2);
512 }
513 rcu_read_unlock();
514 return result;
515 }
Eric Dumazet271b72c2008-10-29 02:11:14 -0700516begin:
517 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000518 badness = 0;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800519 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
Eric Dumazet645ca702008-10-29 01:41:45 -0700520 score = compute_score(sk, net, saddr, hnum, sport,
521 daddr, dport, dif);
522 if (score > badness) {
523 result = sk;
524 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000525 reuseport = sk->sk_reuseport;
526 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200527 hash = udp_ehashfn(net, daddr, hnum,
528 saddr, sport);
Tom Herbertba418fa2013-01-22 09:50:32 +0000529 matches = 1;
530 }
531 } else if (score == badness && reuseport) {
532 matches++;
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200533 if (reciprocal_scale(hash, matches) == 0)
Tom Herbertba418fa2013-01-22 09:50:32 +0000534 result = sk;
535 hash = next_pseudo_random32(hash);
David S. Millerdb8dac22008-03-06 16:22:02 -0800536 }
537 }
Eric Dumazet88ab1932008-11-16 19:39:21 -0800538 /*
539 * if the nulls value we got at the end of this lookup is
540 * not the expected one, we must restart lookup.
541 * We probably met an item that was moved to another chain.
542 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000543 if (get_nulls_value(node) != slot)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800544 goto begin;
545
Eric Dumazet271b72c2008-10-29 02:11:14 -0700546 if (result) {
Eric Dumazetc31504d2010-11-15 19:58:26 +0000547 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet271b72c2008-10-29 02:11:14 -0700548 result = NULL;
549 else if (unlikely(compute_score(result, net, saddr, hnum, sport,
550 daddr, dport, dif) < badness)) {
551 sock_put(result);
552 goto begin;
553 }
554 }
555 rcu_read_unlock();
David S. Millerdb8dac22008-03-06 16:22:02 -0800556 return result;
557}
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000558EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
David S. Millerdb8dac22008-03-06 16:22:02 -0800559
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700560static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
561 __be16 sport, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700562 struct udp_table *udptable)
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700563{
564 const struct iphdr *iph = ip_hdr(skb);
565
Eric Dumazet8afdd992013-12-10 18:07:23 -0800566 return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport,
567 iph->daddr, dport, inet_iif(skb),
568 udptable);
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700569}
570
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700571struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
572 __be32 daddr, __be16 dport, int dif)
573{
Eric Dumazet645ca702008-10-29 01:41:45 -0700574 return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700575}
576EXPORT_SYMBOL_GPL(udp4_lib_lookup);
577
Shawn Bohrer421b3882013-10-07 11:01:39 -0500578static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk,
579 __be16 loc_port, __be32 loc_addr,
580 __be16 rmt_port, __be32 rmt_addr,
581 int dif, unsigned short hnum)
582{
583 struct inet_sock *inet = inet_sk(sk);
584
585 if (!net_eq(sock_net(sk), net) ||
586 udp_sk(sk)->udp_port_hash != hnum ||
587 (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
588 (inet->inet_dport != rmt_port && inet->inet_dport) ||
589 (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) ||
590 ipv6_only_sock(sk) ||
591 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
592 return false;
593 if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif))
594 return false;
595 return true;
596}
597
David S. Millerdb8dac22008-03-06 16:22:02 -0800598/*
599 * This routine is called by the ICMP module when it gets some
600 * sort of error condition. If err < 0 then the socket should
601 * be closed and the error returned to the user. If err > 0
602 * it's just the icmp type << 8 | icmp code.
603 * Header points to the ip header of the error packet. We move
604 * on past this. Then (as it used to claim before adjustment)
605 * header points to the first 8 bytes of the udp header. We need
606 * to find the appropriate port.
607 */
608
Eric Dumazet645ca702008-10-29 01:41:45 -0700609void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800610{
611 struct inet_sock *inet;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000612 const struct iphdr *iph = (const struct iphdr *)skb->data;
Eric Dumazetc482c562009-07-17 00:26:32 +0000613 struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
David S. Millerdb8dac22008-03-06 16:22:02 -0800614 const int type = icmp_hdr(skb)->type;
615 const int code = icmp_hdr(skb)->code;
616 struct sock *sk;
617 int harderr;
618 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700619 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800620
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700621 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
David S. Millerdb8dac22008-03-06 16:22:02 -0800622 iph->saddr, uh->source, skb->dev->ifindex, udptable);
623 if (sk == NULL) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700624 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800625 return; /* No socket for error */
626 }
627
628 err = 0;
629 harderr = 0;
630 inet = inet_sk(sk);
631
632 switch (type) {
633 default:
634 case ICMP_TIME_EXCEEDED:
635 err = EHOSTUNREACH;
636 break;
637 case ICMP_SOURCE_QUENCH:
638 goto out;
639 case ICMP_PARAMETERPROB:
640 err = EPROTO;
641 harderr = 1;
642 break;
643 case ICMP_DEST_UNREACH:
644 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
David S. Miller36393392012-06-14 22:21:46 -0700645 ipv4_sk_update_pmtu(skb, sk, info);
David S. Millerdb8dac22008-03-06 16:22:02 -0800646 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
647 err = EMSGSIZE;
648 harderr = 1;
649 break;
650 }
651 goto out;
652 }
653 err = EHOSTUNREACH;
654 if (code <= NR_ICMP_UNREACH) {
655 harderr = icmp_err_convert[code].fatal;
656 err = icmp_err_convert[code].errno;
657 }
658 break;
David S. Miller55be7a92012-07-11 21:27:49 -0700659 case ICMP_REDIRECT:
660 ipv4_sk_redirect(skb, sk);
Duan Jiong1a462d12013-09-20 18:20:28 +0800661 goto out;
David S. Millerdb8dac22008-03-06 16:22:02 -0800662 }
663
664 /*
665 * RFC1122: OK. Passes ICMP errors back to application, as per
666 * 4.1.3.3.
667 */
668 if (!inet->recverr) {
669 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
670 goto out;
Eric Dumazetb1faf562010-05-31 23:44:05 -0700671 } else
Eric Dumazetc482c562009-07-17 00:26:32 +0000672 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
Eric Dumazetb1faf562010-05-31 23:44:05 -0700673
David S. Millerdb8dac22008-03-06 16:22:02 -0800674 sk->sk_err = err;
675 sk->sk_error_report(sk);
676out:
677 sock_put(sk);
678}
679
680void udp_err(struct sk_buff *skb, u32 info)
681{
Eric Dumazet645ca702008-10-29 01:41:45 -0700682 __udp4_lib_err(skb, info, &udp_table);
David S. Millerdb8dac22008-03-06 16:22:02 -0800683}
684
685/*
686 * Throw away all pending data and cancel the corking. Socket is locked.
687 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400688void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800689{
690 struct udp_sock *up = udp_sk(sk);
691
692 if (up->pending) {
693 up->len = 0;
694 up->pending = 0;
695 ip_flush_pending_frames(sk);
696 }
697}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400698EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800699
700/**
Herbert Xuf6b96642011-03-01 02:36:48 +0000701 * udp4_hwcsum - handle outgoing HW checksumming
David S. Millerdb8dac22008-03-06 16:22:02 -0800702 * @skb: sk_buff containing the filled-in UDP header
703 * (checksum field must be zeroed out)
Herbert Xuf6b96642011-03-01 02:36:48 +0000704 * @src: source IP address
705 * @dst: destination IP address
David S. Millerdb8dac22008-03-06 16:22:02 -0800706 */
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200707void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
David S. Millerdb8dac22008-03-06 16:22:02 -0800708{
David S. Millerdb8dac22008-03-06 16:22:02 -0800709 struct udphdr *uh = udp_hdr(skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000710 int offset = skb_transport_offset(skb);
711 int len = skb->len - offset;
712 int hlen = len;
David S. Millerdb8dac22008-03-06 16:22:02 -0800713 __wsum csum = 0;
714
WANG Congebbe4952014-06-02 16:12:02 -0700715 if (!skb_has_frag_list(skb)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800716 /*
717 * Only one fragment on the socket.
718 */
719 skb->csum_start = skb_transport_header(skb) - skb->head;
720 skb->csum_offset = offsetof(struct udphdr, check);
Herbert Xuf6b96642011-03-01 02:36:48 +0000721 uh->check = ~csum_tcpudp_magic(src, dst, len,
722 IPPROTO_UDP, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800723 } else {
WANG Congebbe4952014-06-02 16:12:02 -0700724 struct sk_buff *frags;
725
David S. Millerdb8dac22008-03-06 16:22:02 -0800726 /*
727 * HW-checksum won't work as there are two or more
728 * fragments on the socket so that all csums of sk_buffs
729 * should be together
730 */
WANG Congebbe4952014-06-02 16:12:02 -0700731 skb_walk_frags(skb, frags) {
Herbert Xuf6b96642011-03-01 02:36:48 +0000732 csum = csum_add(csum, frags->csum);
733 hlen -= frags->len;
WANG Congebbe4952014-06-02 16:12:02 -0700734 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800735
Herbert Xuf6b96642011-03-01 02:36:48 +0000736 csum = skb_checksum(skb, offset, hlen, csum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800737 skb->ip_summed = CHECKSUM_NONE;
738
David S. Millerdb8dac22008-03-06 16:22:02 -0800739 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
740 if (uh->check == 0)
741 uh->check = CSUM_MANGLED_0;
742 }
743}
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200744EXPORT_SYMBOL_GPL(udp4_hwcsum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800745
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700746/* Function to set UDP checksum for an IPv4 UDP packet. This is intended
747 * for the simple case like when setting the checksum for a UDP tunnel.
748 */
749void udp_set_csum(bool nocheck, struct sk_buff *skb,
750 __be32 saddr, __be32 daddr, int len)
751{
752 struct udphdr *uh = udp_hdr(skb);
753
754 if (nocheck)
755 uh->check = 0;
756 else if (skb_is_gso(skb))
757 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
758 else if (skb_dst(skb) && skb_dst(skb)->dev &&
759 (skb_dst(skb)->dev->features & NETIF_F_V4_CSUM)) {
760
761 BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL);
762
763 skb->ip_summed = CHECKSUM_PARTIAL;
764 skb->csum_start = skb_transport_header(skb) - skb->head;
765 skb->csum_offset = offsetof(struct udphdr, check);
766 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
767 } else {
768 __wsum csum;
769
770 BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL);
771
772 uh->check = 0;
773 csum = skb_checksum(skb, 0, len, 0);
774 uh->check = udp_v4_check(len, saddr, daddr, csum);
775 if (uh->check == 0)
776 uh->check = CSUM_MANGLED_0;
777
778 skb->ip_summed = CHECKSUM_UNNECESSARY;
779 }
780}
781EXPORT_SYMBOL(udp_set_csum);
782
David S. Miller79ab0532011-05-09 13:31:04 -0700783static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
Herbert Xuf6b96642011-03-01 02:36:48 +0000784{
785 struct sock *sk = skb->sk;
786 struct inet_sock *inet = inet_sk(sk);
787 struct udphdr *uh;
Herbert Xuf6b96642011-03-01 02:36:48 +0000788 int err = 0;
789 int is_udplite = IS_UDPLITE(sk);
790 int offset = skb_transport_offset(skb);
791 int len = skb->len - offset;
792 __wsum csum = 0;
793
794 /*
795 * Create a UDP header
796 */
797 uh = udp_hdr(skb);
798 uh->source = inet->inet_sport;
David S. Miller79ab0532011-05-09 13:31:04 -0700799 uh->dest = fl4->fl4_dport;
Herbert Xuf6b96642011-03-01 02:36:48 +0000800 uh->len = htons(len);
801 uh->check = 0;
802
803 if (is_udplite) /* UDP-Lite */
804 csum = udplite_csum(skb);
805
Tom Herbert28448b82014-05-23 08:47:19 -0700806 else if (sk->sk_no_check_tx) { /* UDP csum disabled */
Herbert Xuf6b96642011-03-01 02:36:48 +0000807
808 skb->ip_summed = CHECKSUM_NONE;
809 goto send;
810
811 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
812
David S. Miller79ab0532011-05-09 13:31:04 -0700813 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
Herbert Xuf6b96642011-03-01 02:36:48 +0000814 goto send;
815
816 } else
817 csum = udp_csum(skb);
818
819 /* add protocol-dependent pseudo-header */
David S. Miller79ab0532011-05-09 13:31:04 -0700820 uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
Herbert Xuf6b96642011-03-01 02:36:48 +0000821 sk->sk_protocol, csum);
822 if (uh->check == 0)
823 uh->check = CSUM_MANGLED_0;
824
825send:
Eric Dumazetb5ec8ee2012-08-10 02:22:47 +0000826 err = ip_send_skb(sock_net(sk), skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000827 if (err) {
828 if (err == -ENOBUFS && !inet->recverr) {
829 UDP_INC_STATS_USER(sock_net(sk),
830 UDP_MIB_SNDBUFERRORS, is_udplite);
831 err = 0;
832 }
833 } else
834 UDP_INC_STATS_USER(sock_net(sk),
835 UDP_MIB_OUTDATAGRAMS, is_udplite);
836 return err;
837}
838
David S. Millerdb8dac22008-03-06 16:22:02 -0800839/*
840 * Push out all pending data as one UDP datagram. Socket is locked.
841 */
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200842int udp_push_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800843{
844 struct udp_sock *up = udp_sk(sk);
845 struct inet_sock *inet = inet_sk(sk);
David S. Millerb6f21b22011-03-12 02:09:18 -0500846 struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800847 struct sk_buff *skb;
David S. Millerdb8dac22008-03-06 16:22:02 -0800848 int err = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800849
David S. Miller77968b72011-05-08 17:12:19 -0700850 skb = ip_finish_skb(sk, fl4);
Herbert Xuf6b96642011-03-01 02:36:48 +0000851 if (!skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800852 goto out;
853
David S. Miller79ab0532011-05-09 13:31:04 -0700854 err = udp_send_skb(skb, fl4);
David S. Millerdb8dac22008-03-06 16:22:02 -0800855
David S. Millerdb8dac22008-03-06 16:22:02 -0800856out:
857 up->len = 0;
858 up->pending = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800859 return err;
860}
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200861EXPORT_SYMBOL(udp_push_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800862
863int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
864 size_t len)
865{
866 struct inet_sock *inet = inet_sk(sk);
867 struct udp_sock *up = udp_sk(sk);
David S. Millere4749952011-05-08 16:38:45 -0700868 struct flowi4 fl4_stack;
David S. Millerb6f21b22011-03-12 02:09:18 -0500869 struct flowi4 *fl4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800870 int ulen = len;
871 struct ipcm_cookie ipc;
872 struct rtable *rt = NULL;
873 int free = 0;
874 int connected = 0;
875 __be32 daddr, faddr, saddr;
876 __be16 dport;
877 u8 tos;
878 int err, is_udplite = IS_UDPLITE(sk);
879 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
880 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
Herbert Xu903ab862011-03-01 02:36:48 +0000881 struct sk_buff *skb;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000882 struct ip_options_data opt_copy;
David S. Millerdb8dac22008-03-06 16:22:02 -0800883
884 if (len > 0xFFFF)
885 return -EMSGSIZE;
886
887 /*
888 * Check the flags.
889 */
890
Eric Dumazetc482c562009-07-17 00:26:32 +0000891 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
David S. Millerdb8dac22008-03-06 16:22:02 -0800892 return -EOPNOTSUPP;
893
894 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000895 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200896 ipc.ttl = 0;
897 ipc.tos = -1;
David S. Millerdb8dac22008-03-06 16:22:02 -0800898
Herbert Xu903ab862011-03-01 02:36:48 +0000899 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
900
David S. Millerf5fca602011-05-08 17:24:10 -0700901 fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800902 if (up->pending) {
903 /*
904 * There are pending frames.
905 * The socket lock must be held while it's corked.
906 */
907 lock_sock(sk);
908 if (likely(up->pending)) {
909 if (unlikely(up->pending != AF_INET)) {
910 release_sock(sk);
911 return -EINVAL;
912 }
913 goto do_append_data;
914 }
915 release_sock(sk);
916 }
917 ulen += sizeof(struct udphdr);
918
919 /*
920 * Get and verify the address.
921 */
922 if (msg->msg_name) {
Steffen Hurrle342dfc32014-01-17 22:53:15 +0100923 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -0800924 if (msg->msg_namelen < sizeof(*usin))
925 return -EINVAL;
926 if (usin->sin_family != AF_INET) {
927 if (usin->sin_family != AF_UNSPEC)
928 return -EAFNOSUPPORT;
929 }
930
931 daddr = usin->sin_addr.s_addr;
932 dport = usin->sin_port;
933 if (dport == 0)
934 return -EINVAL;
935 } else {
936 if (sk->sk_state != TCP_ESTABLISHED)
937 return -EDESTADDRREQ;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000938 daddr = inet->inet_daddr;
939 dport = inet->inet_dport;
David S. Millerdb8dac22008-03-06 16:22:02 -0800940 /* Open fast path for connected socket.
941 Route will not be used, if at least one option is set.
942 */
943 connected = 1;
944 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000945 ipc.addr = inet->inet_saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800946
947 ipc.oif = sk->sk_bound_dev_if;
Daniel Borkmannbf84a012013-04-14 08:08:13 +0000948
949 sock_tx_timestamp(sk, &ipc.tx_flags);
950
David S. Millerdb8dac22008-03-06 16:22:02 -0800951 if (msg->msg_controllen) {
Hannes Frederic Sowac8e6ad02014-02-18 21:38:08 +0100952 err = ip_cmsg_send(sock_net(sk), msg, &ipc,
953 sk->sk_family == AF_INET6);
David S. Millerdb8dac22008-03-06 16:22:02 -0800954 if (err)
955 return err;
956 if (ipc.opt)
957 free = 1;
958 connected = 0;
959 }
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000960 if (!ipc.opt) {
961 struct ip_options_rcu *inet_opt;
962
963 rcu_read_lock();
964 inet_opt = rcu_dereference(inet->inet_opt);
965 if (inet_opt) {
966 memcpy(&opt_copy, inet_opt,
967 sizeof(*inet_opt) + inet_opt->opt.optlen);
968 ipc.opt = &opt_copy.opt;
969 }
970 rcu_read_unlock();
971 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800972
973 saddr = ipc.addr;
974 ipc.addr = faddr = daddr;
975
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000976 if (ipc.opt && ipc.opt->opt.srr) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800977 if (!daddr)
978 return -EINVAL;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000979 faddr = ipc.opt->opt.faddr;
David S. Millerdb8dac22008-03-06 16:22:02 -0800980 connected = 0;
981 }
Francesco Fuscoaa661582013-09-24 15:43:09 +0200982 tos = get_rttos(&ipc, inet);
David S. Millerdb8dac22008-03-06 16:22:02 -0800983 if (sock_flag(sk, SOCK_LOCALROUTE) ||
984 (msg->msg_flags & MSG_DONTROUTE) ||
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000985 (ipc.opt && ipc.opt->opt.is_strictroute)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800986 tos |= RTO_ONLINK;
987 connected = 0;
988 }
989
990 if (ipv4_is_multicast(daddr)) {
991 if (!ipc.oif)
992 ipc.oif = inet->mc_index;
993 if (!saddr)
994 saddr = inet->mc_addr;
995 connected = 0;
Erich E. Hoover76e21052012-02-08 09:11:07 +0000996 } else if (!ipc.oif)
997 ipc.oif = inet->uc_index;
David S. Millerdb8dac22008-03-06 16:22:02 -0800998
999 if (connected)
Eric Dumazetc482c562009-07-17 00:26:32 +00001000 rt = (struct rtable *)sk_dst_check(sk, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -08001001
1002 if (rt == NULL) {
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -07001003 struct net *net = sock_net(sk);
1004
David S. Millere4749952011-05-08 16:38:45 -07001005 fl4 = &fl4_stack;
1006 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
David S. Millerc0951cb2011-03-31 04:54:27 -07001007 RT_SCOPE_UNIVERSE, sk->sk_protocol,
Steffen Klassert0e0d44a2013-08-28 08:04:14 +02001008 inet_sk_flowi_flags(sk),
David S. Millerc0951cb2011-03-31 04:54:27 -07001009 faddr, saddr, dport, inet->inet_sport);
1010
David S. Millere4749952011-05-08 16:38:45 -07001011 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
1012 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001013 if (IS_ERR(rt)) {
1014 err = PTR_ERR(rt);
David S. Miller06dc94b2011-03-03 10:38:01 -08001015 rt = NULL;
David S. Millerdb8dac22008-03-06 16:22:02 -08001016 if (err == -ENETUNREACH)
Eric Dumazetf1d8cba2013-11-28 09:51:22 -08001017 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -08001018 goto out;
1019 }
1020
1021 err = -EACCES;
1022 if ((rt->rt_flags & RTCF_BROADCAST) &&
1023 !sock_flag(sk, SOCK_BROADCAST))
1024 goto out;
1025 if (connected)
Changli Gaod8d1f302010-06-10 23:31:35 -07001026 sk_dst_set(sk, dst_clone(&rt->dst));
David S. Millerdb8dac22008-03-06 16:22:02 -08001027 }
1028
1029 if (msg->msg_flags&MSG_CONFIRM)
1030 goto do_confirm;
1031back_from_confirm:
1032
David S. Millere4749952011-05-08 16:38:45 -07001033 saddr = fl4->saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001034 if (!ipc.addr)
David S. Millere4749952011-05-08 16:38:45 -07001035 daddr = ipc.addr = fl4->daddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001036
Herbert Xu903ab862011-03-01 02:36:48 +00001037 /* Lockless fast path for the non-corking case. */
1038 if (!corkreq) {
David S. Miller77968b72011-05-08 17:12:19 -07001039 skb = ip_make_skb(sk, fl4, getfrag, msg->msg_iov, ulen,
Herbert Xu903ab862011-03-01 02:36:48 +00001040 sizeof(struct udphdr), &ipc, &rt,
1041 msg->msg_flags);
1042 err = PTR_ERR(skb);
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001043 if (!IS_ERR_OR_NULL(skb))
David S. Miller79ab0532011-05-09 13:31:04 -07001044 err = udp_send_skb(skb, fl4);
Herbert Xu903ab862011-03-01 02:36:48 +00001045 goto out;
1046 }
1047
David S. Millerdb8dac22008-03-06 16:22:02 -08001048 lock_sock(sk);
1049 if (unlikely(up->pending)) {
1050 /* The socket is already corked while preparing it. */
1051 /* ... which is an evident application bug. --ANK */
1052 release_sock(sk);
1053
Joe Perchesba7a46f2014-11-11 10:59:17 -08001054 net_dbg_ratelimited("cork app bug 2\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001055 err = -EINVAL;
1056 goto out;
1057 }
1058 /*
1059 * Now cork the socket to pend data.
1060 */
David S. Millerb6f21b22011-03-12 02:09:18 -05001061 fl4 = &inet->cork.fl.u.ip4;
1062 fl4->daddr = daddr;
1063 fl4->saddr = saddr;
David S. Miller9cce96d2011-03-12 03:00:33 -05001064 fl4->fl4_dport = dport;
1065 fl4->fl4_sport = inet->inet_sport;
David S. Millerdb8dac22008-03-06 16:22:02 -08001066 up->pending = AF_INET;
1067
1068do_append_data:
1069 up->len += ulen;
David S. Millerf5fca602011-05-08 17:24:10 -07001070 err = ip_append_data(sk, fl4, getfrag, msg->msg_iov, ulen,
1071 sizeof(struct udphdr), &ipc, &rt,
1072 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001073 if (err)
1074 udp_flush_pending_frames(sk);
1075 else if (!corkreq)
1076 err = udp_push_pending_frames(sk);
1077 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1078 up->pending = 0;
1079 release_sock(sk);
1080
1081out:
1082 ip_rt_put(rt);
1083 if (free)
1084 kfree(ipc.opt);
1085 if (!err)
1086 return len;
1087 /*
1088 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1089 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1090 * we don't have a good statistic (IpOutDiscards but it can be too many
1091 * things). We could add another new stat but at least for now that
1092 * seems like overkill.
1093 */
1094 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001095 UDP_INC_STATS_USER(sock_net(sk),
1096 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001097 }
1098 return err;
1099
1100do_confirm:
Changli Gaod8d1f302010-06-10 23:31:35 -07001101 dst_confirm(&rt->dst);
David S. Millerdb8dac22008-03-06 16:22:02 -08001102 if (!(msg->msg_flags&MSG_PROBE) || len)
1103 goto back_from_confirm;
1104 err = 0;
1105 goto out;
1106}
Eric Dumazetc482c562009-07-17 00:26:32 +00001107EXPORT_SYMBOL(udp_sendmsg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001108
1109int udp_sendpage(struct sock *sk, struct page *page, int offset,
1110 size_t size, int flags)
1111{
David S. Millerf5fca602011-05-08 17:24:10 -07001112 struct inet_sock *inet = inet_sk(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001113 struct udp_sock *up = udp_sk(sk);
1114 int ret;
1115
Shawn Landdend3f7d562013-11-24 22:36:28 -08001116 if (flags & MSG_SENDPAGE_NOTLAST)
1117 flags |= MSG_MORE;
1118
David S. Millerdb8dac22008-03-06 16:22:02 -08001119 if (!up->pending) {
1120 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
1121
1122 /* Call udp_sendmsg to specify destination address which
1123 * sendpage interface can't pass.
1124 * This will succeed only when the socket is connected.
1125 */
1126 ret = udp_sendmsg(NULL, sk, &msg, 0);
1127 if (ret < 0)
1128 return ret;
1129 }
1130
1131 lock_sock(sk);
1132
1133 if (unlikely(!up->pending)) {
1134 release_sock(sk);
1135
Joe Perchesba7a46f2014-11-11 10:59:17 -08001136 net_dbg_ratelimited("udp cork app bug 3\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001137 return -EINVAL;
1138 }
1139
David S. Millerf5fca602011-05-08 17:24:10 -07001140 ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1141 page, offset, size, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001142 if (ret == -EOPNOTSUPP) {
1143 release_sock(sk);
1144 return sock_no_sendpage(sk->sk_socket, page, offset,
1145 size, flags);
1146 }
1147 if (ret < 0) {
1148 udp_flush_pending_frames(sk);
1149 goto out;
1150 }
1151
1152 up->len += size;
1153 if (!(up->corkflag || (flags&MSG_MORE)))
1154 ret = udp_push_pending_frames(sk);
1155 if (!ret)
1156 ret = size;
1157out:
1158 release_sock(sk);
1159 return ret;
1160}
1161
Eric Dumazet85584672009-10-09 04:43:40 +00001162
1163/**
1164 * first_packet_length - return length of first packet in receive queue
1165 * @sk: socket
1166 *
1167 * Drops all bad checksum frames, until a valid one is found.
1168 * Returns the length of found skb, or 0 if none is found.
1169 */
1170static unsigned int first_packet_length(struct sock *sk)
1171{
1172 struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1173 struct sk_buff *skb;
1174 unsigned int res;
1175
1176 __skb_queue_head_init(&list_kill);
1177
1178 spin_lock_bh(&rcvq->lock);
1179 while ((skb = skb_peek(rcvq)) != NULL &&
1180 udp_lib_checksum_complete(skb)) {
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001181 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS,
1182 IS_UDPLITE(sk));
Eric Dumazet85584672009-10-09 04:43:40 +00001183 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1184 IS_UDPLITE(sk));
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001185 atomic_inc(&sk->sk_drops);
Eric Dumazet85584672009-10-09 04:43:40 +00001186 __skb_unlink(skb, rcvq);
1187 __skb_queue_tail(&list_kill, skb);
1188 }
1189 res = skb ? skb->len : 0;
1190 spin_unlock_bh(&rcvq->lock);
1191
1192 if (!skb_queue_empty(&list_kill)) {
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001193 bool slow = lock_sock_fast(sk);
1194
Eric Dumazet85584672009-10-09 04:43:40 +00001195 __skb_queue_purge(&list_kill);
1196 sk_mem_reclaim_partial(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001197 unlock_sock_fast(sk, slow);
Eric Dumazet85584672009-10-09 04:43:40 +00001198 }
1199 return res;
1200}
1201
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202/*
1203 * IOCTL requests applicable to the UDP protocol
1204 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001205
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1207{
Stephen Hemminger6516c652007-03-08 20:41:55 -08001208 switch (cmd) {
1209 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 {
Eric Dumazet31e6d362009-06-17 19:05:41 -07001211 int amount = sk_wmem_alloc_get(sk);
1212
Stephen Hemminger6516c652007-03-08 20:41:55 -08001213 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 }
Stephen Hemminger6516c652007-03-08 20:41:55 -08001215
1216 case SIOCINQ:
1217 {
Eric Dumazet85584672009-10-09 04:43:40 +00001218 unsigned int amount = first_packet_length(sk);
Stephen Hemminger6516c652007-03-08 20:41:55 -08001219
Eric Dumazet85584672009-10-09 04:43:40 +00001220 if (amount)
Stephen Hemminger6516c652007-03-08 20:41:55 -08001221 /*
1222 * We will only return the amount
1223 * of this packet since that is all
1224 * that will be read.
1225 */
Eric Dumazet85584672009-10-09 04:43:40 +00001226 amount -= sizeof(struct udphdr);
1227
Stephen Hemminger6516c652007-03-08 20:41:55 -08001228 return put_user(amount, (int __user *)arg);
1229 }
1230
1231 default:
1232 return -ENOIOCTLCMD;
1233 }
1234
1235 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236}
Eric Dumazetc482c562009-07-17 00:26:32 +00001237EXPORT_SYMBOL(udp_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238
David S. Millerdb8dac22008-03-06 16:22:02 -08001239/*
1240 * This should be easy, if there is something there we
1241 * return it, otherwise we block.
1242 */
1243
1244int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
1245 size_t len, int noblock, int flags, int *addr_len)
1246{
1247 struct inet_sock *inet = inet_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001248 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -08001249 struct sk_buff *skb;
David S. Miller59c2cda2011-12-01 14:12:55 -05001250 unsigned int ulen, copied;
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001251 int peeked, off = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001252 int err;
1253 int is_udplite = IS_UDPLITE(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001254 bool slow;
David S. Millerdb8dac22008-03-06 16:22:02 -08001255
David S. Millerdb8dac22008-03-06 16:22:02 -08001256 if (flags & MSG_ERRQUEUE)
Hannes Frederic Sowa85fbaa72013-11-23 00:46:12 +01001257 return ip_recv_error(sk, msg, len, addr_len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001258
1259try_again:
1260 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001261 &peeked, &off, &err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001262 if (!skb)
1263 goto out;
1264
1265 ulen = skb->len - sizeof(struct udphdr);
David S. Miller59c2cda2011-12-01 14:12:55 -05001266 copied = len;
1267 if (copied > ulen)
1268 copied = ulen;
1269 else if (copied < ulen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001270 msg->msg_flags |= MSG_TRUNC;
1271
1272 /*
1273 * If checksum is needed at all, try to do it while copying the
1274 * data. If the data is truncated, or if we only want a partial
1275 * coverage checksum (UDP-Lite), do it before the copy.
1276 */
1277
David S. Miller59c2cda2011-12-01 14:12:55 -05001278 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001279 if (udp_lib_checksum_complete(skb))
1280 goto csum_copy_err;
1281 }
1282
1283 if (skb_csum_unnecessary(skb))
David S. Miller51f3d022014-11-05 16:46:40 -05001284 err = skb_copy_datagram_msg(skb, sizeof(struct udphdr),
1285 msg, copied);
David S. Millerdb8dac22008-03-06 16:22:02 -08001286 else {
Eric Dumazetc482c562009-07-17 00:26:32 +00001287 err = skb_copy_and_csum_datagram_iovec(skb,
1288 sizeof(struct udphdr),
1289 msg->msg_iov);
David S. Millerdb8dac22008-03-06 16:22:02 -08001290
1291 if (err == -EINVAL)
1292 goto csum_copy_err;
1293 }
1294
Eric Dumazet22911fc2012-06-27 00:23:44 +00001295 if (unlikely(err)) {
1296 trace_kfree_skb(skb, udp_recvmsg);
Eric Dumazet979402b2012-09-05 23:34:44 +00001297 if (!peeked) {
1298 atomic_inc(&sk->sk_drops);
1299 UDP_INC_STATS_USER(sock_net(sk),
1300 UDP_MIB_INERRORS, is_udplite);
1301 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001302 goto out_free;
Eric Dumazet22911fc2012-06-27 00:23:44 +00001303 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001304
1305 if (!peeked)
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001306 UDP_INC_STATS_USER(sock_net(sk),
1307 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001308
Neil Horman3b885782009-10-12 13:26:31 -07001309 sock_recv_ts_and_drops(msg, sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001310
1311 /* Copy the address. */
Eric Dumazetc482c562009-07-17 00:26:32 +00001312 if (sin) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001313 sin->sin_family = AF_INET;
1314 sin->sin_port = udp_hdr(skb)->source;
1315 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1316 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
Hannes Frederic Sowabceaa902013-11-18 04:20:45 +01001317 *addr_len = sizeof(*sin);
David S. Millerdb8dac22008-03-06 16:22:02 -08001318 }
1319 if (inet->cmsg_flags)
1320 ip_cmsg_recv(msg, skb);
1321
David S. Miller59c2cda2011-12-01 14:12:55 -05001322 err = copied;
David S. Millerdb8dac22008-03-06 16:22:02 -08001323 if (flags & MSG_TRUNC)
1324 err = ulen;
1325
1326out_free:
Eric Dumazet9d410c72009-10-30 05:03:53 +00001327 skb_free_datagram_locked(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001328out:
1329 return err;
1330
1331csum_copy_err:
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001332 slow = lock_sock_fast(sk);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001333 if (!skb_kill_datagram(sk, skb, flags)) {
1334 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001335 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001336 }
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001337 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001338
1339 if (noblock)
1340 return -EAGAIN;
Xufeng Zhang9cfaa8d2011-06-21 10:43:40 +00001341
1342 /* starting over for a new packet */
1343 msg->msg_flags &= ~MSG_TRUNC;
David S. Millerdb8dac22008-03-06 16:22:02 -08001344 goto try_again;
1345}
1346
1347
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348int udp_disconnect(struct sock *sk, int flags)
1349{
1350 struct inet_sock *inet = inet_sk(sk);
1351 /*
1352 * 1003.1g - break association.
1353 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001354
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355 sk->sk_state = TCP_CLOSE;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001356 inet->inet_daddr = 0;
1357 inet->inet_dport = 0;
Tom Herbertbdeab992011-08-14 19:45:55 +00001358 sock_rps_reset_rxhash(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 sk->sk_bound_dev_if = 0;
1360 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1361 inet_reset_saddr(sk);
1362
1363 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1364 sk->sk_prot->unhash(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001365 inet->inet_sport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366 }
1367 sk_dst_reset(sk);
1368 return 0;
1369}
Eric Dumazetc482c562009-07-17 00:26:32 +00001370EXPORT_SYMBOL(udp_disconnect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371
Eric Dumazet645ca702008-10-29 01:41:45 -07001372void udp_lib_unhash(struct sock *sk)
1373{
Eric Dumazet723b4612008-11-25 13:55:15 -08001374 if (sk_hashed(sk)) {
1375 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Eric Dumazet512615b2009-11-08 10:17:58 +00001376 struct udp_hslot *hslot, *hslot2;
1377
1378 hslot = udp_hashslot(udptable, sock_net(sk),
1379 udp_sk(sk)->udp_port_hash);
1380 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
Eric Dumazet645ca702008-10-29 01:41:45 -07001381
Eric Dumazet723b4612008-11-25 13:55:15 -08001382 spin_lock_bh(&hslot->lock);
1383 if (sk_nulls_del_node_init_rcu(sk)) {
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00001384 hslot->count--;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001385 inet_sk(sk)->inet_num = 0;
Eric Dumazet723b4612008-11-25 13:55:15 -08001386 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet512615b2009-11-08 10:17:58 +00001387
1388 spin_lock(&hslot2->lock);
1389 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1390 hslot2->count--;
1391 spin_unlock(&hslot2->lock);
Eric Dumazet723b4612008-11-25 13:55:15 -08001392 }
1393 spin_unlock_bh(&hslot->lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001394 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001395}
1396EXPORT_SYMBOL(udp_lib_unhash);
1397
Eric Dumazet719f8352010-09-08 05:08:44 +00001398/*
1399 * inet_rcv_saddr was changed, we must rehash secondary hash
1400 */
1401void udp_lib_rehash(struct sock *sk, u16 newhash)
1402{
1403 if (sk_hashed(sk)) {
1404 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1405 struct udp_hslot *hslot, *hslot2, *nhslot2;
1406
1407 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1408 nhslot2 = udp_hashslot2(udptable, newhash);
1409 udp_sk(sk)->udp_portaddr_hash = newhash;
1410 if (hslot2 != nhslot2) {
1411 hslot = udp_hashslot(udptable, sock_net(sk),
1412 udp_sk(sk)->udp_port_hash);
1413 /* we must lock primary chain too */
1414 spin_lock_bh(&hslot->lock);
1415
1416 spin_lock(&hslot2->lock);
1417 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1418 hslot2->count--;
1419 spin_unlock(&hslot2->lock);
1420
1421 spin_lock(&nhslot2->lock);
1422 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
1423 &nhslot2->head);
1424 nhslot2->count++;
1425 spin_unlock(&nhslot2->lock);
1426
1427 spin_unlock_bh(&hslot->lock);
1428 }
1429 }
1430}
1431EXPORT_SYMBOL(udp_lib_rehash);
1432
1433static void udp_v4_rehash(struct sock *sk)
1434{
1435 u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1436 inet_sk(sk)->inet_rcv_saddr,
1437 inet_sk(sk)->inet_num);
1438 udp_lib_rehash(sk, new_hash);
1439}
1440
Herbert Xu93821772008-09-15 11:48:46 -07001441static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1442{
Tom Herbertfec5e652010-04-16 16:01:27 -07001443 int rc;
Herbert Xu93821772008-09-15 11:48:46 -07001444
Shawn Bohrer005ec972013-10-07 11:01:38 -05001445 if (inet_sk(sk)->inet_daddr) {
Tom Herbertbdeab992011-08-14 19:45:55 +00001446 sock_rps_save_rxhash(sk, skb);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001447 sk_mark_napi_id(sk, skb);
Eric Dumazet2c8c56e2014-11-11 05:54:28 -08001448 sk_incoming_cpu_update(sk);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001449 }
Tom Herbertfec5e652010-04-16 16:01:27 -07001450
Eric Dumazetd826eb12011-11-09 07:24:35 +00001451 rc = sock_queue_rcv_skb(sk, skb);
Eric Dumazet766e90372009-10-14 20:40:11 -07001452 if (rc < 0) {
1453 int is_udplite = IS_UDPLITE(sk);
1454
Herbert Xu93821772008-09-15 11:48:46 -07001455 /* Note that an ENOMEM error is charged twice */
Eric Dumazet766e90372009-10-14 20:40:11 -07001456 if (rc == -ENOMEM)
Herbert Xu93821772008-09-15 11:48:46 -07001457 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1458 is_udplite);
Eric Dumazet766e90372009-10-14 20:40:11 -07001459 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1460 kfree_skb(skb);
Satoru Moriya296f7ea2011-06-17 11:58:39 +00001461 trace_udp_fail_queue_rcv_skb(rc, sk);
Eric Dumazet766e90372009-10-14 20:40:11 -07001462 return -1;
Herbert Xu93821772008-09-15 11:48:46 -07001463 }
1464
1465 return 0;
1466
Herbert Xu93821772008-09-15 11:48:46 -07001467}
1468
Eric Dumazet447167b2012-04-11 23:05:28 +00001469static struct static_key udp_encap_needed __read_mostly;
1470void udp_encap_enable(void)
1471{
1472 if (!static_key_enabled(&udp_encap_needed))
1473 static_key_slow_inc(&udp_encap_needed);
1474}
1475EXPORT_SYMBOL(udp_encap_enable);
1476
David S. Millerdb8dac22008-03-06 16:22:02 -08001477/* returns:
1478 * -1: error
1479 * 0: success
1480 * >0: "udp encap" protocol resubmission
1481 *
1482 * Note that in the success and error cases, the skb is assumed to
1483 * have either been requeued or freed.
1484 */
Eric Dumazetc482c562009-07-17 00:26:32 +00001485int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -08001486{
1487 struct udp_sock *up = udp_sk(sk);
1488 int rc;
1489 int is_udplite = IS_UDPLITE(sk);
1490
1491 /*
1492 * Charge it to the socket, dropping if the queue is full.
1493 */
1494 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1495 goto drop;
1496 nf_reset(skb);
1497
Eric Dumazet447167b2012-04-11 23:05:28 +00001498 if (static_key_false(&udp_encap_needed) && up->encap_type) {
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001499 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1500
David S. Millerdb8dac22008-03-06 16:22:02 -08001501 /*
1502 * This is an encapsulation socket so pass the skb to
1503 * the socket's udp_encap_rcv() hook. Otherwise, just
1504 * fall through and pass this up the UDP socket.
1505 * up->encap_rcv() returns the following value:
1506 * =0 if skb was successfully passed to the encap
1507 * handler or was discarded by it.
1508 * >0 if skb should be passed on to UDP.
1509 * <0 if skb should be resubmitted as proto -N
1510 */
1511
1512 /* if we're overly short, let UDP handle it */
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001513 encap_rcv = ACCESS_ONCE(up->encap_rcv);
1514 if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001515 int ret;
1516
Tom Herbert0a809662014-05-07 16:52:39 -07001517 /* Verify checksum before giving to encap */
1518 if (udp_lib_checksum_complete(skb))
1519 goto csum_error;
1520
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001521 ret = encap_rcv(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001522 if (ret <= 0) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001523 UDP_INC_STATS_BH(sock_net(sk),
1524 UDP_MIB_INDATAGRAMS,
David S. Millerdb8dac22008-03-06 16:22:02 -08001525 is_udplite);
1526 return -ret;
1527 }
1528 }
1529
1530 /* FALLTHROUGH -- it's a UDP Packet */
1531 }
1532
1533 /*
1534 * UDP-Lite specific tests, ignored on UDP sockets
1535 */
1536 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1537
1538 /*
1539 * MIB statistics other than incrementing the error count are
1540 * disabled for the following two types of errors: these depend
1541 * on the application settings, not on the functioning of the
1542 * protocol stack as such.
1543 *
1544 * RFC 3828 here recommends (sec 3.3): "There should also be a
1545 * way ... to ... at least let the receiving application block
1546 * delivery of packets with coverage values less than a value
1547 * provided by the application."
1548 */
1549 if (up->pcrlen == 0) { /* full coverage was set */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001550 net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n",
1551 UDP_SKB_CB(skb)->cscov, skb->len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001552 goto drop;
1553 }
1554 /* The next case involves violating the min. coverage requested
1555 * by the receiver. This is subtle: if receiver wants x and x is
1556 * greater than the buffersize/MTU then receiver will complain
1557 * that it wants x while sender emits packets of smaller size y.
1558 * Therefore the above ...()->partial_cov statement is essential.
1559 */
1560 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
Joe Perchesba7a46f2014-11-11 10:59:17 -08001561 net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n",
1562 UDP_SKB_CB(skb)->cscov, up->pcrlen);
David S. Millerdb8dac22008-03-06 16:22:02 -08001563 goto drop;
1564 }
1565 }
1566
Eric Dumazet33d480c2011-08-11 19:30:52 +00001567 if (rcu_access_pointer(sk->sk_filter) &&
1568 udp_lib_checksum_complete(skb))
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001569 goto csum_error;
David S. Millerdb8dac22008-03-06 16:22:02 -08001570
Eric Dumazetc3774112010-04-27 15:13:20 -07001571
Sorin Dumitru274f4822014-07-22 21:16:51 +03001572 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
James M Leddy3e215c82014-06-25 17:38:13 -04001573 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1574 is_udplite);
Eric Dumazetc3774112010-04-27 15:13:20 -07001575 goto drop;
James M Leddy3e215c82014-06-25 17:38:13 -04001576 }
Eric Dumazetc3774112010-04-27 15:13:20 -07001577
Herbert Xu93821772008-09-15 11:48:46 -07001578 rc = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001579
Shawn Bohrerfbf88662013-10-07 11:01:40 -05001580 ipv4_pktinfo_prepare(sk, skb);
Herbert Xu93821772008-09-15 11:48:46 -07001581 bh_lock_sock(sk);
1582 if (!sock_owned_by_user(sk))
1583 rc = __udp_queue_rcv_skb(sk, skb);
Eric Dumazetf545a382012-04-22 23:34:26 +00001584 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
Zhu Yi55349792010-03-04 18:01:42 +00001585 bh_unlock_sock(sk);
1586 goto drop;
1587 }
Herbert Xu93821772008-09-15 11:48:46 -07001588 bh_unlock_sock(sk);
1589
1590 return rc;
David S. Millerdb8dac22008-03-06 16:22:02 -08001591
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001592csum_error:
1593 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001594drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001595 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001596 atomic_inc(&sk->sk_drops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001597 kfree_skb(skb);
1598 return -1;
1599}
1600
Eric Dumazet1240d132009-11-08 10:18:44 +00001601
1602static void flush_stack(struct sock **stack, unsigned int count,
1603 struct sk_buff *skb, unsigned int final)
1604{
1605 unsigned int i;
1606 struct sk_buff *skb1 = NULL;
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001607 struct sock *sk;
Eric Dumazet1240d132009-11-08 10:18:44 +00001608
1609 for (i = 0; i < count; i++) {
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001610 sk = stack[i];
Eric Dumazet1240d132009-11-08 10:18:44 +00001611 if (likely(skb1 == NULL))
1612 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
1613
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001614 if (!skb1) {
1615 atomic_inc(&sk->sk_drops);
1616 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1617 IS_UDPLITE(sk));
1618 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1619 IS_UDPLITE(sk));
1620 }
1621
1622 if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0)
Eric Dumazet1240d132009-11-08 10:18:44 +00001623 skb1 = NULL;
David Held2dc41cf2014-07-15 23:28:32 -04001624
1625 sock_put(sk);
Eric Dumazet1240d132009-11-08 10:18:44 +00001626 }
1627 if (unlikely(skb1))
1628 kfree_skb(skb1);
1629}
1630
Eric Dumazet97502232013-12-11 14:46:51 -08001631/* For TCP sockets, sk_rx_dst is protected by socket lock
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001632 * For UDP, we use xchg() to guard against concurrent changes.
Eric Dumazet97502232013-12-11 14:46:51 -08001633 */
1634static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
Shawn Bohrer421b3882013-10-07 11:01:39 -05001635{
Eric Dumazet97502232013-12-11 14:46:51 -08001636 struct dst_entry *old;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001637
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001638 dst_hold(dst);
1639 old = xchg(&sk->sk_rx_dst, dst);
1640 dst_release(old);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001641}
1642
David S. Millerdb8dac22008-03-06 16:22:02 -08001643/*
1644 * Multicasts and broadcasts go to each listener.
1645 *
Eric Dumazet1240d132009-11-08 10:18:44 +00001646 * Note: called only from the BH handler context.
David S. Millerdb8dac22008-03-06 16:22:02 -08001647 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001648static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001649 struct udphdr *uh,
1650 __be32 saddr, __be32 daddr,
Rick Jones36cbb242014-11-06 10:37:54 -08001651 struct udp_table *udptable,
1652 int proto)
David S. Millerdb8dac22008-03-06 16:22:02 -08001653{
Eric Dumazet1240d132009-11-08 10:18:44 +00001654 struct sock *sk, *stack[256 / sizeof(struct sock *)];
David Held5cf3d462014-07-15 23:28:31 -04001655 struct hlist_nulls_node *node;
1656 unsigned short hnum = ntohs(uh->dest);
1657 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
1658 int dif = skb->dev->ifindex;
David Held2dc41cf2014-07-15 23:28:32 -04001659 unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node);
1660 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
Rick Jones36cbb242014-11-06 10:37:54 -08001661 bool inner_flushed = false;
David Held2dc41cf2014-07-15 23:28:32 -04001662
1663 if (use_hash2) {
1664 hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) &
1665 udp_table.mask;
1666 hash2 = udp4_portaddr_hash(net, daddr, hnum) & udp_table.mask;
1667start_lookup:
1668 hslot = &udp_table.hash2[hash2];
1669 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
1670 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001671
Eric Dumazet645ca702008-10-29 01:41:45 -07001672 spin_lock(&hslot->lock);
David Held2dc41cf2014-07-15 23:28:32 -04001673 sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) {
David Held5cf3d462014-07-15 23:28:31 -04001674 if (__udp_is_mcast_sock(net, sk,
1675 uh->dest, daddr,
1676 uh->source, saddr,
1677 dif, hnum)) {
1678 if (unlikely(count == ARRAY_SIZE(stack))) {
1679 flush_stack(stack, count, skb, ~0);
Rick Jones36cbb242014-11-06 10:37:54 -08001680 inner_flushed = true;
David Held5cf3d462014-07-15 23:28:31 -04001681 count = 0;
1682 }
1683 stack[count++] = sk;
David Held2dc41cf2014-07-15 23:28:32 -04001684 sock_hold(sk);
Eric Dumazet1240d132009-11-08 10:18:44 +00001685 }
1686 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001687
Eric Dumazet645ca702008-10-29 01:41:45 -07001688 spin_unlock(&hslot->lock);
Eric Dumazet1240d132009-11-08 10:18:44 +00001689
David Held2dc41cf2014-07-15 23:28:32 -04001690 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
1691 if (use_hash2 && hash2 != hash2_any) {
1692 hash2 = hash2_any;
1693 goto start_lookup;
1694 }
1695
Eric Dumazet1240d132009-11-08 10:18:44 +00001696 /*
1697 * do the slow work with no lock held
1698 */
1699 if (count) {
1700 flush_stack(stack, count, skb, count - 1);
Eric Dumazet1240d132009-11-08 10:18:44 +00001701 } else {
Rick Jones36cbb242014-11-06 10:37:54 -08001702 if (!inner_flushed)
1703 UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
1704 proto == IPPROTO_UDPLITE);
1705 consume_skb(skb);
Eric Dumazet1240d132009-11-08 10:18:44 +00001706 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001707 return 0;
1708}
1709
1710/* Initialize UDP checksum. If exited with zero value (success),
1711 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1712 * Otherwise, csum completion requires chacksumming packet body,
1713 * including udp header and folding it to skb->csum.
1714 */
1715static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1716 int proto)
1717{
David S. Millerdb8dac22008-03-06 16:22:02 -08001718 int err;
1719
1720 UDP_SKB_CB(skb)->partial_cov = 0;
1721 UDP_SKB_CB(skb)->cscov = skb->len;
1722
1723 if (proto == IPPROTO_UDPLITE) {
1724 err = udplite_checksum_init(skb, uh);
1725 if (err)
1726 return err;
1727 }
1728
Tom Herberted70fcf2014-05-02 16:29:38 -07001729 return skb_checksum_init_zero_check(skb, proto, uh->check,
1730 inet_compute_pseudo);
David S. Millerdb8dac22008-03-06 16:22:02 -08001731}
1732
1733/*
1734 * All we need to do is get the socket, and then do a checksum.
1735 */
1736
Eric Dumazet645ca702008-10-29 01:41:45 -07001737int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
David S. Millerdb8dac22008-03-06 16:22:02 -08001738 int proto)
1739{
1740 struct sock *sk;
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001741 struct udphdr *uh;
David S. Millerdb8dac22008-03-06 16:22:02 -08001742 unsigned short ulen;
Eric Dumazetadf30902009-06-02 05:19:30 +00001743 struct rtable *rt = skb_rtable(skb);
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001744 __be32 saddr, daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001745 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001746
1747 /*
1748 * Validate the packet.
1749 */
1750 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1751 goto drop; /* No space for header. */
1752
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001753 uh = udp_hdr(skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001754 ulen = ntohs(uh->len);
Bjørn Morkccc2d972010-05-06 03:44:34 +00001755 saddr = ip_hdr(skb)->saddr;
1756 daddr = ip_hdr(skb)->daddr;
1757
David S. Millerdb8dac22008-03-06 16:22:02 -08001758 if (ulen > skb->len)
1759 goto short_packet;
1760
1761 if (proto == IPPROTO_UDP) {
1762 /* UDP validates ulen. */
1763 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1764 goto short_packet;
1765 uh = udp_hdr(skb);
1766 }
1767
1768 if (udp4_csum_init(skb, uh, proto))
1769 goto csum_error;
1770
Eric Dumazet8afdd992013-12-10 18:07:23 -08001771 sk = skb_steal_sock(skb);
1772 if (sk) {
Eric Dumazet97502232013-12-11 14:46:51 -08001773 struct dst_entry *dst = skb_dst(skb);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001774 int ret;
David S. Millerdb8dac22008-03-06 16:22:02 -08001775
Eric Dumazet97502232013-12-11 14:46:51 -08001776 if (unlikely(sk->sk_rx_dst != dst))
1777 udp_sk_rx_dst_set(sk, dst);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001778
1779 ret = udp_queue_rcv_skb(sk, skb);
Eric Dumazet8afdd992013-12-10 18:07:23 -08001780 sock_put(sk);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001781 /* a return value > 0 means to resubmit the input, but
1782 * it wants the return to be -protocol, or 0
1783 */
1784 if (ret > 0)
1785 return -ret;
1786 return 0;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001787 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001788
Fabian Frederickc18450a2014-11-04 20:48:41 +01001789 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1790 return __udp4_lib_mcast_deliver(net, skb, uh,
Rick Jones36cbb242014-11-06 10:37:54 -08001791 saddr, daddr, udptable, proto);
Fabian Frederickc18450a2014-11-04 20:48:41 +01001792
1793 sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
David S. Millerdb8dac22008-03-06 16:22:02 -08001794 if (sk != NULL) {
Eliezer Tamira5b50472013-06-10 11:40:00 +03001795 int ret;
1796
Tom Herbert2abb7cd2014-08-31 15:12:43 -07001797 if (udp_sk(sk)->convert_csum && uh->check && !IS_UDPLITE(sk))
1798 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
1799 inet_compute_pseudo);
1800
Eliezer Tamira5b50472013-06-10 11:40:00 +03001801 ret = udp_queue_rcv_skb(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001802 sock_put(sk);
1803
1804 /* a return value > 0 means to resubmit the input, but
1805 * it wants the return to be -protocol, or 0
1806 */
1807 if (ret > 0)
1808 return -ret;
1809 return 0;
1810 }
1811
1812 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1813 goto drop;
1814 nf_reset(skb);
1815
1816 /* No socket. Drop packet silently, if checksum is wrong */
1817 if (udp_lib_checksum_complete(skb))
1818 goto csum_error;
1819
Pavel Emelyanov02833282008-07-05 21:18:48 -07001820 UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001821 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1822
1823 /*
1824 * Hmm. We got an UDP packet to a port to which we
1825 * don't wanna listen. Ignore it.
1826 */
1827 kfree_skb(skb);
1828 return 0;
1829
1830short_packet:
Joe Perchesba7a46f2014-11-11 10:59:17 -08001831 net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
1832 proto == IPPROTO_UDPLITE ? "Lite" : "",
1833 &saddr, ntohs(uh->source),
1834 ulen, skb->len,
1835 &daddr, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001836 goto drop;
1837
1838csum_error:
1839 /*
1840 * RFC1122: OK. Discards the bad packet silently (as far as
1841 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1842 */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001843 net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
1844 proto == IPPROTO_UDPLITE ? "Lite" : "",
1845 &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
1846 ulen);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001847 UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001848drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001849 UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001850 kfree_skb(skb);
1851 return 0;
1852}
1853
Shawn Bohrer421b3882013-10-07 11:01:39 -05001854/* We can only early demux multicast if there is a single matching socket.
1855 * If more than one socket found returns NULL
1856 */
1857static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net,
1858 __be16 loc_port, __be32 loc_addr,
1859 __be16 rmt_port, __be32 rmt_addr,
1860 int dif)
1861{
1862 struct sock *sk, *result;
1863 struct hlist_nulls_node *node;
1864 unsigned short hnum = ntohs(loc_port);
1865 unsigned int count, slot = udp_hashfn(net, hnum, udp_table.mask);
1866 struct udp_hslot *hslot = &udp_table.hash[slot];
1867
Eric Dumazet63c6f812014-06-12 16:13:06 -07001868 /* Do not bother scanning a too big list */
1869 if (hslot->count > 10)
1870 return NULL;
1871
Shawn Bohrer421b3882013-10-07 11:01:39 -05001872 rcu_read_lock();
1873begin:
1874 count = 0;
1875 result = NULL;
1876 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
1877 if (__udp_is_mcast_sock(net, sk,
1878 loc_port, loc_addr,
1879 rmt_port, rmt_addr,
1880 dif, hnum)) {
1881 result = sk;
1882 ++count;
1883 }
1884 }
1885 /*
1886 * if the nulls value we got at the end of this lookup is
1887 * not the expected one, we must restart lookup.
1888 * We probably met an item that was moved to another chain.
1889 */
1890 if (get_nulls_value(node) != slot)
1891 goto begin;
1892
1893 if (result) {
1894 if (count != 1 ||
1895 unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
1896 result = NULL;
Eric Dumazetf69b9232013-10-08 21:47:29 -07001897 else if (unlikely(!__udp_is_mcast_sock(net, result,
Shawn Bohrer421b3882013-10-07 11:01:39 -05001898 loc_port, loc_addr,
1899 rmt_port, rmt_addr,
1900 dif, hnum))) {
1901 sock_put(result);
1902 result = NULL;
1903 }
1904 }
1905 rcu_read_unlock();
1906 return result;
1907}
1908
1909/* For unicast we should only early demux connected sockets or we can
1910 * break forwarding setups. The chains here can be long so only check
1911 * if the first socket is an exact match and if not move on.
1912 */
1913static struct sock *__udp4_lib_demux_lookup(struct net *net,
1914 __be16 loc_port, __be32 loc_addr,
1915 __be16 rmt_port, __be32 rmt_addr,
1916 int dif)
1917{
1918 struct sock *sk, *result;
1919 struct hlist_nulls_node *node;
1920 unsigned short hnum = ntohs(loc_port);
1921 unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum);
1922 unsigned int slot2 = hash2 & udp_table.mask;
1923 struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
Joe Perchesc7228312014-05-13 20:30:07 -07001924 INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001925 const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
1926
1927 rcu_read_lock();
1928 result = NULL;
1929 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
1930 if (INET_MATCH(sk, net, acookie,
1931 rmt_addr, loc_addr, ports, dif))
1932 result = sk;
1933 /* Only check first socket in chain */
1934 break;
1935 }
1936
1937 if (result) {
1938 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
1939 result = NULL;
1940 else if (unlikely(!INET_MATCH(sk, net, acookie,
1941 rmt_addr, loc_addr,
1942 ports, dif))) {
1943 sock_put(result);
1944 result = NULL;
1945 }
1946 }
1947 rcu_read_unlock();
1948 return result;
1949}
1950
1951void udp_v4_early_demux(struct sk_buff *skb)
1952{
Eric Dumazet610438b2013-12-11 08:10:05 -08001953 struct net *net = dev_net(skb->dev);
1954 const struct iphdr *iph;
1955 const struct udphdr *uh;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001956 struct sock *sk;
1957 struct dst_entry *dst;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001958 int dif = skb->dev->ifindex;
1959
1960 /* validate the packet */
1961 if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr)))
1962 return;
1963
Eric Dumazet610438b2013-12-11 08:10:05 -08001964 iph = ip_hdr(skb);
1965 uh = udp_hdr(skb);
1966
Shawn Bohrer421b3882013-10-07 11:01:39 -05001967 if (skb->pkt_type == PACKET_BROADCAST ||
1968 skb->pkt_type == PACKET_MULTICAST)
1969 sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr,
1970 uh->source, iph->saddr, dif);
1971 else if (skb->pkt_type == PACKET_HOST)
1972 sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr,
1973 uh->source, iph->saddr, dif);
1974 else
1975 return;
1976
1977 if (!sk)
1978 return;
1979
1980 skb->sk = sk;
Alexander Duyck82eabd92014-09-04 13:32:11 -04001981 skb->destructor = sock_efree;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001982 dst = sk->sk_rx_dst;
1983
1984 if (dst)
1985 dst = dst_check(dst, 0);
1986 if (dst)
1987 skb_dst_set_noref(skb, dst);
1988}
1989
David S. Millerdb8dac22008-03-06 16:22:02 -08001990int udp_rcv(struct sk_buff *skb)
1991{
Eric Dumazet645ca702008-10-29 01:41:45 -07001992 return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
David S. Millerdb8dac22008-03-06 16:22:02 -08001993}
1994
Brian Haley7d06b2e2008-06-14 17:04:49 -07001995void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08001996{
Tom Parkin44046a52013-03-19 06:11:12 +00001997 struct udp_sock *up = udp_sk(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001998 bool slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001999 udp_flush_pending_frames(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00002000 unlock_sock_fast(sk, slow);
Tom Parkin44046a52013-03-19 06:11:12 +00002001 if (static_key_false(&udp_encap_needed) && up->encap_type) {
2002 void (*encap_destroy)(struct sock *sk);
2003 encap_destroy = ACCESS_ONCE(up->encap_destroy);
2004 if (encap_destroy)
2005 encap_destroy(sk);
2006 }
David S. Millerdb8dac22008-03-06 16:22:02 -08002007}
2008
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009/*
2010 * Socket option code for UDP
2011 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002012int udp_lib_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002013 char __user *optval, unsigned int optlen,
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002014 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015{
2016 struct udp_sock *up = udp_sk(sk);
Tom Herbert1c194482014-05-23 08:47:32 -07002017 int val, valbool;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11002019 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020
Eric Dumazetc482c562009-07-17 00:26:32 +00002021 if (optlen < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 return -EINVAL;
2023
2024 if (get_user(val, (int __user *)optval))
2025 return -EFAULT;
2026
Tom Herbert1c194482014-05-23 08:47:32 -07002027 valbool = val ? 1 : 0;
2028
Stephen Hemminger6516c652007-03-08 20:41:55 -08002029 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 case UDP_CORK:
2031 if (val != 0) {
2032 up->corkflag = 1;
2033 } else {
2034 up->corkflag = 0;
2035 lock_sock(sk);
Joe Perches4243cdc2014-11-11 21:59:20 -08002036 push_pending_frames(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 release_sock(sk);
2038 }
2039 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002040
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 case UDP_ENCAP:
2042 switch (val) {
2043 case 0:
2044 case UDP_ENCAP_ESPINUDP:
2045 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07002046 up->encap_rcv = xfrm4_udp_encap_rcv;
2047 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07002048 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 up->encap_type = val;
Eric Dumazet447167b2012-04-11 23:05:28 +00002050 udp_encap_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051 break;
2052 default:
2053 err = -ENOPROTOOPT;
2054 break;
2055 }
2056 break;
2057
Tom Herbert1c194482014-05-23 08:47:32 -07002058 case UDP_NO_CHECK6_TX:
2059 up->no_check6_tx = valbool;
2060 break;
2061
2062 case UDP_NO_CHECK6_RX:
2063 up->no_check6_rx = valbool;
2064 break;
2065
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002066 /*
2067 * UDP-Lite's partial checksum coverage (RFC 3828).
2068 */
2069 /* The sender sets actual checksum coverage length via this option.
2070 * The case coverage > packet length is handled by send module. */
2071 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002072 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002073 return -ENOPROTOOPT;
2074 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
2075 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002076 else if (val > USHRT_MAX)
2077 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002078 up->pcslen = val;
2079 up->pcflag |= UDPLITE_SEND_CC;
2080 break;
2081
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002082 /* The receiver specifies a minimum checksum coverage value. To make
2083 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002084 * used, this again means full checksum coverage. */
2085 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002086 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002087 return -ENOPROTOOPT;
2088 if (val != 0 && val < 8) /* Avoid silly minimal values. */
2089 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002090 else if (val > USHRT_MAX)
2091 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002092 up->pcrlen = val;
2093 up->pcflag |= UDPLITE_RECV_CC;
2094 break;
2095
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 default:
2097 err = -ENOPROTOOPT;
2098 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002099 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100
2101 return err;
2102}
Eric Dumazetc482c562009-07-17 00:26:32 +00002103EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104
David S. Millerdb8dac22008-03-06 16:22:02 -08002105int udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002106 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002107{
2108 if (level == SOL_UDP || level == SOL_UDPLITE)
2109 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2110 udp_push_pending_frames);
2111 return ip_setsockopt(sk, level, optname, optval, optlen);
2112}
2113
2114#ifdef CONFIG_COMPAT
2115int compat_udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002116 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002117{
2118 if (level == SOL_UDP || level == SOL_UDPLITE)
2119 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2120 udp_push_pending_frames);
2121 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
2122}
2123#endif
2124
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002125int udp_lib_getsockopt(struct sock *sk, int level, int optname,
2126 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127{
2128 struct udp_sock *up = udp_sk(sk);
2129 int val, len;
2130
Eric Dumazetc482c562009-07-17 00:26:32 +00002131 if (get_user(len, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002132 return -EFAULT;
2133
2134 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002135
Stephen Hemminger6516c652007-03-08 20:41:55 -08002136 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002137 return -EINVAL;
2138
Stephen Hemminger6516c652007-03-08 20:41:55 -08002139 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140 case UDP_CORK:
2141 val = up->corkflag;
2142 break;
2143
2144 case UDP_ENCAP:
2145 val = up->encap_type;
2146 break;
2147
Tom Herbert1c194482014-05-23 08:47:32 -07002148 case UDP_NO_CHECK6_TX:
2149 val = up->no_check6_tx;
2150 break;
2151
2152 case UDP_NO_CHECK6_RX:
2153 val = up->no_check6_rx;
2154 break;
2155
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002156 /* The following two cannot be changed on UDP sockets, the return is
2157 * always 0 (which corresponds to the full checksum coverage of UDP). */
2158 case UDPLITE_SEND_CSCOV:
2159 val = up->pcslen;
2160 break;
2161
2162 case UDPLITE_RECV_CSCOV:
2163 val = up->pcrlen;
2164 break;
2165
Linus Torvalds1da177e2005-04-16 15:20:36 -07002166 default:
2167 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002168 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002169
Stephen Hemminger6516c652007-03-08 20:41:55 -08002170 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002171 return -EFAULT;
Eric Dumazetc482c562009-07-17 00:26:32 +00002172 if (copy_to_user(optval, &val, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002174 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002175}
Eric Dumazetc482c562009-07-17 00:26:32 +00002176EXPORT_SYMBOL(udp_lib_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177
David S. Millerdb8dac22008-03-06 16:22:02 -08002178int udp_getsockopt(struct sock *sk, int level, int optname,
2179 char __user *optval, int __user *optlen)
2180{
2181 if (level == SOL_UDP || level == SOL_UDPLITE)
2182 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2183 return ip_getsockopt(sk, level, optname, optval, optlen);
2184}
2185
2186#ifdef CONFIG_COMPAT
2187int compat_udp_getsockopt(struct sock *sk, int level, int optname,
2188 char __user *optval, int __user *optlen)
2189{
2190 if (level == SOL_UDP || level == SOL_UDPLITE)
2191 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2192 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
2193}
2194#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195/**
2196 * udp_poll - wait for a UDP event.
2197 * @file - file struct
2198 * @sock - socket
2199 * @wait - poll table
2200 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002201 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202 * blocking sockets. If application is using a blocking fd
2203 * and a packet with checksum error is in the queue;
2204 * then it could get return from select indicating data available
2205 * but then block when reading it. Add special case code
2206 * to work around these arguably broken applications.
2207 */
2208unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
2209{
2210 unsigned int mask = datagram_poll(file, sock, wait);
2211 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002212
David Majnemerc3f1dba2013-05-31 13:15:38 +00002213 sock_rps_record_flow(sk);
2214
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215 /* Check for false positives due to checksum errors */
Eric Dumazet85584672009-10-09 04:43:40 +00002216 if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
2217 !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk))
2218 mask &= ~(POLLIN | POLLRDNORM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219
2220 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002221
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222}
Eric Dumazetc482c562009-07-17 00:26:32 +00002223EXPORT_SYMBOL(udp_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002224
David S. Millerdb8dac22008-03-06 16:22:02 -08002225struct proto udp_prot = {
2226 .name = "UDP",
2227 .owner = THIS_MODULE,
2228 .close = udp_lib_close,
2229 .connect = ip4_datagram_connect,
2230 .disconnect = udp_disconnect,
2231 .ioctl = udp_ioctl,
2232 .destroy = udp_destroy_sock,
2233 .setsockopt = udp_setsockopt,
2234 .getsockopt = udp_getsockopt,
2235 .sendmsg = udp_sendmsg,
2236 .recvmsg = udp_recvmsg,
2237 .sendpage = udp_sendpage,
Herbert Xu93821772008-09-15 11:48:46 -07002238 .backlog_rcv = __udp_queue_rcv_skb,
Steffen Klassert8141ed92013-01-21 02:00:03 +00002239 .release_cb = ip4_datagram_release_cb,
David S. Millerdb8dac22008-03-06 16:22:02 -08002240 .hash = udp_lib_hash,
2241 .unhash = udp_lib_unhash,
Eric Dumazet719f8352010-09-08 05:08:44 +00002242 .rehash = udp_v4_rehash,
David S. Millerdb8dac22008-03-06 16:22:02 -08002243 .get_port = udp_v4_get_port,
2244 .memory_allocated = &udp_memory_allocated,
2245 .sysctl_mem = sysctl_udp_mem,
2246 .sysctl_wmem = &sysctl_udp_wmem_min,
2247 .sysctl_rmem = &sysctl_udp_rmem_min,
2248 .obj_size = sizeof(struct udp_sock),
Eric Dumazet271b72c2008-10-29 02:11:14 -07002249 .slab_flags = SLAB_DESTROY_BY_RCU,
Eric Dumazet645ca702008-10-29 01:41:45 -07002250 .h.udp_table = &udp_table,
David S. Millerdb8dac22008-03-06 16:22:02 -08002251#ifdef CONFIG_COMPAT
2252 .compat_setsockopt = compat_udp_setsockopt,
2253 .compat_getsockopt = compat_udp_getsockopt,
2254#endif
Octavian Purdilafcbdf092010-12-16 14:26:56 -08002255 .clear_sk = sk_prot_clear_portaddr_nulls,
David S. Millerdb8dac22008-03-06 16:22:02 -08002256};
Eric Dumazetc482c562009-07-17 00:26:32 +00002257EXPORT_SYMBOL(udp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002258
2259/* ------------------------------------------------------------------------ */
2260#ifdef CONFIG_PROC_FS
2261
Eric Dumazet645ca702008-10-29 01:41:45 -07002262static struct sock *udp_get_first(struct seq_file *seq, int start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263{
2264 struct sock *sk;
2265 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002266 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002268 for (state->bucket = start; state->bucket <= state->udp_table->mask;
2269 ++state->bucket) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002270 struct hlist_nulls_node *node;
Eric Dumazet645ca702008-10-29 01:41:45 -07002271 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002272
2273 if (hlist_nulls_empty(&hslot->head))
2274 continue;
2275
Eric Dumazet645ca702008-10-29 01:41:45 -07002276 spin_lock_bh(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08002277 sk_nulls_for_each(sk, node, &hslot->head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002278 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002279 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002280 if (sk->sk_family == state->family)
2281 goto found;
2282 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002283 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284 }
2285 sk = NULL;
2286found:
2287 return sk;
2288}
2289
2290static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
2291{
2292 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002293 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294
2295 do {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002296 sk = sk_nulls_next(sk);
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002297 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298
Eric Dumazet645ca702008-10-29 01:41:45 -07002299 if (!sk) {
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002300 if (state->bucket <= state->udp_table->mask)
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002301 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07002302 return udp_get_first(seq, state->bucket + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002303 }
2304 return sk;
2305}
2306
2307static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2308{
Eric Dumazet645ca702008-10-29 01:41:45 -07002309 struct sock *sk = udp_get_first(seq, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002310
2311 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08002312 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313 --pos;
2314 return pos ? NULL : sk;
2315}
2316
2317static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2318{
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002319 struct udp_iter_state *state = seq->private;
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002320 state->bucket = MAX_UDP_PORTS;
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002321
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002322 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323}
2324
2325static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2326{
2327 struct sock *sk;
2328
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002329 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002330 sk = udp_get_idx(seq, 0);
2331 else
2332 sk = udp_get_next(seq, v);
2333
2334 ++*pos;
2335 return sk;
2336}
2337
2338static void udp_seq_stop(struct seq_file *seq, void *v)
2339{
Eric Dumazet645ca702008-10-29 01:41:45 -07002340 struct udp_iter_state *state = seq->private;
2341
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002342 if (state->bucket <= state->udp_table->mask)
Eric Dumazet645ca702008-10-29 01:41:45 -07002343 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344}
2345
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002346int udp_seq_open(struct inode *inode, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347{
Al Virod9dda782013-03-31 18:16:14 -04002348 struct udp_seq_afinfo *afinfo = PDE_DATA(inode);
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002349 struct udp_iter_state *s;
2350 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002352 err = seq_open_net(inode, file, &afinfo->seq_ops,
2353 sizeof(struct udp_iter_state));
2354 if (err < 0)
2355 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002356
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002357 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358 s->family = afinfo->family;
Eric Dumazet645ca702008-10-29 01:41:45 -07002359 s->udp_table = afinfo->udp_table;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002360 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002361}
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002362EXPORT_SYMBOL(udp_seq_open);
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002363
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002365int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366{
2367 struct proc_dir_entry *p;
2368 int rc = 0;
2369
Denis V. Lunevdda61922008-03-28 18:24:26 -07002370 afinfo->seq_ops.start = udp_seq_start;
2371 afinfo->seq_ops.next = udp_seq_next;
2372 afinfo->seq_ops.stop = udp_seq_stop;
2373
Denis V. Lunev84841c32008-05-02 04:10:08 -07002374 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002375 afinfo->seq_fops, afinfo);
Denis V. Lunev84841c32008-05-02 04:10:08 -07002376 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377 rc = -ENOMEM;
2378 return rc;
2379}
Eric Dumazetc482c562009-07-17 00:26:32 +00002380EXPORT_SYMBOL(udp_proc_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002382void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002383{
Gao fengece31ff2013-02-18 01:34:56 +00002384 remove_proc_entry(afinfo->name, net->proc_net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002385}
Eric Dumazetc482c562009-07-17 00:26:32 +00002386EXPORT_SYMBOL(udp_proc_unregister);
David S. Millerdb8dac22008-03-06 16:22:02 -08002387
2388/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002389static void udp4_format_sock(struct sock *sp, struct seq_file *f,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002390 int bucket)
David S. Millerdb8dac22008-03-06 16:22:02 -08002391{
2392 struct inet_sock *inet = inet_sk(sp);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002393 __be32 dest = inet->inet_daddr;
2394 __be32 src = inet->inet_rcv_saddr;
2395 __u16 destp = ntohs(inet->inet_dport);
2396 __u16 srcp = ntohs(inet->inet_sport);
David S. Millerdb8dac22008-03-06 16:22:02 -08002397
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002398 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
Tetsuo Handa652586d2013-11-14 14:31:57 -08002399 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
David S. Millerdb8dac22008-03-06 16:22:02 -08002400 bucket, src, srcp, dest, destp, sp->sk_state,
Eric Dumazet31e6d362009-06-17 19:05:41 -07002401 sk_wmem_alloc_get(sp),
2402 sk_rmem_alloc_get(sp),
Eric W. Biedermana7cb5a42012-05-24 01:10:10 -06002403 0, 0L, 0,
2404 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
2405 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002406 atomic_read(&sp->sk_refcnt), sp,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002407 atomic_read(&sp->sk_drops));
David S. Millerdb8dac22008-03-06 16:22:02 -08002408}
2409
2410int udp4_seq_show(struct seq_file *seq, void *v)
2411{
Tetsuo Handa652586d2013-11-14 14:31:57 -08002412 seq_setwidth(seq, 127);
David S. Millerdb8dac22008-03-06 16:22:02 -08002413 if (v == SEQ_START_TOKEN)
Tetsuo Handa652586d2013-11-14 14:31:57 -08002414 seq_puts(seq, " sl local_address rem_address st tx_queue "
David S. Millerdb8dac22008-03-06 16:22:02 -08002415 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002416 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08002417 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08002418 struct udp_iter_state *state = seq->private;
2419
Tetsuo Handa652586d2013-11-14 14:31:57 -08002420 udp4_format_sock(v, seq, state->bucket);
David S. Millerdb8dac22008-03-06 16:22:02 -08002421 }
Tetsuo Handa652586d2013-11-14 14:31:57 -08002422 seq_pad(seq, '\n');
David S. Millerdb8dac22008-03-06 16:22:02 -08002423 return 0;
2424}
2425
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002426static const struct file_operations udp_afinfo_seq_fops = {
2427 .owner = THIS_MODULE,
2428 .open = udp_seq_open,
2429 .read = seq_read,
2430 .llseek = seq_lseek,
2431 .release = seq_release_net
2432};
2433
David S. Millerdb8dac22008-03-06 16:22:02 -08002434/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08002435static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08002436 .name = "udp",
2437 .family = AF_INET,
Eric Dumazet645ca702008-10-29 01:41:45 -07002438 .udp_table = &udp_table,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002439 .seq_fops = &udp_afinfo_seq_fops,
Denis V. Lunevdda61922008-03-28 18:24:26 -07002440 .seq_ops = {
2441 .show = udp4_seq_show,
2442 },
David S. Millerdb8dac22008-03-06 16:22:02 -08002443};
2444
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002445static int __net_init udp4_proc_init_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002446{
2447 return udp_proc_register(net, &udp4_seq_afinfo);
2448}
2449
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002450static void __net_exit udp4_proc_exit_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002451{
2452 udp_proc_unregister(net, &udp4_seq_afinfo);
2453}
2454
2455static struct pernet_operations udp4_net_ops = {
2456 .init = udp4_proc_init_net,
2457 .exit = udp4_proc_exit_net,
2458};
2459
David S. Millerdb8dac22008-03-06 16:22:02 -08002460int __init udp4_proc_init(void)
2461{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002462 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002463}
2464
2465void udp4_proc_exit(void)
2466{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002467 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002468}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469#endif /* CONFIG_PROC_FS */
2470
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002471static __initdata unsigned long uhash_entries;
2472static int __init set_uhash_entries(char *str)
Eric Dumazet645ca702008-10-29 01:41:45 -07002473{
Eldad Zack413c27d2012-05-19 14:13:18 +00002474 ssize_t ret;
2475
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002476 if (!str)
2477 return 0;
Eldad Zack413c27d2012-05-19 14:13:18 +00002478
2479 ret = kstrtoul(str, 0, &uhash_entries);
2480 if (ret)
2481 return 0;
2482
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002483 if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2484 uhash_entries = UDP_HTABLE_SIZE_MIN;
2485 return 1;
2486}
2487__setup("uhash_entries=", set_uhash_entries);
Eric Dumazet645ca702008-10-29 01:41:45 -07002488
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002489void __init udp_table_init(struct udp_table *table, const char *name)
2490{
2491 unsigned int i;
2492
Tim Bird31fe62b2012-05-23 13:33:35 +00002493 table->hash = alloc_large_system_hash(name,
2494 2 * sizeof(struct udp_hslot),
2495 uhash_entries,
2496 21, /* one slot per 2 MB */
2497 0,
2498 &table->log,
2499 &table->mask,
2500 UDP_HTABLE_SIZE_MIN,
2501 64 * 1024);
2502
Eric Dumazet512615b2009-11-08 10:17:58 +00002503 table->hash2 = table->hash + (table->mask + 1);
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002504 for (i = 0; i <= table->mask; i++) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002505 INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00002506 table->hash[i].count = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -07002507 spin_lock_init(&table->hash[i].lock);
2508 }
Eric Dumazet512615b2009-11-08 10:17:58 +00002509 for (i = 0; i <= table->mask; i++) {
2510 INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i);
2511 table->hash2[i].count = 0;
2512 spin_lock_init(&table->hash2[i].lock);
2513 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002514}
2515
Hideo Aoki95766ff2007-12-31 00:29:24 -08002516void __init udp_init(void)
2517{
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002518 unsigned long limit;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002519
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002520 udp_table_init(&udp_table, "UDP");
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002521 limit = nr_free_buffer_pages() / 8;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002522 limit = max(limit, 128UL);
2523 sysctl_udp_mem[0] = limit / 4 * 3;
2524 sysctl_udp_mem[1] = limit;
2525 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2526
2527 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2528 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2529}