blob: 95d2f198017ef1a6db687ddfe405a13830a37f3f [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>
Shawn Bohrer6e540302015-06-03 16:27:38 -050093#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094#include <linux/in.h>
95#include <linux/errno.h>
96#include <linux/timer.h>
97#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include <linux/inet.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099#include <linux/netdevice.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +0900100#include <linux/slab.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -0700101#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102#include <linux/skbuff.h>
103#include <linux/proc_fs.h>
104#include <linux/seq_file.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200105#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106#include <net/icmp.h>
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"
Craig Galleke32ea7e2016-01-04 17:41:46 -0500116#include <net/sock_reuseport.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000118struct udp_table udp_table __read_mostly;
Eric Dumazet645ca702008-10-29 01:41:45 -0700119EXPORT_SYMBOL(udp_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
Eric Dumazet8d987e52010-11-09 23:24:26 +0000121long sysctl_udp_mem[3] __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800122EXPORT_SYMBOL(sysctl_udp_mem);
Eric Dumazetc482c562009-07-17 00:26:32 +0000123
124int sysctl_udp_rmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800125EXPORT_SYMBOL(sysctl_udp_rmem_min);
Eric Dumazetc482c562009-07-17 00:26:32 +0000126
127int sysctl_udp_wmem_min __read_mostly;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800128EXPORT_SYMBOL(sysctl_udp_wmem_min);
129
Eric Dumazet8d987e52010-11-09 23:24:26 +0000130atomic_long_t udp_memory_allocated;
Hideo Aoki95766ff2007-12-31 00:29:24 -0800131EXPORT_SYMBOL(udp_memory_allocated);
132
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000133#define MAX_UDP_PORTS 65536
134#define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN)
Eric Dumazet98322f22009-01-26 21:35:35 -0800135
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700136static int udp_lib_lport_inuse(struct net *net, __u16 num,
Eric Dumazet645ca702008-10-29 01:41:45 -0700137 const struct udp_hslot *hslot,
Eric Dumazet98322f22009-01-26 21:35:35 -0800138 unsigned long *bitmap,
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700139 struct sock *sk,
140 int (*saddr_comp)(const struct sock *sk1,
Craig Galleke32ea7e2016-01-04 17:41:46 -0500141 const struct sock *sk2,
142 bool match_wildcard),
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000143 unsigned int log)
Gerrit Renker25030a72006-08-26 20:06:05 -0700144{
Eric Dumazetf24d43c2008-10-09 14:51:27 -0700145 struct sock *sk2;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800146 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000147 kuid_t uid = sock_i_uid(sk);
Gerrit Renker25030a72006-08-26 20:06:05 -0700148
Joe Perches4243cdc2014-11-11 21:59:20 -0800149 sk_nulls_for_each(sk2, node, &hslot->head) {
Joe Perches9d4fb272009-11-23 10:41:23 -0800150 if (net_eq(sock_net(sk2), net) &&
151 sk2 != sk &&
Eric Dumazetd4cada42009-11-08 10:17:30 +0000152 (bitmap || udp_sk(sk2)->udp_port_hash == num) &&
Joe Perches9d4fb272009-11-23 10:41:23 -0800153 (!sk2->sk_reuse || !sk->sk_reuse) &&
154 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
155 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000156 (!sk2->sk_reuseport || !sk->sk_reuseport ||
Craig Galleke32ea7e2016-01-04 17:41:46 -0500157 rcu_access_pointer(sk->sk_reuseport_cb) ||
Joe Perches4243cdc2014-11-11 21:59:20 -0800158 !uid_eq(uid, sock_i_uid(sk2))) &&
Craig Galleke32ea7e2016-01-04 17:41:46 -0500159 saddr_comp(sk, sk2, true)) {
Joe Perches4243cdc2014-11-11 21:59:20 -0800160 if (!bitmap)
Eric Dumazet98322f22009-01-26 21:35:35 -0800161 return 1;
Joe Perches4243cdc2014-11-11 21:59:20 -0800162 __set_bit(udp_sk(sk2)->udp_port_hash >> log, bitmap);
Eric Dumazet98322f22009-01-26 21:35:35 -0800163 }
Joe Perches4243cdc2014-11-11 21:59:20 -0800164 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700165 return 0;
166}
167
Eric Dumazet30fff922009-11-09 05:26:33 +0000168/*
169 * Note: we still hold spinlock of primary hash chain, so no other writer
170 * can insert/delete a socket with local_port == num
171 */
172static int udp_lib_lport_inuse2(struct net *net, __u16 num,
Joe Perches4243cdc2014-11-11 21:59:20 -0800173 struct udp_hslot *hslot2,
174 struct sock *sk,
175 int (*saddr_comp)(const struct sock *sk1,
Craig Galleke32ea7e2016-01-04 17:41:46 -0500176 const struct sock *sk2,
177 bool match_wildcard))
Eric Dumazet30fff922009-11-09 05:26:33 +0000178{
179 struct sock *sk2;
180 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000181 kuid_t uid = sock_i_uid(sk);
Eric Dumazet30fff922009-11-09 05:26:33 +0000182 int res = 0;
183
184 spin_lock(&hslot2->lock);
Joe Perches4243cdc2014-11-11 21:59:20 -0800185 udp_portaddr_for_each_entry(sk2, node, &hslot2->head) {
Joe Perches9d4fb272009-11-23 10:41:23 -0800186 if (net_eq(sock_net(sk2), net) &&
187 sk2 != sk &&
188 (udp_sk(sk2)->udp_port_hash == num) &&
189 (!sk2->sk_reuse || !sk->sk_reuse) &&
190 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
191 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
Tom Herbertba418fa2013-01-22 09:50:32 +0000192 (!sk2->sk_reuseport || !sk->sk_reuseport ||
Craig Galleke32ea7e2016-01-04 17:41:46 -0500193 rcu_access_pointer(sk->sk_reuseport_cb) ||
Joe Perches4243cdc2014-11-11 21:59:20 -0800194 !uid_eq(uid, sock_i_uid(sk2))) &&
Craig Galleke32ea7e2016-01-04 17:41:46 -0500195 saddr_comp(sk, sk2, true)) {
Eric Dumazet30fff922009-11-09 05:26:33 +0000196 res = 1;
197 break;
198 }
Joe Perches4243cdc2014-11-11 21:59:20 -0800199 }
Eric Dumazet30fff922009-11-09 05:26:33 +0000200 spin_unlock(&hslot2->lock);
201 return res;
202}
203
Craig Galleke32ea7e2016-01-04 17:41:46 -0500204static int udp_reuseport_add_sock(struct sock *sk, struct udp_hslot *hslot,
205 int (*saddr_same)(const struct sock *sk1,
206 const struct sock *sk2,
207 bool match_wildcard))
208{
209 struct net *net = sock_net(sk);
210 struct hlist_nulls_node *node;
211 kuid_t uid = sock_i_uid(sk);
212 struct sock *sk2;
213
214 sk_nulls_for_each(sk2, node, &hslot->head) {
215 if (net_eq(sock_net(sk2), net) &&
216 sk2 != sk &&
217 sk2->sk_family == sk->sk_family &&
218 ipv6_only_sock(sk2) == ipv6_only_sock(sk) &&
219 (udp_sk(sk2)->udp_port_hash == udp_sk(sk)->udp_port_hash) &&
220 (sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
221 sk2->sk_reuseport && uid_eq(uid, sock_i_uid(sk2)) &&
222 (*saddr_same)(sk, sk2, false)) {
223 return reuseport_add_sock(sk, sk2);
224 }
225 }
226
227 /* Initial allocation may have already happened via setsockopt */
228 if (!rcu_access_pointer(sk->sk_reuseport_cb))
229 return reuseport_alloc(sk);
230 return 0;
231}
232
Gerrit Renker25030a72006-08-26 20:06:05 -0700233/**
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700234 * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
Gerrit Renker25030a72006-08-26 20:06:05 -0700235 *
236 * @sk: socket struct in question
237 * @snum: port number to look up
David S. Millerdf2bc452007-06-05 15:18:43 -0700238 * @saddr_comp: AF-dependent comparison of bound local IP addresses
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300239 * @hash2_nulladdr: AF-dependent hash value in secondary hash chains,
Eric Dumazet30fff922009-11-09 05:26:33 +0000240 * with NULL address
Gerrit Renker25030a72006-08-26 20:06:05 -0700241 */
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700242int udp_lib_get_port(struct sock *sk, unsigned short snum,
Joe Perches4243cdc2014-11-11 21:59:20 -0800243 int (*saddr_comp)(const struct sock *sk1,
Craig Galleke32ea7e2016-01-04 17:41:46 -0500244 const struct sock *sk2,
245 bool match_wildcard),
Eric Dumazet30fff922009-11-09 05:26:33 +0000246 unsigned int hash2_nulladdr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247{
Eric Dumazet512615b2009-11-08 10:17:58 +0000248 struct udp_hslot *hslot, *hslot2;
Eric Dumazet645ca702008-10-29 01:41:45 -0700249 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Gerrit Renker25030a72006-08-26 20:06:05 -0700250 int error = 1;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900251 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252
Stephen Hemminger32c1da72007-08-24 23:09:41 -0700253 if (!snum) {
Eric Dumazet9088c562008-10-08 11:44:17 -0700254 int low, high, remaining;
Eric Dumazet95c96172012-04-15 05:58:06 +0000255 unsigned int rand;
Eric Dumazet98322f22009-01-26 21:35:35 -0800256 unsigned short first, last;
257 DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
Eric W. Biederman0bbf87d2013-09-28 14:10:59 -0700259 inet_get_local_port_range(net, &low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700260 remaining = (high - low) + 1;
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700261
Aruna-Hewapathirane63862b52014-01-11 07:15:59 -0500262 rand = prandom_u32();
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200263 first = reciprocal_scale(rand, remaining) + low;
Eric Dumazet98322f22009-01-26 21:35:35 -0800264 /*
265 * force rand to be an odd multiple of UDP_HTABLE_SIZE
266 */
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000267 rand = (rand | 1) * (udptable->mask + 1);
Eric Dumazet5781b232009-12-13 19:32:39 -0800268 last = first + udptable->mask + 1;
269 do {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000270 hslot = udp_hashslot(udptable, net, first);
Eric Dumazet98322f22009-01-26 21:35:35 -0800271 bitmap_zero(bitmap, PORTS_PER_CHAIN);
Eric Dumazet645ca702008-10-29 01:41:45 -0700272 spin_lock_bh(&hslot->lock);
Eric Dumazet98322f22009-01-26 21:35:35 -0800273 udp_lib_lport_inuse(net, snum, hslot, bitmap, sk,
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000274 saddr_comp, udptable->log);
Eric Dumazet98322f22009-01-26 21:35:35 -0800275
276 snum = first;
277 /*
278 * Iterate on all possible values of snum for this hash.
279 * Using steps of an odd multiple of UDP_HTABLE_SIZE
280 * give us randomization and full range coverage.
281 */
Eric Dumazet9088c562008-10-08 11:44:17 -0700282 do {
Eric Dumazet98322f22009-01-26 21:35:35 -0800283 if (low <= snum && snum <= high &&
Amerigo Wange3826f12010-05-05 00:27:06 +0000284 !test_bit(snum >> udptable->log, bitmap) &&
WANG Cong122ff242014-05-12 16:04:53 -0700285 !inet_is_local_reserved_port(net, snum))
Eric Dumazet98322f22009-01-26 21:35:35 -0800286 goto found;
287 snum += rand;
288 } while (snum != first);
289 spin_unlock_bh(&hslot->lock);
Eric Dumazet5781b232009-12-13 19:32:39 -0800290 } while (++first != last);
Eric Dumazet98322f22009-01-26 21:35:35 -0800291 goto fail;
Eric Dumazet645ca702008-10-29 01:41:45 -0700292 } else {
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000293 hslot = udp_hashslot(udptable, net, snum);
Eric Dumazet645ca702008-10-29 01:41:45 -0700294 spin_lock_bh(&hslot->lock);
Eric Dumazet30fff922009-11-09 05:26:33 +0000295 if (hslot->count > 10) {
296 int exist;
297 unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum;
298
299 slot2 &= udptable->mask;
300 hash2_nulladdr &= udptable->mask;
301
302 hslot2 = udp_hashslot2(udptable, slot2);
303 if (hslot->count < hslot2->count)
304 goto scan_primary_hash;
305
306 exist = udp_lib_lport_inuse2(net, snum, hslot2,
307 sk, saddr_comp);
308 if (!exist && (hash2_nulladdr != slot2)) {
309 hslot2 = udp_hashslot2(udptable, hash2_nulladdr);
310 exist = udp_lib_lport_inuse2(net, snum, hslot2,
311 sk, saddr_comp);
312 }
313 if (exist)
314 goto fail_unlock;
315 else
316 goto found;
317 }
318scan_primary_hash:
Eric Dumazetf86dcc52009-10-07 00:37:59 +0000319 if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk,
320 saddr_comp, 0))
Eric Dumazet645ca702008-10-29 01:41:45 -0700321 goto fail_unlock;
322 }
Eric Dumazet98322f22009-01-26 21:35:35 -0800323found:
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000324 inet_sk(sk)->inet_num = snum;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000325 udp_sk(sk)->udp_port_hash = snum;
326 udp_sk(sk)->udp_portaddr_hash ^= snum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 if (sk_unhashed(sk)) {
Craig Galleke32ea7e2016-01-04 17:41:46 -0500328 if (sk->sk_reuseport &&
329 udp_reuseport_add_sock(sk, hslot, saddr_comp)) {
330 inet_sk(sk)->inet_num = 0;
331 udp_sk(sk)->udp_port_hash = 0;
332 udp_sk(sk)->udp_portaddr_hash ^= snum;
333 goto fail_unlock;
334 }
335
Eric Dumazet88ab1932008-11-16 19:39:21 -0800336 sk_nulls_add_node_rcu(sk, &hslot->head);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +0000337 hslot->count++;
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -0700338 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet512615b2009-11-08 10:17:58 +0000339
340 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
341 spin_lock(&hslot2->lock);
342 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
343 &hslot2->head);
344 hslot2->count++;
345 spin_unlock(&hslot2->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 }
Gerrit Renker25030a72006-08-26 20:06:05 -0700347 error = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -0700348fail_unlock:
349 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350fail:
Gerrit Renker25030a72006-08-26 20:06:05 -0700351 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352}
Eric Dumazetc482c562009-07-17 00:26:32 +0000353EXPORT_SYMBOL(udp_lib_get_port);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354
Craig Galleke32ea7e2016-01-04 17:41:46 -0500355/* match_wildcard == true: 0.0.0.0 equals to any IPv4 addresses
356 * match_wildcard == false: addresses must be exactly the same, i.e.
357 * 0.0.0.0 only equals to 0.0.0.0
358 */
359static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2,
360 bool match_wildcard)
David S. Millerdb8dac22008-03-06 16:22:02 -0800361{
362 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
363
Craig Galleke32ea7e2016-01-04 17:41:46 -0500364 if (!ipv6_only_sock(sk2)) {
365 if (inet1->inet_rcv_saddr == inet2->inet_rcv_saddr)
366 return 1;
367 if (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr)
368 return match_wildcard;
369 }
370 return 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800371}
372
Eric Dumazet6eada012015-03-18 14:05:33 -0700373static u32 udp4_portaddr_hash(const struct net *net, __be32 saddr,
374 unsigned int port)
Eric Dumazetd4cada42009-11-08 10:17:30 +0000375{
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700376 return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
Eric Dumazetd4cada42009-11-08 10:17:30 +0000377}
378
Pavel Emelyanov6ba5a3c2008-03-22 16:51:21 -0700379int udp_v4_get_port(struct sock *sk, unsigned short snum)
David S. Millerdb8dac22008-03-06 16:22:02 -0800380{
Eric Dumazet30fff922009-11-09 05:26:33 +0000381 unsigned int hash2_nulladdr =
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700382 udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum);
Eric Dumazet30fff922009-11-09 05:26:33 +0000383 unsigned int hash2_partial =
384 udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0);
385
Eric Dumazetd4cada42009-11-08 10:17:30 +0000386 /* precompute partial secondary hash */
Eric Dumazet30fff922009-11-09 05:26:33 +0000387 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
388 return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr);
David S. Millerdb8dac22008-03-06 16:22:02 -0800389}
390
Joe Perches60c04ae2014-12-01 20:29:06 -0800391static inline int compute_score(struct sock *sk, struct net *net,
392 __be32 saddr, unsigned short hnum, __be16 sport,
393 __be32 daddr, __be16 dport, int dif)
Eric Dumazet645ca702008-10-29 01:41:45 -0700394{
Joe Perches60c04ae2014-12-01 20:29:06 -0800395 int score;
396 struct inet_sock *inet;
Eric Dumazet645ca702008-10-29 01:41:45 -0700397
Joe Perches60c04ae2014-12-01 20:29:06 -0800398 if (!net_eq(sock_net(sk), net) ||
399 udp_sk(sk)->udp_port_hash != hnum ||
400 ipv6_only_sock(sk))
401 return -1;
Eric Dumazet645ca702008-10-29 01:41:45 -0700402
Joe Perches60c04ae2014-12-01 20:29:06 -0800403 score = (sk->sk_family == PF_INET) ? 2 : 1;
404 inet = inet_sk(sk);
405
406 if (inet->inet_rcv_saddr) {
407 if (inet->inet_rcv_saddr != daddr)
408 return -1;
409 score += 4;
Eric Dumazet645ca702008-10-29 01:41:45 -0700410 }
Joe Perches60c04ae2014-12-01 20:29:06 -0800411
412 if (inet->inet_daddr) {
413 if (inet->inet_daddr != saddr)
414 return -1;
415 score += 4;
416 }
417
418 if (inet->inet_dport) {
419 if (inet->inet_dport != sport)
420 return -1;
421 score += 4;
422 }
423
424 if (sk->sk_bound_dev_if) {
425 if (sk->sk_bound_dev_if != dif)
426 return -1;
427 score += 4;
428 }
Eric Dumazet70da2682015-10-08 19:33:21 -0700429 if (sk->sk_incoming_cpu == raw_smp_processor_id())
430 score++;
Eric Dumazet645ca702008-10-29 01:41:45 -0700431 return score;
432}
433
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000434/*
435 * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num)
436 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000437static inline int compute_score2(struct sock *sk, struct net *net,
438 __be32 saddr, __be16 sport,
439 __be32 daddr, unsigned int hnum, int dif)
440{
Joe Perches60c04ae2014-12-01 20:29:06 -0800441 int score;
442 struct inet_sock *inet;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000443
Joe Perches60c04ae2014-12-01 20:29:06 -0800444 if (!net_eq(sock_net(sk), net) ||
445 ipv6_only_sock(sk))
446 return -1;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000447
Joe Perches60c04ae2014-12-01 20:29:06 -0800448 inet = inet_sk(sk);
449
450 if (inet->inet_rcv_saddr != daddr ||
451 inet->inet_num != hnum)
452 return -1;
453
454 score = (sk->sk_family == PF_INET) ? 2 : 1;
455
456 if (inet->inet_daddr) {
457 if (inet->inet_daddr != saddr)
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000458 return -1;
Joe Perches60c04ae2014-12-01 20:29:06 -0800459 score += 4;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000460 }
Joe Perches60c04ae2014-12-01 20:29:06 -0800461
462 if (inet->inet_dport) {
463 if (inet->inet_dport != sport)
464 return -1;
465 score += 4;
466 }
467
468 if (sk->sk_bound_dev_if) {
469 if (sk->sk_bound_dev_if != dif)
470 return -1;
471 score += 4;
472 }
473
Eric Dumazet70da2682015-10-08 19:33:21 -0700474 if (sk->sk_incoming_cpu == raw_smp_processor_id())
475 score++;
476
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000477 return score;
478}
479
Eric Dumazet6eada012015-03-18 14:05:33 -0700480static u32 udp_ehashfn(const struct net *net, const __be32 laddr,
481 const __u16 lport, const __be32 faddr,
482 const __be16 fport)
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200483{
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200484 static u32 udp_ehash_secret __read_mostly;
485
486 net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret));
487
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200488 return __inet_ehashfn(laddr, lport, faddr, fport,
Hannes Frederic Sowa1bbdcee2013-10-19 21:48:57 +0200489 udp_ehash_secret + net_hash_mix(net));
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200490}
491
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000492/* called with read_rcu_lock() */
493static struct sock *udp4_lib_lookup2(struct net *net,
494 __be32 saddr, __be16 sport,
495 __be32 daddr, unsigned int hnum, int dif,
Craig Gallek11341582016-01-05 15:08:07 -0500496 struct udp_hslot *hslot2, unsigned int slot2,
497 struct sk_buff *skb)
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000498{
499 struct sock *sk, *result;
500 struct hlist_nulls_node *node;
Tom Herbertba418fa2013-01-22 09:50:32 +0000501 int score, badness, matches = 0, reuseport = 0;
Eric Dumazeted0dfff2016-01-19 08:36:43 -0800502 bool select_ok = true;
Tom Herbertba418fa2013-01-22 09:50:32 +0000503 u32 hash = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000504
505begin:
506 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000507 badness = 0;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000508 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
509 score = compute_score2(sk, net, saddr, sport,
510 daddr, hnum, dif);
511 if (score > badness) {
512 result = sk;
513 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000514 reuseport = sk->sk_reuseport;
515 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200516 hash = udp_ehashfn(net, daddr, hnum,
517 saddr, sport);
Eric Dumazeted0dfff2016-01-19 08:36:43 -0800518 if (select_ok) {
519 struct sock *sk2;
520
521 sk2 = reuseport_select_sock(sk, hash, skb,
522 sizeof(struct udphdr));
523 if (sk2) {
524 result = sk2;
525 select_ok = false;
526 goto found;
527 }
Craig Galleke32ea7e2016-01-04 17:41:46 -0500528 }
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);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000536 }
537 }
538 /*
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 */
543 if (get_nulls_value(node) != slot2)
544 goto begin;
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000545 if (result) {
Craig Galleke32ea7e2016-01-04 17:41:46 -0500546found:
Eric Dumazetc31504d2010-11-15 19:58:26 +0000547 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000548 result = NULL;
549 else if (unlikely(compute_score2(result, net, saddr, sport,
550 daddr, hnum, dif) < badness)) {
551 sock_put(result);
552 goto begin;
553 }
554 }
555 return result;
556}
557
David S. Millerdb8dac22008-03-06 16:22:02 -0800558/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
559 * harder than this. -DaveM
560 */
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000561struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
David S. Millerdb8dac22008-03-06 16:22:02 -0800562 __be16 sport, __be32 daddr, __be16 dport,
Craig Gallek538950a2016-01-04 17:41:47 -0500563 int dif, struct udp_table *udptable, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800564{
Eric Dumazet271b72c2008-10-29 02:11:14 -0700565 struct sock *sk, *result;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800566 struct hlist_nulls_node *node;
David S. Millerdb8dac22008-03-06 16:22:02 -0800567 unsigned short hnum = ntohs(dport);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000568 unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
569 struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
Tom Herbertba418fa2013-01-22 09:50:32 +0000570 int score, badness, matches = 0, reuseport = 0;
Eric Dumazeted0dfff2016-01-19 08:36:43 -0800571 bool select_ok = true;
Tom Herbertba418fa2013-01-22 09:50:32 +0000572 u32 hash = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800573
Eric Dumazet271b72c2008-10-29 02:11:14 -0700574 rcu_read_lock();
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000575 if (hslot->count > 10) {
576 hash2 = udp4_portaddr_hash(net, daddr, hnum);
577 slot2 = hash2 & udptable->mask;
578 hslot2 = &udptable->hash2[slot2];
579 if (hslot->count < hslot2->count)
580 goto begin;
581
582 result = udp4_lib_lookup2(net, saddr, sport,
583 daddr, hnum, dif,
Craig Gallek11341582016-01-05 15:08:07 -0500584 hslot2, slot2, skb);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000585 if (!result) {
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700586 hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000587 slot2 = hash2 & udptable->mask;
588 hslot2 = &udptable->hash2[slot2];
589 if (hslot->count < hslot2->count)
590 goto begin;
591
Jorge Boncompte [DTI2]1223c672010-04-08 04:56:48 +0000592 result = udp4_lib_lookup2(net, saddr, sport,
Eric Dumazet0eae88f2010-04-20 19:06:52 -0700593 htonl(INADDR_ANY), hnum, dif,
Craig Gallek11341582016-01-05 15:08:07 -0500594 hslot2, slot2, skb);
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000595 }
596 rcu_read_unlock();
597 return result;
598 }
Eric Dumazet271b72c2008-10-29 02:11:14 -0700599begin:
600 result = NULL;
Tom Herbertba418fa2013-01-22 09:50:32 +0000601 badness = 0;
Eric Dumazet88ab1932008-11-16 19:39:21 -0800602 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
Eric Dumazet645ca702008-10-29 01:41:45 -0700603 score = compute_score(sk, net, saddr, hnum, sport,
604 daddr, dport, dif);
605 if (score > badness) {
606 result = sk;
607 badness = score;
Tom Herbertba418fa2013-01-22 09:50:32 +0000608 reuseport = sk->sk_reuseport;
609 if (reuseport) {
Hannes Frederic Sowa65cd8032013-10-19 21:48:51 +0200610 hash = udp_ehashfn(net, daddr, hnum,
611 saddr, sport);
Eric Dumazeted0dfff2016-01-19 08:36:43 -0800612 if (select_ok) {
613 struct sock *sk2;
614
615 sk2 = reuseport_select_sock(sk, hash, skb,
Craig Gallek538950a2016-01-04 17:41:47 -0500616 sizeof(struct udphdr));
Eric Dumazeted0dfff2016-01-19 08:36:43 -0800617 if (sk2) {
618 result = sk2;
619 select_ok = false;
620 goto found;
621 }
Craig Galleke32ea7e2016-01-04 17:41:46 -0500622 }
Tom Herbertba418fa2013-01-22 09:50:32 +0000623 matches = 1;
624 }
625 } else if (score == badness && reuseport) {
626 matches++;
Daniel Borkmann8fc54f62014-08-23 20:58:54 +0200627 if (reciprocal_scale(hash, matches) == 0)
Tom Herbertba418fa2013-01-22 09:50:32 +0000628 result = sk;
629 hash = next_pseudo_random32(hash);
David S. Millerdb8dac22008-03-06 16:22:02 -0800630 }
631 }
Eric Dumazet88ab1932008-11-16 19:39:21 -0800632 /*
633 * if the nulls value we got at the end of this lookup is
634 * not the expected one, we must restart lookup.
635 * We probably met an item that was moved to another chain.
636 */
Eric Dumazet5051ebd2009-11-08 10:18:11 +0000637 if (get_nulls_value(node) != slot)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800638 goto begin;
639
Eric Dumazet271b72c2008-10-29 02:11:14 -0700640 if (result) {
Craig Galleke32ea7e2016-01-04 17:41:46 -0500641found:
Eric Dumazetc31504d2010-11-15 19:58:26 +0000642 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
Eric Dumazet271b72c2008-10-29 02:11:14 -0700643 result = NULL;
644 else if (unlikely(compute_score(result, net, saddr, hnum, sport,
645 daddr, dport, dif) < badness)) {
646 sock_put(result);
647 goto begin;
648 }
649 }
650 rcu_read_unlock();
David S. Millerdb8dac22008-03-06 16:22:02 -0800651 return result;
652}
Pavel Emelyanovfce82332011-12-09 06:23:34 +0000653EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
David S. Millerdb8dac22008-03-06 16:22:02 -0800654
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700655static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
656 __be16 sport, __be16 dport,
Eric Dumazet645ca702008-10-29 01:41:45 -0700657 struct udp_table *udptable)
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700658{
659 const struct iphdr *iph = ip_hdr(skb);
660
Eric Dumazet8afdd992013-12-10 18:07:23 -0800661 return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport,
662 iph->daddr, dport, inet_iif(skb),
Craig Gallek538950a2016-01-04 17:41:47 -0500663 udptable, skb);
KOVACS Krisztian607c4aa2008-10-07 12:38:32 -0700664}
665
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700666struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
667 __be32 daddr, __be16 dport, int dif)
668{
Craig Galleke32ea7e2016-01-04 17:41:46 -0500669 return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif,
Craig Gallek538950a2016-01-04 17:41:47 -0500670 &udp_table, NULL);
KOVACS Krisztianbcd41302008-10-01 07:48:10 -0700671}
672EXPORT_SYMBOL_GPL(udp4_lib_lookup);
673
Shawn Bohrer421b3882013-10-07 11:01:39 -0500674static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk,
675 __be16 loc_port, __be32 loc_addr,
676 __be16 rmt_port, __be32 rmt_addr,
677 int dif, unsigned short hnum)
678{
679 struct inet_sock *inet = inet_sk(sk);
680
681 if (!net_eq(sock_net(sk), net) ||
682 udp_sk(sk)->udp_port_hash != hnum ||
683 (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
684 (inet->inet_dport != rmt_port && inet->inet_dport) ||
685 (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) ||
686 ipv6_only_sock(sk) ||
687 (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
688 return false;
689 if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif))
690 return false;
691 return true;
692}
693
David S. Millerdb8dac22008-03-06 16:22:02 -0800694/*
695 * This routine is called by the ICMP module when it gets some
696 * sort of error condition. If err < 0 then the socket should
697 * be closed and the error returned to the user. If err > 0
698 * it's just the icmp type << 8 | icmp code.
699 * Header points to the ip header of the error packet. We move
700 * on past this. Then (as it used to claim before adjustment)
701 * header points to the first 8 bytes of the udp header. We need
702 * to find the appropriate port.
703 */
704
Eric Dumazet645ca702008-10-29 01:41:45 -0700705void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
David S. Millerdb8dac22008-03-06 16:22:02 -0800706{
707 struct inet_sock *inet;
Eric Dumazetb71d1d42011-04-22 04:53:02 +0000708 const struct iphdr *iph = (const struct iphdr *)skb->data;
Eric Dumazetc482c562009-07-17 00:26:32 +0000709 struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
David S. Millerdb8dac22008-03-06 16:22:02 -0800710 const int type = icmp_hdr(skb)->type;
711 const int code = icmp_hdr(skb)->code;
712 struct sock *sk;
713 int harderr;
714 int err;
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700715 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -0800716
Pavel Emelyanovfd54d712008-07-14 23:01:40 -0700717 sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
Craig Gallek538950a2016-01-04 17:41:47 -0500718 iph->saddr, uh->source, skb->dev->ifindex, udptable,
719 NULL);
Ian Morris51456b22015-04-03 09:17:26 +0100720 if (!sk) {
Pavel Emelyanovdcfc23c2008-07-14 23:03:00 -0700721 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
David S. Millerdb8dac22008-03-06 16:22:02 -0800722 return; /* No socket for error */
723 }
724
725 err = 0;
726 harderr = 0;
727 inet = inet_sk(sk);
728
729 switch (type) {
730 default:
731 case ICMP_TIME_EXCEEDED:
732 err = EHOSTUNREACH;
733 break;
734 case ICMP_SOURCE_QUENCH:
735 goto out;
736 case ICMP_PARAMETERPROB:
737 err = EPROTO;
738 harderr = 1;
739 break;
740 case ICMP_DEST_UNREACH:
741 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
David S. Miller36393392012-06-14 22:21:46 -0700742 ipv4_sk_update_pmtu(skb, sk, info);
David S. Millerdb8dac22008-03-06 16:22:02 -0800743 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
744 err = EMSGSIZE;
745 harderr = 1;
746 break;
747 }
748 goto out;
749 }
750 err = EHOSTUNREACH;
751 if (code <= NR_ICMP_UNREACH) {
752 harderr = icmp_err_convert[code].fatal;
753 err = icmp_err_convert[code].errno;
754 }
755 break;
David S. Miller55be7a92012-07-11 21:27:49 -0700756 case ICMP_REDIRECT:
757 ipv4_sk_redirect(skb, sk);
Duan Jiong1a462d12013-09-20 18:20:28 +0800758 goto out;
David S. Millerdb8dac22008-03-06 16:22:02 -0800759 }
760
761 /*
762 * RFC1122: OK. Passes ICMP errors back to application, as per
763 * 4.1.3.3.
764 */
765 if (!inet->recverr) {
766 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
767 goto out;
Eric Dumazetb1faf562010-05-31 23:44:05 -0700768 } else
Eric Dumazetc482c562009-07-17 00:26:32 +0000769 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
Eric Dumazetb1faf562010-05-31 23:44:05 -0700770
David S. Millerdb8dac22008-03-06 16:22:02 -0800771 sk->sk_err = err;
772 sk->sk_error_report(sk);
773out:
774 sock_put(sk);
775}
776
777void udp_err(struct sk_buff *skb, u32 info)
778{
Eric Dumazet645ca702008-10-29 01:41:45 -0700779 __udp4_lib_err(skb, info, &udp_table);
David S. Millerdb8dac22008-03-06 16:22:02 -0800780}
781
782/*
783 * Throw away all pending data and cancel the corking. Socket is locked.
784 */
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400785void udp_flush_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800786{
787 struct udp_sock *up = udp_sk(sk);
788
789 if (up->pending) {
790 up->len = 0;
791 up->pending = 0;
792 ip_flush_pending_frames(sk);
793 }
794}
Denis V. Lunev36d926b2008-06-04 15:49:07 +0400795EXPORT_SYMBOL(udp_flush_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800796
797/**
Herbert Xuf6b96642011-03-01 02:36:48 +0000798 * udp4_hwcsum - handle outgoing HW checksumming
David S. Millerdb8dac22008-03-06 16:22:02 -0800799 * @skb: sk_buff containing the filled-in UDP header
800 * (checksum field must be zeroed out)
Herbert Xuf6b96642011-03-01 02:36:48 +0000801 * @src: source IP address
802 * @dst: destination IP address
David S. Millerdb8dac22008-03-06 16:22:02 -0800803 */
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200804void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
David S. Millerdb8dac22008-03-06 16:22:02 -0800805{
David S. Millerdb8dac22008-03-06 16:22:02 -0800806 struct udphdr *uh = udp_hdr(skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000807 int offset = skb_transport_offset(skb);
808 int len = skb->len - offset;
809 int hlen = len;
David S. Millerdb8dac22008-03-06 16:22:02 -0800810 __wsum csum = 0;
811
WANG Congebbe4952014-06-02 16:12:02 -0700812 if (!skb_has_frag_list(skb)) {
David S. Millerdb8dac22008-03-06 16:22:02 -0800813 /*
814 * Only one fragment on the socket.
815 */
816 skb->csum_start = skb_transport_header(skb) - skb->head;
817 skb->csum_offset = offsetof(struct udphdr, check);
Herbert Xuf6b96642011-03-01 02:36:48 +0000818 uh->check = ~csum_tcpudp_magic(src, dst, len,
819 IPPROTO_UDP, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -0800820 } else {
WANG Congebbe4952014-06-02 16:12:02 -0700821 struct sk_buff *frags;
822
David S. Millerdb8dac22008-03-06 16:22:02 -0800823 /*
824 * HW-checksum won't work as there are two or more
825 * fragments on the socket so that all csums of sk_buffs
826 * should be together
827 */
WANG Congebbe4952014-06-02 16:12:02 -0700828 skb_walk_frags(skb, frags) {
Herbert Xuf6b96642011-03-01 02:36:48 +0000829 csum = csum_add(csum, frags->csum);
830 hlen -= frags->len;
WANG Congebbe4952014-06-02 16:12:02 -0700831 }
David S. Millerdb8dac22008-03-06 16:22:02 -0800832
Herbert Xuf6b96642011-03-01 02:36:48 +0000833 csum = skb_checksum(skb, offset, hlen, csum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800834 skb->ip_summed = CHECKSUM_NONE;
835
David S. Millerdb8dac22008-03-06 16:22:02 -0800836 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
837 if (uh->check == 0)
838 uh->check = CSUM_MANGLED_0;
839 }
840}
Thomas Grafc26bf4a2013-07-25 18:12:18 +0200841EXPORT_SYMBOL_GPL(udp4_hwcsum);
David S. Millerdb8dac22008-03-06 16:22:02 -0800842
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700843/* Function to set UDP checksum for an IPv4 UDP packet. This is intended
844 * for the simple case like when setting the checksum for a UDP tunnel.
845 */
846void udp_set_csum(bool nocheck, struct sk_buff *skb,
847 __be32 saddr, __be32 daddr, int len)
848{
849 struct udphdr *uh = udp_hdr(skb);
850
851 if (nocheck)
852 uh->check = 0;
853 else if (skb_is_gso(skb))
854 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
855 else if (skb_dst(skb) && skb_dst(skb)->dev &&
Tom Herbertc8cd0982015-12-14 11:19:44 -0800856 (skb_dst(skb)->dev->features &
857 (NETIF_F_IP_CSUM | NETIF_F_HW_CSUM))) {
Tom Herbertaf5fcba2014-06-04 17:19:48 -0700858
859 BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL);
860
861 skb->ip_summed = CHECKSUM_PARTIAL;
862 skb->csum_start = skb_transport_header(skb) - skb->head;
863 skb->csum_offset = offsetof(struct udphdr, check);
864 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
865 } else {
866 __wsum csum;
867
868 BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL);
869
870 uh->check = 0;
871 csum = skb_checksum(skb, 0, len, 0);
872 uh->check = udp_v4_check(len, saddr, daddr, csum);
873 if (uh->check == 0)
874 uh->check = CSUM_MANGLED_0;
875
876 skb->ip_summed = CHECKSUM_UNNECESSARY;
877 }
878}
879EXPORT_SYMBOL(udp_set_csum);
880
David S. Miller79ab0532011-05-09 13:31:04 -0700881static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
Herbert Xuf6b96642011-03-01 02:36:48 +0000882{
883 struct sock *sk = skb->sk;
884 struct inet_sock *inet = inet_sk(sk);
885 struct udphdr *uh;
Herbert Xuf6b96642011-03-01 02:36:48 +0000886 int err = 0;
887 int is_udplite = IS_UDPLITE(sk);
888 int offset = skb_transport_offset(skb);
889 int len = skb->len - offset;
890 __wsum csum = 0;
891
892 /*
893 * Create a UDP header
894 */
895 uh = udp_hdr(skb);
896 uh->source = inet->inet_sport;
David S. Miller79ab0532011-05-09 13:31:04 -0700897 uh->dest = fl4->fl4_dport;
Herbert Xuf6b96642011-03-01 02:36:48 +0000898 uh->len = htons(len);
899 uh->check = 0;
900
901 if (is_udplite) /* UDP-Lite */
902 csum = udplite_csum(skb);
903
Tom Herbert28448b82014-05-23 08:47:19 -0700904 else if (sk->sk_no_check_tx) { /* UDP csum disabled */
Herbert Xuf6b96642011-03-01 02:36:48 +0000905
906 skb->ip_summed = CHECKSUM_NONE;
907 goto send;
908
909 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
910
David S. Miller79ab0532011-05-09 13:31:04 -0700911 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
Herbert Xuf6b96642011-03-01 02:36:48 +0000912 goto send;
913
914 } else
915 csum = udp_csum(skb);
916
917 /* add protocol-dependent pseudo-header */
David S. Miller79ab0532011-05-09 13:31:04 -0700918 uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
Herbert Xuf6b96642011-03-01 02:36:48 +0000919 sk->sk_protocol, csum);
920 if (uh->check == 0)
921 uh->check = CSUM_MANGLED_0;
922
923send:
Eric Dumazetb5ec8ee2012-08-10 02:22:47 +0000924 err = ip_send_skb(sock_net(sk), skb);
Herbert Xuf6b96642011-03-01 02:36:48 +0000925 if (err) {
926 if (err == -ENOBUFS && !inet->recverr) {
927 UDP_INC_STATS_USER(sock_net(sk),
928 UDP_MIB_SNDBUFERRORS, is_udplite);
929 err = 0;
930 }
931 } else
932 UDP_INC_STATS_USER(sock_net(sk),
933 UDP_MIB_OUTDATAGRAMS, is_udplite);
934 return err;
935}
936
David S. Millerdb8dac22008-03-06 16:22:02 -0800937/*
938 * Push out all pending data as one UDP datagram. Socket is locked.
939 */
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200940int udp_push_pending_frames(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -0800941{
942 struct udp_sock *up = udp_sk(sk);
943 struct inet_sock *inet = inet_sk(sk);
David S. Millerb6f21b22011-03-12 02:09:18 -0500944 struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800945 struct sk_buff *skb;
David S. Millerdb8dac22008-03-06 16:22:02 -0800946 int err = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800947
David S. Miller77968b72011-05-08 17:12:19 -0700948 skb = ip_finish_skb(sk, fl4);
Herbert Xuf6b96642011-03-01 02:36:48 +0000949 if (!skb)
David S. Millerdb8dac22008-03-06 16:22:02 -0800950 goto out;
951
David S. Miller79ab0532011-05-09 13:31:04 -0700952 err = udp_send_skb(skb, fl4);
David S. Millerdb8dac22008-03-06 16:22:02 -0800953
David S. Millerdb8dac22008-03-06 16:22:02 -0800954out:
955 up->len = 0;
956 up->pending = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -0800957 return err;
958}
Hannes Frederic Sowa8822b642013-07-01 20:21:30 +0200959EXPORT_SYMBOL(udp_push_pending_frames);
David S. Millerdb8dac22008-03-06 16:22:02 -0800960
Ying Xue1b784142015-03-02 15:37:48 +0800961int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
David S. Millerdb8dac22008-03-06 16:22:02 -0800962{
963 struct inet_sock *inet = inet_sk(sk);
964 struct udp_sock *up = udp_sk(sk);
David S. Millere4749952011-05-08 16:38:45 -0700965 struct flowi4 fl4_stack;
David S. Millerb6f21b22011-03-12 02:09:18 -0500966 struct flowi4 *fl4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800967 int ulen = len;
968 struct ipcm_cookie ipc;
969 struct rtable *rt = NULL;
970 int free = 0;
971 int connected = 0;
972 __be32 daddr, faddr, saddr;
973 __be16 dport;
974 u8 tos;
975 int err, is_udplite = IS_UDPLITE(sk);
976 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
977 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
Herbert Xu903ab862011-03-01 02:36:48 +0000978 struct sk_buff *skb;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000979 struct ip_options_data opt_copy;
David S. Millerdb8dac22008-03-06 16:22:02 -0800980
981 if (len > 0xFFFF)
982 return -EMSGSIZE;
983
984 /*
985 * Check the flags.
986 */
987
Eric Dumazetc482c562009-07-17 00:26:32 +0000988 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
David S. Millerdb8dac22008-03-06 16:22:02 -0800989 return -EOPNOTSUPP;
990
991 ipc.opt = NULL;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000992 ipc.tx_flags = 0;
Francesco Fuscoaa661582013-09-24 15:43:09 +0200993 ipc.ttl = 0;
994 ipc.tos = -1;
David S. Millerdb8dac22008-03-06 16:22:02 -0800995
Herbert Xu903ab862011-03-01 02:36:48 +0000996 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
997
David S. Millerf5fca602011-05-08 17:24:10 -0700998 fl4 = &inet->cork.fl.u.ip4;
David S. Millerdb8dac22008-03-06 16:22:02 -0800999 if (up->pending) {
1000 /*
1001 * There are pending frames.
1002 * The socket lock must be held while it's corked.
1003 */
1004 lock_sock(sk);
1005 if (likely(up->pending)) {
1006 if (unlikely(up->pending != AF_INET)) {
1007 release_sock(sk);
1008 return -EINVAL;
1009 }
1010 goto do_append_data;
1011 }
1012 release_sock(sk);
1013 }
1014 ulen += sizeof(struct udphdr);
1015
1016 /*
1017 * Get and verify the address.
1018 */
1019 if (msg->msg_name) {
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001020 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -08001021 if (msg->msg_namelen < sizeof(*usin))
1022 return -EINVAL;
1023 if (usin->sin_family != AF_INET) {
1024 if (usin->sin_family != AF_UNSPEC)
1025 return -EAFNOSUPPORT;
1026 }
1027
1028 daddr = usin->sin_addr.s_addr;
1029 dport = usin->sin_port;
1030 if (dport == 0)
1031 return -EINVAL;
1032 } else {
1033 if (sk->sk_state != TCP_ESTABLISHED)
1034 return -EDESTADDRREQ;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001035 daddr = inet->inet_daddr;
1036 dport = inet->inet_dport;
David S. Millerdb8dac22008-03-06 16:22:02 -08001037 /* Open fast path for connected socket.
1038 Route will not be used, if at least one option is set.
1039 */
1040 connected = 1;
1041 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001042 ipc.addr = inet->inet_saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001043
1044 ipc.oif = sk->sk_bound_dev_if;
Daniel Borkmannbf84a012013-04-14 08:08:13 +00001045
1046 sock_tx_timestamp(sk, &ipc.tx_flags);
1047
David S. Millerdb8dac22008-03-06 16:22:02 -08001048 if (msg->msg_controllen) {
Hannes Frederic Sowac8e6ad02014-02-18 21:38:08 +01001049 err = ip_cmsg_send(sock_net(sk), msg, &ipc,
1050 sk->sk_family == AF_INET6);
Eric Dumazet91948302016-02-04 06:23:28 -08001051 if (unlikely(err)) {
1052 kfree(ipc.opt);
David S. Millerdb8dac22008-03-06 16:22:02 -08001053 return err;
Eric Dumazet91948302016-02-04 06:23:28 -08001054 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001055 if (ipc.opt)
1056 free = 1;
1057 connected = 0;
1058 }
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001059 if (!ipc.opt) {
1060 struct ip_options_rcu *inet_opt;
1061
1062 rcu_read_lock();
1063 inet_opt = rcu_dereference(inet->inet_opt);
1064 if (inet_opt) {
1065 memcpy(&opt_copy, inet_opt,
1066 sizeof(*inet_opt) + inet_opt->opt.optlen);
1067 ipc.opt = &opt_copy.opt;
1068 }
1069 rcu_read_unlock();
1070 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001071
1072 saddr = ipc.addr;
1073 ipc.addr = faddr = daddr;
1074
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001075 if (ipc.opt && ipc.opt->opt.srr) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001076 if (!daddr)
1077 return -EINVAL;
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001078 faddr = ipc.opt->opt.faddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001079 connected = 0;
1080 }
Francesco Fuscoaa661582013-09-24 15:43:09 +02001081 tos = get_rttos(&ipc, inet);
David S. Millerdb8dac22008-03-06 16:22:02 -08001082 if (sock_flag(sk, SOCK_LOCALROUTE) ||
1083 (msg->msg_flags & MSG_DONTROUTE) ||
Eric Dumazetf6d8bd02011-04-21 09:45:37 +00001084 (ipc.opt && ipc.opt->opt.is_strictroute)) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001085 tos |= RTO_ONLINK;
1086 connected = 0;
1087 }
1088
1089 if (ipv4_is_multicast(daddr)) {
1090 if (!ipc.oif)
1091 ipc.oif = inet->mc_index;
1092 if (!saddr)
1093 saddr = inet->mc_addr;
1094 connected = 0;
Erich E. Hoover76e21052012-02-08 09:11:07 +00001095 } else if (!ipc.oif)
1096 ipc.oif = inet->uc_index;
David S. Millerdb8dac22008-03-06 16:22:02 -08001097
1098 if (connected)
Eric Dumazetc482c562009-07-17 00:26:32 +00001099 rt = (struct rtable *)sk_dst_check(sk, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -08001100
Ian Morris51456b22015-04-03 09:17:26 +01001101 if (!rt) {
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -07001102 struct net *net = sock_net(sk);
David Ahern9a24abf2015-08-13 14:59:03 -06001103 __u8 flow_flags = inet_sk_flowi_flags(sk);
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -07001104
David S. Millere4749952011-05-08 16:38:45 -07001105 fl4 = &fl4_stack;
David Ahern9a24abf2015-08-13 14:59:03 -06001106
David S. Millere4749952011-05-08 16:38:45 -07001107 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
David S. Millerc0951cb2011-03-31 04:54:27 -07001108 RT_SCOPE_UNIVERSE, sk->sk_protocol,
David Ahern9a24abf2015-08-13 14:59:03 -06001109 flow_flags,
David S. Millerc0951cb2011-03-31 04:54:27 -07001110 faddr, saddr, dport, inet->inet_sport);
1111
David Ahernb5bdacf2016-01-04 09:09:27 -08001112 if (!saddr && ipc.oif) {
1113 err = l3mdev_get_saddr(net, ipc.oif, fl4);
1114 if (err < 0)
1115 goto out;
1116 }
David Ahern8cbb512c2015-10-05 08:51:26 -07001117
David S. Millere4749952011-05-08 16:38:45 -07001118 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
1119 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -08001120 if (IS_ERR(rt)) {
1121 err = PTR_ERR(rt);
David S. Miller06dc94b2011-03-03 10:38:01 -08001122 rt = NULL;
David S. Millerdb8dac22008-03-06 16:22:02 -08001123 if (err == -ENETUNREACH)
Eric Dumazetf1d8cba2013-11-28 09:51:22 -08001124 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
David S. Millerdb8dac22008-03-06 16:22:02 -08001125 goto out;
1126 }
1127
1128 err = -EACCES;
1129 if ((rt->rt_flags & RTCF_BROADCAST) &&
1130 !sock_flag(sk, SOCK_BROADCAST))
1131 goto out;
1132 if (connected)
Changli Gaod8d1f302010-06-10 23:31:35 -07001133 sk_dst_set(sk, dst_clone(&rt->dst));
David S. Millerdb8dac22008-03-06 16:22:02 -08001134 }
1135
1136 if (msg->msg_flags&MSG_CONFIRM)
1137 goto do_confirm;
1138back_from_confirm:
1139
David S. Millere4749952011-05-08 16:38:45 -07001140 saddr = fl4->saddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001141 if (!ipc.addr)
David S. Millere4749952011-05-08 16:38:45 -07001142 daddr = ipc.addr = fl4->daddr;
David S. Millerdb8dac22008-03-06 16:22:02 -08001143
Herbert Xu903ab862011-03-01 02:36:48 +00001144 /* Lockless fast path for the non-corking case. */
1145 if (!corkreq) {
Al Virof69e6d12014-11-24 13:23:40 -05001146 skb = ip_make_skb(sk, fl4, getfrag, msg, ulen,
Herbert Xu903ab862011-03-01 02:36:48 +00001147 sizeof(struct udphdr), &ipc, &rt,
1148 msg->msg_flags);
1149 err = PTR_ERR(skb);
YOSHIFUJI Hideaki / 吉藤英明50c3a482013-01-22 06:32:49 +00001150 if (!IS_ERR_OR_NULL(skb))
David S. Miller79ab0532011-05-09 13:31:04 -07001151 err = udp_send_skb(skb, fl4);
Herbert Xu903ab862011-03-01 02:36:48 +00001152 goto out;
1153 }
1154
David S. Millerdb8dac22008-03-06 16:22:02 -08001155 lock_sock(sk);
1156 if (unlikely(up->pending)) {
1157 /* The socket is already corked while preparing it. */
1158 /* ... which is an evident application bug. --ANK */
1159 release_sock(sk);
1160
Joe Perchesba7a46f2014-11-11 10:59:17 -08001161 net_dbg_ratelimited("cork app bug 2\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001162 err = -EINVAL;
1163 goto out;
1164 }
1165 /*
1166 * Now cork the socket to pend data.
1167 */
David S. Millerb6f21b22011-03-12 02:09:18 -05001168 fl4 = &inet->cork.fl.u.ip4;
1169 fl4->daddr = daddr;
1170 fl4->saddr = saddr;
David S. Miller9cce96d2011-03-12 03:00:33 -05001171 fl4->fl4_dport = dport;
1172 fl4->fl4_sport = inet->inet_sport;
David S. Millerdb8dac22008-03-06 16:22:02 -08001173 up->pending = AF_INET;
1174
1175do_append_data:
1176 up->len += ulen;
Al Virof69e6d12014-11-24 13:23:40 -05001177 err = ip_append_data(sk, fl4, getfrag, msg, ulen,
David S. Millerf5fca602011-05-08 17:24:10 -07001178 sizeof(struct udphdr), &ipc, &rt,
1179 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001180 if (err)
1181 udp_flush_pending_frames(sk);
1182 else if (!corkreq)
1183 err = udp_push_pending_frames(sk);
1184 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1185 up->pending = 0;
1186 release_sock(sk);
1187
1188out:
1189 ip_rt_put(rt);
1190 if (free)
1191 kfree(ipc.opt);
1192 if (!err)
1193 return len;
1194 /*
1195 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1196 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1197 * we don't have a good statistic (IpOutDiscards but it can be too many
1198 * things). We could add another new stat but at least for now that
1199 * seems like overkill.
1200 */
1201 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001202 UDP_INC_STATS_USER(sock_net(sk),
1203 UDP_MIB_SNDBUFERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001204 }
1205 return err;
1206
1207do_confirm:
Changli Gaod8d1f302010-06-10 23:31:35 -07001208 dst_confirm(&rt->dst);
David S. Millerdb8dac22008-03-06 16:22:02 -08001209 if (!(msg->msg_flags&MSG_PROBE) || len)
1210 goto back_from_confirm;
1211 err = 0;
1212 goto out;
1213}
Eric Dumazetc482c562009-07-17 00:26:32 +00001214EXPORT_SYMBOL(udp_sendmsg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001215
1216int udp_sendpage(struct sock *sk, struct page *page, int offset,
1217 size_t size, int flags)
1218{
David S. Millerf5fca602011-05-08 17:24:10 -07001219 struct inet_sock *inet = inet_sk(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08001220 struct udp_sock *up = udp_sk(sk);
1221 int ret;
1222
Shawn Landdend3f7d562013-11-24 22:36:28 -08001223 if (flags & MSG_SENDPAGE_NOTLAST)
1224 flags |= MSG_MORE;
1225
David S. Millerdb8dac22008-03-06 16:22:02 -08001226 if (!up->pending) {
1227 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
1228
1229 /* Call udp_sendmsg to specify destination address which
1230 * sendpage interface can't pass.
1231 * This will succeed only when the socket is connected.
1232 */
Ying Xue1b784142015-03-02 15:37:48 +08001233 ret = udp_sendmsg(sk, &msg, 0);
David S. Millerdb8dac22008-03-06 16:22:02 -08001234 if (ret < 0)
1235 return ret;
1236 }
1237
1238 lock_sock(sk);
1239
1240 if (unlikely(!up->pending)) {
1241 release_sock(sk);
1242
Joe Perchesba7a46f2014-11-11 10:59:17 -08001243 net_dbg_ratelimited("udp cork app bug 3\n");
David S. Millerdb8dac22008-03-06 16:22:02 -08001244 return -EINVAL;
1245 }
1246
David S. Millerf5fca602011-05-08 17:24:10 -07001247 ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1248 page, offset, size, flags);
David S. Millerdb8dac22008-03-06 16:22:02 -08001249 if (ret == -EOPNOTSUPP) {
1250 release_sock(sk);
1251 return sock_no_sendpage(sk->sk_socket, page, offset,
1252 size, flags);
1253 }
1254 if (ret < 0) {
1255 udp_flush_pending_frames(sk);
1256 goto out;
1257 }
1258
1259 up->len += size;
1260 if (!(up->corkflag || (flags&MSG_MORE)))
1261 ret = udp_push_pending_frames(sk);
1262 if (!ret)
1263 ret = size;
1264out:
1265 release_sock(sk);
1266 return ret;
1267}
1268
Eric Dumazet85584672009-10-09 04:43:40 +00001269/**
1270 * first_packet_length - return length of first packet in receive queue
1271 * @sk: socket
1272 *
1273 * Drops all bad checksum frames, until a valid one is found.
1274 * Returns the length of found skb, or 0 if none is found.
1275 */
1276static unsigned int first_packet_length(struct sock *sk)
1277{
1278 struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1279 struct sk_buff *skb;
1280 unsigned int res;
1281
1282 __skb_queue_head_init(&list_kill);
1283
1284 spin_lock_bh(&rcvq->lock);
1285 while ((skb = skb_peek(rcvq)) != NULL &&
1286 udp_lib_checksum_complete(skb)) {
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001287 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS,
1288 IS_UDPLITE(sk));
Eric Dumazet85584672009-10-09 04:43:40 +00001289 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1290 IS_UDPLITE(sk));
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001291 atomic_inc(&sk->sk_drops);
Eric Dumazet85584672009-10-09 04:43:40 +00001292 __skb_unlink(skb, rcvq);
1293 __skb_queue_tail(&list_kill, skb);
1294 }
1295 res = skb ? skb->len : 0;
1296 spin_unlock_bh(&rcvq->lock);
1297
1298 if (!skb_queue_empty(&list_kill)) {
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001299 bool slow = lock_sock_fast(sk);
1300
Eric Dumazet85584672009-10-09 04:43:40 +00001301 __skb_queue_purge(&list_kill);
1302 sk_mem_reclaim_partial(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001303 unlock_sock_fast(sk, slow);
Eric Dumazet85584672009-10-09 04:43:40 +00001304 }
1305 return res;
1306}
1307
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308/*
1309 * IOCTL requests applicable to the UDP protocol
1310 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001311
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1313{
Stephen Hemminger6516c652007-03-08 20:41:55 -08001314 switch (cmd) {
1315 case SIOCOUTQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 {
Eric Dumazet31e6d362009-06-17 19:05:41 -07001317 int amount = sk_wmem_alloc_get(sk);
1318
Stephen Hemminger6516c652007-03-08 20:41:55 -08001319 return put_user(amount, (int __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 }
Stephen Hemminger6516c652007-03-08 20:41:55 -08001321
1322 case SIOCINQ:
1323 {
Eric Dumazet85584672009-10-09 04:43:40 +00001324 unsigned int amount = first_packet_length(sk);
Stephen Hemminger6516c652007-03-08 20:41:55 -08001325
Eric Dumazet85584672009-10-09 04:43:40 +00001326 if (amount)
Stephen Hemminger6516c652007-03-08 20:41:55 -08001327 /*
1328 * We will only return the amount
1329 * of this packet since that is all
1330 * that will be read.
1331 */
Eric Dumazet85584672009-10-09 04:43:40 +00001332 amount -= sizeof(struct udphdr);
1333
Stephen Hemminger6516c652007-03-08 20:41:55 -08001334 return put_user(amount, (int __user *)arg);
1335 }
1336
1337 default:
1338 return -ENOIOCTLCMD;
1339 }
1340
1341 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342}
Eric Dumazetc482c562009-07-17 00:26:32 +00001343EXPORT_SYMBOL(udp_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344
David S. Millerdb8dac22008-03-06 16:22:02 -08001345/*
1346 * This should be easy, if there is something there we
1347 * return it, otherwise we block.
1348 */
1349
Ying Xue1b784142015-03-02 15:37:48 +08001350int udp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
1351 int flags, int *addr_len)
David S. Millerdb8dac22008-03-06 16:22:02 -08001352{
1353 struct inet_sock *inet = inet_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001354 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
David S. Millerdb8dac22008-03-06 16:22:02 -08001355 struct sk_buff *skb;
David S. Miller59c2cda2011-12-01 14:12:55 -05001356 unsigned int ulen, copied;
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001357 int peeked, off = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001358 int err;
1359 int is_udplite = IS_UDPLITE(sk);
Eric Dumazet197c9492015-12-30 08:51:12 -05001360 bool checksum_valid = false;
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001361 bool slow;
David S. Millerdb8dac22008-03-06 16:22:02 -08001362
David S. Millerdb8dac22008-03-06 16:22:02 -08001363 if (flags & MSG_ERRQUEUE)
Hannes Frederic Sowa85fbaa72013-11-23 00:46:12 +01001364 return ip_recv_error(sk, msg, len, addr_len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001365
1366try_again:
1367 skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00001368 &peeked, &off, &err);
David S. Millerdb8dac22008-03-06 16:22:02 -08001369 if (!skb)
1370 goto out;
1371
1372 ulen = skb->len - sizeof(struct udphdr);
David S. Miller59c2cda2011-12-01 14:12:55 -05001373 copied = len;
1374 if (copied > ulen)
1375 copied = ulen;
1376 else if (copied < ulen)
David S. Millerdb8dac22008-03-06 16:22:02 -08001377 msg->msg_flags |= MSG_TRUNC;
1378
1379 /*
1380 * If checksum is needed at all, try to do it while copying the
1381 * data. If the data is truncated, or if we only want a partial
1382 * coverage checksum (UDP-Lite), do it before the copy.
1383 */
1384
David S. Miller59c2cda2011-12-01 14:12:55 -05001385 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
Eric Dumazet197c9492015-12-30 08:51:12 -05001386 checksum_valid = !udp_lib_checksum_complete(skb);
1387 if (!checksum_valid)
David S. Millerdb8dac22008-03-06 16:22:02 -08001388 goto csum_copy_err;
1389 }
1390
Eric Dumazet197c9492015-12-30 08:51:12 -05001391 if (checksum_valid || skb_csum_unnecessary(skb))
David S. Miller51f3d022014-11-05 16:46:40 -05001392 err = skb_copy_datagram_msg(skb, sizeof(struct udphdr),
1393 msg, copied);
David S. Millerdb8dac22008-03-06 16:22:02 -08001394 else {
Al Viro227158d2014-04-06 18:47:38 -04001395 err = skb_copy_and_csum_datagram_msg(skb, sizeof(struct udphdr),
1396 msg);
David S. Millerdb8dac22008-03-06 16:22:02 -08001397
1398 if (err == -EINVAL)
1399 goto csum_copy_err;
1400 }
1401
Eric Dumazet22911fc2012-06-27 00:23:44 +00001402 if (unlikely(err)) {
1403 trace_kfree_skb(skb, udp_recvmsg);
Eric Dumazet979402b2012-09-05 23:34:44 +00001404 if (!peeked) {
1405 atomic_inc(&sk->sk_drops);
1406 UDP_INC_STATS_USER(sock_net(sk),
1407 UDP_MIB_INERRORS, is_udplite);
1408 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001409 goto out_free;
Eric Dumazet22911fc2012-06-27 00:23:44 +00001410 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001411
1412 if (!peeked)
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001413 UDP_INC_STATS_USER(sock_net(sk),
1414 UDP_MIB_INDATAGRAMS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001415
Neil Horman3b885782009-10-12 13:26:31 -07001416 sock_recv_ts_and_drops(msg, sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001417
1418 /* Copy the address. */
Eric Dumazetc482c562009-07-17 00:26:32 +00001419 if (sin) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001420 sin->sin_family = AF_INET;
1421 sin->sin_port = udp_hdr(skb)->source;
1422 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1423 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
Hannes Frederic Sowabceaa902013-11-18 04:20:45 +01001424 *addr_len = sizeof(*sin);
David S. Millerdb8dac22008-03-06 16:22:02 -08001425 }
1426 if (inet->cmsg_flags)
Tom Herbertad6f9392015-01-05 13:56:17 -08001427 ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr));
David S. Millerdb8dac22008-03-06 16:22:02 -08001428
David S. Miller59c2cda2011-12-01 14:12:55 -05001429 err = copied;
David S. Millerdb8dac22008-03-06 16:22:02 -08001430 if (flags & MSG_TRUNC)
1431 err = ulen;
1432
1433out_free:
Eric Dumazet9d410c72009-10-30 05:03:53 +00001434 skb_free_datagram_locked(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001435out:
1436 return err;
1437
1438csum_copy_err:
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001439 slow = lock_sock_fast(sk);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001440 if (!skb_kill_datagram(sk, skb, flags)) {
1441 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
Pavel Emelyanov629ca232008-07-05 21:18:07 -07001442 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001443 }
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001444 unlock_sock_fast(sk, slow);
David S. Millerdb8dac22008-03-06 16:22:02 -08001445
Eric Dumazetbeb39db2015-05-30 09:16:53 -07001446 /* starting over for a new packet, but check if we need to yield */
1447 cond_resched();
Xufeng Zhang9cfaa8d2011-06-21 10:43:40 +00001448 msg->msg_flags &= ~MSG_TRUNC;
David S. Millerdb8dac22008-03-06 16:22:02 -08001449 goto try_again;
1450}
1451
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452int udp_disconnect(struct sock *sk, int flags)
1453{
1454 struct inet_sock *inet = inet_sk(sk);
1455 /*
1456 * 1003.1g - break association.
1457 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001458
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 sk->sk_state = TCP_CLOSE;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001460 inet->inet_daddr = 0;
1461 inet->inet_dport = 0;
Tom Herbertbdeab992011-08-14 19:45:55 +00001462 sock_rps_reset_rxhash(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 sk->sk_bound_dev_if = 0;
1464 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1465 inet_reset_saddr(sk);
1466
1467 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1468 sk->sk_prot->unhash(sk);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001469 inet->inet_sport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 }
1471 sk_dst_reset(sk);
1472 return 0;
1473}
Eric Dumazetc482c562009-07-17 00:26:32 +00001474EXPORT_SYMBOL(udp_disconnect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475
Eric Dumazet645ca702008-10-29 01:41:45 -07001476void udp_lib_unhash(struct sock *sk)
1477{
Eric Dumazet723b4612008-11-25 13:55:15 -08001478 if (sk_hashed(sk)) {
1479 struct udp_table *udptable = sk->sk_prot->h.udp_table;
Eric Dumazet512615b2009-11-08 10:17:58 +00001480 struct udp_hslot *hslot, *hslot2;
1481
1482 hslot = udp_hashslot(udptable, sock_net(sk),
1483 udp_sk(sk)->udp_port_hash);
1484 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
Eric Dumazet645ca702008-10-29 01:41:45 -07001485
Eric Dumazet723b4612008-11-25 13:55:15 -08001486 spin_lock_bh(&hslot->lock);
Craig Galleke32ea7e2016-01-04 17:41:46 -05001487 if (rcu_access_pointer(sk->sk_reuseport_cb))
1488 reuseport_detach_sock(sk);
Eric Dumazet723b4612008-11-25 13:55:15 -08001489 if (sk_nulls_del_node_init_rcu(sk)) {
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00001490 hslot->count--;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001491 inet_sk(sk)->inet_num = 0;
Eric Dumazet723b4612008-11-25 13:55:15 -08001492 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
Eric Dumazet512615b2009-11-08 10:17:58 +00001493
1494 spin_lock(&hslot2->lock);
1495 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1496 hslot2->count--;
1497 spin_unlock(&hslot2->lock);
Eric Dumazet723b4612008-11-25 13:55:15 -08001498 }
1499 spin_unlock_bh(&hslot->lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07001500 }
Eric Dumazet645ca702008-10-29 01:41:45 -07001501}
1502EXPORT_SYMBOL(udp_lib_unhash);
1503
Eric Dumazet719f8352010-09-08 05:08:44 +00001504/*
1505 * inet_rcv_saddr was changed, we must rehash secondary hash
1506 */
1507void udp_lib_rehash(struct sock *sk, u16 newhash)
1508{
1509 if (sk_hashed(sk)) {
1510 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1511 struct udp_hslot *hslot, *hslot2, *nhslot2;
1512
1513 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1514 nhslot2 = udp_hashslot2(udptable, newhash);
1515 udp_sk(sk)->udp_portaddr_hash = newhash;
Craig Galleke32ea7e2016-01-04 17:41:46 -05001516
1517 if (hslot2 != nhslot2 ||
1518 rcu_access_pointer(sk->sk_reuseport_cb)) {
Eric Dumazet719f8352010-09-08 05:08:44 +00001519 hslot = udp_hashslot(udptable, sock_net(sk),
1520 udp_sk(sk)->udp_port_hash);
1521 /* we must lock primary chain too */
1522 spin_lock_bh(&hslot->lock);
Craig Galleke32ea7e2016-01-04 17:41:46 -05001523 if (rcu_access_pointer(sk->sk_reuseport_cb))
1524 reuseport_detach_sock(sk);
Eric Dumazet719f8352010-09-08 05:08:44 +00001525
Craig Galleke32ea7e2016-01-04 17:41:46 -05001526 if (hslot2 != nhslot2) {
1527 spin_lock(&hslot2->lock);
1528 hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1529 hslot2->count--;
1530 spin_unlock(&hslot2->lock);
Eric Dumazet719f8352010-09-08 05:08:44 +00001531
Craig Galleke32ea7e2016-01-04 17:41:46 -05001532 spin_lock(&nhslot2->lock);
1533 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
1534 &nhslot2->head);
1535 nhslot2->count++;
1536 spin_unlock(&nhslot2->lock);
1537 }
Eric Dumazet719f8352010-09-08 05:08:44 +00001538
1539 spin_unlock_bh(&hslot->lock);
1540 }
1541 }
1542}
1543EXPORT_SYMBOL(udp_lib_rehash);
1544
1545static void udp_v4_rehash(struct sock *sk)
1546{
1547 u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1548 inet_sk(sk)->inet_rcv_saddr,
1549 inet_sk(sk)->inet_num);
1550 udp_lib_rehash(sk, new_hash);
1551}
1552
Herbert Xu93821772008-09-15 11:48:46 -07001553static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1554{
Tom Herbertfec5e652010-04-16 16:01:27 -07001555 int rc;
Herbert Xu93821772008-09-15 11:48:46 -07001556
Shawn Bohrer005ec972013-10-07 11:01:38 -05001557 if (inet_sk(sk)->inet_daddr) {
Tom Herbertbdeab992011-08-14 19:45:55 +00001558 sock_rps_save_rxhash(sk, skb);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001559 sk_mark_napi_id(sk, skb);
Eric Dumazet2c8c56e2014-11-11 05:54:28 -08001560 sk_incoming_cpu_update(sk);
Shawn Bohrer005ec972013-10-07 11:01:38 -05001561 }
Tom Herbertfec5e652010-04-16 16:01:27 -07001562
Eric Dumazetd826eb12011-11-09 07:24:35 +00001563 rc = sock_queue_rcv_skb(sk, skb);
Eric Dumazet766e90372009-10-14 20:40:11 -07001564 if (rc < 0) {
1565 int is_udplite = IS_UDPLITE(sk);
1566
Herbert Xu93821772008-09-15 11:48:46 -07001567 /* Note that an ENOMEM error is charged twice */
Eric Dumazet766e90372009-10-14 20:40:11 -07001568 if (rc == -ENOMEM)
Herbert Xu93821772008-09-15 11:48:46 -07001569 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1570 is_udplite);
Eric Dumazet766e90372009-10-14 20:40:11 -07001571 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1572 kfree_skb(skb);
Satoru Moriya296f7ea2011-06-17 11:58:39 +00001573 trace_udp_fail_queue_rcv_skb(rc, sk);
Eric Dumazet766e90372009-10-14 20:40:11 -07001574 return -1;
Herbert Xu93821772008-09-15 11:48:46 -07001575 }
1576
1577 return 0;
1578
Herbert Xu93821772008-09-15 11:48:46 -07001579}
1580
Eric Dumazet447167b2012-04-11 23:05:28 +00001581static struct static_key udp_encap_needed __read_mostly;
1582void udp_encap_enable(void)
1583{
1584 if (!static_key_enabled(&udp_encap_needed))
1585 static_key_slow_inc(&udp_encap_needed);
1586}
1587EXPORT_SYMBOL(udp_encap_enable);
1588
David S. Millerdb8dac22008-03-06 16:22:02 -08001589/* returns:
1590 * -1: error
1591 * 0: success
1592 * >0: "udp encap" protocol resubmission
1593 *
1594 * Note that in the success and error cases, the skb is assumed to
1595 * have either been requeued or freed.
1596 */
Eric Dumazetc482c562009-07-17 00:26:32 +00001597int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
David S. Millerdb8dac22008-03-06 16:22:02 -08001598{
1599 struct udp_sock *up = udp_sk(sk);
1600 int rc;
1601 int is_udplite = IS_UDPLITE(sk);
1602
1603 /*
1604 * Charge it to the socket, dropping if the queue is full.
1605 */
1606 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1607 goto drop;
1608 nf_reset(skb);
1609
Eric Dumazet447167b2012-04-11 23:05:28 +00001610 if (static_key_false(&udp_encap_needed) && up->encap_type) {
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001611 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1612
David S. Millerdb8dac22008-03-06 16:22:02 -08001613 /*
1614 * This is an encapsulation socket so pass the skb to
1615 * the socket's udp_encap_rcv() hook. Otherwise, just
1616 * fall through and pass this up the UDP socket.
1617 * up->encap_rcv() returns the following value:
1618 * =0 if skb was successfully passed to the encap
1619 * handler or was discarded by it.
1620 * >0 if skb should be passed on to UDP.
1621 * <0 if skb should be resubmitted as proto -N
1622 */
1623
1624 /* if we're overly short, let UDP handle it */
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001625 encap_rcv = ACCESS_ONCE(up->encap_rcv);
Ian Morris00db4122015-04-03 09:17:27 +01001626 if (skb->len > sizeof(struct udphdr) && encap_rcv) {
David S. Millerdb8dac22008-03-06 16:22:02 -08001627 int ret;
1628
Tom Herbert0a809662014-05-07 16:52:39 -07001629 /* Verify checksum before giving to encap */
1630 if (udp_lib_checksum_complete(skb))
1631 goto csum_error;
1632
Eric Dumazet0ad92ad2011-11-01 12:56:59 +00001633 ret = encap_rcv(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001634 if (ret <= 0) {
Pavel Emelyanov02833282008-07-05 21:18:48 -07001635 UDP_INC_STATS_BH(sock_net(sk),
1636 UDP_MIB_INDATAGRAMS,
David S. Millerdb8dac22008-03-06 16:22:02 -08001637 is_udplite);
1638 return -ret;
1639 }
1640 }
1641
1642 /* FALLTHROUGH -- it's a UDP Packet */
1643 }
1644
1645 /*
1646 * UDP-Lite specific tests, ignored on UDP sockets
1647 */
1648 if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1649
1650 /*
1651 * MIB statistics other than incrementing the error count are
1652 * disabled for the following two types of errors: these depend
1653 * on the application settings, not on the functioning of the
1654 * protocol stack as such.
1655 *
1656 * RFC 3828 here recommends (sec 3.3): "There should also be a
1657 * way ... to ... at least let the receiving application block
1658 * delivery of packets with coverage values less than a value
1659 * provided by the application."
1660 */
1661 if (up->pcrlen == 0) { /* full coverage was set */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001662 net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n",
1663 UDP_SKB_CB(skb)->cscov, skb->len);
David S. Millerdb8dac22008-03-06 16:22:02 -08001664 goto drop;
1665 }
1666 /* The next case involves violating the min. coverage requested
1667 * by the receiver. This is subtle: if receiver wants x and x is
1668 * greater than the buffersize/MTU then receiver will complain
1669 * that it wants x while sender emits packets of smaller size y.
1670 * Therefore the above ...()->partial_cov statement is essential.
1671 */
1672 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
Joe Perchesba7a46f2014-11-11 10:59:17 -08001673 net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n",
1674 UDP_SKB_CB(skb)->cscov, up->pcrlen);
David S. Millerdb8dac22008-03-06 16:22:02 -08001675 goto drop;
1676 }
1677 }
1678
Eric Dumazet33d480c2011-08-11 19:30:52 +00001679 if (rcu_access_pointer(sk->sk_filter) &&
1680 udp_lib_checksum_complete(skb))
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001681 goto csum_error;
David S. Millerdb8dac22008-03-06 16:22:02 -08001682
Sorin Dumitru274f4822014-07-22 21:16:51 +03001683 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
James M Leddy3e215c82014-06-25 17:38:13 -04001684 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1685 is_udplite);
Eric Dumazetc3774112010-04-27 15:13:20 -07001686 goto drop;
James M Leddy3e215c82014-06-25 17:38:13 -04001687 }
Eric Dumazetc3774112010-04-27 15:13:20 -07001688
Herbert Xu93821772008-09-15 11:48:46 -07001689 rc = 0;
David S. Millerdb8dac22008-03-06 16:22:02 -08001690
Shawn Bohrerfbf88662013-10-07 11:01:40 -05001691 ipv4_pktinfo_prepare(sk, skb);
Herbert Xu93821772008-09-15 11:48:46 -07001692 bh_lock_sock(sk);
1693 if (!sock_owned_by_user(sk))
1694 rc = __udp_queue_rcv_skb(sk, skb);
Eric Dumazetf545a382012-04-22 23:34:26 +00001695 else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
Zhu Yi55349792010-03-04 18:01:42 +00001696 bh_unlock_sock(sk);
1697 goto drop;
1698 }
Herbert Xu93821772008-09-15 11:48:46 -07001699 bh_unlock_sock(sk);
1700
1701 return rc;
David S. Millerdb8dac22008-03-06 16:22:02 -08001702
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001703csum_error:
1704 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
David S. Millerdb8dac22008-03-06 16:22:02 -08001705drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001706 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
Eric Dumazet8edf19c2009-10-15 00:12:40 +00001707 atomic_inc(&sk->sk_drops);
David S. Millerdb8dac22008-03-06 16:22:02 -08001708 kfree_skb(skb);
1709 return -1;
1710}
1711
Eric Dumazet1240d132009-11-08 10:18:44 +00001712static void flush_stack(struct sock **stack, unsigned int count,
1713 struct sk_buff *skb, unsigned int final)
1714{
1715 unsigned int i;
1716 struct sk_buff *skb1 = NULL;
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001717 struct sock *sk;
Eric Dumazet1240d132009-11-08 10:18:44 +00001718
1719 for (i = 0; i < count; i++) {
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001720 sk = stack[i];
Ian Morris51456b22015-04-03 09:17:26 +01001721 if (likely(!skb1))
Eric Dumazet1240d132009-11-08 10:18:44 +00001722 skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
1723
Eric Dumazetf6b8f322009-11-08 10:20:19 +00001724 if (!skb1) {
1725 atomic_inc(&sk->sk_drops);
1726 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1727 IS_UDPLITE(sk));
1728 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1729 IS_UDPLITE(sk));
1730 }
1731
1732 if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0)
Eric Dumazet1240d132009-11-08 10:18:44 +00001733 skb1 = NULL;
David Held2dc41cf2014-07-15 23:28:32 -04001734
1735 sock_put(sk);
Eric Dumazet1240d132009-11-08 10:18:44 +00001736 }
1737 if (unlikely(skb1))
1738 kfree_skb(skb1);
1739}
1740
Eric Dumazet97502232013-12-11 14:46:51 -08001741/* For TCP sockets, sk_rx_dst is protected by socket lock
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001742 * For UDP, we use xchg() to guard against concurrent changes.
Eric Dumazet97502232013-12-11 14:46:51 -08001743 */
1744static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
Shawn Bohrer421b3882013-10-07 11:01:39 -05001745{
Eric Dumazet97502232013-12-11 14:46:51 -08001746 struct dst_entry *old;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001747
Eric Dumazete47eb5d2013-12-15 10:53:46 -08001748 dst_hold(dst);
1749 old = xchg(&sk->sk_rx_dst, dst);
1750 dst_release(old);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001751}
1752
David S. Millerdb8dac22008-03-06 16:22:02 -08001753/*
1754 * Multicasts and broadcasts go to each listener.
1755 *
Eric Dumazet1240d132009-11-08 10:18:44 +00001756 * Note: called only from the BH handler context.
David S. Millerdb8dac22008-03-06 16:22:02 -08001757 */
Pavel Emelyanove3163492008-06-16 17:12:11 -07001758static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
David S. Millerdb8dac22008-03-06 16:22:02 -08001759 struct udphdr *uh,
1760 __be32 saddr, __be32 daddr,
Rick Jones36cbb242014-11-06 10:37:54 -08001761 struct udp_table *udptable,
1762 int proto)
David S. Millerdb8dac22008-03-06 16:22:02 -08001763{
Eric Dumazet1240d132009-11-08 10:18:44 +00001764 struct sock *sk, *stack[256 / sizeof(struct sock *)];
David Held5cf3d462014-07-15 23:28:31 -04001765 struct hlist_nulls_node *node;
1766 unsigned short hnum = ntohs(uh->dest);
1767 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
1768 int dif = skb->dev->ifindex;
David Held2dc41cf2014-07-15 23:28:32 -04001769 unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node);
1770 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
Rick Jones36cbb242014-11-06 10:37:54 -08001771 bool inner_flushed = false;
David Held2dc41cf2014-07-15 23:28:32 -04001772
1773 if (use_hash2) {
1774 hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) &
1775 udp_table.mask;
1776 hash2 = udp4_portaddr_hash(net, daddr, hnum) & udp_table.mask;
1777start_lookup:
1778 hslot = &udp_table.hash2[hash2];
1779 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
1780 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001781
Eric Dumazet645ca702008-10-29 01:41:45 -07001782 spin_lock(&hslot->lock);
David Held2dc41cf2014-07-15 23:28:32 -04001783 sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) {
David Held5cf3d462014-07-15 23:28:31 -04001784 if (__udp_is_mcast_sock(net, sk,
1785 uh->dest, daddr,
1786 uh->source, saddr,
1787 dif, hnum)) {
1788 if (unlikely(count == ARRAY_SIZE(stack))) {
1789 flush_stack(stack, count, skb, ~0);
Rick Jones36cbb242014-11-06 10:37:54 -08001790 inner_flushed = true;
David Held5cf3d462014-07-15 23:28:31 -04001791 count = 0;
1792 }
1793 stack[count++] = sk;
David Held2dc41cf2014-07-15 23:28:32 -04001794 sock_hold(sk);
Eric Dumazet1240d132009-11-08 10:18:44 +00001795 }
1796 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001797
Eric Dumazet645ca702008-10-29 01:41:45 -07001798 spin_unlock(&hslot->lock);
Eric Dumazet1240d132009-11-08 10:18:44 +00001799
David Held2dc41cf2014-07-15 23:28:32 -04001800 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
1801 if (use_hash2 && hash2 != hash2_any) {
1802 hash2 = hash2_any;
1803 goto start_lookup;
1804 }
1805
Eric Dumazet1240d132009-11-08 10:18:44 +00001806 /*
1807 * do the slow work with no lock held
1808 */
1809 if (count) {
1810 flush_stack(stack, count, skb, count - 1);
Eric Dumazet1240d132009-11-08 10:18:44 +00001811 } else {
Rick Jones36cbb242014-11-06 10:37:54 -08001812 if (!inner_flushed)
1813 UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI,
1814 proto == IPPROTO_UDPLITE);
1815 consume_skb(skb);
Eric Dumazet1240d132009-11-08 10:18:44 +00001816 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001817 return 0;
1818}
1819
1820/* Initialize UDP checksum. If exited with zero value (success),
1821 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1822 * Otherwise, csum completion requires chacksumming packet body,
1823 * including udp header and folding it to skb->csum.
1824 */
1825static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1826 int proto)
1827{
David S. Millerdb8dac22008-03-06 16:22:02 -08001828 int err;
1829
1830 UDP_SKB_CB(skb)->partial_cov = 0;
1831 UDP_SKB_CB(skb)->cscov = skb->len;
1832
1833 if (proto == IPPROTO_UDPLITE) {
1834 err = udplite_checksum_init(skb, uh);
1835 if (err)
1836 return err;
1837 }
1838
Tom Herberted70fcf2014-05-02 16:29:38 -07001839 return skb_checksum_init_zero_check(skb, proto, uh->check,
1840 inet_compute_pseudo);
David S. Millerdb8dac22008-03-06 16:22:02 -08001841}
1842
1843/*
1844 * All we need to do is get the socket, and then do a checksum.
1845 */
1846
Eric Dumazet645ca702008-10-29 01:41:45 -07001847int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
David S. Millerdb8dac22008-03-06 16:22:02 -08001848 int proto)
1849{
1850 struct sock *sk;
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001851 struct udphdr *uh;
David S. Millerdb8dac22008-03-06 16:22:02 -08001852 unsigned short ulen;
Eric Dumazetadf30902009-06-02 05:19:30 +00001853 struct rtable *rt = skb_rtable(skb);
Jesper Dangaard Brouer2783ef22009-02-06 01:59:12 -08001854 __be32 saddr, daddr;
Pavel Emelyanov02833282008-07-05 21:18:48 -07001855 struct net *net = dev_net(skb->dev);
David S. Millerdb8dac22008-03-06 16:22:02 -08001856
1857 /*
1858 * Validate the packet.
1859 */
1860 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1861 goto drop; /* No space for header. */
1862
Jesper Dangaard Brouer7b5e56f2009-02-05 15:05:45 -08001863 uh = udp_hdr(skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001864 ulen = ntohs(uh->len);
Bjørn Morkccc2d972010-05-06 03:44:34 +00001865 saddr = ip_hdr(skb)->saddr;
1866 daddr = ip_hdr(skb)->daddr;
1867
David S. Millerdb8dac22008-03-06 16:22:02 -08001868 if (ulen > skb->len)
1869 goto short_packet;
1870
1871 if (proto == IPPROTO_UDP) {
1872 /* UDP validates ulen. */
1873 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1874 goto short_packet;
1875 uh = udp_hdr(skb);
1876 }
1877
1878 if (udp4_csum_init(skb, uh, proto))
1879 goto csum_error;
1880
Eric Dumazet8afdd992013-12-10 18:07:23 -08001881 sk = skb_steal_sock(skb);
1882 if (sk) {
Eric Dumazet97502232013-12-11 14:46:51 -08001883 struct dst_entry *dst = skb_dst(skb);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001884 int ret;
David S. Millerdb8dac22008-03-06 16:22:02 -08001885
Eric Dumazet97502232013-12-11 14:46:51 -08001886 if (unlikely(sk->sk_rx_dst != dst))
1887 udp_sk_rx_dst_set(sk, dst);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001888
1889 ret = udp_queue_rcv_skb(sk, skb);
Eric Dumazet8afdd992013-12-10 18:07:23 -08001890 sock_put(sk);
Shawn Bohrer421b3882013-10-07 11:01:39 -05001891 /* a return value > 0 means to resubmit the input, but
1892 * it wants the return to be -protocol, or 0
1893 */
1894 if (ret > 0)
1895 return -ret;
1896 return 0;
Shawn Bohrer421b3882013-10-07 11:01:39 -05001897 }
David S. Millerdb8dac22008-03-06 16:22:02 -08001898
Fabian Frederickc18450a2014-11-04 20:48:41 +01001899 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1900 return __udp4_lib_mcast_deliver(net, skb, uh,
Rick Jones36cbb242014-11-06 10:37:54 -08001901 saddr, daddr, udptable, proto);
Fabian Frederickc18450a2014-11-04 20:48:41 +01001902
1903 sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
Ian Morris00db4122015-04-03 09:17:27 +01001904 if (sk) {
Eliezer Tamira5b50472013-06-10 11:40:00 +03001905 int ret;
1906
Tom Herbert224d0192015-01-05 13:56:14 -08001907 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
Tom Herbert2abb7cd2014-08-31 15:12:43 -07001908 skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
1909 inet_compute_pseudo);
1910
Eliezer Tamira5b50472013-06-10 11:40:00 +03001911 ret = udp_queue_rcv_skb(sk, skb);
David S. Millerdb8dac22008-03-06 16:22:02 -08001912 sock_put(sk);
1913
1914 /* a return value > 0 means to resubmit the input, but
1915 * it wants the return to be -protocol, or 0
1916 */
1917 if (ret > 0)
1918 return -ret;
1919 return 0;
1920 }
1921
1922 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1923 goto drop;
1924 nf_reset(skb);
1925
1926 /* No socket. Drop packet silently, if checksum is wrong */
1927 if (udp_lib_checksum_complete(skb))
1928 goto csum_error;
1929
Pavel Emelyanov02833282008-07-05 21:18:48 -07001930 UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001931 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1932
1933 /*
1934 * Hmm. We got an UDP packet to a port to which we
1935 * don't wanna listen. Ignore it.
1936 */
1937 kfree_skb(skb);
1938 return 0;
1939
1940short_packet:
Joe Perchesba7a46f2014-11-11 10:59:17 -08001941 net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
1942 proto == IPPROTO_UDPLITE ? "Lite" : "",
1943 &saddr, ntohs(uh->source),
1944 ulen, skb->len,
1945 &daddr, ntohs(uh->dest));
David S. Millerdb8dac22008-03-06 16:22:02 -08001946 goto drop;
1947
1948csum_error:
1949 /*
1950 * RFC1122: OK. Discards the bad packet silently (as far as
1951 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1952 */
Joe Perchesba7a46f2014-11-11 10:59:17 -08001953 net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
1954 proto == IPPROTO_UDPLITE ? "Lite" : "",
1955 &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
1956 ulen);
Eric Dumazet6a5dc9e2013-04-29 08:39:56 +00001957 UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001958drop:
Pavel Emelyanov02833282008-07-05 21:18:48 -07001959 UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
David S. Millerdb8dac22008-03-06 16:22:02 -08001960 kfree_skb(skb);
1961 return 0;
1962}
1963
Shawn Bohrer421b3882013-10-07 11:01:39 -05001964/* We can only early demux multicast if there is a single matching socket.
1965 * If more than one socket found returns NULL
1966 */
1967static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net,
1968 __be16 loc_port, __be32 loc_addr,
1969 __be16 rmt_port, __be32 rmt_addr,
1970 int dif)
1971{
1972 struct sock *sk, *result;
1973 struct hlist_nulls_node *node;
1974 unsigned short hnum = ntohs(loc_port);
1975 unsigned int count, slot = udp_hashfn(net, hnum, udp_table.mask);
1976 struct udp_hslot *hslot = &udp_table.hash[slot];
1977
Eric Dumazet63c6f812014-06-12 16:13:06 -07001978 /* Do not bother scanning a too big list */
1979 if (hslot->count > 10)
1980 return NULL;
1981
Shawn Bohrer421b3882013-10-07 11:01:39 -05001982 rcu_read_lock();
1983begin:
1984 count = 0;
1985 result = NULL;
1986 sk_nulls_for_each_rcu(sk, node, &hslot->head) {
1987 if (__udp_is_mcast_sock(net, sk,
1988 loc_port, loc_addr,
1989 rmt_port, rmt_addr,
1990 dif, hnum)) {
1991 result = sk;
1992 ++count;
1993 }
1994 }
1995 /*
1996 * if the nulls value we got at the end of this lookup is
1997 * not the expected one, we must restart lookup.
1998 * We probably met an item that was moved to another chain.
1999 */
2000 if (get_nulls_value(node) != slot)
2001 goto begin;
2002
2003 if (result) {
2004 if (count != 1 ||
2005 unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
2006 result = NULL;
Eric Dumazetf69b9232013-10-08 21:47:29 -07002007 else if (unlikely(!__udp_is_mcast_sock(net, result,
Shawn Bohrer421b3882013-10-07 11:01:39 -05002008 loc_port, loc_addr,
2009 rmt_port, rmt_addr,
2010 dif, hnum))) {
2011 sock_put(result);
2012 result = NULL;
2013 }
2014 }
2015 rcu_read_unlock();
2016 return result;
2017}
2018
2019/* For unicast we should only early demux connected sockets or we can
2020 * break forwarding setups. The chains here can be long so only check
2021 * if the first socket is an exact match and if not move on.
2022 */
2023static struct sock *__udp4_lib_demux_lookup(struct net *net,
2024 __be16 loc_port, __be32 loc_addr,
2025 __be16 rmt_port, __be32 rmt_addr,
2026 int dif)
2027{
2028 struct sock *sk, *result;
2029 struct hlist_nulls_node *node;
2030 unsigned short hnum = ntohs(loc_port);
2031 unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum);
2032 unsigned int slot2 = hash2 & udp_table.mask;
2033 struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
Joe Perchesc7228312014-05-13 20:30:07 -07002034 INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr);
Shawn Bohrer421b3882013-10-07 11:01:39 -05002035 const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
2036
2037 rcu_read_lock();
2038 result = NULL;
2039 udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
2040 if (INET_MATCH(sk, net, acookie,
2041 rmt_addr, loc_addr, ports, dif))
2042 result = sk;
2043 /* Only check first socket in chain */
2044 break;
2045 }
2046
2047 if (result) {
2048 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
2049 result = NULL;
2050 else if (unlikely(!INET_MATCH(sk, net, acookie,
2051 rmt_addr, loc_addr,
2052 ports, dif))) {
2053 sock_put(result);
2054 result = NULL;
2055 }
2056 }
2057 rcu_read_unlock();
2058 return result;
2059}
2060
2061void udp_v4_early_demux(struct sk_buff *skb)
2062{
Eric Dumazet610438b2013-12-11 08:10:05 -08002063 struct net *net = dev_net(skb->dev);
2064 const struct iphdr *iph;
2065 const struct udphdr *uh;
Shawn Bohrer421b3882013-10-07 11:01:39 -05002066 struct sock *sk;
2067 struct dst_entry *dst;
Shawn Bohrer421b3882013-10-07 11:01:39 -05002068 int dif = skb->dev->ifindex;
Shawn Bohrer6e540302015-06-03 16:27:38 -05002069 int ours;
Shawn Bohrer421b3882013-10-07 11:01:39 -05002070
2071 /* validate the packet */
2072 if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr)))
2073 return;
2074
Eric Dumazet610438b2013-12-11 08:10:05 -08002075 iph = ip_hdr(skb);
2076 uh = udp_hdr(skb);
2077
Shawn Bohrer421b3882013-10-07 11:01:39 -05002078 if (skb->pkt_type == PACKET_BROADCAST ||
Shawn Bohrer6e540302015-06-03 16:27:38 -05002079 skb->pkt_type == PACKET_MULTICAST) {
2080 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
2081
2082 if (!in_dev)
2083 return;
2084
2085 ours = ip_check_mc_rcu(in_dev, iph->daddr, iph->saddr,
2086 iph->protocol);
2087 if (!ours)
2088 return;
Shawn Bohrer421b3882013-10-07 11:01:39 -05002089 sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr,
2090 uh->source, iph->saddr, dif);
Shawn Bohrer6e540302015-06-03 16:27:38 -05002091 } else if (skb->pkt_type == PACKET_HOST) {
Shawn Bohrer421b3882013-10-07 11:01:39 -05002092 sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr,
2093 uh->source, iph->saddr, dif);
Shawn Bohrer6e540302015-06-03 16:27:38 -05002094 } else {
Shawn Bohrer421b3882013-10-07 11:01:39 -05002095 return;
Shawn Bohrer6e540302015-06-03 16:27:38 -05002096 }
Shawn Bohrer421b3882013-10-07 11:01:39 -05002097
2098 if (!sk)
2099 return;
2100
2101 skb->sk = sk;
Alexander Duyck82eabd92014-09-04 13:32:11 -04002102 skb->destructor = sock_efree;
Eric Dumazet10e2eb82015-08-01 12:14:33 +02002103 dst = READ_ONCE(sk->sk_rx_dst);
Shawn Bohrer421b3882013-10-07 11:01:39 -05002104
2105 if (dst)
2106 dst = dst_check(dst, 0);
Eric Dumazet10e2eb82015-08-01 12:14:33 +02002107 if (dst) {
2108 /* DST_NOCACHE can not be used without taking a reference */
2109 if (dst->flags & DST_NOCACHE) {
2110 if (likely(atomic_inc_not_zero(&dst->__refcnt)))
2111 skb_dst_set(skb, dst);
2112 } else {
2113 skb_dst_set_noref(skb, dst);
2114 }
2115 }
Shawn Bohrer421b3882013-10-07 11:01:39 -05002116}
2117
David S. Millerdb8dac22008-03-06 16:22:02 -08002118int udp_rcv(struct sk_buff *skb)
2119{
Eric Dumazet645ca702008-10-29 01:41:45 -07002120 return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
David S. Millerdb8dac22008-03-06 16:22:02 -08002121}
2122
Brian Haley7d06b2e2008-06-14 17:04:49 -07002123void udp_destroy_sock(struct sock *sk)
David S. Millerdb8dac22008-03-06 16:22:02 -08002124{
Tom Parkin44046a52013-03-19 06:11:12 +00002125 struct udp_sock *up = udp_sk(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00002126 bool slow = lock_sock_fast(sk);
David S. Millerdb8dac22008-03-06 16:22:02 -08002127 udp_flush_pending_frames(sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00002128 unlock_sock_fast(sk, slow);
Tom Parkin44046a52013-03-19 06:11:12 +00002129 if (static_key_false(&udp_encap_needed) && up->encap_type) {
2130 void (*encap_destroy)(struct sock *sk);
2131 encap_destroy = ACCESS_ONCE(up->encap_destroy);
2132 if (encap_destroy)
2133 encap_destroy(sk);
2134 }
David S. Millerdb8dac22008-03-06 16:22:02 -08002135}
2136
Linus Torvalds1da177e2005-04-16 15:20:36 -07002137/*
2138 * Socket option code for UDP
2139 */
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002140int udp_lib_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002141 char __user *optval, unsigned int optlen,
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002142 int (*push_pending_frames)(struct sock *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143{
2144 struct udp_sock *up = udp_sk(sk);
Tom Herbert1c194482014-05-23 08:47:32 -07002145 int val, valbool;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146 int err = 0;
Wang Chenb2bf1e22007-12-03 22:34:16 +11002147 int is_udplite = IS_UDPLITE(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148
Eric Dumazetc482c562009-07-17 00:26:32 +00002149 if (optlen < sizeof(int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150 return -EINVAL;
2151
2152 if (get_user(val, (int __user *)optval))
2153 return -EFAULT;
2154
Tom Herbert1c194482014-05-23 08:47:32 -07002155 valbool = val ? 1 : 0;
2156
Stephen Hemminger6516c652007-03-08 20:41:55 -08002157 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002158 case UDP_CORK:
2159 if (val != 0) {
2160 up->corkflag = 1;
2161 } else {
2162 up->corkflag = 0;
2163 lock_sock(sk);
Joe Perches4243cdc2014-11-11 21:59:20 -08002164 push_pending_frames(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165 release_sock(sk);
2166 }
2167 break;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002168
Linus Torvalds1da177e2005-04-16 15:20:36 -07002169 case UDP_ENCAP:
2170 switch (val) {
2171 case 0:
2172 case UDP_ENCAP_ESPINUDP:
2173 case UDP_ENCAP_ESPINUDP_NON_IKE:
James Chapman067b2072007-07-05 17:08:05 -07002174 up->encap_rcv = xfrm4_udp_encap_rcv;
2175 /* FALLTHROUGH */
James Chapman342f0232007-06-27 15:37:46 -07002176 case UDP_ENCAP_L2TPINUDP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177 up->encap_type = val;
Eric Dumazet447167b2012-04-11 23:05:28 +00002178 udp_encap_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179 break;
2180 default:
2181 err = -ENOPROTOOPT;
2182 break;
2183 }
2184 break;
2185
Tom Herbert1c194482014-05-23 08:47:32 -07002186 case UDP_NO_CHECK6_TX:
2187 up->no_check6_tx = valbool;
2188 break;
2189
2190 case UDP_NO_CHECK6_RX:
2191 up->no_check6_rx = valbool;
2192 break;
2193
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002194 /*
2195 * UDP-Lite's partial checksum coverage (RFC 3828).
2196 */
2197 /* The sender sets actual checksum coverage length via this option.
2198 * The case coverage > packet length is handled by send module. */
2199 case UDPLITE_SEND_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002200 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002201 return -ENOPROTOOPT;
2202 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
2203 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002204 else if (val > USHRT_MAX)
2205 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002206 up->pcslen = val;
2207 up->pcflag |= UDPLITE_SEND_CC;
2208 break;
2209
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002210 /* The receiver specifies a minimum checksum coverage value. To make
2211 * sense, this should be set to at least 8 (as done below). If zero is
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002212 * used, this again means full checksum coverage. */
2213 case UDPLITE_RECV_CSCOV:
Wang Chenb2bf1e22007-12-03 22:34:16 +11002214 if (!is_udplite) /* Disable the option on UDP sockets */
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002215 return -ENOPROTOOPT;
2216 if (val != 0 && val < 8) /* Avoid silly minimal values. */
2217 val = 8;
Alexey Dobriyan4be929b2010-05-24 14:33:03 -07002218 else if (val > USHRT_MAX)
2219 val = USHRT_MAX;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002220 up->pcrlen = val;
2221 up->pcflag |= UDPLITE_RECV_CC;
2222 break;
2223
Linus Torvalds1da177e2005-04-16 15:20:36 -07002224 default:
2225 err = -ENOPROTOOPT;
2226 break;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002227 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002228
2229 return err;
2230}
Eric Dumazetc482c562009-07-17 00:26:32 +00002231EXPORT_SYMBOL(udp_lib_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232
David S. Millerdb8dac22008-03-06 16:22:02 -08002233int udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002234 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002235{
2236 if (level == SOL_UDP || level == SOL_UDPLITE)
2237 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2238 udp_push_pending_frames);
2239 return ip_setsockopt(sk, level, optname, optval, optlen);
2240}
2241
2242#ifdef CONFIG_COMPAT
2243int compat_udp_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002244 char __user *optval, unsigned int optlen)
David S. Millerdb8dac22008-03-06 16:22:02 -08002245{
2246 if (level == SOL_UDP || level == SOL_UDPLITE)
2247 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2248 udp_push_pending_frames);
2249 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
2250}
2251#endif
2252
Gerrit Renker4c0a6cb2006-11-27 09:29:59 -08002253int udp_lib_getsockopt(struct sock *sk, int level, int optname,
2254 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002255{
2256 struct udp_sock *up = udp_sk(sk);
2257 int val, len;
2258
Eric Dumazetc482c562009-07-17 00:26:32 +00002259 if (get_user(len, optlen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002260 return -EFAULT;
2261
2262 len = min_t(unsigned int, len, sizeof(int));
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002263
Stephen Hemminger6516c652007-03-08 20:41:55 -08002264 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002265 return -EINVAL;
2266
Stephen Hemminger6516c652007-03-08 20:41:55 -08002267 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268 case UDP_CORK:
2269 val = up->corkflag;
2270 break;
2271
2272 case UDP_ENCAP:
2273 val = up->encap_type;
2274 break;
2275
Tom Herbert1c194482014-05-23 08:47:32 -07002276 case UDP_NO_CHECK6_TX:
2277 val = up->no_check6_tx;
2278 break;
2279
2280 case UDP_NO_CHECK6_RX:
2281 val = up->no_check6_rx;
2282 break;
2283
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002284 /* The following two cannot be changed on UDP sockets, the return is
2285 * always 0 (which corresponds to the full checksum coverage of UDP). */
2286 case UDPLITE_SEND_CSCOV:
2287 val = up->pcslen;
2288 break;
2289
2290 case UDPLITE_RECV_CSCOV:
2291 val = up->pcrlen;
2292 break;
2293
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294 default:
2295 return -ENOPROTOOPT;
Stephen Hemminger6516c652007-03-08 20:41:55 -08002296 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297
Stephen Hemminger6516c652007-03-08 20:41:55 -08002298 if (put_user(len, optlen))
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002299 return -EFAULT;
Eric Dumazetc482c562009-07-17 00:26:32 +00002300 if (copy_to_user(optval, &val, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002301 return -EFAULT;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002302 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002303}
Eric Dumazetc482c562009-07-17 00:26:32 +00002304EXPORT_SYMBOL(udp_lib_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305
David S. Millerdb8dac22008-03-06 16:22:02 -08002306int udp_getsockopt(struct sock *sk, int level, int optname,
2307 char __user *optval, int __user *optlen)
2308{
2309 if (level == SOL_UDP || level == SOL_UDPLITE)
2310 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2311 return ip_getsockopt(sk, level, optname, optval, optlen);
2312}
2313
2314#ifdef CONFIG_COMPAT
2315int compat_udp_getsockopt(struct sock *sk, int level, int optname,
2316 char __user *optval, int __user *optlen)
2317{
2318 if (level == SOL_UDP || level == SOL_UDPLITE)
2319 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2320 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
2321}
2322#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323/**
2324 * udp_poll - wait for a UDP event.
2325 * @file - file struct
2326 * @sock - socket
2327 * @wait - poll table
2328 *
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002329 * This is same as datagram poll, except for the special case of
Linus Torvalds1da177e2005-04-16 15:20:36 -07002330 * blocking sockets. If application is using a blocking fd
2331 * and a packet with checksum error is in the queue;
2332 * then it could get return from select indicating data available
2333 * but then block when reading it. Add special case code
2334 * to work around these arguably broken applications.
2335 */
2336unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
2337{
2338 unsigned int mask = datagram_poll(file, sock, wait);
2339 struct sock *sk = sock->sk;
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08002340
David Majnemerc3f1dba2013-05-31 13:15:38 +00002341 sock_rps_record_flow(sk);
2342
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343 /* Check for false positives due to checksum errors */
Eric Dumazet85584672009-10-09 04:43:40 +00002344 if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
2345 !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk))
2346 mask &= ~(POLLIN | POLLRDNORM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347
2348 return mask;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09002349
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350}
Eric Dumazetc482c562009-07-17 00:26:32 +00002351EXPORT_SYMBOL(udp_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002352
David S. Millerdb8dac22008-03-06 16:22:02 -08002353struct proto udp_prot = {
2354 .name = "UDP",
2355 .owner = THIS_MODULE,
2356 .close = udp_lib_close,
2357 .connect = ip4_datagram_connect,
2358 .disconnect = udp_disconnect,
2359 .ioctl = udp_ioctl,
2360 .destroy = udp_destroy_sock,
2361 .setsockopt = udp_setsockopt,
2362 .getsockopt = udp_getsockopt,
2363 .sendmsg = udp_sendmsg,
2364 .recvmsg = udp_recvmsg,
2365 .sendpage = udp_sendpage,
Herbert Xu93821772008-09-15 11:48:46 -07002366 .backlog_rcv = __udp_queue_rcv_skb,
Steffen Klassert8141ed92013-01-21 02:00:03 +00002367 .release_cb = ip4_datagram_release_cb,
David S. Millerdb8dac22008-03-06 16:22:02 -08002368 .hash = udp_lib_hash,
2369 .unhash = udp_lib_unhash,
Eric Dumazet719f8352010-09-08 05:08:44 +00002370 .rehash = udp_v4_rehash,
David S. Millerdb8dac22008-03-06 16:22:02 -08002371 .get_port = udp_v4_get_port,
2372 .memory_allocated = &udp_memory_allocated,
2373 .sysctl_mem = sysctl_udp_mem,
2374 .sysctl_wmem = &sysctl_udp_wmem_min,
2375 .sysctl_rmem = &sysctl_udp_rmem_min,
2376 .obj_size = sizeof(struct udp_sock),
Eric Dumazet271b72c2008-10-29 02:11:14 -07002377 .slab_flags = SLAB_DESTROY_BY_RCU,
Eric Dumazet645ca702008-10-29 01:41:45 -07002378 .h.udp_table = &udp_table,
David S. Millerdb8dac22008-03-06 16:22:02 -08002379#ifdef CONFIG_COMPAT
2380 .compat_setsockopt = compat_udp_setsockopt,
2381 .compat_getsockopt = compat_udp_getsockopt,
2382#endif
Octavian Purdilafcbdf092010-12-16 14:26:56 -08002383 .clear_sk = sk_prot_clear_portaddr_nulls,
David S. Millerdb8dac22008-03-06 16:22:02 -08002384};
Eric Dumazetc482c562009-07-17 00:26:32 +00002385EXPORT_SYMBOL(udp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002386
2387/* ------------------------------------------------------------------------ */
2388#ifdef CONFIG_PROC_FS
2389
Eric Dumazet645ca702008-10-29 01:41:45 -07002390static struct sock *udp_get_first(struct seq_file *seq, int start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391{
2392 struct sock *sk;
2393 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002394 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002396 for (state->bucket = start; state->bucket <= state->udp_table->mask;
2397 ++state->bucket) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002398 struct hlist_nulls_node *node;
Eric Dumazet645ca702008-10-29 01:41:45 -07002399 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002400
2401 if (hlist_nulls_empty(&hslot->head))
2402 continue;
2403
Eric Dumazet645ca702008-10-29 01:41:45 -07002404 spin_lock_bh(&hslot->lock);
Eric Dumazet88ab1932008-11-16 19:39:21 -08002405 sk_nulls_for_each(sk, node, &hslot->head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002406 if (!net_eq(sock_net(sk), net))
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002407 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408 if (sk->sk_family == state->family)
2409 goto found;
2410 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002411 spin_unlock_bh(&hslot->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002412 }
2413 sk = NULL;
2414found:
2415 return sk;
2416}
2417
2418static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
2419{
2420 struct udp_iter_state *state = seq->private;
Denis V. Lunev6f191ef2008-03-28 18:23:33 -07002421 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002422
2423 do {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002424 sk = sk_nulls_next(sk);
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09002425 } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426
Eric Dumazet645ca702008-10-29 01:41:45 -07002427 if (!sk) {
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002428 if (state->bucket <= state->udp_table->mask)
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002429 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Eric Dumazet645ca702008-10-29 01:41:45 -07002430 return udp_get_first(seq, state->bucket + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002431 }
2432 return sk;
2433}
2434
2435static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2436{
Eric Dumazet645ca702008-10-29 01:41:45 -07002437 struct sock *sk = udp_get_first(seq, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002438
2439 if (sk)
Stephen Hemminger6516c652007-03-08 20:41:55 -08002440 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441 --pos;
2442 return pos ? NULL : sk;
2443}
2444
2445static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2446{
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002447 struct udp_iter_state *state = seq->private;
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002448 state->bucket = MAX_UDP_PORTS;
Vitaly Mayatskikh30842f22009-03-23 15:22:33 -07002449
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002450 return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451}
2452
2453static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2454{
2455 struct sock *sk;
2456
YOSHIFUJI Hideakib50660f2008-03-31 19:38:15 -07002457 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002458 sk = udp_get_idx(seq, 0);
2459 else
2460 sk = udp_get_next(seq, v);
2461
2462 ++*pos;
2463 return sk;
2464}
2465
2466static void udp_seq_stop(struct seq_file *seq, void *v)
2467{
Eric Dumazet645ca702008-10-29 01:41:45 -07002468 struct udp_iter_state *state = seq->private;
2469
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002470 if (state->bucket <= state->udp_table->mask)
Eric Dumazet645ca702008-10-29 01:41:45 -07002471 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002472}
2473
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002474int udp_seq_open(struct inode *inode, struct file *file)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002475{
Al Virod9dda782013-03-31 18:16:14 -04002476 struct udp_seq_afinfo *afinfo = PDE_DATA(inode);
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002477 struct udp_iter_state *s;
2478 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002479
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002480 err = seq_open_net(inode, file, &afinfo->seq_ops,
2481 sizeof(struct udp_iter_state));
2482 if (err < 0)
2483 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002484
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002485 s = ((struct seq_file *)file->private_data)->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002486 s->family = afinfo->family;
Eric Dumazet645ca702008-10-29 01:41:45 -07002487 s->udp_table = afinfo->udp_table;
Denis V. Luneva2be75c2008-03-28 18:25:06 -07002488 return err;
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002489}
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002490EXPORT_SYMBOL(udp_seq_open);
Daniel Lezcanoa91275e2008-03-21 04:11:58 -07002491
Linus Torvalds1da177e2005-04-16 15:20:36 -07002492/* ------------------------------------------------------------------------ */
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002493int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002494{
2495 struct proc_dir_entry *p;
2496 int rc = 0;
2497
Denis V. Lunevdda61922008-03-28 18:24:26 -07002498 afinfo->seq_ops.start = udp_seq_start;
2499 afinfo->seq_ops.next = udp_seq_next;
2500 afinfo->seq_ops.stop = udp_seq_stop;
2501
Denis V. Lunev84841c32008-05-02 04:10:08 -07002502 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002503 afinfo->seq_fops, afinfo);
Denis V. Lunev84841c32008-05-02 04:10:08 -07002504 if (!p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002505 rc = -ENOMEM;
2506 return rc;
2507}
Eric Dumazetc482c562009-07-17 00:26:32 +00002508EXPORT_SYMBOL(udp_proc_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002509
Daniel Lezcano0c96d8c2008-03-21 04:14:17 -07002510void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002511{
Gao fengece31ff2013-02-18 01:34:56 +00002512 remove_proc_entry(afinfo->name, net->proc_net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002513}
Eric Dumazetc482c562009-07-17 00:26:32 +00002514EXPORT_SYMBOL(udp_proc_unregister);
David S. Millerdb8dac22008-03-06 16:22:02 -08002515
2516/* ------------------------------------------------------------------------ */
Pavel Emelyanov5e659e42008-04-24 01:02:16 -07002517static void udp4_format_sock(struct sock *sp, struct seq_file *f,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002518 int bucket)
David S. Millerdb8dac22008-03-06 16:22:02 -08002519{
2520 struct inet_sock *inet = inet_sk(sp);
Eric Dumazetc720c7e2009-10-15 06:30:45 +00002521 __be32 dest = inet->inet_daddr;
2522 __be32 src = inet->inet_rcv_saddr;
2523 __u16 destp = ntohs(inet->inet_dport);
2524 __u16 srcp = ntohs(inet->inet_sport);
David S. Millerdb8dac22008-03-06 16:22:02 -08002525
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002526 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
Tetsuo Handa652586d2013-11-14 14:31:57 -08002527 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
David S. Millerdb8dac22008-03-06 16:22:02 -08002528 bucket, src, srcp, dest, destp, sp->sk_state,
Eric Dumazet31e6d362009-06-17 19:05:41 -07002529 sk_wmem_alloc_get(sp),
2530 sk_rmem_alloc_get(sp),
Eric W. Biedermana7cb5a42012-05-24 01:10:10 -06002531 0, 0L, 0,
2532 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
2533 0, sock_i_ino(sp),
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002534 atomic_read(&sp->sk_refcnt), sp,
Tetsuo Handa652586d2013-11-14 14:31:57 -08002535 atomic_read(&sp->sk_drops));
David S. Millerdb8dac22008-03-06 16:22:02 -08002536}
2537
2538int udp4_seq_show(struct seq_file *seq, void *v)
2539{
Tetsuo Handa652586d2013-11-14 14:31:57 -08002540 seq_setwidth(seq, 127);
David S. Millerdb8dac22008-03-06 16:22:02 -08002541 if (v == SEQ_START_TOKEN)
Tetsuo Handa652586d2013-11-14 14:31:57 -08002542 seq_puts(seq, " sl local_address rem_address st tx_queue "
David S. Millerdb8dac22008-03-06 16:22:02 -08002543 "rx_queue tr tm->when retrnsmt uid timeout "
Eric Dumazetcb61cb92008-06-17 21:04:56 -07002544 "inode ref pointer drops");
David S. Millerdb8dac22008-03-06 16:22:02 -08002545 else {
David S. Millerdb8dac22008-03-06 16:22:02 -08002546 struct udp_iter_state *state = seq->private;
2547
Tetsuo Handa652586d2013-11-14 14:31:57 -08002548 udp4_format_sock(v, seq, state->bucket);
David S. Millerdb8dac22008-03-06 16:22:02 -08002549 }
Tetsuo Handa652586d2013-11-14 14:31:57 -08002550 seq_pad(seq, '\n');
David S. Millerdb8dac22008-03-06 16:22:02 -08002551 return 0;
2552}
2553
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002554static const struct file_operations udp_afinfo_seq_fops = {
2555 .owner = THIS_MODULE,
2556 .open = udp_seq_open,
2557 .read = seq_read,
2558 .llseek = seq_lseek,
2559 .release = seq_release_net
2560};
2561
David S. Millerdb8dac22008-03-06 16:22:02 -08002562/* ------------------------------------------------------------------------ */
David S. Millerdb8dac22008-03-06 16:22:02 -08002563static struct udp_seq_afinfo udp4_seq_afinfo = {
David S. Millerdb8dac22008-03-06 16:22:02 -08002564 .name = "udp",
2565 .family = AF_INET,
Eric Dumazet645ca702008-10-29 01:41:45 -07002566 .udp_table = &udp_table,
Arjan van de Ven73cb88e2011-10-30 06:46:30 +00002567 .seq_fops = &udp_afinfo_seq_fops,
Denis V. Lunevdda61922008-03-28 18:24:26 -07002568 .seq_ops = {
2569 .show = udp4_seq_show,
2570 },
David S. Millerdb8dac22008-03-06 16:22:02 -08002571};
2572
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002573static int __net_init udp4_proc_init_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002574{
2575 return udp_proc_register(net, &udp4_seq_afinfo);
2576}
2577
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002578static void __net_exit udp4_proc_exit_net(struct net *net)
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002579{
2580 udp_proc_unregister(net, &udp4_seq_afinfo);
2581}
2582
2583static struct pernet_operations udp4_net_ops = {
2584 .init = udp4_proc_init_net,
2585 .exit = udp4_proc_exit_net,
2586};
2587
David S. Millerdb8dac22008-03-06 16:22:02 -08002588int __init udp4_proc_init(void)
2589{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002590 return register_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002591}
2592
2593void udp4_proc_exit(void)
2594{
Pavel Emelyanov15439fe2008-03-24 14:53:49 -07002595 unregister_pernet_subsys(&udp4_net_ops);
David S. Millerdb8dac22008-03-06 16:22:02 -08002596}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002597#endif /* CONFIG_PROC_FS */
2598
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002599static __initdata unsigned long uhash_entries;
2600static int __init set_uhash_entries(char *str)
Eric Dumazet645ca702008-10-29 01:41:45 -07002601{
Eldad Zack413c27d2012-05-19 14:13:18 +00002602 ssize_t ret;
2603
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002604 if (!str)
2605 return 0;
Eldad Zack413c27d2012-05-19 14:13:18 +00002606
2607 ret = kstrtoul(str, 0, &uhash_entries);
2608 if (ret)
2609 return 0;
2610
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002611 if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2612 uhash_entries = UDP_HTABLE_SIZE_MIN;
2613 return 1;
2614}
2615__setup("uhash_entries=", set_uhash_entries);
Eric Dumazet645ca702008-10-29 01:41:45 -07002616
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002617void __init udp_table_init(struct udp_table *table, const char *name)
2618{
2619 unsigned int i;
2620
Tim Bird31fe62b2012-05-23 13:33:35 +00002621 table->hash = alloc_large_system_hash(name,
2622 2 * sizeof(struct udp_hslot),
2623 uhash_entries,
2624 21, /* one slot per 2 MB */
2625 0,
2626 &table->log,
2627 &table->mask,
2628 UDP_HTABLE_SIZE_MIN,
2629 64 * 1024);
2630
Eric Dumazet512615b2009-11-08 10:17:58 +00002631 table->hash2 = table->hash + (table->mask + 1);
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002632 for (i = 0; i <= table->mask; i++) {
Eric Dumazet88ab1932008-11-16 19:39:21 -08002633 INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i);
Eric Dumazetfdcc8aa92009-11-08 10:17:05 +00002634 table->hash[i].count = 0;
Eric Dumazet645ca702008-10-29 01:41:45 -07002635 spin_lock_init(&table->hash[i].lock);
2636 }
Eric Dumazet512615b2009-11-08 10:17:58 +00002637 for (i = 0; i <= table->mask; i++) {
2638 INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i);
2639 table->hash2[i].count = 0;
2640 spin_lock_init(&table->hash2[i].lock);
2641 }
Eric Dumazet645ca702008-10-29 01:41:45 -07002642}
2643
Tom Herbert723b8e42015-02-24 09:17:31 -08002644u32 udp_flow_hashrnd(void)
2645{
2646 static u32 hashrnd __read_mostly;
2647
2648 net_get_random_once(&hashrnd, sizeof(hashrnd));
2649
2650 return hashrnd;
2651}
2652EXPORT_SYMBOL(udp_flow_hashrnd);
2653
Hideo Aoki95766ff2007-12-31 00:29:24 -08002654void __init udp_init(void)
2655{
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002656 unsigned long limit;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002657
Eric Dumazetf86dcc52009-10-07 00:37:59 +00002658 udp_table_init(&udp_table, "UDP");
Eric Dumazetf03d78d2011-07-07 00:27:05 -07002659 limit = nr_free_buffer_pages() / 8;
Hideo Aoki95766ff2007-12-31 00:29:24 -08002660 limit = max(limit, 128UL);
2661 sysctl_udp_mem[0] = limit / 4 * 3;
2662 sysctl_udp_mem[1] = limit;
2663 sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2664
2665 sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2666 sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2667}